Fire Safety Journal, 12 (1987) 153 - 164
Initiation of G r a i n D u s t E x p l o s i o n s b y H e a t G e n e r a t e d d u r i n g Single Impact
between Solid Bodies
GEIR H. PEDERSEN and ROLF K. ECKHOFF
Chr. Michelsen Institute, Department of Science and Technology, 5036 Fantoft, Bergen (Norway)
(Received March 3, 1987; in final form May 25, 1987)
SUMMARY
Several independent investigations o f past
dust explosions in the grain, feed and flour
industries in the U.S.A. and Europe arrive at
the conclusion that 'friction sparks' m a y have
been the initiator o f a substantial fraction, up
to 50%, o f all the explosions recorded. How-
ever, this conclusion has to a large e x t e n t been
based on indirect evidence, and essential de-
tails o f the suspected ignition process most
often remained unconfirmed.
In the present investigation a comprehen-
sive series o f single-impact ignition experi-
ments were carried out, by generating explo-
sible dust clouds o f dried maize starch, in the
region o f tangential impact between a moving
body and a stagnant horizontal anvil.
The net energy o f the impact was defined
as the loss o f kinetic energy o f the moving
body during impact. The influence o f the
impact energy on the frequency o f ignition
was determined. It was f o u n d that much
higher energies were required for ignition at
high velocities o f impact than at low ones.
The reason is that any moving object colliding
with a solid surface, reduces the ignition sensi-
tivity o f the dust cloud in the vicinity o f the
impact p o i n t by inducing turbulence.
Details o f the impact and subsequent igni-
tion o f dust clouds were studied using a high-
speed movie camera. The temperatures o f the
metal sparks generated from a range o f differ-
ent metals were measured by a four-wave-
length optical pyrometer. Typical spark tem-
peratures were ~ 1500 - 2 7 0 0 C for mild steel
and ~ 2 0 0 0 - 2 9 0 0 C for titanium.
The overall conclusion from the present
investigation is as follows: steel sparks pro-
duced by single impacts o f net energies o f up
to 20 J, between steel and concrete, steel or
rusty steel, are unlikely to ignite clouds o f
0379-7112/87/$3.50
153
dried maize starch. In the case o f maize starch
o f 10% moisture, ignition is unlikely even
with titanium sparks. Maize starch is amongst
the easiest ignitable dusts encountered in the
grain, feed and flour industry. Also, most
dusts f o u n d in practice will contain some
moisture. Therefore, it seems unlikely that
dust explosions involving dusts o f grain, feed
or flour can be initiated by heat from acciden-
tal single impacts between tramp metal, and
anvils o f metal, corroded metal, stone or con-
crete, unless the net impact energies are much
higher than 20 J.
1. BACKGROUND AND MOTIVATION FOR THE
PRESENT WORK
Whether or not metal sparks or hot spots
from single accidental impacts between solid
bodies can initiate dust explosions, has re-
mained a controversial issue for a long time.
Several attemps [1 - 4] have been made at
resolving the puzzle with reference to the
grain, feed and flour industry by analysing
past accidents with the objective to identify
the ignition sources. A summary is given in
Table 1.
As can be seen, 'friction sparks' are claimed
to play a significant part. If it is further taken
into account that it is often tacitly implied
t h a t a substantial part of the 'unknowns' may
have been initiated by some untraceable
sources such as metal sparks and electrostatic
discharges, the friction spark becomes the
most suspect of all the potential ignition
sources.
The situations in which metal sparks can be
generated in an industrial process plant fall
into two main categories. The first is grinding
and cutting operations, by which continuous,
dense showers of sparks are produced. The
second is single accidental impacts.
Elsevier Sequoia/Printed in The Netherlands
154
TABLE ]
Percentage of dust explosions in the grain, feed and food industry assumed t,) be initiated i)y 'friction sparks" or
unknown sources
Ref. Number of % ignited by
explosions friction sparks
1 535 20
1 128 17
2 91 54
3 137 9 62
4 83 28
The ability of metal sparks or hot spots
from grinding and cutting to ignite dust clouds
has been demonstrated by several researchers.
The experiments by Leuschke and Zehr [5],
published in 1962, are probably the first ones
in which dust clouds were ignited by grinding
wheel metal sparks. However, no clouds of
organic dusts ignited. Zuzuki et al. [6] ignited
different coal dusts using both sparks and hot
spots from a piece of steel in contact with a
grinding wheel rotating at 23 - 47 m/s peri-
pheral velocity, Later Allen and Calcote [7]
conducted similar experiments in which metal
sparks were generated by pressing a steel rod
against a rotating grinding wheel. By retarding
and focusing the spark stream, it was possible
to ignite clouds of natural organic dusts such
as maize starch and wheat grain dusts. An-
other recent, and very comprehensive study is
Ritter's [8] investigation. This work confirm-
ed that metal sparks generated by grinding,
rubbing and repeated impacts can ignite
clouds of several dusts, including cellulose and
lycopodium.
The second situation, in which metal sparks
may be generated in industry, is accidental
single impacts between solid bodies. In this
case relevant evidence of the ignition poten-
tial has been lacking. However, as long as the
necessary conditions for initiation of dust
explosions by such impacts remain unidenti-
fied, one is forced to maintain the hypothesis
that such sparks may be hazardous in general.
This in turn forces industry to take precau-
tions that may be superfluous, and causes fear
that may be unnecessary.
This unsatisfactory state of affairs created
the motivation for the present work which
has been reported in greater detail elsewhere
% unknown ';;~friction Period
sparks
+ unknown
46 66 1860-1973
27 44 1949-1973
18 72 1941 -1945
71 1958 -1975
5 30-35 1965 -1980
[9, 10]. The specific objective has been to
answer the following question: under what
circumstances of impact energy, impacting
velocity and type of moving object and anvil,
can single impacts between solid bodies ini-
tiate a dust explosion in a cloud of grain dust
in air?
2. EXPERIMENTAL APPARATUS
2.1. Impact generator
After having considered various principles
for constructing a single-impact generator
suitable for laboratory work, it was decided
to build the apparatus shown in Fig. 1. The
length of the rigid, spring-loaded arm, the
'hammer', from the centre of rotation to the
tip of the test object, can be varied by means
of an adjustment screw. The condition for
obtaining impact is that the arm is slightly
longer than the perpendicular distance from
the arm axis to the anvil. By careful adjust-
ment of this small difference it is possible to
generate impacts of different impact energies
without changing the impact velocity. The
cylindrical tip of the test objects that impact-
ed on the anvil material had a reduced diam-
eter of 4 mm. Details of the end of the arm
are shown in Fig. 2. Springs of different
strengths were used for generating a range of
impacting velocities of approximately 8 - 25
m/s. Fine adjustment of the velocity was pos-
sible by adjusting the spring tension.
When performing an impact test, the arm is
first twisted to the start position, whereby the
helical tension spring is stretched. In this
position the arm is locked by a pneumatically
operated device that can be released electri-
cally. After having been released, the arm
 155

will be further retarded by the spring, and

subsequently arrested by an arm catcher.

2.2. Dust dispersion system

Considerable work was carried o u t to devel-
SPRING op a suitable technique for generating the dust
DRIVEN
iMPACTING clouds to be used in the experiments. A dis-
ARM "----~
INTER-
persion system based on a modified version of
CHANGEABLE
METALOBJECT the dispersion cup of the Hartmann bomb
DUST ~ [11] was adopted. The dispersion cup was
CLOUD
....~.
covered by a cap in order to deflect the dust
cloud being expelled from the cup, from the
vertical to the horizontal direction. F r o m the
dispersion cup, the dust cloud flowed into a
rectangular channel consisting of two parallel
side walls and a b o t t o m plate. The height of
the side walls was 20 cm, the length 50 cm,
30 cm
and the spacing between t h e m 12 cm. Hence,
the volume of the dust cloud was approxi-
mately 12 litres. The front side wall was made
of transparent plastic, which made it possible
to photograph the impacts and the ignitions.
In order to establish the combination of
q u a n t i t y of dust, dispersion pressure and delay
time, that gave the o p t i m u m conditions for
ignition, the minimum electrical ignition ener-
Fig. 1. Sketch of the test rig. gy of the dust clouds generated in the 12-1
channel, was determined by means of cali-
.,,...I.- Atugm'humprofde brated electric sparks. The various parameters
- - 3 0 - 30,, 2
were adjusted until an optimal combination
that gave dust clouds of the lowest minimum
ignition energy had been found.
i
; / ~ljustment screw
M16 ~ 1 2. 3. Dust concentration probes
Steel msert The concentration of the optimal dust
clouds as a function of time and space was
measured using a light attenuation probe con-
structed by The Chr. Michelsen Institute
/~ ~ Lock,rignut (CMI) on the basis of the work by the US
Bureau of Mines [12]. The probe, which is
~l ~ Con~c~tocWmcJnut
shown in Fig. 3, is described in detail else-
Test object where [13]. By repeating the dust-cloud gen-
eration process with the dust concentration
Fig. 2. Detail of the arm of the test rig.

accelerates until it reaches its m a x i m u m an-

gular velocity close to the 'six o'clock' posi-
tion. In this position the spring exerts no force
upon the arm and the velocity remains con-
stant for a short period of time. At this point,
impact occurs between the test object at the
tip of the arm and the plane, horizontal anvil.
The anvil is fixed in a strong vice m o u n t e d
rigidly on the test rig. After impact the arm Fig. 3. Light attenuation dust concentration probe.
156
probe in various locations, the concentration
development of the dust cloud with time, at
any desired location, could be determined.
2.4. Timing system
A microprocessor-based 4-channel timer
was used to control the system. The timer
controlled the triggering of the impacting arm,
dispersion of the dust, as well as high-speed
movie camera and other photographic equip-
ment.
2.5. Equipment for measuring the velocity of
the arm and the impact energy
The kinetic energy of the arm at any instant
was calculated from its mass distribution and
instantaneous velocity. By measuring the velo-
city of the arm immediately before and after
the impact, i.e., immediately before and after
'6 o'clock' position, the kinetic energy lost in
the impact could be calculated. Two sets of
reflective opto-switches were used for this
purpose. The opto-switches were mounted on
the wall of the test rig, with narrow clearance
to the passing arm. A small mirror was fixed
to the arm, and an electrical signal was gener-
ated when the mirror passed the opto-switches.
2.6. Photographic equipment
Different kinds of photographic equipment,
including ordinary single-frame cameras, video
camera, 16 mm movie camera and 16 mm
high-speed movie camera, were used. The high-
speed camera (Hycam 4043020) was primarily
used to make black and white negative film
recordings at 1000 and 2000 frames per sec-
ond. The films were developed and examined
at CMI immediately after each experiment.
2.7. Temperature measurements by a four-
wavelength optical pyrometer
In order to measure the temperature of the
sparks, a four-wavelength optical pyrometer
was used [14] (see 'Acknowledgements').
The method is based on the determination of
the relative magnitude of the radiation inten-
sities from the investigated solid body at four
different wavelengths. The intensity ratios are
compared with the radiation intensity ratios
from black bodies of various temperatures. By
identifying the black body temperature that
gives the best fit, a best estimate of the 'colour
temperature' of the solid body is obtained.
The 'colour temperature' is close to the real
temperature of the body ff the spectral emis-
sivity is approximately constant in the range
of the spectrum of wavelengths used in the
measurement.
2.8. Determination of minimum electric spark
energies for ignition of dust clouds
These energies were determined by means
of the method described by Eckhoff [15].
This method, which is based on using electric
sparks of known energies (integration of
power-versus-time curves), is essentially iden-
tical with the new method proposed recently
in W. Germany [16].
3. DUSTS USED IN THE EXPERIMENTS
Since maize starch is one of the agricultural
materials that have often been involved in
dust explosions in the U.S.A. and is one of
the most sensitive to ignition, it had been de-
cided to use this material in the experiments.
Native maize starch supplied from Amylum,
Aalst, Belgium, was used. On receipt the
starch contained 10 - 11% moisture, which
increased the minimum ignition energy con-
siderably compared with that of dried starch.
In order to ensure maximum sensitivity to
ignition, the starch used in the majority of
the present experiments was first dried for
four hours at about 105 C to remove the
moisture.
A few other dusts of different minimum
electric ignition energies (see Table 2), were
used in order to investigate how the ignition
probability in impact ignition tests ~}aried
with the minimum electric spark ignition
energy.
4. CONCENTRATION DISTRIBUTION AND
ELECTRIC SPARK IGNITION SENSITIVITY OF
THE EXPERIMENTAL DUST CLOUDS
4.1. Dust cloud concentration distribution
The first phase of the experiments concen-
trated on formation of a suitable experimen-
tal dust cloud of sufficient volume and low
minimum ignition energy. This was achieved
by dispersing 8 g of dried maize starch at 2 bar
overpressure in the compressed air line. With
8 g of dust, the theoretical average dust con-
centration of the cloud in the 12-1 channel was
 157

TABLE 2
M i n i m u m i g n i t i o n energies a n d particle sizes of t h e 1oo I i(
dusts used in the e x p e r i m e n t s
g zooI
Dust Min. ign. e n e r g y Particle
(electric sparks) diameter
( m J) (/Jm)

Maize s t a r c h , < 4.5 ~ 5 - 20

dried
Maize s t a r c h , 27 - 36 ~5 -20
'88oi
i , t , t , i , i , ,
10 - 11%
o 0.5 lo 1.5 ;~.0
moisture
TIME (r,.)
Lycopodium < 4.5 ~30 Fig. 5. Dust c o n c e n t r a t i o n d e v e l o p m e n t as a f u n c t i o n
Barley p r o t e i n ]3 o f t i m e at l o c a t i o n s 1 t o 4 (see Fig. 4). S m o o t h e d
average curves based o n t h r e e o r m o r e r e p e a t e d mea-
Barley starch 18 - 22 s u r e m e n t s at each l o c a t i o n .
Barley fibre 47 - 59
W h e a t dust 36 - 47
means of calibrated electric sparks. The elec-
trodes were mounted horizontally a fraction
of a miUimetre above the surface of the plain
anvil, at the same location as a hot spot would
DU5~
DUSI
D|SP::RSIO!~I
Cl~D be expected to occur and metal sparks be gen-
CUe'
erated in an impact test. The electric spark
I 2 3 4
was passed 120 - 300 ms after starting the
+ + + + dust dispersion. Both dried maize starch and
lycopodium ignited with 100% ignition prob-
t
ANVIL ability even at spark energies as low as 4.5 mJ.
Fig. 4. A l t e r n a t i v e l o c a t i o n s o f t h e d u s t c o n c e n t r a t i o n However, any object travelling through a
p r o b e d u r i n g c o n c e n t r a t i o n m e a s u r e m e n t s in t h e ex- dust cloud at a high velocity will create tur-
plosion c h a n n e l .
bulence in its vicinity. This is also the case in
the test rig when the arm passes through the
667 mg/1. However, the dust concentration dust cloud. It would be expected, therefore,
was not uniform. Therefore, the dust concen- to be more difficult to ignite the dust cloud
tration development had to be measured as when the arm is passing than when the cloud
functions o f both time and location, using the is comparatively quiescent.
optical dust concentration probe mounted in The influence of the moving arm on the
different locations of the cloud. The various ignition sensitivity of the dust cloud in the
positions of the dust concentration probe dur- vicinity of the point of impact was studied b y
ing these measurements are shown in Fig. 4. means of electric sparks of known energies.
The results from the measurements are given The spark electrodes were again mounted at
in Fig. 5. the same point as a hot spot would be expect-
The dust concentration measurements ed to occur in case of an impact test. However,
showed that at 120 - 300 ms after opening the in these tests the arm used to generate impact
solenoid valve, the dust concentration was sparks was passed through the dust cloud as it
most uniform. At this m o m e n t the dust con- would be in an impact test, only with a small
centration varied from approximately 400 to clearance to the electric spark electrodes. The
800 g/m 3. For this reason the timing system electric spark was discharged 0.3 - I 0 ms after
was always adjusted so the impact occurred the moving arm had passed the '6 o'clock'
at this point in time. position. Three different peripheral velocities
of the arm were used, viz., 9.5, 15 and 22 m/s.
4.2. Ignition sensitivity for quiescent and Results showing probabilities of ignition of
turbulent dust clouds clouds of lycopodium as functions of electric
The minimum ignition energies o f t h e du.st spark energy at different arm velocities and
clouds in the 12-1 channel were measured b y delay times, are summarized in Fig. 6. When
158
1.0
0.9
I 5m/,~
O.B
0.'/
~ 0.6
E, o.s
>,-
O.l.
g o.~
c)
R~ 0.2
0.1
0.1 I 10 100 1000
SPARK ENERGY , mJ
Fig. 6. Effect o f i m p a c t velocity o n ignition sensi-
tivity o f l y c o p o d i u m . Delay b e t w e e n i m p a c t a n d trig-
gering o f t h e electric spark is 0.3 - 10 ms. E n v e l o p e s
enclose t h e e x p e r i m e n t a l points.
the peripheral velocity of the arm was 22 m/s,
the electric spark energy necessary to ignite
the turbulent dust cloud was two orders of
magnitude higher than the 4.5 mJ necessary
to ignite the quiescent dust cloud. At 9.5 m/s
the electric spark energy required for ignition
was approximately one order of magnitude
higher. This relative ranking of the ignition
sensitivities of the clouds only applies when
ignition occurs close to the point of impact.
For spark~luster ignition away from the
impact point, it would be necessary to perform
separate electric spark ignition tests with the
spark gap at the location where ignition takes
place, and with the appropriate delay between
passing the impacting arm and firing the elec-
tric spark.
5. T E M P E R A T U R E S O F M E T A L S P A R K S F R O M
MECHANICAL IMPACTS AND GRINDING
Temperatures of single-impact sparks from
ordinary mild steel (St37), alloy steel and tita-
nium were measured by means of the four-
wave optical pyrometer. The anvil materials
used were rusty steel and concrete. For
comparison, temperatures of sparks from mild
steel in contact with a rotating grinding wheel
were also measured. In the case of the single
impacts the temperatures were measured on
sparks 40 ram 'downstream' of the impact
point and 10 mm above the anvil surface.
Temperatures of titanium sparks were also
measured 140 mm and 240 mm downstream
of the impact point. The arm velocity just
prior to impact was 24 m/s and the net impact
energy about 20 J in all the experiments. In
the case of grinding-wheel sparks, tempera-
tures were measured 40 mm downstream of
the contact point between the steel rod and
the grinding wheel. The peripheral velocity of
the grinding wheel was 29 m/s.
The results are summarized in Table 3. Ap-
proximately 10 tests were carried out for each
combination of test object and anvil material,
using different combinations of detecting
wavelengths in the pyrometer.
TABLE 3
Results f r o m t h e spark t e m p e r a t u r e m e a s u r e m e n t s b y
the four-wavelength pyrometer
Test o b j e c t Anvil material Spark
temperatures
C~c)
Titanium R u s t y steel 2030 - 2900
Mild steel, St37 R u s t y steel 1730 - 2 3 8 0
Alloy steel R u s t y steel 1530 - 1980
Mild steel, S t 3 7 Concrete 2530 - 2730
Mild steel, St37 Rotating 1600 - 1730
grinding wheel
The results show that the highest tempera-
tures were measured on sparks produced from
impacts of titanium against rusty steel, and
mild steel against concrete. The temperatures
of titanium sparks at 140 - 240 mm down-
stream of the impact point were generally
higher than at 40 mm. There was a clear ten-
dency of increasing temperature with increas-
ing length of travel, i.e., increasing burning
time.
The measured temperatures of sparks from
mild steel against concrete were also high. At
40 mm from the point of impact, some steel
spark temperatures were even higher than
some titanium spark temperatures. However,
the boiling point of titanium is higher than
that of iron, indicating that on average the
titanium sparks should attain higher tempera-
tures than sparks from steel. This was also
observed.
The results from the tests with mild steel
as spark-producing material showed a large

variation in the spark temperatures, depend-
ing u p o n the anvil material. The lowest spark
temperatures were obtained in the tests with
the rotating grinding wheel. The observed
temperatures of 1600 - 1730 C were very
close to those for mild steels measured b y
Hardt [17 ], and also in agreement with Wahl's
[18] t w o ~ o l o u r p y r o m e t e r measurement of
temperatures of steel sparks from grinding
wheels.
The temperature measurements showed
that impact sparks from mild steel attained
higher temperatures when concrete was used
as wall material than when rusty steel was
used. This may in part be due to the heat
conductivity of concrete being much lower
than that of steel, causing more of the fric-
tional heat generated during an impact to be
retained in the particles. This m a y provide
more favourable initial conditions for com-
plete combustion of the particles and hence
contribute indirectly to a significantly higher
spark temperature.
The observation that the temperatures of
single-impact steel sparks were generally
considerably higher than of steel grinding-
wheel sparks, is interesting. One reason for
this effect could be the substantial difference
in the contact pressure between metal and
anvil during spark formation (see next Sec-
tion). A higher contact pressure could lead to
transfer of more heat to the particles during
their formation. Although the maximum pos-
sible difference between the a m o u n t s of heat
transferred to a particle is small compared
with the maximum possible combustion heat,
it could significantly influence the extent to
which the particle will oxidize completely.
This, in turn, could result in significantly
different spark temperatures.
6. ESTIMATION OF THE IMPACT P R E S S U R E
The sparks from single impacts b e t w e e n
carbon steel and concrete or rusty steel, gen-
erated in the present investigation, attained
higher temperatures than the grinding wheel
sparks. A possible explanation could be that
the maximum compression pressure in the
impact apparatus was much higher than the
pressure of the metal against the grinding
wheel. Ritter [8] found that the spark tem-
perature increases with increasing contact
pressure to a certain point where the spark
159
temperature becomes constant and indepen-
dent of the pressure. A similar observation
was made b y Allen and Calcote [7] who
found that low pressure sparks were relatively
small and irregular in shape. The irregular
shape indicated that the sparks had not
reached the melting point. The high pressure
sparks were, however, spherical and coated
with oxide, which means that the particles
had melted and burnt.
In the present investigation, the maximum
contact pressure was estimated b y measuring
the static force exerted on the anvil by the tip
of the impacting arm, when the arm was left in
the vertical position. The measurement was
accomplished b y mounting a mechanical load
cell between the tip of the arm and the anvil.
The force was distributed over a circular area
of diameter 4 ram. The static pressure at the
contact area as a function of excess in arm
length is given in Fig. 7.
Pressure.
[MPa)
200
100
025 050 075 100 T [mml
Excess Am Length
Fig. 7. Static pressure at the c o n t a c t area between the
test object and the w a l l material as a f u n c t i o n o f dif-
ference between arm length and perpendicular dis-
tanee f r o m arm axis to anvil plane.
The dynamic pressures during impacts may
be somewhat different from those measured
under static conditions. It seems reasonable to
assume, however, that the dynamic peak pres-
sures in the present investigation were consid-
erably higher than the contact pressures re-
proted by Ritter [8] (10 - 15 MPa) and Allen
and Calcote [7] {about 1 MPa). It is worth
mentioning that the yield strengths of carbon
and low-alloy steels are approximately 25 MPa.
7. I N F L U E N C E OF IMPACT ENERGY AND
IMPACT VELOCITY ON IGNITION F R E Q U E N C Y
The influence of the net impact energy and
the peripheral velocity of the arm on the fre-
160
quency of ignition of maize starch by tita-
nium sparks, was studied. The peripheral
velocity of the arm was varied using different
springs, whereas the net impact energy was
varied by adjusting the excess arm length.
The results are given in Fig. 8. The frequen-
cy of ignition for each combination of arm
velocity and impact energy is based on obser-
vations from 10 or 20 tests. The scatter (+1 std.
dev.) is indicated. At constant velocity of
approach, the frequency of ignition decreases
systematically with decreasing impact energy,
as would be expected. Low-energy impacts
produce fewer sparks than high-energy im-
pacts, and if the number of sparks is low, the
ignition probability is also low.
10
09
08,
07
c06
o
c,
~ 0~
03
02
0t
I i i i i i i t
2 z. 6 8 10 12 ~z. 16
Impact energy
Fig. 8. F r e q u e n c y o f ignition o f clouds of maize
s t a r c h as ,~ f u n c t i o n o f i m p a c t energy at 16 m / s and
24 m/s, for impaet~ b e t w e e n t i t a n i u m a n d r u s t y steel.
The impacts of similar net energies, but of
different velocities of approach, i.e., different
turbulence levels, gave decreasing frequency
of ignition with increasing turbulence. This is
in agreement with the results from the experi-
ments where the minimum electric spark ener-
gy for ignition was determined when the arm
was moving through the dust cloud at the
m o m e n t of spark discharge (Section 4.2).
8. H I G H - S P E E D F I L M I N G
The 1000 - 2000 frames per second high-
speed films confirmed that impacts with mild
steel against rusty steel or concrete produced
very few sparks compared with titanium im-
pacts. Typical numbers of sparks produced
from mild steel were 5 - 50. Immediately after
impact, no or very few sparks were visible.
But 3 - 5 ms after impact, some sparks became
visible, and after 10 - 20 ms the number of
sparks visible on the f i l l was at maximum.
Impacts with steel never generated a sufficient
number of sparks to ignite clouds of dried
maize starch.
The films of ignition by sparks from tita-
nium against rusty steel showed that the dust
clouds could be ignited in two different ways.
(Reference 9 contains reproductions of some
of the film pictures.) First, ignition frequently
occurred very close to the point of impact,
immediately (i.e., 0 - 5 ms) after the impact
had taken place. At that m o m e n t an extremely
luminous hemispherical 'pocket' close to the
impact point, with a radius of a few centi-
meters and containing 100 - 1000 sparks, was
observed. In addition to the great number of
discrete particles, a 'thermite' flash probably
also contributed to the formation of the lu-
~ m/s minous pocket. Impacts of titanium against
concrete also generated high spark densities
close to the point of impact, but not the addi-
tional continuous, luminous phase as was seen
with impacts against rust. Flame propagation
started somewhere in the region of the 'lumi-
nous pocket', but could neither be referred
specifically to one single metal particle nor to
i i
18 ~0 the hot spot generated at the point of impact.
All the high-speed films of ignition of maize
starch by titanium impacts against rusty steel
showed that ignition and flame propagation
started in this region.
The second mode of ignition was only ob-
served in some of the experiments with lyco-
podium. In this case ignition occurred 1 0 - 50
ms after impact at a distance of 10 - 30 cm
downstream of the point of impact. In this
region the number of sparks per unit volume
of dust cloud was considerably lower than in
the 'luminous pocket' close to the point of
impact. But still it was not possible to trace
the ignition back to any single metal spark.
'Explosion' and fragmentation of single metal
sparks were observed frequently, but never
seemed to cause ignition of the dust cloud.
The high-speed films suggest that a cluster
of a large number of sparks is necessary to
cause ignition of a cloud of grain dust. Im-
pacts with titanium as spark-producing

material generated such conditions, and usual-
ly resulted in ignition.
9. RESULTS FROM IGNITION TRIALS WITH
DIFFERENT DUST CLOUDS AND IMPACTING
MATERIALS
A series o f impact tests with various mate-
rials was carried o u t in clouds o f dried maize
starch in order to identify combinations o f
impacting materials t hat would cause ignition.
The anvil materials used comprised typical
c o n s t r u ctio n steels with either fresh or rusty
surfaces, and different concrete ~ m p l e s . Sup-
p l e men tar y tests were c o n d u c t e d with anvils
o f different minerals and grinding wheels.
T he steel alloys used in the experiments
were obtained f r o m the CMI workshop. Pre-
cise names or compositions o f the alloys were
n o t available, b u t t h e y are likely to be repre-
sentative o f the range of steels in c o m m o n
use. Th e combinations o f steel test objects
and anvil materials investigated, are given in
Table 4.
TABLE 4
Combinations of steel test objects and anvil materials
used in the experiments
Mild steel against
-- mild steel (fresh surface or rusty surface)
-- concrete (incl. two samples supplied by NGFA)
minerals (granite, gneiss, s'andstone, quartz)
- -
- grinding wheel
-
Acid resistant steel, alloyed steel, hardened steel or
stainless steel against
-- mild steel
-- concrete
-- granite
-- grinding wheel
T he peripheral velocity of t he arm in the
test rig just b ef o r e impact was 24 m/s. T he
calculated average impact energy was a b o u t
20 J. After each test, t he anvil was displaced
slightly, so th at every impact t o o k place on
fresh anvil surface.
T he experiments revealed t h a t neither mild
steel against concrete, nor any o t h e r combina-
tions o f steel objects and anvil materials, pro-
duced an y sign o f ignition o f t he dried maize
starch cloud.
F r o m i n t r o d u c t o r y tests with concrete as
anvil material, it was f o u n d t hat the nature
o f the surface was of decisive i m por t ance for
161
the spark f o r m a t i o n process. Impacts by car-
bon steel against a 'fresh' surface containing
a thin layer of cement, produced no visible
sparks at all. However, if t he c e m e n t layer
was removed, so t h a t the impact t o o k place
on t he gravel or crushed stone fraction of t he
concrete, visible sparks were produced. For
this reason experiments using concrete as
anvil material were always carried o u t after
removal of the cem ent layer.
Ignition experiments with impacts between
titanium and several anvil materials were also
carried out. T he combinations o f materials
used, and t he results f r o m the tests, are given
in Table 5. Dust clouds o f dried maize starch
were used in all tests.
TABLE 5
Results from the experiments with impacts between
titanium and different anvil materials
Titanium against Ignition ?
-- rusty, mild steel Yes
-- concrete Yes
-- sandstone Yes
- granite, gneiss, quartz
- No
- grinding wheel
- No
T he experiments showed t h a t visible sparks
were produced f r o m all combinations of im-
pacts between titanium and anvil materials.
T he n u m b e r o f sparks produced d e p e n d e d
u p o n the surface t e x t u r e of t he anvil material.
A rough surface p r o d u c e d m ore sparks t han a
s m o o t h surface, and ignitions were onl y ob-
tained f r o m impacts against anvil materials
having a relatively rough surface.
T he i m port ance of t he surface properties
of t he anvil material was emphasized during
the experiments with titanium against rust y
steel. Surfaces apparently covered by t he
same kind o f rust, would in some cases pro-
duce large numbers o f powerful sparks t h a t
ignited the dust cloud, whereas in o t h e r cases
considerably fewer sparks were produced and
no ignition occurred.
A few experiments with zirconium as spark-
producing material were also carried out.
Rust y steel was used as anvil material. The
impacts produced m ore sparks than titanium
against rusty steel, and ignition of clouds of
dried maize starch were easily obtained.
162
It is worth mentioning that at the very be-
ginning of the present investigation it was
attempted to generate impact sparks by strik-
ing a test object of standard quality alumi-
nium against a rusty steel anvil. However, no
metal sparks or any other luminous thermal
reaction could be observed, only a thin smear
of aluminium was deposited on top of the
rust at the point of impact. This is contrary
to the frequent assumption that impacts of
aluminium against rust will generate energetic
thermite sparks. It should be added, though,
that thermite sparks may be generated if such
a smear o f aluminium on rusty steel is subse-
quently struck by some other object.
In order to produce more steel sparks than
those being produced by single impacts in the
present test rig, a strong rotating cutting disk
was pressed against a piece o f low carbon steel.
A continuous powerful shower of steel sparks
was generated, but dust clouds of lycopodium
and dried maize starch exposed to this heavy
spark shower did not ignite. The reason for
this has not been fully disclosed. It should be
mentioned, though, that the number of sparks
per unit volume of cloud was considerably
lower than in the 'luminous pocket' generated
by titanium impacts.
Finally, the ignition sensitivity of some
further agricultural dusts were compared with
that of dried maize starch and lycopodium
clouds, using sparks from impacts between
titanium and rusty steel. The peripheral velo-
city of the arm was 24 m/s and the net fric-
tional impact energy approximately 20 J. The
minimum electric ignition energy, Emin of the
dusts were measured by calibrated electric
sparks in the Hartmann tube. The results are
given in Table 6.
The results show that the impact sparks
were only able to ignite the dust clouds of
the lowest minimum electric ignition energies.
Apart from dried maize starch and lycopo-
dium, ignition was only obtained with one
dust, with Emin equal to ~ 13 mJ. Maize starch
with 10 - 11% moisture and Emin 27 - 36 mJ
did not ignite. This fairly consistent correla-
tion of minimum electric spark ignition ener-
gies and the frequency of ignition by impact
sparks should probably not be generalized
beyond the group of natural organic sub-
stances to which all the dusts tested belong.
In general, the relationships are likely to be
more complex [8].
TABLE 6
Results from the ignition tests of dust clouds of dif-
ferent 'minimum ignition energy'
Dust Minimum Frequency
electric of ignition
ignition in impact
energy (mJ) tests (%)
Maize starch, dried .~.4.5 100
Lycopodium .: 4.5 100
Barley protein 13 l0
Barley starch 18 - 22 0
Maize starch, 27 - 36 0
10 - 11% moisture
Barley fibre ,17 - 59 0
10. CONCLUSIONS
(1) Metal sparks or hot spots generated in
single impacts between moving steel objects
and anvils of steel, rusty steel, concrete or
minerals failed to ignite clouds of dried maize
starch in air over the entire range of experi-
mental conditions. (Net impact energies of up
to 20 J and velocities of approach from 10 m/s
tp 25 m/s.) Clouds of dried maize starch have
a minimum electric spark ignition energy of
less than 5 mJ, and are probably among the
most ignition-sensitive dusts encountered in
the grain, feed and flour industries.
(2) However, clouds of dried maize starch
could be ignited by single impacts of titanium
on an anvil of either rusty steel, concrete or
sandstone, but if the maize starch contained
about 10% moisture, it could not be ignited
even by titanium impacts.
(3) Single impacts with steel as spark-
producing material produced a very low num-
ber of sparks as compared to the number pro-
duced by titanium under the same impact con-
ditions. The temperatures of individual steel
sparks, however, could reach the same level as
those of titanium sparks (~2500 C). How-
ever, the temperatures of grinding-wheel steel
sparks ate considerably lower, of the order of
1500 - 2000 C.
(4) Impacts of standard quality aluminium
against rusty steel did not generate any sparks
or any other luminous reaction at all, only a
thin smear of aluminium on top of the rust.
Impacts with harder aluminium-containing
alloys were not investigated.

(5) In most cases, ignition by titanium
sparks was observed very close to the point
of impact. However, ignition was occasionally
observed 10 - 30 cm downstream of the im-
pact point. Ignition by one single metal spark
was never observed. A fairly dense cluster of
sparks seemed to be necessary for igniting
the clouds of maize starch.
(6) A n y moving object in the dust cloud
reduces the ignition sensitivity o f the cloud in
the vicinity of the object by inducing turbu-
lence. Experiments with titanium against rusty
steel showed that at a given net impact energy,
the ignition frequency dropped when the im-
pact velocity increased. Thus, at a given net
impact energy, objects generating low turbu-
lence represent a greater ignition hazard than
objects generating high turbulence.
(7) At a given impact velocity, the frequen-
cy of ignition by titanium sparks decreased
systematically with the impact energy, as
would be expected.
(8) The microscopic nature of the anvil sur-
face is decisive for the spark formation pro-
cess. For example, impacts against the gravel
or crushed stone fraction of concrete pro-
duced considerably more sparks than impacts
against a fresh concrete surface covered with
cement.
(9) The overall practical conclusion of the
present investigation is that up to net impact
energies of 20 J, tangential accidental single
impacts between various types of steel, and
between steel and rusty steel or concrete, are
unable to ignite clouds of grain and feed dust,
or flour, even if the dusts are dry. Impacts of
standard quality aluminium against rusty steel
did not even generate any visible sparks. In
the case of titanium or zirconium, the sparks
produced may initiate explosions in clouds of
dried dusts, but not in clouds of dust contain-
ing 10% moisture or more.
ACKNOWLEDGEMENTS
CMI wishes to express its sincere gratitude
to the National Grain and Feed Association
(NGFA), Washington, DC, U.S.A., for the
substantial financial support rendering this
work possible, and for unfailing constructive
advice and support throughout. This paper is
a condensed version o f a comprehensive
report prepared by CMI for NGFA [9].
163
During his stay at CMI in the a u t u m n of
1985, Dr. R. Klemens, Technical University
of Warsaw, gave invaluable help, both theo-
retically and experimentally, with the mea-
surement of metal spark temperatures by a
four-wavelength optical pyrometer borrowed
from TU of Warsaw.
REFERENCES
1 H. de Goijer, Th. M. Groothhuizen and M. E.
Reinders, Literature Investigation into the Dust
Explosion Danger in Industries Storing and Pro-
cessing Cereals and Flour, Report from Tech-
nological Laboratory TNO, October, 1975.
2 D. J. Rasbash, Fires Involving Dusts, Technical
Booklet No. 5, Fire Protection Association,
London, 1949.
3 M. Verkade and P. Chiotti, Literature Survey of
Dust Explosions in Grain Handling Facilities:
Causes and Prevention, Energy and Mineral Re-
sources Institute, Iowa State University, Ames,
IA, 1976.
4 H. Beck and A. Jeske, Documentation Staub-
explosionen, Analyse und Einzelfallderstellung,
Report No. 4/82, Berufsgenossenschaftliches
Institut fiir Arbeitssicherheit, St. Augustin,
F.R.G., 1982.
5 G. Leuschke and J. Zehr, Ziindung yon Staub-
lagerungen und Staub/LuftGemischen dutch me-
chanisch erzeugte Funken, Arbeitsschutz, 6
(1962) 146.
6 T. Zuzuki, S. Takaoka and S. Fujii, The ignition
of coal dust by rubbing, frictional heat and sparks,
Proc. Restricted Int. Conf. of Directors of Safety
in Mines Research, July 1965, Safety in Mines
Research Establishment, Sheffield, England.
7 J. Allen and H. F. Calcote, Grain Dust Ignition
by Friction Sparks, SMS-81-049, National Grain
and Feed Association, Washington, DC 20005,
1981.
8 K. Ritter, Die Ziindwirksamkeit mechanisch
erzeugter Funken gegeniiber Gas/Luft- und
Staub/Luft-Gemischen, Dr.-Ing. Dissertation, Uni-
versit/it Fridericiana, Karlsruhe, 1984.
9 G. H. Pedersen, Initiation of Dust Explosions by
Heat Generated during Single Impact between
Solid Bodies, Fire and Explosion Research Re-
port No. SMS-86-055, National Grain and Feed
Association, Washington, DC, 1986. (Prepared by
Chr. Michelsen Institute, Report No 833310-2,
December 1985.)
10 G. H. Pedersen and R. K. Eckhoff, Initiation of
dust explosions by heat generated during single
impact between solid bodies, Proc. 2nd Int.
Colloquium in Dust Explosions, Jadwisin, Poland,
November, 1986, Report No. 863351-1, CMI,
1986.
11 R. K. Eckhoff, Use of (alP/dr)max from closed-
bomb tests for predicting violence of accidental
dust explosions in industrial plants, Fire Safety
J., 8 (1984/85) 159. (Paper presented at the First
Int. Colloquium on Explosibility of Industrial
164

Dusts in Baranow, Poland, November 8 - 10, 1984.
12 K. L. CashdoUar, I. Liebman and R. S. Conti,
Three Bureau o f Mines Optical Dust Probes, Re-
port o f Investigations No. 8542, U.S. Bureau of
Mines, 1981.
13 R. K. Eckhoff, K. Fuhre and G. tt. Pedersen,
Vented Maize Starch Explosions in a 236 m 3
Experimental Silo, Fire and Explosions Research
Report No. ESV-86-070, National Grain and
Feed Association, Washington, DC, 1986. (Pre-
pared by Chr. Michelsen Institute, Report No.
843307-2, December, 1985.)
14 K. L. Cashdollar and M. Hertzberg, Infrared
Pyrometers for measuring dust explosion temper-
atures, Optical Eng., 21 (1) {1982) 82 - 86.
15 R. K. Eckhoff, Towards absolute minimum igni-
tion energies for dust clouds?, Combust. Flame,
24 (1975) 53 - 64.
16 W. Berthold, Bestimmung der Mindestziindenergie
yon Staub/Luft-Gemischen, Progress Report No.
134, VDI-Verlag, Diisseldorf, 1987.
17 L. Hardt, Temperaturme~ungen an Schleiffun-
ken, Arbeitsschutz, 15 (1954) 430 - 436.
18 H. Wahl, Z. Angew. Phys., 12 {1960) 60 - 62.

Erratum

A simplified characterization of upholstered furniture heat release rates, by V. Babrauskas and

W. D. Walton, published in Fire Safety Journal, 11 (1986) 181 - 192.

In eqn. (4), [style factor] = 1.0 for plain as p r i n t e d o n p. 1 8 9 at t h e t o p o f c o l u m n

primarily rectilinear construction, not 1.5 2.
 Powder Technology 135 136 (2003) 209 215
www.elsevier.com/locate/powtec

A new passive-type electrostatic eliminator used in a powder storage silo

Tsutomu Kodama a,*, Teruo Suzuki b, Tomohumi Mogami b, Satoru Watano c
a
National Institute of Industrial Safety, 1-4-6 Umezono, Kiyose, Tokyo 204-0024, Japan
b
Kasuga Denki Co. Ltd., Tokyo, Japan
c
Department of Chemical Engineering, Osaka Prefecture University, Osaka, Japan
Received 12 November 2002; received in revised form 29 March 2003; accepted 7 April 2003

Abstract

In order to prevent dust explosions caused by electrostatic discharges in a silo during loading of particulates in a pneumatic transport
system, a passive-type electrostatic eliminator without a power supply has been newly developed. The eliminator consists of nozzle-type
ionizers mounted on the outside of an insulating short pipe that is installed at the end of the loading pipe inside the silo. The ionizer is
equipped with a grounded needle electrode within the insulated metal nozzle. The elimination ability depends on the electrification of the
insulating pipe with the same polarity as the charge on the products during the loading. The nozzle is also electrified due to electrostatic
induction and discharge from the charged insulating pipe and induces corona discharges at the needle electrode. The air ions generated by
corona discharges are blown into the pipe to neutralize the static charge of the products before entering the silo. Improvement of the
eliminator was attempted, and its effectiveness was examined using a real-sized pneumatic powder-transporting facility.
D 2003 Elsevier B.V. All rights reserved.

Keywords: Electrostatic eliminator; Pneumatic powder transport; Silo; Nozzle-type ionizer; Incendiary discharge; Dust explosion

1. Introduction granules coexisting with fine powder are loaded. Recently,

Large amounts of electrostatic charge are accumulated on
powder by friction with the pipe wall in pneumatic powder-
transport systems. Loading such powder into a silo causes
the electric field strength within the silo and may generate
incendiary electrostatic discharges and induce an explosion
of flammable powder when the total electric charge accu-
mulated in the silo generates an electric field higher than the
threshold value able to trigger the ignition of powder with a
concentration higher than the lower explosion limit. Two
types of incendiary discharges may be generated in silos.
One is lightning-like discharge, which may occur within a
large-scale charged powder cloud. The other is bulk surface
discharge or cone discharge, which may occur along the bulk
surface of a charged powder heap. Bulk surface discharges
are likely to cause dust explosions rather than lightning-like
discharges and occur when coarse granules with a diameter
of 1 10 mm are loaded at a high flow rate [1 9]. The
discharge energy is estimated to be approximately 10 mJ.
Therefore, dust explosions in a silo can occur when coarse
* Corresponding author. Tel.: +81-424914512; fax: +81-424917846.
E-mail address: kodama@anken.go.jp (T. Kodama).
however, it has been suggested that the discharge energy
increases with the silo diameter, flow rate, and charge-to-
mass ratio of particles, which may result in the ignition of
powder with a minimum ignition energy of over 10 mJ
[10,11]. It has also been reported that bulk surface discharges
may also ignite sensitive fine powder with a low minimum
ignition energy of 10 mJ or less even if coarse granules are
not present [12]. In order to prevent incendiary discharges in
a silo, it is effective to reduce the static charge on powder by
using electrostatic eliminators [13,14].
We developed a nozzle-type eliminator [15 18] that
consists of nozzle-type ionizers mounted on the outside of
the loading pipe inside the silo. A high voltage of DC or AC
is applied to the insulated needle electrodes within the
grounded metal nozzles of the ionizers to generate corona
discharges. The eliminator to which a DC high voltage was
applied had a problem with reverse charging on the products
with opposite polarity to the charge before the elimination
[15,16]. This requires feedback control of the applied
voltage from the electric charge accumulated in the silo
[17]. On the other hand, applying AC high voltage with a
frequency of 50 tended to eliminate the shortage [15,16].
The elimination ability increased with the increase of the
frequency, which requires feedback control as well [18]. To

0032-5910/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0032-5910(03)00161-X
210 T. Kodama et al. / Powder Technology 135 136 (2003) 209215
achieve safe use in a flammable atmosphere and lower cost,
we developed an eliminator equipped with passive-type
ionizers without a power supply [19]. This paper describes
improvements of this type of eliminator demonstrated by
real-size experiments using polypropylene pellets.
2. Passive-type eliminator
The newly developed passive-type eliminator consists of
six or eight passive-type ionizers mounted in a row on the
outside of a short PVC pipe that is installed at the end of the
Fig. 1. Schema of the passive-type ionizers. (a) Type-A ionizer. (b) Type-B
ionizer.
loading pipe inside the silo. The ionizer is equipped with a
needle electrode grounded through a resistor of 50 MV
within the insulated metal nozzle. The resistor was used to
prevent the occurrence of spark discharges. Each nozzle was
fixed to the PVC pipe wall with a metal fitting and four
M4  8 screws, as shown in Fig. 1. The inside wall of the
PVC pipe was electrified with the same polarity as the charge
on the pellets during the loading, probably due to the electric
discharges from the charged pellets. Consequently, the metal
nozzle was electrified with the same polarity due to the
electrostatic induction and discharge from the charged PVC
pipe. The electrification of the nozzle induced corona dis-
charges between the nozzle throat and the needle electrode
and produced air ions with opposite polarity to the charged
pellets. The air ions generated by corona discharges were
blown into the PVC pipe by compressed air with a pressure
of 200 300 kPa to neutralize the charge on the pellets. The
air blow could also prevent powder intruding into the nozzle.
In this study, two types of ionizers, type-A and type-B,
with metal nozzle bodies of different lengths, as shown in
Fig. 1, were tested in the experiments. The metal nozzle of
the type-A electrically contacted the metal fitting and the
metal screws, while the one of the type-B was isolated from
the other conductors. The main difference between the
nozzles was the electrostatic capacitance of the metal nozzle
to the ground. The capacitance in the type-A was 2.8 pF,
which was twice as much as that in the type-B.
3. Experimental setup
A real-sized pneumatic powder-transport facility, shown
in Fig. 2, was used to evaluate the effectiveness of the
developed eliminator. The facility consists of a stainless
steel cylindrical silo with a capacity of 20 m3 (a diameter
of 1.5 m and a straight body length of 2 m), a stainless steel
pipeline with a 4-in. diameter and a length of approximately
20 m, an air blower, and an air-conditioning unit. A 350-kg
sample of polypropylene pellets with the size ranges between
4 and 5 mm were continuously circulated through the
pipeline (discharged from the bottom of the silo through a
rotary valve, and then conveyed back into the silo from the
loading pipe through the pipeline). The developed eliminator
was installed at the end of the loading pipe inside the silo, as
shown in Fig. 3. Photo 1 shows the appearance of the 6-in.
PVC pipe (mentioned later) with mounting six ionizers.
The effectiveness of the eliminator was evaluated from the
electric field strength measured with an electrostatic field
sensor of the air-blow type mounted on the side wall of the
silo at a position of 30 cm above the pellet heap surface. It
was confirmed from experiments that this electric field
depended on the accumulated charge on the piled-up pellets
rather than on the charge on the falling pellets. In order to
confirm the elimination effectiveness, light emissions from
electrostatic discharges along the heap surface were contin-
uously observed using a video camera with an image inten-
 T. Kodama et al. / Powder Technology 135 136 (2003) 209215
Fig. 2. Schema of the real-sized experimental facility.
sifier (Hamamatsu Photonics: Night Viewer C3100) placed
on the window on the silo roof. The threshold of the electric
field measured with the field sensor for incendiary discharge
was determined from the observation of light emissions.
The electric potential of the PVC pipe wall, which would
contribute to the elimination performance, was measured
with an electrostatic potential meter set at a position of 20
mm apart from the pipe wall. The measurement point on the
PVC was separated from the nearest ionizer by 5 cm. The
potential was obtained by calibrating the reading. The
nozzle potential was measured in a similar way.
The nozzle current, which would be directly related to the
elimination performance, was measured with an electrometer
connected between a terminal collecting the other ends of all
the resistors connected to the needle electrodes and the
ground. The elimination was stopped by disconnecting the
terminal from the electrometer to cut off the nozzle current.
Fig. 3. Schema of the silo in which a nozzle-type static eliminator was
installed.
211
Photo 1. Appearance of the 6-in. PVC pipe with mounting six ionizers.
The pipe current, Ip, was measured between the insulated
pipeline and the ground, as shown in Fig. 2, to estimate the
charging tendency of the pellets.
4. Experimental results
When the eliminator did not operate, discharges with
strong light emissions were observed along the pellet heap
surface with a camera/video camera with or without an
image intensifier and even with the naked eye. This type of
discharge, called incendiary discharge in this paper, was
observed when the electric field measured at the silo wall
was higher than 1 kV/cm under the experimental conditions.
Therefore, a value of 1 kV/cm was the threshold for the
incendiary discharge in this paper. Photo 2 shows discharges
taken using a camera with an image intensifier.
Photo 2. Bulk surface discharges taken with an image-intensifier camera;
exposure time = 10 s.
212 T. Kodama et al. / Powder Technology 135 136 (2003) 209215

The charge-elimination performance was evaluated from

the real-size experiment under the following conditions: 30
jC in temperature and 30 60% in relative humidity of the
air, 10 m3/min in air flow rate, and 28 45 kg/min in pellet
flow rate.
Fig. 4 indicates the performance for the eliminators
consisting of a 4-in. PVC pipe (ISO3633, VU 100) with
a length of 15 cm and eight type-A ionizers fixed by
screws with different points or material. The threshold for
the incendiary discharge is shown by the dotted line in
Fig. 4. The metal screws with an as-rolled end or a cone-
point end resulted in better performance than that of
those with a flat end and a plastic screw. Results showed
that corona discharges at the pointed end of the screw
raised the electrification of the metal nozzle of the type-A Fig. 5. Charge elimination performance for the eliminators consisting of a
ionizer, resulting in the improvement of the performance. 4-in. PVC pipe with a length of 45 cm and six type-B ionizers with different
However, the type-A ionizer showed a tendency toward shapes of nozzle fronts fixed by different screws; relative humidity = 30%.
elimination shortage at pellet flow rates over 40 kg/min
even when better screws were used. This seems to be
due to the fact that the nozzle potential does not for type-A ionizers. The nozzle with small chips on its
sufficiently rise because of the rather larger capacitance front showed a tendency toward better performance than
of the nozzle. that with the flat front. Possibly, this is because the field
Fig. 5 shows the performance for the eliminators con- concentration at the chips promotes electrification of the
sisting of a 4-in. PVC pipe with a length of 45 cm and six nozzle due to corona discharges from the charged pellets
type-B ionizers with different shapes of nozzle fronts and and PVC pipe and results in the increase of the nozzle
different screws. The metal screw with a cone-point end current.
showed satisfactory elimination performance, while the Fig. 6 indicates the performance for the eliminators
plastic screw showed rather poor performance. It was consisting of different sizes of PVC pipes and six type-B
found that the field concentration at the pointed end of ionizers fixed by metal screws with an as-rolled end. There
the screw furthers corona discharges at the needle electrode were no significant differences in the performance between
of type-B ionizers and results in the improvement of the 4-in. PVC pipes with different lengths of 15 cm and 45 cm
performance. Type-B ionizers had better performance than because the electric potential of the PVC pipes did not differ
type-A ionizers, since the electric potential of type-A much. On the other hand, the 6-in. PVC pipe shown in
nozzles is higher than that of type-B ones because of the Photo 1 (a 6-in. straight PVC pipe (ISO 3633, VU 150) with
difference in the capacitances. The electric potentials of the a length of 27 cm connected to a 4-in. PVC pipe (ISO 3633,
4-in. PVC pipe wall with type-B ionizers were 16 kV VU 100) with a length of 17 cm by a PVC pipe-increasing
during the nonoperation of the eliminator and 3.7 kV socket (ISO 3633, IN 150x100)) with a total length of 50 cm
during the operation. These potentials were similar to those

Fig. 4. Charge elimination performance for the eliminators consisting of a Fig. 6. Charge elimination performance for the eliminators consisting of
4-in. PVC pipe with a length of 15 cm and eight type-A ionizers fixed by different sizes of PVC pipes and type-B ionizers fixed by metal screws;
different screws; relative humidity = 30%. relative humidity = 30%.
 T. Kodama et al. / Powder Technology 135 136 (2003) 209215
Fig. 7. Humidity dependency of the charge elimination performance for the
eliminators consisting of different sizes of PVC pipes and type-B ionizers
fixed by metal screws; pellet flow rate = 28 kg/min.
showed rather superior performance to the 4-in. PVC pipes.
It was found that the 6-in. PVC pipe is electrified in a
similar way to the 4-in. PVC pipe, probably due to electric
discharges from the charged pellets within the pipe, though
collision of pellets with the 6-in. PVC pipe was less than
that in the 4-in. PVC pipe. To give an example, the electric
potential of the 6-in. PVC pipe was 20 kV during the
nonoperation of the eliminator. The positive sign of the field
strength in Fig. 6 shows reverse charging, which means
excessive supply of the opposite air ions toward the charged
pellets. This results from utilizing the electrification of the
PVC pipe to produce air ions.
Fig. 7 shows the humidity dependency of eliminators
consisting of two types of PVC pipes and six type-B
ionizers fixed by metal screws with an as-rolled end. The
elimination ability tended to decrease with the increase of
the relative humidity of air because the electric potential of
the PVC pipe wall decreased, though the charge generation
on pellets was not much changed. Both the 4-in. PVC pipe
with a length of 15 cm and the 6-in. PVC pipe showed a
Fig. 8. Typical time change of measured values for the eliminator consisting of a 4-in. PVC pipe with a length of 15 cm and type-A ionizers; a: pellet flow start,
b and d: elimination start, c: elimination stop, e: pellet flow stop.
213
Fig. 9. Flow-rate dependency of the measured values for the eliminator
consisting of a 4-in. PVC pipe with a length of 15 cm and type-B ionizers
with chipping on the nozzle front fixed by plastic screws; relative
humidity = 30%.
significant decrease in the elimination performance when
the relative humidity was higher than 50%.
5. Discussion
In the passive-type eliminator, air ions with opposite
polarity to the charged pellets are produced due to corona
discharges around the needle electrode within the ionizer.
The corona current measured as the nozzle current between
the needle electrode and the ground depends on the electric
potentials of both the metal nozzle and PVC pipe wall and
the charge quantity on the pellets within the pipe.
Fig. 8 shows the typical time change of measured values
during both the operation and the nonoperation of the
passive-type eliminator. The nozzle current flows continu-
ously and results in the reduction of the field strength during
the operation of the eliminator. The electric potential of the
PVC pipe measured at a position close to the ionizer
becomes high during the period without the elimination
and remains at several thousand volts during the elimina-
214 T. Kodama et al. / Powder Technology 135 136 (2003) 209215
Fig. 10. Humidity dependency of the measured values for the eliminator
consisting of a 4-in. PVC pipe with a length of 15 cm and type-B ionizers
with chipping on the nozzle front fixed by plastic screws; pellet flow
rate = 28 kg/min.
tion. The measured nozzle potential was almost constant at
approximately 2 kV during a satisfactory elimination. In
this example, since there is almost no residual charge on the
pellets during the elimination, the streaming current at the
end of the loading pipe is almost equal to the nozzle current
of 2.5 AA. The charge-to-mass ratio, which is calculated
from dividing the streaming current by the pellet flow rate,
is 3.7 AC/kg, since the pellet flow rate is 0.68 kg/s.
As shown in Fig. 8, the developed eliminator has stable
charge-elimination performance through a short period of
time. However, it is found throughout a long-term experi-
ment that the performance tends to depend on the charging
tendency of pellets. Therefore, further improvements are
required so that satisfactory performance can be attained in
spite of the change in the charging tendency of products.
Fig. 9 indicates examples of measurement results when
changing the pellet flow rate. The pipe current, Ip, shown in
Fig. 2, shows the charging tendency on the pellets due to
friction with the pipe wall. The streaming current at the
lower end of the insulated pipe can be calculated from
subtracting the pipe current from the streaming current at the
upper end of the insulated pipe, which depends on the initial
charge on the pellets. With increasing the pellet flow rate,
the nozzle current and the pipe current slightly increase in a
similar way. The electric potential of the PVC pipe does not
change much in the range of the pellet flow rate. The field
strength shows the magnitude of the residual charge on the
pellets during the elimination. In this experiment, the
potential and the field strength without elimination were
16 kV and 1.6 kV/cm, respectively. For use of the
eliminator in real cases, the residual charge must be kept at
steadily less than a sufficiently low criterion under real
handling conditions.
Fig. 10 shows examples of measurement results when
changing the relative humidity of air. The nozzle current
decreases with the increase of the humidity at over 50%
due to the decrease of the electric potential of the PVC
pipe wall. Accordingly, the elimination ability decreases,
since the streaming current does not change much with
the increase of humidity. For improvement, much better
quality insulating material is required for the insulating
pipe.
6. Conclusions
For the purpose of developing electrostatic eliminators to
prevent dust explosions in silos, improvements on the
passive-type eliminator that neutralizes static charge on
particulate products before entering a silo were attempted
through real-size experiments using polypropylene pellets.
The results obtained are as follows:
(1) The passive-type eliminator consisting of nozzle-type
ionizers without a power supply mounted on a PVC
short pipe could reduce the static charge on pellets to
prevent incendiary bulk surface discharges inside the
silo under the experimental conditions, except at
relative humidity higher than 50%.
(2) The performance of the eliminator was improved by
using an insulated metal nozzle with a capacitance of
1.4 pF, a 6-in. PVC short pipe connected with a 4-in.
PVC short pipe by an tapered pipe fitting, and metal
screws with a sharp end for fixing ionizers to the 6-in.
PVC pipe.
For use of this type of eliminator to real cases in
industries, it must be confirmed that a satisfactory elimina-
tion performance can be steadily attained under real han-
dling conditions.
References
[1] B. Maurer, Discharges due to electrostatic charging of particles in
large storage silos, Ger. Chem. Eng. 2 (1979) 189 195.
[2] A.R. Blythe, W. Reddish, Charging on powders and bulking effects,
Inst. Phys. Conf. Ser. 48 (1979) 107 114.
[3] B. Maurer, Proceedings of Int. Conference on Safe Handling of Flam-
mable Dusts, Organized by VDI, Nurnberg, VDI-Berichte, vol. 494,
1984, p. 119.
[4] M. Glor, Condition for the appearance of discharges from the gravita-
tional compaction of powders, J. Electrost. 15 (1984) 223 235.
[5] M. Glor, Hazards due to electrostatic charging of powders, J. Elec-
trost. 16 (1985) 175 191.
[6] M. Glor, Electrostatic Hazards in Powder Handling, Research Studies
Press, Letchworth, 1988, pp. 93 102.
[7] G. Luttgens, M. Glor, Understanding and Controlling Static Electric-
ity, Expert Verlag, Ehringen bei Bblingen, 1989.
[8] B. Maurer, M. Glor, G. Luttgens, L. Post, Test rig for reproducible
generation of discharges from bulked polymeric granules, J. Electrost.
23 (1989) 25 34.
[9] M. Glor, G. Luttgens, B. Maurer, L. Post, Discharges from bulked
polymeric granules during the filling of siloscharacterization by
measurements and influencing factors, J. Electrost. 23 (1989) 35 43.
[10] M. Glor, B. Maurer, Ignition tests with discharges from bulked poly-
meric granules in silos (cone discharges), J. Electrost. 30 (1993)
123 134.
[11] M. Glor, Proceedings of Eighth Int. Symposium on Loss Prevention
 T. Kodama et al. / Powder Technology 135 136 (2003) 209215
and Safety Promotion in the Process Industries, Antwerp, Elsevier,
Amsterdam, 1995.
[12] M. Glor, K. Schwenzfeuer, Occurrence of cone discharges in produc-
tion silos, J. Electrost. 40&41 (1997) 511 516.
[13] J. Taillet, P. Poidras, M. Glor, B. Maurer, Testing a new device for
suppressing electrostatic charges on polymer granules, J. Loss Prev.
Process Ind. 5 (1992) 235 237.
[14] J. Taillet, Elimination of static charges in the processing of bulk
material, J. Electrost. 30 (1993) 181 190.
[15] T. Kodama, T. Suzuki, K. Nishimura, S. Yagi, S. Watano, Prevention
of incendiary discharge in a silo using a newly developed electrostatic
eliminator, Conf. Record (CD) of 2000 IEEE/IAS Annual Meeting,
October 2000, Rome, 2000, pp. 652 656.
[16] T. Kodama, T. Suzuki, K. Nishimura, S. Yagi, S. Watano, Static
charge elimination on pellets in a silo using a new nozzle-type elim-
inator, J. Electrost. 55 (2002) 289 297.
215
[17] T. Suzuki, T. Mogami, T. Kodama, A. Ohsawa, S. Watano, Online
measuring and controlling method for electrostatic charged powder in
pneumatic handling process, Proceedings of Annual Meeting on
Safety Engineering, November, Japan Society for Safety Engineering,
2001, pp. 15 18, in Japanese.
[18] T. Suzuki, T. Mogami, T. Kodama, S. Watano, Charge elimination
performance of nozzle-type eliminator for powdereffect of wave-
form and frequency of applied AC voltage, Proceedings of Annual
Meeting of the Institute of Electrostatics Japan, August, 2002, pp.
231 232, in Japanese.
[19] T. Kodama, T. Suzuki, K. Nishimura, T. Mogami, S. Watano, A new
nozzle-type static charge eliminator used in a silo during pneumatic
powder transport, Conf. Record (CD) of World Congress on Particle
Technology 4, July, Sydney, 2002, p. 326.
 Powder Technology 135 136 (2003) 223 233
www.elsevier.com/locate/powtec

Ignition hazard due to static electricity in particulate processes

Martin Glor *
Swiss Institute for the Promotion of Safety and Security, CH-4002 Basel, Switzerland

Abstract

The paper gives a review of the state of the art in the field of powder ignition by discharges due to static electricity. After a short
introduction, emphasis is given to most recent results from research on powder ignition by different types of discharges occurring during
powder handling and processing. It is explained why minor changes of the equipment or of the particulate processes have a huge effect on the
energy of spark discharges and thus on the ignition hazard. Discharges associated with charge accumulation on high resistivity powders and
granules and on insulating walls of equipment are discussed, and the most recent results concerning their incendivity are reviewed. In this
context, cone discharges, brush discharges, propagating brush discharges and corona discharges are dealt with. An assessment of the ignition
hazard due to the build up of static electricity cannot occur without taking into account the relevant parameters characterizing the initiation
and the course of dust explosions. The progress achieved in this field (measurement and interpretation of minimum ignition energy data, role
of bulk resistivity, presence of flammable gases or vapoursso called hybrid mixtures, etc.) is also shortly reviewed.
D 2003 Elsevier B.V. All rights reserved.

Keywords: Ignition; Static electricity; Particulate; Dust explosion

1. Introduction In industry, bulk materials are handled in always larger

The formation of a flammable dust cloud is inherently
related to most particulate processes with organic products.
This is even true for many processes with metallic products.
In addition, the phenomenon of tribocharging and thus the
buildup of charges on the product as well as on equipment
and installations can hardly be excluded. Most organic
products become readily charged even when handled and
processed in metallic, earthed and bonded equipment. If the
charge buildup grows high enough to initiate a so-called
discharge by ionisation of the surrounding atmosphere, the
hot plasma within the discharge channel may ignite the
flammable dust cloud. In the past, the coincidence of a
sensitive dust cloud in air with a discharge of sufficient
energy has led to many serious fires and explosions in
different branches of industry.
Due to the fact that static electricity is intrinsically related
to most particulate processes, a serious risk analysis has to
be performed to assess the explosion hazard. The crucial
elements in such a risk analysis are the assessment of the
probability of the occurrence of a sensitive dust cloud and
the assessment of the probability and energy of a discharge
occurring at the same time and place.
* Fax: +41-61-696-70-72.
E-mail address: martin.glor@swissi.ch (M. Glor).
quantities at always larger transfer rates. Very often, filling
and emptying operations are no longer performed with 25-
kg bags. These have been replaced by the so-called FIBCs
(flexible intermediate bulk containers) or even by silo
trucks. FIBCs typically have a volume of 1 2 m3 and are
normally made from insulating plastics. All of these factors
enhance the buildup of a high level of charge accumulation
increasing the probability of an incendive discharge.
According to German statistics, about one dust explosion
occurs each day in Germany and every tenth explosion is
caused by static electricity [1].
2. Dust explosions
2.1. Characteristics
Dust explosions may occur if flammable solid material is
subdivided into fine particles dispersed in air. Particles with
a diameter of more than 0.5 mm do not take part in the
explosion. Fine particles may, however, also be formed
during the handling and processing of granules due to
abrasion. The finer the particles are, the more violent will
be the explosion and the more sensitive will be the dust
cloud with regard to ignition sources. For a worst-case
assessment, the fraction below a mesh size of 63 Am is

0032-5910/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.powtec.2003.08.017
224 M. Glor / Powder Technology 135 136 (2003) 223233

usually chosen. At atmospheric conditions, dust explosions

normally show the following characteristics [2,3]explo-
sion range: from 20 g/m3 to several kg/m3; maximum
explosion overpressure: 9 bar for organic material and up
to 13 bar for metal powder; maximum rate of pressure rise
in a 1-m3 vessel: 100 300 bar/s for organic material and up
to approx. 1000 bar/s for metal powder. Fig. 1 shows the
course of an unconfined dust explosion with 6 kg of maize
starch homogeneously dispersed in air. With regard to
people, dust explosions are normally even more hazardous
than gas explosions since the exposure to the heat radiation
lasts much longer.
With respect to ignition by static electricity, the ignition
sensitivity of a dust cloud is of most interest. It is charac-
terized by the so-called minimum ignition energy (MIE).
The MIE is the minimum energy that can ignite a mixture of
a specified flammable material with air, measured by a
standard procedure (spark discharges produced in a special-
ly designed circuit and apparatus) [2,4]. For a given mate-
rial, the ignition energy is lowest for very fine powder
homogeneously dispersed in a nearly quiescent dust cloud.
Since the energy of the electrostatic discharges most
frequently occurring in industry is normally less than about
10 mJ, materials with a MIE of 10 mJ or less are considered
most critical with respect to ignition by static electricity [5].
It must, however, be kept in mind that this is only true in the
absence of any flammable gases or vapours. As soon as
flammable gases or vapours may in addition be present, the
so-called hybrid mixtures are formed and the MIE may drop
from above 100 mJ to below 10 mJ or even into the range of
the MIE of pure gases or vapours (typically 0.2 0.4 mJ for
hydrocarbons) [2].

2.2. Prevention and protection

A fire or explosion may occur if a coincidence in space

and time of a flammable material, oxygen and an ignition
source takes place. The flammable material, the oxygen and
the ignition source represent the three components of the
well-known fire triangle (see Fig. 2). To prevent an explo-
sion, one of these components must be excluded. Though in
this paper the main focus will be on the exclusion of the
ignition source static electricity, the most important objec-
tive in industry is to avoid an explosive atmosphere at all
[6]. This objective can only be achieved by modifying either
the nature or the concentration of the flammable substance
to a value outside the explosion range or by modifying the
oxygen concentration to a value below the limiting oxygen
concentration (LOC) [2,3]. If, however, the only possible
ignition source is a low-energy discharge due to static
electricity, a lower reduction of the oxygen concentration
as compared with the LOC may be sufficient, since the MIE
strongly increases with decreasing oxygen concentration.
This particularly applies to dust/air mixtures, the LOC of
which is determined with very energetic (2 10 kJ) ignition
sources [7]. Fig. 1. (a e) Course of an unconfined dust explosion (6 kg of maize starch).
 M. Glor / Powder Technology 135 136 (2003) 223233 225

ing will be predominant. Other possibilities are induction

Fig. 2. Fire triangle representing the prerequisites for the occurrence of a
fire or explosion [9].
However, despite the demand to avoid explosive atmos-
pheres in industrial processes, many operations and pro-
cesses with powders and granules are still performed under
atmospheric conditions. The exclusion of effective ignition
sources such as static electricity then represents the only
basis of safety. Even if the effects of an explosion are
limited to an acceptable extent by constructional protective
measures, the exclusion of all effective ignition sources is of
utmost commercial interest.
and corona charging. Surface contact, separation and move-
ment involving poorly conducting materials are an intrinsic
property of most industrial processes with powders and
granules. Examples are grinding, micronisation, blending,
sieving, dust separation and pneumatic transfer. It must be
kept in mind that in all these processes, both the equipment
and the product may become charged, as indicated in Fig. 3.
This is even true if an organic product is processed in a
stainless-steel equipment. These facts have to be considered
in every assessment of electrostatic hazards.
Charge separation does not by itself automatically lead to
a hazardous situation. The amount of charge accumulation
and the resulting distribution of the electric field represent
the decisive factor. Charge accumulation is determined by
the rate of charge separation and charge dissipation. In
practice, charge may accumulate on electrically insulated
conductors, on insulating surfaces of the equipment and
installations, or on highly insulating powders, granules or
liquids.
Fortunately, charge dissipation occurs even at relatively
high resistances to earth. The electric currents that usually
occur in practice due to separation processes are very small.
Typical values are 10
6 A or less. Under extreme condi-
tions (e.g. pneumatic transfer), values up to 10
4 A may be
reached. For such low charging currents, leakage of charge
via resistances to earth in the range of 106 to 108 V is
sufficient to prevent a hazardous level of charge accumula-

3. From charge buildup to ignition

Charge buildup as such does not cause ignition. In all

cases where a fire or explosion in industrial environments is
caused by static electricity, the sequence of events passes
through the same stages. These stages are shown in Fig. 3.
Though the scheme looks rather trivial at first sight, it is not
always easy to localize each step within the diagram for a
given process with respect to space and time. Of course, it
often happens that several steps occur simultaneously. For
example, the level of charge accumulation is determined by
the equilibrium between both, the rate of charge separation
and the rate of charge dissipation. For both, the assessment
of an electrostatic ignition hazard and the analysis of
whether static electricity could have caused a fire or
explosion, the scheme shown in Fig. 3 is extremely helpful.
Charge separation may occur by different mechanisms. Fig. 3. Scheme showing the different steps from charge buildup to ignition
In particulate processes, contact or the so-called tribocharg- [9].
226 M. Glor / Powder Technology 135 136 (2003) 223233
tion. However, it must be kept in mind that not only high-
resistivity materials such as polyethylene, polypropylene,
etc. lead to resistances to earth far above this order of
magnitude. Most pigments, dyestuffs, additives, drugs,
foodstuffs, etc. show resistivities above 1010 V m.
As accumulation of charges grows higher, the resulting
electric field in the air may reach its limit value. This limit
value is also called the dielectric strength of air. Under
normal conditions, it amounts to about 3 MV/m. At this
limit a discharge may occur. All or only part of the energy
stored in the charge accumulation may be released in a hot
discharge channel that may ignite a given flammable
atmosphere.
The energy released in the discharge and the sensitivity
of a given flammable atmosphere measured in terms of its
minimum ignition energy determine whether ignition will
occur or not. The assessment of the occurrence and
incendivity of discharges in different situations in practice
represents the most important, but also the most difficult,
step in the analysis of electrostatic hazards. The basic
principles and typical examples are given in the following
paragraphs.
4. Electrostatic discharges in particulate processes
Because of the difficulty to predict the occurrence and
incendivity of discharges in industry by the laws of plasma
physics, a more or less phenomenological approach is
commonly used. The discharges occurring in practice are
classified into different discharge types. These types of
discharges have different incendivities. The electrical prop-
erties of the products and installations, their geometrical
arrangement and the operation determine which discharge
type will occur in a given situation. In a first approxima-
tion, the incendivity of a discharge is estimated by its total
energy and the ignition probability assessed by comparing
this total energy with the minimum ignition energy of the
flammable atmosphere as determined with a spark dis-
charge from a capacitive circuit [4]. The first problem,
which arises in such an approach, is the calculation of the
energy of the discharge. This is far from being trivial for
most discharges other than sparks. In addition, due to the
complexity of the ignition mechanisms, this approach is
too simplified since the incendivity of a discharge depends
not only upon its total energy but also upon the energy
distribution with respect to space and time. This spatial
power density varies widely among the different discharge
types. Due to these problems, the energy released in a
discharge and its incendivity is best determined in ignition
tests with flammable atmospheres. The energy determined
in this way is called equivalent energy of the discharge and
is defined as follows: A discharge has the equivalent
energy W if it is just able to ignite a flammable mixture
with a minimum ignition energy W, as determined with
capacitor spark discharges.
5. Discharges associated with installations and
equipment
5.1. Spark discharges
Spark discharges occur between two conductors at
different potential as soon as the electrical field in the
gap reaches the breakdown value of about 3 MV/m under
normal atmospheric conditions. In practice, one of these
conductors (including personnel) reaches a high potential
since it is not connected to earth; the other conductor is
normally at earth potential. Nearly the total energy stored
in such systems (capacitors) is released in a single spark,
which generates a single discharge channel of rather high
current density (Fig. 4). Therefore, these discharges are
rather incendive and their energy release can simply be
calculated by the energy W stored in the capacitor using
the potential U (in Volts) and the capacitance C (in
Farads):
1 2
W CU 1
2
Due to the increasing use of nonconducting plastics in the
construction of apparatus and equipment, the probability
that part of the system is electrically insulated from earth
becomes always higher. Charge buildup on isolated con-
ductors is responsible for the majority of all ignitions of
flammable atmospheres in industry caused by static elec-
tricity. Typical examples are:
 Metal funnel mounted on top of a pipe made from
insulating material.
 Ball valve with the metal ball isolated from the pipe of
the valve by an insulating coating.
 Piece of metal pipe insulated from earth by nonconduct-
ing gaskets.
 Metal drum on a trolley with insulating tyres.
 Metal shovel held by an operator wearing insulating
gloves.
 Human body insulated from earth via insulating shoes
and/or insulating floor.
Fig. 4. Spark discharge [9].
 M. Glor / Powder Technology 135 136 (2003) 223233
Fig. 5. Electrical circuit simulating the charge buildup on a conductor.
In principle, gases, vapours and dust clouds can be
ignited by spark discharges. Earthing and bonding of all
conductive parts, which may become charged, is therefore
mandatory in areas where flammable atmospheres formed
by these substances have to be expected. It is commonly
agreed [8,9] that a resistance to earth of less than 106 V is
generally sufficient in the case of equipment and 108 V in
the case of personnel to avoid spark discharges. Even
higher resistances to ground may not necessarily lead to
hazardous potentials under moderate charging conditions.
If, however, only minor changes in the operating proce-
dures (slightly different resistance to ground and/or charg-
ing current) occur, this may have a drastic effect on the
potential acquired. The potential buildup on a container
filled with charged product or on a piece of pipe in a
pneumatic transport system or on any capacitor, which
becomes charged by static electricity, can be calculated by
using the electrical circuit shown in Fig. 5. For example,
in the case of filling a container with powder, the charging
current I is calculated from the charge to mass ratio q of
Fig. 6. Maximum energy of a spark from an electrically isolated plant item (conductor) exposed to different charging currents I as a function of the leakage
resistance to ground.
227
the powder and the filling rate dm/dt by the simple
relation
I q  dm=dt 2
If R is the overall leakage resistance in ohms, C the
capacitance in Farads, U the potential in volts and Q the
total charge in Coulombs in/on the container when equi-
librium is reached, which means that the capacitor is
charged to potential U, the following simple relations
apply
U IR 3
Q UC 4
This is the case if all the current flows to ground via the
leakage resistor R. Based on Eqs. (1) (4), the maximum
energy of spark discharges generated by a small plant item
(conductor) not safely connected to ground can be calcu-
lated. Fig. 6 shows the results for such a plant item
having a capacitance of 20 pF. Since the potential is
limited by the geometrical arrangement and the dielectric
strength of air, the limit value of the spark energy for
different potentials is indicated.
Since the energy increases with the square of the charg-
ing current and with the square of the leakage resistance,
minor changes of these quantities have a huge effect on the
ignition hazard. These considerations explain why many
operations may run without any problems for a long time
and all of a sudden an explosion occurs.
5.2. Brush discharges
Charges of one polarity accumulated on the surface of
nonconducting installations and equipment may produce
very high electric fields in the surroundings. As soon as
228 M. Glor / Powder Technology 135 136 (2003) 223233
an earthed conductive electrode is approached to the sur-
face, the electric field is locally distorted and the dielectric
strength of air may be reached (3 MV/m at atmospheric
conditions) at the surface of the electrode. As a result, the
so-called brush discharges will occur. An exact mathemat-
ical treatment for a special arrangement of the field gener-
ating charges and of the electrode (cylindrical symmetry) is
given in Ref. [5]. Fig. 7 shows a sequence of several brush
discharges generated at the surface of an earthed spherical
electrode.
It depends on many parameters such as on the radius of
curvature of the electrode, on the speed at which the
electrode is introduced into the electric field and on the
polarity, whether a corona or a brush type discharge will
actually occur in such situations. In practice, it can generally
be assumed that brush discharges occur, if the radius of
curvature of the electrode is larger than 5 mm.
In addition to the situation with a charged insulating
surface, brush discharges may also occur, if an earthed
conductive electrode is introduced into a high electric field
of any origin. The high electric field may, for example,
result from a highly charged insulating liquid or suspen-
sion, from a highly charged mist, from highly charged
insulating powder in bulk or from a highly charged dust
cloud. In what follows, examples of plant operations that
are the most frequent sources of brush discharges are
summarized:
 Approach of an earthed conductive electrode such as a
tool or a finger tip to highly charged insulating surfaces
(e.g., plastic pipe for the conveyance of liquids or dusts,
plastic bags, common flexible intermediate bulk contain-
ers, plastic packing drums or filter cloths as well as film
webs, nonconductive conveyor belts or V-belts).
 Discharging of solids from plastic bags, or shaking out of
plastic bags in the vicinity of metal fittings (e.g. above an
access port of a reaction vessel).
Fig. 7. Brush discharge [9].
 Projection of internal fittings into a highly charged dust
cloud or a highly charged droplet cloud.
 Charging of nonconductive powder into packing drums,
containers or silos, approach of the highly charged dust
surface to the internal fittings or lowering of a conductive
earthed sample baker for sampling or a level probe to
determine the level.
 Projection of flagpoles, ships masts, antennae or ice axes
into powerful atmospheric electric fields (St. Elmos fire
during thunderstorms).
The characteristics and incendivity of brush discharges
have been investigated by different authors (see Refs. [10,11]
for details). The values reported for the equivalent energy
have been determined with gas/air-mixtures and lie in the
range of 1 to 3.6 mJ. The parameters affecting the incendivity
of brush discharges are the radius of curvature of the
electrode, the polarity of the electric field andif the electric
field is generated by a highly charged nonconducting sur-
facethe surface charge density and the area of the surface.
Brush discharges from an earthed, metallic electrode opposite
to a negatively charged surface are by far more incendive than
from an electrode opposite to a positively charged surface.
On the basis of the values of the experimentally deter-
mined equivalent energy of brush discharges, it is expected
that most flammable gas or solvent vapour atmospheres and
hybrid mixtures can be ignited by brush discharges.
The question, whether brush discharges from highly
charged insulating surfaces of equipment or packaging are
able to ignite powders in the absence of flammable gases or
solvent vapours is still the subject of an ongoing debate. The
reason being contradictory results obtained from gases and
vapours on the one hand and from powders on the other
hand. Although the minimum ignition energy of certain
powders lie in the range of 1 10 mJ, no ignition of a dust
cloud clearly caused by a brush discharge has yet been
reported. Experiments performed in the laboratory [12 14]
have so far also shown only negative results. Therefore,
according to the present state of knowledge, brush dis-
charges are very unlikely to ignite pure dust clouds con-
taining no flammable gases or vapours.
In conclusion, brush discharges must be excluded in
areas where flammable atmospheres formed by flammable
gases or vapours may be present but they may be tolerated
when only flammable powders are present.
Brush and corona discharges associated with installa-
tions, containers and packing materials can be avoided by
the use of conductive material or by a limitation of the area
of insulating surfaces as specified in the different codes of
practice and guidelines, e.g. Ref. [8].
In this context, the term antistatic is often used. In the
German literature, the antistatic property of a surface is
defined in terms of the surface resistance. If the surface
resistance lies below 1011 V and above 108 V, neither brush
nor spark discharges can be expected. Apart from the
surface resistance, measurement of charge decay times is
 M. Glor / Powder Technology 135 136 (2003) 223233
Fig. 8. Corona discharge [9].
also used to characterize the charge retention behaviour of
nonconducting surfaces.
The incorporation of the so-called antistatic additives
into the polymer is often used to decrease the charge
retention of nonconducting surfaces. In this way, the surface
resistance can be reduced to the limits mentioned above.
Problems associated with this measure are that the antistatic
effect often depends on relative humidity of the environment
and that the antistatic additives may become absorbed by the
product in contact with the treated surface.
The incorporation of carbon black into the polymer may
drastically increase the conductivity (if sufficient carbon is
added). Equipment made from such material is conducting
from the point of view of electrostatics and must be earthed
in practice to prevent the occurrence of spark discharges.
Because of the effect of electrical induction, brush
discharges do not occur from insulating coatings of con-
ductive surfaces or from insulating walls having, at least on
one side, an antistatic treatment. This only applies if the
thickness of the insulating coating of the wall or the wall
itself is no more than a few millimetres. In these cases, the
electric field is directed through the dielectric layer (see also
Section 5.4).
5.3. Corona discharges
Corona discharges (see Fig. 8) occur under similar
conditions to brush discharges, but preferably at sharply
pointed electrodes (radius of curvature less than 0.5 mm).
The energy of corona discharges is much lower than that of
brush discharges and the energy distribution with respect to
space and time is quite different. Ignition of flammable
atmospheres with a minimum ignition energy above 0.2 mJ
by corona discharges is not to be expected nor is the ignition
of flammable dust/air mixtures. The value of the equivalent
ignition energy of corona discharges is not known and it
may be well below 0.2 mJ. Ignition of very sensitive gases
with a minimum ignition energy below 0.025 mJ can,
however, not be excluded.
5.4. Propagating brush discharges
If the charges are not arranged in the form of one single
layer of one polarity on a nonconducting surface, but in the
229
form of a double layer of charges of opposite polarity on
the opposite surfaces of a nonconducting material in form
of a sheet, propagating brush discharges may occur. The
level to which a nonconducting surface exposed to air can
become charged in the form of a charge layer of one single
polarity is limited to about 2.7  10
5 C/m2. At this value
of the surface charge density, the field strength above the
surface reaches the dielectric strength of the air and any
additional charges would initiate a discharge into the air.
However, when a double layer of charges of opposite
polarity is generated across a nonconducting sheet of a
thickness of a few millimetres, more charges can be placed
on each surface. This is explained by the fact that the
electric field in the air due to one layer of charges is, to a
certain extent, compensated by the field of the other layer
of charges. Clearly, the electric field across the noncon-
ducting sheet will be very high. A quantitative treatment of
such double layers of charges across dielectric materials is
found in Ref. [5].
Propagating brush discharges are initiated by an electrical
short circuit of the two surfaces of the highly charged
dielectric sheet. This electrical short circuit can be achieved
either externally by the approach of electrically connected
electrodes to each of the two surfaces, or internally by a
mechanical or electrical perforation of the dielectric layer. If
one surface is already in contact with an earthed conductive
plate, the discharge is initiated by the approach of one
electrode to the other surface. The discharge pattern is
always the same: Many discharge channels branch off to
collect the charge from much of the surface. They all end in
a bright discharge channel at the location of the actual short
circuit. This central discharge channel bridges the gap
between the approaching electrode and the insulating sur-
face in the case of an external short circuit, or runs through
the perforated sheet in the case of an internal short circuit.
Fig. 9 shows a propagating brush discharge initiated by the
Fig. 9. Propagating brush discharge [9].
230 M. Glor / Powder Technology 135 136 (2003) 223233
approach of an earthed electrode to a highly charged
insulating surface on top of a metal plate.
For a long time, it has been assumed that these discharges
can only occur if one surface of the insulating layer is in
intimate contact with an earthed metallic plate. In this case,
the second charge layer of opposite polarity is accomplished
by the induced charges on the metallic plate. It can, however,
be demonstrated, by spraying charges of opposite polarity to
the two surfaces of a dielectric sheet, that the metallic plate in
intimate contact with one surface of the insulating sheet is
not essential for the buildup of the double layer. If a dielectric
sheet in free space (e.g., the insulating wall of a container)
becomes highly charged on one surface and the electric field
is mainly directed through the dielectric sheet towards earth,
the other surface of this sheet may become charged with
charges of opposite polarity by corona or brush discharges.
Such a charging mechanism is observed, for example, when
filling an insulating container such as a FIBC with highly
charged polymeric powder.
From practical experience and experimental evidence, it
is generally agreed that the high surface charge densities
necessary for propagating brush discharges cannot be
achieved by separation processes in manual operations such
as wiping of insulating surfaces or discharging a powder
from a plastic bag. Charging mechanisms associated with
high separation velocity are necessary to build up the high
surface charge densities necessary for propagating brush
discharges. Such high-charge buildup may be observed
during the following industrial operations:
 High-velocity pneumatic transfer of powder through an
insulating pipe or a conductive pipe with an insulating
internal coating.
 Use of inspection windows made from glass or Plexiglas
in pneumatic transfer pipes.
 High-velocity transportation of a highly insulating
suspension through an insulating pipe or a conductive
pipe with an insulating internal coating.
 Continuous impact of powder particles onto an insulating
surface (e.g., a coated dust deflector plate in the cyclone
of a dust separator).
 Fast rotation of conveyor or transmission belts made
from insulating material or from conductive material,
which is coated with an insulating layer of high dielectric
strength.
 Filling of large containers or silos made from insulating
material (e.g., flexible intermediate bulk containers) or of
metallic containers or silos internally coated with an
insulating layer of high dielectric strength.
According to practical experience and experiments in the
laboratory, it is not expected that layers of insulating pow-
ders that are, for example, formed in metallic pipes during
pneumatic transfer, or paints, will give rise to propagating
brush discharges unless the powder is subjected to a sintering
or melting process.
The energy released in propagating brush discharges is
high enough to ignite most flammable gases and vapours as
well as most combustible powders. Persons may suffer a
serious shock when the human body acts as the initiating
electrode for such a discharge. Propagating brush discharges
must be excluded in areas where flammable atmospheres
formed by flammable gases or vapours or by combustible
powders may be present.
Propagating brush discharges can be excluded by the use
of conductive materials or insulating materials of low di-
electric strength at all locations where the buildup of high
surface charge densities may occur. If the breakdown voltage
across a nonconducting layer or sheet is less than 4 kV,
propagating brush discharges will not occur.
6. Discharges associated with powder
6.1. Brush and corona discharges
In addition to the situation with a charged insulating
surface described in Section 5.2, brush discharges may
always occur if an earthed conductive electrode is intro-
duced into a high electric field resulting from a highly
charged insulating powder in bulk or from a highly
charged dust cloud. With respect to energy and incendivity,
the same arguments and figures apply as mentioned in
Section 5.2.
6.2. Lightning-like discharges
Before the phenomenon of cone discharges described in
the next section was discovered, the charged dust cloud
was considered to be the most dangerous source of elec-
trostatic ignition in a silo or container. This hypothesis was
originally based on the lightning flashes observed in the
clouds of dust and ashes generated by erupting volcanoes.
Since then, systematic tests have been carried out in a 60-
m3 container [15]. The results obtained from these tests
indicate that in silos with a volume of 60 m3, or in those
with any other volume but a maximum diameter of 3 m,
discharges of the lightning type are not to be expected. For
larger silos, no predictions based on firm experimental
evidence can be made. Nevertheless, the prevailing opinion
is that lightning-type discharges in industrial-size silos are
highly improbable. The investigations now described have
shown that much more importance must be paid to cone
discharges.
6.3. Cone discharges
Cone or bulking discharges were first observed in 1979
by Maurer [16] while silos were being filled with polyeth-
ylene granules. Results published in the same year [17] on
the electrostatic discharges occurring during the filling of
FIBCs with plastic granules suggest that cone discharges
 M. Glor / Powder Technology 135 136 (2003) 223233 231

were involved here as well. Following serious explosions in of the bed of powder is greater than the silo radius, or
the German plastics industry in the late 1970s, a large-scale downward through the bed if its height is less than the silo
research project was commissioned, its objectives being to radius (see Fig. 10).
investigate the incidence and incendivity of cone discharges The initial supposition that cone discharges only occur
in tests performed under realistic conditions on a 1:1 scale. when low-conductivity, coarse-grained products (polymer
The project was financed from German research funds and granules) are handled could not be confirmed as well as the
by contributions from major German and Swiss chemical supposition put forward in earlier publications that the
companies, and the results of the investigations were pub- ignition energy of cone discharges is not greater than 10
lished in numerous technical reports and journals [18 22]. mJ. It was found in the large-scale research project already
Parallel to this experimental work, theoretical and compu- mentioned [18 22] that cone discharges can have an
tational model calculations on the occurrence of cone incendivity (equivalent ignition energy) that increases with
discharges and the energy involved were carried out the silo diameter and the size of the particles in the bed of
[23,24]. Supplementary results were also obtained from bulk material. The magnitude of this equivalent energy is
experimental work carried out on a laboratory scale [25]. such that the ignition of explosive gas, vapour and dust
To understand the physics of cone discharges, a phe- atmospheres is to be expected from cone discharges.
nomenon that is most succinctly termed charge compac- The upper limit for the equivalent energy of cone
tion [23] must first be described. When feeding a product discharges can be estimated with the following empirical
in powder form into large containers or silos by a pneumatic equation (see also Fig. 11):
conveying system, orto a limited extentalso under
gravity, the flowing product is dispersed in air. Because of
WAe 5:22  D3:36  d 1:462 6
this, the product density when in motion is low in compar-
ison with its bulk density under storage conditions. In this
where WAe is the upper limit of the equivalent energy of
dispersed condition, the product accumulates varying
cone discharges in mJ; D is the diameter of the conductive,
degrees of electrostatic charge due to triboelectrification
grounded silo in m; and d is the median of product particle
against the walls of the feed piping. This can result in high
size distribution in the bed in mm.
space charge densities in the conveying system, depending
This equation is based on results obtained in a test
on the specific charge and degree of dispersion of the
system on a 1:1 scale using silos with a diameter D between
product. When filling large containers and silos or, stating
1.0 and 3.0 m and products with a median particle size
it more accurately, when dropping the conveyed product
distribution, d, between 0.8 and 3.0 mm. In addition, values
into the bed of bulk material, at the same time as the
from the literature [25] for D = 0.5 m and d = 3 mm, together
mechanical compression and compaction of the particles,
with measurement results from production silos with D = 2.5
an extreme compaction of the electrostatic charge also takes
m and d = 0.1 mm, were taken into account [26]. The
place. If the surface conductivity of the powder particles is
very low, the charge they bear cannot dissipate to ground
quickly enough. This means that a much higher space
charge density accumulates in the bed of powder than in
the conveying system or the charged dust cloud. Strong
electric fields therefore result at the edges of the powder
bed. In the case of a cylindrical bed, the electric field
intensity E can be estimated from the simple formula
qR
E 5
2ee0

where q is the space charge density in the bed, R the

radius of the cylindrical bed (in metres), e0 the permittivity
of free space (e0 = 8.8510
12 As/V m) and e the relative
permittivity of the powder bed. Even with a moderate
product charge of some 10
7 C/kg in a silo only 1.5 m in
diameter, this relation gives an electric field intensity in the
order of magnitude of the dielectric strength of air. Once
the height of the bed is greater than the radius of the silo,
the electric field has a predominantly radial direction, with
the maximum electric field intensity being reached at the
inner wall of the silo [24]. Following the path of least
resistance, cone discharges take place in the form of
discharge channels, pointing radially outward if the height Fig. 10. Cone discharge in a silo [9].
232 M. Glor / Powder Technology 135 136 (2003) 223233
Fig. 11. Upper limit of equivalent energy of cone discharges as a function of particle size (median) and silo diameter in a grounded metal silo.
applicability of this estimation method is thus limited to the
range:
0:5 m < D < 3:0 m
and
0:1 mm < d < 3:0 mm
This does not mean that the empirical equation is a priori
no longer valid outside the stated range. It is, however, no
longer backed up by measured results. Extrapolations can
therefore only be recommended after consultation with an
expert.
The occurrence of cone discharges depends on whether
such a strong electric field can be formed in the silo from
the product being filled into it such that the dielectric
strength is reached. Experience indicates that this is espe-
cially the case when silos are filled from pneumatic con-
veying systems and if the charge introduced into the silo
from the system cannot be dissipated back to ground
quickly enough.
At the present time, 1010 Vm is stated as the limit for the
resistivity of bulk material above which cone discharges are
likely to occur. Owing to the lack of more detailed knowl-
edge, this is probably a rather conservative value. Initial
model calculations have been carried out for orientation
purposes, taking into account the silo geometry, the product
specific charge and the filling rate. These have shown that,
depending on the situation, values up to two orders of 10
greater than this limit may possibly be regarded as still not
being critical. More research is needed on this point.
In general, it is hardly possible to increase the conduc-
tivity of bulk materials in industrial practice. In situations
where cone discharges cannot be excluded, and the antici-
pated equivalent energy is the same or greater than the
minimum ignition energy of the product to be processed,
other measures therefore need to be taken, such as prevent-
ing potentially explosive atmospheres arising (for example
by excluding flammable gases, vapours and fine dust or by
reducing the oxygen content), or designing the plant with
explosion protection elements.
7. Conclusions
Based on experience and industrial research, the ignition
hazards associated with static electricity can nowadays be
assessed rather well. The present knowledge in this field is
sufficient to prescribe safety measures for most common
industrial processes. It must, however, be kept in mind that
electrostatic phenomena are encountered so widely that it is
not possible to cover all cases and that research in the field
continues providing new information. In addition, new
processes are developed, common processes are run under
more hazardous conditions or conflicts of interests do not
allow a standard solution. Thus, there are and there will be
open questions concerning the assessment of electrostatic
ignition hazards.
References
[1] H. Beck, VDI Ber. Nr. 975 (1992) 47.
[2] W. Bartknecht, Explosionsschutz, Grundlagen und Anwendung,
Springer Verlag, Berlin, 1993.
[3] R.K. Eckhoff, Dust Explosions in the Process Industries, Butterworth-
Heinemann, Oxford, 1991.
[4] R. Siwek, C. Cesana, Process Saf. Prog. 14 (1995) 107.
[5] M. Glor, Electrostatic Hazards in Powder Handling, Research Studies
Press, Letchworth, Hertfordshire, 1988.
[6] European Standard EN 1127-1, 1997.
[7] A. Gitzi, K. Schwenzfeuer, M. Glor, Proceedings of the 10th Interna-
tional Symposium on Loss Prevention and Safety Promotion in the
Process Industries, 19 21 June 2001, Stockholm, Elsevier, Amster-
dam, 2001, p. 909.
[8] CENELEC Report R044-001 Safety of machineryGuidance and
 M. Glor / Powder Technology 135 136 (2003) 223233
recommendations for the avoidance of hazards due to static electric-
ity 1999. The document is available from BSI-British Standards
Institution as published document PD R044-001, 1999.
[9] ISSA Prevention Series No. 2017 (E), Static Electricity-Ignition Haz-
ards and protection measures, Heidelberg, 1996.
[10] N. Gibson, F.C. Lloyd, Br. J. Appl. Phys. 16 (1965) 1619.
[11] M. Glor, J. Electrost. 10 (1981) 327.
[12] K. Schwenzfeuer, M. Glor, J. Electrost. 30 (1993) 115.
[13] K. Schwenzfeuer, M. Glor, J. Electrost. 40 and 41 (1997) 383.
[14] O. Larsen, H.J. Hagen, K. van Wingerden, J. Loss Prev. 14 (2001)
111.
[15] P. Boschung, W. Hilgner, G. Luttgens, B. Maurer, A. Wimer, J. Elec-
trost. 3 (1977) 303.
233
[16] B. Maurer, Chem. Eng. Technol. 51 (1979) 98.
[17] A.R. Blythe, W. Reddish, Inst. Phys. Conf. Ser. No. 48, The Institute
of Physics, Bristol and London, 1979, p. 107, Oxford.
[18] B. Maurer, VDI Rep. 494 (1984) 119.
[19] B. Maurer, M. Glor, G. Luttgens, L. Post, J. Electrost. 23 (1989) 25.
[20] M. Glor, B. Maurer, G. Luttgens, L. Post, J. Electrost. 23 (1989) 35.
[21] M. Glor, B. Maurer, J. Electrost. 30 (1993) 123.
[22] M. Glor, B. Maurer, VDI Rep. 1272 (1996) 87.
[23] M. Glor, J. Electrost. 15 (1984) 223.
[24] A. Kuttler, M. Glor, J. Electrost. 30 (1993) 285.
[25] V. Ebadat, Thesis, University of Southampton, 1989.
[26] M. Glor, K. Schwenzfeuer, J. Electrost. 40 and 41 (1997) 511.
 Journal of Loss Prevention in the Process Industries 16 (2003) 333340
www.elsevier.com/locate/jlp

Simple devices to prevent dust explosion propagation in charge

chutes and pipes
P. Holbrow a,, A. Tyldesley b
a
Health and Safety Laboratory, Harpur Hill, Buxton, Derbyshire, SK17 9JN, UK
b
Health and Safety Executive, Magdalen House, Bootle, Merseyside L20 3QZ, UK

Abstract

This paper is a report of an experimental programme of work to develop inexpensive dust explosion propagation prevention
devices. Two types of device were investigated: one is intended for use with charge chutes attached to process vessels and one is
intended for use with pipes attached to process vessels. Three dusts were used with KSt values up to 308 bar m s-1. Both devices
were found to be effective and resulted in the elimination or significant reduction in the flame from the charge chute or the pipeline.
Crown copyright 2003 Published by Elsevier Science Ltd. All rights reserved.

Keywords: Explosion propagation; Dust explosion; Venting

1. Introduction thermal radiation from a vented flame is effectively

The risk of a dust explosion is difficult to avoid in
processes where combustible dusts and powders are
handled. Precautions against dust explosions fall into
two categories, prevention and protection. Prevention
methods aim to ensure that conditions under which an
explosion becomes possible never occur and protection
methods aim to minimise the effects of an explosion.
In the following two situations, common in the powder
handling industries, there is the potential to make sig-
nificant improvements to explosion protection using
relatively inexpensive explosion propagation protection
devices.
1. Charge chutes are often used to feed powder into pro-
cess vessels by manually tipping the powder from
sacks or bags into the charge chute. The powder then
falls under gravity into the process vessel. In the event
of a dust explosion occurring in the vessel during the
filling procedure, flame could propagate up the charge
chute, impinge on the operator and cause serious
injury. The area in which the operator is at risk from
Corresponding author. Tel.: +44-114-2892121; fax: +44-114-
2892199.
E-mail address: paul.holbrow@hsl.gov.uk (P. Holbrow).
no bigger than the vented flame itself (Holbrow,
Hawksworth, & Tyldesley, 2000). Consequently, a
reduction in the size of the flame will result in a direct
reduction of the danger area. A simple device to pre-
vent flame propagation or reduce the size of the
vented flame from the chute is therefore needed to
protect the operator against burn and blast injury. Cor-
rectly placed baffles in the charge chute could protect
the operator from flame. The device described in this
paper is intended for use where, in the event of an
explosion, the resulting explosion pressure will be
low, typically less than 500 mbar. This work did not
cover the situation where powder is added to a vessel
that is not vented, for example a stirred chemical reac-
tor.
2. A dust explosion in one part of a powder handling
system can cause pressure and/or flames to propagate
to other vessels via connecting pipes. For example, in
an explosion in a vented bag filter, where typically
the reduced explosion pressure (Pred) is less than 500
mbar, tests have shown that the explosion can propa-
gate into the inlet pipe. This could lead to an
explosion propagating with increasing violence
throughout the system. Isolation devices are rec-
ommended, and in this paper, simple inexpensive
devices that may inhibit flames and pressure propa-
gation are described.

0950-4230/03/$ - see front matter Crown copyright 2003 Published by Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0950-4230(03)00017-2
334 P. Holbrow, A. Tyldesley / Journal of Loss Prevention in the Process Industries 16 (2003) 333340

Table 1
Test dusts

Dust Milk powder Coal Anthraquinone

KSt (bar m/s) 86 155 308

Pmax (bar g) 7.4 7.7 8.4

The Health and Safety Executive has sponsored a pro-

ject with the Health and Safety Laboratory to investigate
simple mechanical devices that can be installed rela-
tively cheaply but provide effective protection in the
situations described above.

2. Experimental

2.1. The test dusts

Three dusts were used in the test programme. The

explosion characteristics of the dusts were measured in Fig. 2. Charge chute propagation prevention device (viewed from
the top).
the standard 20 litre sphere dust explosibility apparatus
and are shown in Table 1.

2.2. Propagation prevention in charge chutes chute (Figs. 2 and 3). It incorporates two inclined baffle
plates welded at the leading edge of the chute and
2.2.1. Test device closely fitted to the chute at the sides. The charge chute
A simple device to prevent flame propagation was was manufactured from square section steel tube 220
designed for installation at the inlet to a charge chute.
The prototype device (Fig. 1) was manufactured from
mild steel plate and was bolted to the top of a charge

Fig. 3. Charge chute propagation prevention device (removed from

Fig. 1. Propagation prevention device for charge chutes. the chute).
 P. Holbrow, A. Tyldesley / Journal of Loss Prevention in the Process Industries 16 (2003) 333340
mm 220 mm and was 1000 mm high and was bolted
directly to the top of the test vessel (Fig. 4).
The principle of using small gaps to prevent ignition
breakthrough, for example in rotary valves, is well
recognised. Typically, in a 250 mm rotary valve the gap
is likely to be 1 mm. However, whilst small gaps are
desired to prevent ignition breakthrough, the gaps in the
propagation prevention device need to be significantly
greater because of the requirement for the free flow of
material. In vented vessels where the explosion pressures
are expected to be low (500 mbar) a larger gap is likely
to be suitable. In the design used for this work a gap of
40 mm at the tips of the baffle plates was employed.
The free flow of material over the inclined plates also
needed to be considered and the angle of repose of bulk
granular materials was taken as evidence of its likely
flow behaviour. Woodcock (Woodcock & Mason, 1987)
considered 2530 degrees to be very free flowing and
3038 degrees to be free flowing. In this device, the
short plate was inclined at 30 degrees and the long plate
inclined at 20 degrees.
2.2.2. Test vessel
In order to produce low explosion pressures (500
mbar) a vented explosion vessel was used. The 2 m3 test
vessel (Fig. 4) was designed to accept a range of
explosion vent openings which were fitted with vent
covers having a burst pressure of 0.1 bar. The dust injec-
tion system was installed to the side of the vessel and a
number of ports were located to the opposite side to
accept pressure transducers and the igniter. Dust was dis-
persed within the explosion vessel by direct injection
Fig. 4. Test vessel with charge chute.
335
under pressure from an external injection system and
was ignited either at the centre or rear of the vessel using
a 50 kJ ignition source. The charge chute was attached
to the vessel at a central position on the top of the cylin-
drical body of the vessel. The propagation prevention
device was bolted directly to the top of the charge chute.
2.2.3. Test procedure
The prototype device was tested under a range of
explosion conditions. Initial tests were carried out to
establish the test conditions that gave the correct reduced
explosion pressure for the KSt of the dust in accordance
with vent sizing equation described in VDI 3673 Part 1
for homogeneous dust distribution. These conditions
were then used in the tests with and without the propa-
gation device attached to the chute. The effectiveness of
the device with respect to its ability to prevent flame
propagation was assessed by comparing the length of the
flame from the chute without the device with the length
of the flame (if any) when the device was installed in
the chute.
2.3. Propagation prevention in pipes
2.3.1. Test device
A propagation prevention device was designed for
installation in a pipe in close proximity to the connection
between the dust inlet pipe and a vessel.
The device was designed for use in a 250 mm diam-
eter pipe/duct and incorporated one, two or three cones,
closely spaced within the duct at 300 mm pitch. The first
cone was 600 mm from the wall of the vessel. Fig. 5
shows a typical arrangement with two cones and Fig. 6
shows the device installed in the side of the test vessel. It
is anticipated that the device would be used with process
vessels such as a bag filter; in such vessels there are
minimal obstructions in the duct to promote the free flow
of the dust laden air into the bag filter and to guard
against blockages. The orientation of the cones was
Fig. 5. Propagation prevention device for pipes.
336 P. Holbrow, A. Tyldesley / Journal of Loss Prevention in the Process Industries 16 (2003) 333340
Fig. 6. Two cone device attached to test vessel.
therefore directed towards the vessel to aid the flow of
dust laden air into the vessel. During a dust explosion,
the opposing faces of the cones will resist and quench
the impinging flame and thus prevent the flame from
propagating further into the pipe.
2.3.2. Test vessel
The 6.3 m3 vented explosion test vessel (Fig. 7) had
a similar vent panel, dust injection and ignition arrange-
Fig. 7. Test vessel with propagation prevention device attached.
ment to vessel used in the charge chute tests. A flanged
opening was positioned halfway along the vessel to
which the 250 mm diameter pipe and propagation pre-
vention device were attached.
2.3.3. Test procedure
In common with the charge chute tests, the initial tests
established the test conditions that gave the correct
reduced explosion pressure for the KSt of the dust. Tests
were then done with and without the propagation device
attached. The 500 mm flange opening on the side of the
vessel was used to attach the pipes and the device. The
effectiveness of the device with respect to its ability to
prevent flame propagation was assessed by comparing
the length of the flame from the pipe without the device
with the length of the flame (if any) when the device
was installed in the pipe.
Tests were also carried out with a 0.25 m diameter
5 m long pipe attached to the vessel to assess whether
or not the device would prevent flame propagation along
the length of the pipe. In these tests, in addition to dust
injection into the vessel, the dust was also deposited in
the pipe in the same concentration to that in the vessel.
3. Experimental resultspropagation prevention in
charge chutes
The results are summarised in Table 2.
3.1. Coal dust tests
3.1.1. Effect of device on Pred
The device partially obstructed the chute and the
effect of this was to increase the reduced explosion
pressures when compared with an open chute. For
example with a vent area of 0.13 m2, the Pred was 135
mbar without the device, and increased to 303 mbar with
the device.
3.1.2. Effect of device on flame lengthvent area 0.13
m2
Without the device installed in the chute, and with an
explosion relief vent area of 0.13 m2, a jet of flame of
approximately 2 m in height was released from the
chute; the length was comparable with the length of
flame from the explosion relief vent. When the test was
repeated with the device installed, the jet of flame from
the chute was completely eliminated. Fig. 8 shows the
extent of flame discharged from the chute in and demon-
strates the extent of flame propagation compared with
Fig. 9 which shows the complete elimination of flame
from the chute.
 P. Holbrow, A. Tyldesley / Journal of Loss Prevention in the Process Industries 16 (2003) 333340 337

Table 2
Summary of results for explosion propagation prevention devices in charge chutes

Dust Vessel volume (m3) Vent area (m2) Device not installed Device installed

Flame length from chute (m) Pred (mbar) Flame length from chute (m) Pred (mbar)

Coal 2 0.26 1.5 111 No flame 129

Coal 2 0.13 2 135 No flame 303
Anthraquinone 2 0.26 2 145 No flame 230
Anthraquinone 2 0.6 2 126 0.5 183
Milk 2 0.13 0.25 128 No flame 170

3.2. Anthraquinone dust tests

3.2.1. Effect of device on Pred

Again, the obstruction effect of the device in the chute
resulted in an increase in the reduced explosion pressure.
For example, with a vent area of 0.26 m2, the Pred was
145 mbar without the device and increased to 230 mbar
with the device.

3.2.2. Effect of device on flame lengthvent area 0.26

m2
Without the device installed, and with an explosion
relief vent area of 0.26 m2, a jet of flame of approxi-
mately 2 m in height was released from the chute. The
flame length was again comparable to the flame length
from the explosion relief vent. When the test was
repeated with the device installed, the jet of flame from
Fig. 8. Coal dust test without the device installed.
the chute was completely eliminated.

3.2.3. Effect of device on flame lengthvent area 0.6

m2
With a larger vent area of 0.6 m2, the jet of flame
from the chute was again 2 m in height. When the test
was repeated with the device installed, a 75% reduction
in the flame height was achieved with a small jet of
flame approximately 0.5 m high propagated from the
chute. The comparative length of the flame from the
chute in these tests is demonstrated in Figs. 10 and 11.

3.3. Milk powder tests

3.3.1. Effect of device on Pred

In common with the coal and anthraquinone tests the
device partially obstructed the chute and increased the
reduced explosion pressures when compared with an
Fig. 9. Coal dust test with the device installed.
open chute.

3.1.3. Effect of device on flame lengthvent area 0.26
m2
With a larger relief vent of 0.26 m2, the jet of flame
from the chute was slightly smaller at 1.5 m. When the
test was repeated with the device installed, the jet of
flame from the chute was completely eliminated.
3.3.2. Effect of device on flame lengthvent area
0.126 m2
Without the device installed in the chute, and with an
explosion relief vent area of 0.13 m2, a small tongue of
flame approximately 0.25 m in height was released from
the chute. No flame was vented from the explosion relief
338 P. Holbrow, A. Tyldesley / Journal of Loss Prevention in the Process Industries 16 (2003) 333340
Fig. 10. Anthraquinone dust without the device installed.
Fig. 11. Anthraquinone dust with the device installed.
vent. When the test was repeated with the device
installed, the jet of flame from the chute was completely
eliminated. A 2 m long flame was vented from the front
face of the vessel.
4. Experimental resultspropagation prevention in
pipes
The results are summarised in Table 3.
4.1. Coal dust tests
4.1.1. Effect of device on Pred
The device acted as an obstruction in the pipe and
tended to increase the reduced explosion pressures when
compared with an open pipe. For example, with a vent
area of 0.96 m2, the Pred was 150 mbar without the
device and increased to 161 mbar with the device.
4.1.2. Effect of device on flame lengthvent area 0.96
m2
Tests were carried out with an explosion vent area of
0.96 m2. With a 250 mm diameter 900 mm long pipe
attached but without the device installed, a jet of flame
approximately 3 m long was ejected from the pipe.
Under the same conditions but with the two-cone propa-
gation prevention device installed, flame transmission
from the pipe was completely eliminated. The compara-
tive flames from the pipe in these tests is demonstrated
in Figs. 12 and 13.
4.1.3. Effect of flame lengthvent area 0.41 m2
The tests were repeated but with a smaller vent area
of 0.41 m2. With a 250 mm diameter 900 mm long
pipe attached but without the device installed, a 2 m long
jet of flame was ejected from the pipe. Under the same
conditions but with the two cone propagation prevention
device installed, flame transmission from the pipe was
completely eliminated.
4.1.4. Effect of flame lengthvent area 0.27 m2
The vent area was decreased further to 0.27 m2. With-
out the device installed, a 1 m long jet of flame was
ejected from the pipe. Under the same conditions but
with the two-cone propagation prevention device
installed, flame transmission from the pipe was com-
pletely eliminated.
4.2. Anthraquinone tests
4.2.1. Effect of device on Pred
The device acted as an obstruction in the pipe and
tended to increase the reduced explosion pressures when
compared with an open pipe. With a vent area of 0.96
m2, the Pred was 395 mbar without the device and
increased to 451 mbar with a two-cone device fitted.
4.2.2. Effect of flame lengthvent area 0.96 m2
With a vent area of 0.96 m2, a jet of flame approxi-
mately 5 m long was ejected from the 0.9 m long pipe.
Under the same test conditions, but with the two-cone
device fitted, the flame length was reduced to a length
of 2 m. In a further test with three cones fitted to the
device, flame transmission was completely eliminated
from the pipe. The comparative flames from the pipe in
these tests is demonstrated in Figs. 14 and 15.
4.2.3. Effect of flame lengthvent area 0.66 m2
With a vent area of 0.66 m2, a flame of approximately
4 m long was ejected from the 1.2 m long pipe. The
two-cone and three-cone devices were tested, with a
reduction in the flame length to 1 m using each of the
devices.
 P. Holbrow, A. Tyldesley / Journal of Loss Prevention in the Process Industries 16 (2003) 333340 339

Table 3
Summary of results for explosion propagation prevention devices in pipes

Dust Vessel volume (m3) Vent area (m2) Flame length from pipe

Device not installed Device installed

Coal 6.3 0.96 3 m No flame (2 cones)

Coal 6.3 0.41 2 m No flame (2 cones)
Coal 6.3 0.27 1 m No flame (2 cones)
Anthraquinone 6.3 0.96 5 m 2 m (2 cones)
Anthraquinone 6.3 0.66 4 m 1 m (2 cones)
Anthraquinone 6.3 0.96 5 m No flame (3 cones)
Anthraquinone 6.3 0.66 4 m 1 m (3 cones)
Milk 6.3 0.27 4 m 0.25 m (1 cone)
Milk 6.3 0.27 4 m No flame (2 cones)

Fig. 14. Anthraquinone dust without the device installed.

Fig. 12. Coal dust test without the device installed.

Fig. 15. Anthraquinone dust with the device installed.

Fig. 13. Coal dust test with the device installed.

4.3. Milk powder tests

4.3.1. Effect of device on Pred

With a vent area of 0.27 m2, the peak Pred was 130
mbar mbar without the device. The Pred increased to 181
340 P. Holbrow, A. Tyldesley / Journal of Loss Prevention in the Process Industries 16 (2003) 333340
mbar with the one-cone device fitted. The Pred did not
exceed 100 mbar with the two-cone device fitted.
4.3.2. Effect of flame lengthvent area 0.27 m2
With a vent area of 0.27 m2, a jet of flame approxi-
mately 4 m long was ejected from the 0.6 m long pipe.
When this test was repeated with a one-cone device fit-
ted, the flame length was reduced to 0.25 m and when
the test was repeated with a two-cone device fitted, the
flame was completely eliminated.
5. Conclusions
5.1. Propagation prevention device for use in charge
chutes
The tests have demonstrated that dust explosions in a
2 m3 vessel, with low explosion pressures, will propagate
significant flames through an attached open charge chute.
Coal and anthraquinone flames from the open charge
chute were up to 2 m long and had the potential to cause
serious injury to any operator close to the chute. Milk
powder, the least reactive of the three test dusts, was the
exception and had a flame of only 0.25 m long.
The attachment of the explosion propagation preven-
tion device to the charge chute resulted in the total elim-
ination or significant reduction in the flame from the
charge chute. The results demonstrated the effectiveness
of the device with explosion pressures within the range
111 mbar to 303 mbar. The device was particularly
effective with coal dust, the results showing a complete
elimination of the propagating jet of flame from the top
of the charge chute. With the more reactive anthraqui-
none dust, the device was again very effective, but a
small flame of 0.5 m was observed. The milk powder
flame was completely eliminated with the device.
In the event of a dust explosion in a 2 m3 process
vessel (or similar size vessel) where low explosion press-
ures are anticipated, this type of device would provide
a high level of protection to personnel. The dusts tested
had KSt values of 86 bar m/s, 155 bar m/s and 308 bar
m/s and consequently the device would provide protec-
tion from most dusts within the St1 and St2 dust groups.
Although the device described in this paper has been
very effective in these tests, it may require further
refinement and development to suit other situations.
Without the device, the open chute was effectively an
explosion vent. However, when the device was installed
in the chute, the venting capability of the chute was
reduced and this increased the explosion pressure.
5.2. Propagation prevention device for use in
pipelines
The device was successful in either eliminating or
achieving a significant reduction in the flame from the
vessel over a range of vent areas. Without the device
installed, it was demonstrated that dust explosions in the
vented 6.3 m3 vessel can propagate extensive flames
through an open pipe stub attached to the side of the
vessel. The pipe stub was approximately the same length
as the propagation prevention device (0.30.9 m). Flame
lengths measured from the pipe stub, without the device
installed, were approximately 4 m using milk powder,
13 m using coal dust and 45 m using anthraquinone.
When the propagation device, comprising two or three
cones, was installed the flame from the pipe stub was
generally eliminated or greatly attenuated.
Flame from the coal and milk powder was completely
eliminated using a two-cone device. A three cone device
was successful in the complete elimination of the anthra-
quinone flame when a relatively large vent was installed
(0.96 m2), and 75% reduction in flame length was achi-
eved when a smaller vent configuration was used (0.66
m2). Even though it was not possible to completely elim-
inate the anthraquinone flame, the 75% reduction in
flame length would reduce the potential for ignition of
dust in the pipe. However, the attenuated anthraquinone
flame was still capable of igniting the dust in a 5 m long
pipe attached to the device and causing propagation of
flame down the length of the pipe.
The device, in its current form is therefore not suitable
for reactive dusts similar to anthraquinone (KSt 308 bar
m/s). However, with further development it may be poss-
ible to achieve a more effective device for use with the
more reactive dusts.
In general, the presence of the device in the pipe acted
as an obstruction and resulted in an increase in the
reduced explosion pressure within the vessel.
References
Holbrow, P., Hawksworth, S. J., & Tyldesley, A. (2000). Thermal radi-
ation from vented dust explosions. Journal of Loss Prevention in
the Process Industries, 13, 467476.
Woodcock, C. R., & Mason, J. S. (1987). Bulk solids handling: An
introduction to the practice and technology. Chapman and Hall.
 Journal of Loss Prevention in the Process Industries 16 (2003) 165172
www.elsevier.com/locate/jlp

Silo vulnerability: structural aspects

Guy Mavrot a, Isabelle Sochet b,, Patrice Bailly b, Michel Moisescot c
a
INERIS, DRA/REST, B.P. 2, 60550 Verneuil-en-Halatte, France
b
ENSIB, 10 boulevard Lahitolle, 18020 Bourges, France
c
Structural advisory engineer, 13 Avenue du Cedre, 38 700 Corenc, France

Abstract

This document sets out a diagnostic approach regarding the mechanical resistance of silos to an internal explosion that could
occur within them. Even though prevention is logically a priority, risk analysis is about studying contingencies and vulnerability.
Presented herein are risk assessment methods including methods for quantifying the effects of an explosion on structure, as well
as methods for assessing structure vulnerability. The data required for the vulnerability research approach are defined. The location
of sampling holes, which may prove necessary for diagnosis purposes, is more fully developed for tower silos. A stability analysis
for the different silo components is established. This approach is useful for risk analysts and allows them to highlight the sensitive
elements in order to integrate the protection aspect into an existing facility or a facility to be constructed.
2003 Elsevier Science Ltd. All rights reserved.

Keywords: Structure vulnerability; Silo vulnerability; Structure rupture; Dust explosion

1. Explosion hazards the dust must be in suspension; and

the concentration of dust must be within the explos-
1.1. A significant risk ible domain.

Studying silo vulnerability involves conducting
research into all possible scenarios. That is why analyz-
ing accidents occurring in silos (Fig. 1) for instance,
demonstrates that in addition to a diagnostic approach
to the explosion hazard, it is imperative that a diagnostic
approach to the structures behavior be conducted.
1.2. Contingencies: dust explosion
1.2.1. Explosible atmosphere
The explosion hazard diagnosis approach should be
organized by following the process of an explosion. In
order for an explosible atmosphere to occur, the follow-
ing four conditions should be present simultaneously:
the dust must be flammable;
the atmosphere must contain combustible material;
Corresponding author. Tel.: +33-2-48-48-40-46;fax: +33-2-48-48-
40-40.
E-mail addresses: guy.mavrot@ensi-bourges.fr (G. Mavrot); isab-
elle.sochet@ensi-bourges.fr (I. Sochet); patrice.bailly@ensi-bourges.fr
(P. Bailly).
Evidently, for an explosion to actually occur, the
explosible atmosphere must encounter a source of
inflammation such as a hot point, an electrostatic dis-
charge, etc. Confinement is important because it causes
pressure effects to increase significantly.
Dust explosibility is a difficult matter to study because
of the high number of properties the dust has, unlike
gases. These characteristics are (Eckhoff, 1991): dust
granulometry and humidity rate, concentration, oxygen
rate, temperature, pressure, turbulence, confinement, vol-
ume and dusty atmosphere flammability minimum
energy.
1.2.2. Enclosure wall pressure
An enclosure rupture mode is contingent upon the
pressure signal received by it. This signal is, at some
point, the evolution of pressure over time P(x,y,z,t). Two
situations should be considered:
the combustion mechanism occurs at low speed in
relation to the celerity of sound waves through the gases
contained in the enclosure; in this case one has to con-
sider a uniform pressure in space P(t); more often than

0950-4230/03/$ - see front matter 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0950-4230(02)00092-X
166 G. Mavrot et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 165172

Fig. 1. Accident in Blaye, France, August 1997 (11 dead) (a) before the accident; (b) after the accident.

not this assumption is supported for a dust explosion
which, in a number of instances, is a relatively slow
deflagration,
the burning rate may increase for different reasons,
i.e. because of turbulence or overpressure effects, and
the explosion may move toward a supersonic deton-
ation rate.
Let us consider a reinforced concrete wall receiving
a dynamic load to one side.
The phenomena observed (Fig. 2) are as follows:
compaction of concrete near loaded side,
scaling of non-loaded side, and
shear failure or bending of the structure.

It is usually accepted that a pressure increase rate not
exceeding 1 kPa/s leads to a loading that may be mod-
eled almost statically, while a rate of the order of 100
kPa/s induces a dynamic that should not be neglected
(Baker, Cox, Kulescz, & Strehlow, 1983).
The rupture occurring as a result of the explosion cre-
ates breaks and throws fragments to the exterior. Frag-
mentation is accompanied by a blast wave through air
which propagates through the environment. Secondary
explosions may occur as a result of the wave propagation
(Crowhurst, Colwells, & Hoare, 1995). The internal
overpressure then evolves according to the reactants
being discharged through the breaks resulting from frag-
mentation and through natural openings.
It should be underscored that all of these loading
modes have been taken into account in the EFFEX
software (Proust, 1997; Proust, Roux, & Chhuon, 2000).
The mechanisms are herein described separately, but
they are actually related and simultaneous.
1.3.2. Compaction of loaded side
A few micro-seconds after the explosion and in the
case of high-intensity strains, the loaded side is subjected
to high pressure; the concrete may get superficially pul-
verized. A decrease in volume occurs as a result of the
collapse of the porous structure of the material. Pressure
levels in silo explosions are generally too low for this
to happen.
1.3.3. Wave reflection scaling to non-loaded side
Scaling of a reinforced concrete element results from
a break caused by a traction occurring perpendicularly
to its free surface. This very rapid phenomenon is linked

1.3. Vulnerability and structure rupture

1.3.1. Elementary rupture phenomena in a reinforced

concrete wall
This presentation aims to set out a diagnostic approach
as regards the resistance of a reinforced concrete struc-
ture to an explosion which may occur within it. In this
context, the primary matter of interest is the current situ-
ation, one should outline the type of approach requiring
a subsequent appraisal aiming, in this instance, to protect
these structures to the extent possible against acci-
dental overpressures which may occur within them, i.e.
aiming to safeguard the greatest number of human lives
in the immediate vicinity of the facilities and their
environment and, alternatively, to keep damage to the
buildings to a minimum. Fig. 2. Wall rupture mechanisms.
 G. Mavrot et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 165172 167

to the propagation of the wave through the material.

Typically, in a 20-cm (7.9 in) thick concrete wall, the
time for a wave to reach the free surface is about 60 s.
When a wavefront hits the surface of an element, com-
pression loads are passed onto the element. These strains
propagate through the element by a serrated com-
pressional wave which lasts as long as the strain does.
Upon reaching the free side opposite the strain, a
reflected traction wave is generated, identical in shape
and in amplitude to the compressional wave. When this
reflected wave hits the end of the compression slot,
expansion occurs within the material, which may lead to
rupture if it exceeds the expansion capability of concrete.
The material found between the rear free surface and the
fracture plan will be ejected. Part of the stress wave is Fig. 3. Fragment shapes on a plate subjected to a shock tube
(Pontiroli, 1995).
picked up by the fracture and is instrumental in its pro-
gression the other part continues to propagate, pro-
ducing reflections and fractures within the concrete until 2. Case of an existing building
damping is achieved.
2.1. Structure analysis
The fragments ejected through scaling are small. Sca-
ling, like fragment speed, is linked to the pressure signal.
The approach requires the knowledge of at least a few
Experience has shown that the part ejected through direct
characteristics of the structure under consideration. The
scaling corresponds to the envelope found between the
characteristics are as follows:
free surface and the outermost layer. This phenomenon
has been observed and validated experimentally through mechanical and relative to the nature of the materials
dynamic tensile tests (Brara, 1999). used (tensile strength of the concretes used, nature of the
steels used and yield strength),
geometrically constructive (apparent contour lines of
1.3.4. Wall bending the various elements) as well as operationally.
Bending takes much longer than scaling; it is more
dependent upon the structure and associated boundary Regarding the original structure, it should be pointed
conditions than upon the local behavior of the material. out that this set of characteristics should be recorded in
Typically, the time referred to is the time of the vibration the documents pertaining to the contract entered into at
mode periods. It should be underscored that shear failure the time of the construction, such as:
in slabs under the action of shear force may occur
dynamically, but hardly ever under static loading. the description of the work,
the specifications,
the stability calculations that support the adopted sec-
1.3.5. Frame influence tions, and
The influence of frames on the shape of ejected frag- the work plans (formwork, frames and steel
ments is worthy of note; the concrete cube included in nomenclature).
a grid is detached (Fig. 3). In works specially designed
to withstand shocks, some frames, also referred to as In practice, however, these documents remain unavail-
lacing frames, keep chip projection to a minimum. able, for a number of (legal) reasons. Therefore, a prag-
Fragment typology is shown in Table 1. matic attitude consists in having a general outlook on
This paper is not intended to verify whether or not the structure and to divide it up into various components.
the structures are in compliance with the regulation in Each component should be sampled in order to record
force at the time of their construction. This article deals the characteristics that are relevant to the subject of
exclusively with the structural aspect, without attempting our discussion.
to typify the explosion phenomenon itself. This is an
absolutely essential first step for risk analysts who may 2.2. Data required for diagnosis
take action on the original structure in order to provide
protection against the effects of a blast. Of special con- 2.2.1. Sampling
cern here are concrete grain storage tower silos. In the absence of all or part of the mechanical charac-
teristics, sampling must be conducted; the central pur-
168 G. Mavrot et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 165172

Table 1
Fragment typology

Types of Load elements related to Elements affecting fragment size Elements affecting fragment projection speed Ve
fragments this rupture

Scales Overpressure peak Duration of overpressure peak sr

V e
rCe
Reinforced concrete coating r = density
Ce = wave celerity
sr = traction resistance
Blocks Pressure impulse Reinforcement mesh Ve # wall velocity
Bending response Transient response (plate)
Gravel and dust Material characteristics Ve # blast particle velocity
Dynamic multifissuration

pose of this is to provide information on concrete quality
and on the nature and geometrical layout of the frames.
In principle, it should be pointed out that sampling
locations are to be determined in a timely manner, so that
relevant information may be obtained from a minimum
number of samplings.
It is assumed that the determination of sampling
locations depends upon: (i) the activity and functions
related thereto, within the structure; (ii) the nature of the
structure and of the various components. There will be
a focus on structure resistance under the combined action
of the contract loads and of the loads resulting from acci-
dental overpressure that may occur within the structure.
When accidental overpressure occurs within the struc-
ture it should be underscored, first and foremost, that
horizontal structural elements (floors and associated
components) are usually and exclusively calculated to
carry loads, hence to withstand descending actions.
2.2.2. Tower silo
2.2.2.1. Design characteristics The structure of a con-
crete storage tower silo (Fig. 4) has some variations and
other general arrangements. Tower silos are sets of
contiguous cells whose fundamental aspect is that they
are much higher than they are wide, regardless of the
geometric design of each tower. The following are com-
monplace:
cylindrical cells with a square or rectangular section,
circular cylindrical cells which, when laid flat, may
either be fitted into one another in a staggered pattern,
or arranged mutually tangentially four by four, thereby
enclosing a space which may or may not be in the shape
of a wedge whose cross-section is that of an ace of
diamonds and is usually referred to as such. There usu-
ally are two or three rows enclosing one or two rows of
aces of diamonds. Using an ace of diamonds for stor-
age purposes poses specific strength problems. Again, it
should be underscored that in some very rare cases, these
cells may be divided up into two, three or four segments
by the use of vertical webs whose cross-sections are
diametral, Y-shaped or cross-shaped. These webs may
be built at the time of silo construction or added at a
later stage. Below are statements concerning this latter
type of situation which poses specific strength problems.

Usually a gallery called a top gallery is mounted

above these sets of cells. The main purpose of this gal-
lery is to house the horizontal handling devices, such as
riddlers or elevating conveyors. The geometry and struc-
ture of that gallery can vary considerably. The gallerys
roofs also come in very diverse designs: onduline (wave
tile) roofs, reinforced concrete slab roofs, and so forth.
Shingled steel framings can also be found.
Grain is usually hoisted up to the top of the cells by
means of elevators. These elevators almost always
bucket elevators are contained in a handling tower
which ordinarily only houses the elevators; however, a
great number of possible configurations and structures
may be encountered.

2.2.2.2. Sampling locations It is known from experi-

Fig. 4. Storage silo, tower type structure. ence that samplings are first and foremost performed:
 G. Mavrot et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 165172
at various heights on the cell wall itself. A distinction
should be drawn here between circular and rectangular
cross-section cells. Therefore, in practice, three sam-
plings should be performed: one in the higher section,
one in the lower section and at least another one at mid-
height. It is absolutely imperative that a sampling be per-
formed slightly below the cell roof. Under normal oper-
ational conditions, this area is not awash in grain. The
analysis of some concrete examples demonstrated that
those sections not subject to ensiled matter load under
normal operational conditions have a very low level of
reinforcement and are incapable of resisting any signifi-
cant internal overpressure;
in the drawbell, designating the funnel-shaped device
designed to facilitate grain flow by gravity. It is
important to know how the drawbell is attached to the
post. Samplings therefore concern the securing arrange-
ment as well as the drawbell wall. An additional sam-
pling should be performed to search for radial reinforce-
ment.
The following samplings concern the cell roof slab,
insofar as the slab includes all or part of the reinforced
concrete. Also, the samplings to be performed should
allow the characterization of the roof slab design with
regard to its capability to withstand ascending loads. It
should be clearly understood that those ascending efforts
result from accidental internal overpressure.
The types of wall designs and top gallery roofing
should then be examined.
3. Vulnerability study
3.1. Procedure
The aim of the approach outlined hereafter is to evalu-
ate only statically the accidental internal overpres-
sures the existing structure under study would be capable
of resisting without sustaining important damage.
It is of primary importance to determine, by way of
example, if in the case of an internal explosion:
total structure destruction is to be expected, with
other adjacent or neighboring sections collapsing and
being buried, and various projectiles being thrown into
the air;
the structure, albeit severely damaged and unsuitable
for later use, remains in place, thereby limiting the
effects of the explosion on the environment.
To do this, on the one hand the weakest parts of the
structure should be identified, how these parts are inte-
gral with the framework or not, their capability of being
blown apart or ripped open under pressure, and, on the
other hand, the nature, shape and speed of the resultant
169
projectiles, as well as where these projectiles may be
thrown.
The purpose of this article is to provide diagnostic
elements only statically in response to the first
issues just discussed. In other words, the matter of prime
concern is the Ultimate Limit State (ULS), defined to
mean that the works or an element thereof has been dam-
aged as a result of loss of static balance, rupture, buck-
ling, etc. (Perchat & Roux, 1999). Evidently, this ulti-
mate limit state should in no way be confused with the
Serviceability Limit State (SLS), defined as the limit
state for normal operational and durability conditions.
This approach is obviously analytical and takes the
various structural elements into consideration in suc-
cession. A conclusion can be drawn with regard to the
entire structure under study and by highlighting the
weakest parts and the actions likely to cause damage. It
is only after such diagnosis and having regard to the
various explosion modeling patterns irrelevant for
discussion here that the analyst can, in practice, define
the possible measures to be taken, keeping in mind the
protection of a storage unit against the effects of an
explosion which could originate therein.
3.2. Data to be considered
The data to be considered primarily comprise the
characteristics of the ensiled matter: its volume weight
g kg/m3 and angle of repose f in degrees (Fig. 5). Also
subsequently considered are the geometrical data typical
of the container: its hydraulic radius (m), defined as the
quotient of its cross-section (m2) through the perimeter
(m) of this section, the dimension of its maximum
medium filling plane. It is therefore inferred that when
the ensiled matter is flowing from an opening in the bot-
tom of the cell roof, it follows the shape of a cone typi-
fied by the internal friction angle. This maximum
Fig. 5. Elements for determining grain thrust.
170 G. Mavrot et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 165172
medium filling plane is by definition the horizontal plane
including the center of gravity of the said cone. Through-
out the following text, the dimensions of this plane are
regarded as the origin of the depth vertical axis.
3.3. Stability analysis of various types of cells
For each type of cell, the approach is about evaluating
the side thrusts resulting from the ensiled matter and,
subsequently, comparing them with the total pressure
compatible with the reinforcement.
From experience the DIN 1055 approach should be
accepted with some degree of caution among all the
approaches put forward by Reimbert & Reimbert, 1959,
Albiges & Lumbroso, 1964, Laredo, 1964, Guerrin,
1969, Lumbroso, 1970, 1977. In most cases, its use pro-
vides the most significant thrust values and as a result
leaves the narrowest possible margin to statically evalu-
ate the accidental overpressure the structure can with-
stand without sustaining important damage.
3.3.1. Circular cylindrical cell
The DIN 1055 standard allows us to define the two
corresponding curves:
for the first curve, to the extreme shell of the side
thrusts to which the wall is subjected while filling. The
points on this curve correspond to a completely full cell;
for the second curve, to the extreme shell of the same
thrusts, only while draining.
After evaluating the side thrusts, the resistance offered
by the cylindrical cell must be calculated on the basis
of its reinforcement level.
Insofar as such a cylindrical cell is only under traction
load, a simple calculation, depending upon ring
reinforcement, allows us to plot on the same diagram
as previously and according to the various steel sections
highlighted at the various sampled dimensions the
total side thrusts (ensiled matter+overpressure) the struc-
ture is capable of withstanding (Fig. 6).
Fig. 6. Side thrust diagrams filling & draining (DIN 1055).
Comparing both curves allows us to determine the
area of the cylindrical cell post where the difference
between the capable total pressure and the pressure
resulting from the ensiled matter load is the weakest.
Regarding this, the authors found out that in some
cases the said area of concern was located just below
the roof bottom, especially as a result of very weak
reinforcement in the area which is not awash in grain.
A number of methods are available to evaluate side
thrusts on the walls, while filling or draining. These cal-
culations are of interest only when verifying that the
drawbell is securely attached. From experience, in most
cases due consideration is given to this aspect at the time
of construction.
3.3.2. Aces of diamonds: a specific issue
The horizontal and vertical components resulting from
ensiled matter thrust are analogous to the foregoing (see:
circular cylindrical cell).
The walls are composed of four quarter-circle shaped
webs. As a result, the loads are strictly evaluated accord-
ing to the type of links between the webs and the adjac-
ent circular cells.
It should be underscored that under the load of the
ensiled matter, each of these four webs applies a certain
degree of tangential thrust (p.r) to the cells with the pur-
pose of separating them.
A number of authors, including Lumbroso, 1970,
1977, Albiges & Lumbroso, 1964, Laredo, 1964 et al.
have attempted to theoretically identify the value of that
thrust taking into account the unquestionable aspect of
redundancy. However, they have based themselves upon
assumptions that have never been applied in the field.
In practice, the area of contact between two circular
cells even if its outline resembles a beam does
not feature the specific framing allowing it to be verti-
cally flexible.
Consequently, the structure is balanced through the
use of adjacent circular cells that should reasonably be
considered empty in order to deal with the most unfavor-
able of cases. These cells can withstand thrusts from
ensiled matter within the ace of diamonds by working
like a vertically bent tube, which in practice is embedded
at the foot and almost always anchored at the top.
The calculation consists in checking tensile stresses
which should remain below the values currently con-
sidered as admissible for that material. Thus, from
experience and for very tall silos (more than 20 m or 65
ft in height) this latest condition is hard to achieve.
3.3.3. Cylindrical cells with rectangular cross-sections
Side thrusts are evaluated by considering the rectangu-
lar cross-section cylindrical cell of the same height and
with a diameter that is equal to the diameter of the circle
at the section rectangle. Again, this is a cautious
approach since holding back the marked-out circle is like
 G. Mavrot et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 165172
increasing the thrust. This results in the most unfavor-
able situation in terms of structure or full angle cell,
while adjacent cells are empty. Then, the cell resistance
to grain thrust is characterized, and one has to reflect on
an isolated, contained enclosure given the low inertia of
the walls of the empty adjacent cells.
For this purpose, one must take into account that each
side is subjected to spanwise positive bending moments,
negative at angles, and to normal tensions resulting from
adjacent thrusts.
3.3.4. Subdivided cells
In rectangular cells, where subdivision has not been
anticipated and was performed later, it is important to
know whether the added wall, in certain load cases, does
not cause stress that might jeopardize the wall spanwise
stability against which it is supposed to lean.
Because of its geometry and the values that commonly
apply to the thickness of its concrete walls, and because
of the section of its horizontal frames, a circular cell
is designed to withstand radial loads that are uniformly
distributed around the circumference. This occurs with
the horizontal loads resulting from the ensiled matter or
loads resulting from cell overpressure.
In addition, unavoidable punctual actions due to the
presence of separating partitions should be considered;
stresses occur through their full height at junction points
of the partitions and walls. Under such conditions, it is
easily understood that the structure cannot withstand this
type of stress, under various load cases (full or empty
volumes).
Consequently, it should be underscored that the sub-
division of circular cells through the use of vertical webs
is not at all an option and should not be considered in
the design of the cells.
3.3.5. Abutment slabs of cells
Unlike the perspective used for the cell framework,
the risk analyst should consider to what extent cell col-
lapse can be avoided and how they can be protected from
explosive effects, for instance, through the weakening of
the abutment slab or the provision (on some of its
openings) of explosion vents and restraints. This should
be construed to mean the avoidance of projection of
fragments into the environment.
3.3.6. Fragment projection
A method to assess the projection velocity of large
fragments is proposed in the regulation TM 5-1300.
Fragment velocity depends upon the excessive impulse
defined as the difference between blast impulse and the
section ripped away as a result of the rupture. This is
the area limited by the time resistance curve (Fig. 7).
Although different fragments can travel at different
speeds, the average debris projection speed may be
approximately assessed from excessive impulse ie and
wall unit weight m:
171
Fig. 7. Bending resistance diagram for a wall according to with-
standable pressure.
ie m.vf(avg).
The assessment is based upon the following:
element is at rest prior to loading,
the duration of the applied blast charge and the time
to rupture are low if compared to the time required for
ultimate deviation.
More precisely, if a post-ultimate resistance is taken
into account, this assessment becomes:
ib ru.t1 rup.(tut1)
vf ,
mu mu mup
where ib=applied charge unit impulse (ib=pressuretime),
mu and mup=represent the equivalent weights of a system
with one degree of freedom (representing the bent
element) in ultimate and post-ultimate states, ru=unit
ultimate resistance, rup=unit post-ultimate resistance,
t0=time of blast pressure positive phase, t1=partial break
moment, tu=ultimate deviation moment, vf=average frag-
ment speed after rupture.
4. Conclusion
This approach basically allows us to:
show that the examination of the cell framework
resistance brings us around to considering admissible
static overpressure values for accidental explosions, in
ultimate limit state, by a maximum of a few hundreds
of millibars,
look at the attachment and strength of the drawbell
and how the abutment slab is constructed. This step is
part of a general arrangement which includes the top
gallery, the handling tower, the undercell crawlspace as
well as the ace of diamonds, if any,
show that to provide balance in a relatively high stor-
172 G. Mavrot et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 165172
age unit (in excess of 20 m in height) the contact areas
between the adjacent circular cells, providing an appar-
ent beam outline vertically, shall be reinforced.
inform risk analysts of the opening possibilities in
the cell covering floor, undercell gallery, handling tower,
show that to provide protection against explosions,
all newly erected concrete storage units shall technically
include: (i) frames whose section provides enough mar-
gin for overpressure, especially in the area not awash in
grain, and under vertical bending; (ii) large openings on
top of each cell and in the top gallery, handling tower
and undercell crawlspace, equipped with explosion vents
and restraints; (iii) efficient separations between each
space and adjoining spaces to avoid spreading and, poss-
ibly, building up of the explosion to neighboring attach-
ments.
References
Albiges M., & Lumbroso A. (1964). Silos a cellules principales circul-
aires et intermediaires en as de carreau. Annales de lInstitut Tech-
nique du batiment et des travaux publics, no. 204.
Baker, W. E., Cox, P. A., Kulescz, J. J., & Strehlow, R. A. (1983).
Explosion hazards and evaluation. New York, USA: Elsevier.
Brara, A. (1999). Etude experimentale de la traction dynamique du
beton par ecaillage. These, Universite de Metz, France.
Crowhurst D., Colwells S.A., & Hoare D.P. (1995). The external
explosion characteristics of vented dust explosions. IchemE Symp.
Series no. 139.
Guerrin, A. (1969). Traite de beton arme. France: Dunod.
Eckhoff, R. K. (1991). Dust explosions in the process industries.
Oxford: ButterworthHeinemann.
Laredo M. (1964). Etude tridimensionnelle des cellules as de carreau
des grands silos cylindriques. Memoire enregistre au Secretariat
des Annales des Ponts et Chaussees, France.
Lumbroso A. (1970). Determination numerique des sollicitations exer-
cees par la matiere ensilee dans les silos. Annales de lInstitut Tech-
nique du batiment et des travaux publics no. 267268, France.
Lumbroso A. (1977). Bases scientifiques du calcul des enceintes
renfermant des massifs pulverulents et du calcul des silos. Annales
de lInstitut Techniques du batiment et des travaux publics no.
346, France.
Perchat, J., & Roux, J. (1999). Pratique du B.A.E.L. 91. France: Edi-
tions Eyrolles.
Pontiroli C. (1995). Comportement au souffle des structures en beton
arme analyse experimentale et modelisation. These de doctorat
ENS Cachan, France.
Proust Ch. (1997). Resistance des structures de lindustrie agroalimen-
taire aux effets des explosions de poussieres. Communication pre-
sentee a la journee DRIRE organisee par le ministere de lenviron-
nement, Paris.
Proust Ch., Roux P., & Chhuon B. (2000). Prevoir les effets des
explosions de poussieres sur lenvironnement, EFFEX, un outil de
simulation. Rapport INERIS, France.
Reimbert, A., & Reimbert, M. (1959). Silos, theorie et pratique.
France: Editions Eyrolles.
TM 5-1300. Structures to resist the effects of accidental explosions.
Department of the Army, the Navy and the Air Force, Technical
manual NAFVAC 397/AFM 88.
 Journal of Loss Prevention in the Process Industries 16 (2003) 2732
www.elsevier.com/locate/jlp

Ignitions of explosive dust clouds by smouldering and flaming

agglomerates
J. Gummer, G.A. Lunn
Health and Safety Laboratory, Harpur Hill, Buxton, Derbyshire, SK17 9JN, UK

Abstract

Surveys show that burning nests of dusts have been ignition source for dust clouds involved in industrial explosions. The literature
indicates, however, that hot nests are a poor ignition source and are difficult to convey through powder handling plant. This paper
describes some test in which clouds of dusts with a range of Minimum Ignition Temperatures (MITs) were dispersed around dust
agglomerations smouldering and flaming at various temperatures. Smouldering nests of dusts prove to be poor ignition sources for
most dust clouds, failing to ignite dusts even when there is a large difference between the nest temperature and the MIT of the
dust cloud. Smouldering nests with temperatures above approximately 700800 C were, however, able to ignite sulphur clouds.
Flaming nests, on the other hand, were able to ignite clouds of dusts up to the maximum MIT used, 600675 C.
Crown copyright 2003 Published by Elsevier Science Ltd. All rights reserved.

Keywords: Dust explosion; Agglomerates; Ignitions

1. Introduction deal of work done on the potential for ignition by the

Surveys of industrial dust explosion incidents show
that, in a substantial percentage, friction and mechanical
failure and flames and flaming material are known
ignition sources. Surveys for the UK (Abbott, 1985;
Porter, 1989) covering 19791988, and reviewing 303
events, showed friction and mechanical failure to be the
reason for ignition in 18% of these incidents, and flames
and flaming material to be responsible in another 15%.
Overheating and spontaneous heating featured in a
further 17%. Similarly, a survey by the Berufsgenossen-
schaftliches Institut fur Arbeitssicherheit (BIA) (Jeske
and Beck, 1989) showed mechanical causes to be the
most frequent source of ignition, with smouldering nests
the second most frequent. The relevant percentages from
the BIA survey were 26% for mechanical sparks, 11%
for smouldering nests and 9% for mechanical heating.
Hazardous mechanical friction in dust handling plant
is usually accidental: misalignment of components as in
fans, mixers or mills, the presence of material trapped
in conveying equipment. Although there has been a great
Corresponding author. Tel.: +44-114-2892005; fax: +44-114-
2892010.
E-mail address: geoff.lunn@hsl.gov.uk (G.A. Lunn).
effects of mechanical sparks and frictional heating, there
is no generally accepted method of estimating the likeli-
hood of ignition from mechanical sources in relation to
dusts. The information available in the literature is not
sufficiently wide to give a guidance framework that
would have wide use.
Hot surfaces are capable, if the temperature is suf-
ficiently high, of igniting surrounding dust accumu-
lations. The layer ignition temperature is measured in a
standard test for a depth of 5 mm (ISSA, 1998), but
because of the insulating effect of dusts, thicker deposits
can ignite at lower temperatures (Testing methods for
electrical apparatus installed in a dusty environment with
a potential risk of explosion, 2001). The practical dang-
ers are that a smouldering or burning layer can act either
directly as an ignition source for a dust cloud or by
means of agglomerations or nests of burning material
that break away from deposits and ignite a dust cloud
in another part of the plant.
Although explosion incidents are attributed to
ignitions by nests of burning dust, it has been difficult
experimentally to produce long-lasting, coherent clumps
of smouldering powder that can travel through a powder
handling plant (Pinkwasser, 1985). In addition, the litera-
ture indicates that nests are a relatively poor ignition
source and only if they either break up in flight or hit the

0950-4230/03/$ - see front matter. Crown copyright 2003 Published by Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0950-4230(02)00087-6
28 J. Gummer, G.A. Lunn / Journal of Loss Prevention in the Process Industries 16 (2003) 2732
floor does the risk of an ignition substantially increase
(Zockoll, 1989; Alfert et al., 1988).
The risk of an explosion due to hot nests encountering
a dust cloud has been studied in this project. Burning
nests with different characteristics have been introduced
into clouds of dusts covering a range of Minimum
Ignition Temperatures (MITs). Dusts with different
Train Firing properties have been used as the nest
material and the dust clouds have MIT values of
approximately 400, 500 and 600 C.
2. Literature review
Harper, Plain and Gibson (1997) have discussed the
burning behaviour of powder accumulations on hot sur-
faces. The stages of ignition and the form of the combus-
tion zone can be complex. Some powders burn directly
in the solid phase either with a flame or by smouldering,
others melt and burn as a liquid, whilst some burn with
a large amount of flame. Some dusts can evolve large
amounts of flammable gas when subjected to heat. A
change from solid to liquid or agglomeration/expansion
of dust particles to give an extended mass of material
can block burning if diffusion of oxygen to the seat of
burning is prevented.
A measure of the ignitability of a dust layer and inten-
sity of burning of a dust layer is the Combustion Class
(CC) (ISSA, 1998). This classification is based on the
behaviour of a defined dust heap when subjected to a
gas flame or hot platinum wire:
1. CC1: No ignition; no self-sustained combustion
2. CC2: Short ignition and quick extinguishing; local
combustion of short duration
3. CC3: Local burning or glowing without spreading;
local sustained combustion but no propagation
4. CC4: Spreading of a glowing fire; propagation
smouldering combustion
5. CC5: Spreading of an open fire; propagating open
flame
6. CC6: Explosible burning; explosive combustion.
The train firing test assesses flammability with refer-
ence to different ignition sources, and the Combustion
Class is allotted based on the result. Not all dusts have
the ability to form a coherent burning nest.
Once a smouldering or burning deposit has developed,
nests, ranging in size from several millimetres to several
centimetres, may break off and, carried along by an air
stream until they reach an extensive dust cloud, then act
as an ignition source.
However, the train firing test does not, at first sight,
give an indication of the likelihood that a burning nest
will form. The powder or dust needs to coagulate and
any accumulations that detach need to travel for some
distance as a unit. Furthermore, combustion behaviour
in plant with flowing air will be different to when the
air is stationary, as Zockoll (1989) has shown.
At present, however, it is unclear which properties of
a dust nest make it an effective ignition source. Despite
all the reports of ignition incidents in industrial plant,
experimental studies have in the main indicated that
ignition of dust clouds by hot nests is not easy. Pink-
wasser (1985) showed that smouldering material
entering a pneumatic conveying line was soon extingu-
ishedthe distance to extinguishment depending on the
dust concentration.
Pinkwasser used an 80 m length of 100/110 mm i.d.
pipe, with six 90 elbows, which ended in a cyclone.
Nests of smouldering material were introduced through
an air-lock at the end of the pipe remote from the cyc-
lone. The powder conveying rate was measured from the
weight of powder introduced into a known volume of
conveying air over a given time, and the mean powder
concentration calculated. The temperature of the smould-
ering material was measured by thermocouples, and the
distance of travel of the smouldering material by spark
detectors and flame detectors. The powders examined
were three grades of flour, with Kst-values below 100
bar m s1, and CC ratings of 5.
Only one of the powders was capable of producing
smouldering nests. The powder properties were: moist-
ure content 8.9%, bulk density 290 kg/m3, median par-
ticle size 120 microns and minimum ignition energy
approximately 100 mJ. A much coarser-grained flour
with a bulk density of 510 kg/m3 failed to produce nests,
as did a finer grained flour with the properties: moisture
content 13.1%, bulk density 440 kg/m3, median particle
size 55 microns and minimum ignition energy approxi-
mately 500 mJ.
Glowing clumps up to 15 mm diameter, with tempera-
tures of 500550 C, were fed into the line. In dust-
free air, glowing particles were transported, in conveying
velocities of 10 and 20 m/s, as far as 68 m. But as the
powder loading in the airstream increased the distance
to extinction of nests of approximately 10 g decreased
substantially. Extinction was promoted by breaking up
of a smouldering nest into individual glowing particles.
No dust explosions were detected when powder-loading
was within the explosive range.
Pinkwasser (1986) concluded that it was impossible,
with the dusts tested, for smouldering nests of approxi-
mately 10 g weight to be conveyed in a powder loading
of greater than 1 kg/m3 of air, but that in conditions
where powder loading is lowexhaust systems and under
startupsmouldering material could be conveyed over
relatively long distances.
Alfert et al. (1988) used a pneumatic transport system
ending in a 5.8 m3 filter unit. At an air speed of 35 m/s,
only very strong nests could be transported. Charcoal
nests of 50 cm3 volume entered the filter only as small
 J. Gummer, G.A. Lunn / Journal of Loss Prevention in the Process Industries 16 (2003) 2732
agglomerations ( 1 cm3) even after a relatively short
distance (11 m), and with maize starch as the explosive
atmosphere in the filter even large nests (0.5 litre) pro-
duced no ignitions in the system. When fine wood dust
was used to give the explosive atmosphere, ignition
occurred in the filter, but not in the pipeline.
Pinkwasser (1986) showed those smouldering nests
with a temperature of 700 C, free-falling into dust
clouds, did not produce an ignition in explosive atmos-
pheres of wheat flour or wheat starch. Only when nests
of at least 25 mm diameter and weight of at least 15 g
landed on the bottom of the 1 m tall test column did
some ignitions occur. Jaeger (1989) found that smould-
ering nests could be produced only with dusts having a
Combustion Class greater than 3. A minimum nest area
of 75 cm2 and surface temperature of 900 C were
required for igniting clouds of dust with Minimum Ignit-
ing Temperatures less than 600 C. Alfert et al. (1988)
noted that nests of low mechanical strength disintegrated
during a fall and generated a large fireball that acted as
an ignition source. Mechanically stable nests were cap-
able of igniting the cloud only when they reached the
silo floor, but could get covered with dust before an
explosion had time to start. In these tests, nests of known
size were dropped through dust free air in a silo with a
height of 22 m. The dusts were charcoal, cork dust and
wood dust. Charcoal has a strong nesting structure; no
burning of the charcoal particles occurred, and when the
nest reached the floor of the silo, a shower of glowing
particles was produced. Cork dust forms stable nests;
flaring up of the nests was noted at approximately half
the height of the silo, and on the silo floor. Wood dust
produces unstable nests; these could break up in the very
top part of the silo creating a fireball, remain intact and
flare up in the upper half of the silo or reach the floor
and flare up on impact or not at all.
When nests were dropped into an explosive atmos-
phere of maize starch, nests of wood dust with a size of
0.51 litre produced no ignition in 40% of the tests. Nest
with volumes of 11.5 litres ignited equally in the upper
part of the silo or on the floor. Nests of cork dust pro-
duced no ignitions, but could start fires in settled maize
starch powder. Charcoal nests could produce ignitions if
they were mechanically broken up, and also a short time
after reaching the floor.
Zokoll (1989) has reported some tests using milk pow-
ders as both the nest material and the explosive dust
cloud. Initial tests in which both dust and smouldering
nest were dropped into the test chamber simultaneously
showed that fist-sized nests at temperatures approxi-
mately 100 C or more above bulk powder Minimum
Ignition Temperature as measured in the BAM furnace
test did not ignite ground corn and milk powder clouds.
Ignitions did not occur even when the nests, while fall-
ing, were broken up by blades. Ignition of dust clouds
could not be achieved with nests that did not burn but
29
had internal temperatures of 700800 C. Nests at 1200
C did ignite the dust clouds but only after impact on
the floor of the explosion vessel. When dust was dropped
over smouldering nests on the vessel floor, cloud
explosions could occur at temperatures of about 860 C.
A flaming nest could, however, be practically extingu-
ished by the dispersal of milk powder around it in the
explosion vessel. Tests on the development of smoulder-
ing in nests under the influence of a 0.5m s1 air stream
showed that smouldering developed differently
depending on the type of milk powder. At higher air
speeds open fires occurred in relatively large quantities
of skimmed milk powder. At air speeds of about 10 m/s
compact smouldering nests reached temperatures of
1200 C in the hottest spots. The transition from smould-
ering into open fire occurs around 800850 C,
depending on the type of milk.
Work by Bailey and Walker at Syngenta (Bailey &
Walker, 2000) has shown that clouds can be ignited by
various burning or smouldering ignition sources. Three
ignition sourcesparaformaldehyde, which burns with
a flame, Fe3 (H2), which smoulders, and incandescent
particles of sawdustwere used. Sulphur and lyco-
podium dust clouds of various concentrations were
blown over the first two of the ignition sources, and both
dusts ignited. The incandescent particles were introduced
into the dust clouds soon after the clouds had been pro-
duced. The sulphur clouds ignited, but the lycopodium
did not.
These tests were repeated with dusts of various MIT
values, from 270 C to above 1000 C, as measured in
the Godbert Greenwald furnace. With the burning layer,
dusts with MITs above 600800 C did not ignite; with
the smouldering layer, dusts with MITs above 340 C,
approximately, did not ignite; with the incandescent par-
ticles, dusts with MITs above about 330 C did not
ignite. Some tests using layer ignition sources of various
areas and temperatures showed that as the area
decreased, for a given temperature, the dust MIT above
which a dust cloud did not ignite increased.
In summary, the likelihood of ignition of a dust cloud
by a hot nest depends on the temperature of the nest, its
residence time in the dust cloud and the availability of
oxygen to the burning area. Griesche and Brandt (1976)
have shown that the Minimum Ignition Temperature of
a dust-cloud decreases substantially when the residence-
time of a dust in a Godbert-Greenwald furnace increases.
The longer a local part of a dust cloud remains in contact
with a smouldering nest, the more likely it is that an
ignition will occur. It appears, also, that dust clouds can
extinguish burning nests and an ignition be prevented,
if conditions are right.
30 J. Gummer, G.A. Lunn / Journal of Loss Prevention in the Process Industries 16 (2003) 2732

3. Experimental Table 1
Ignitability characteristicsc
A vertical tube apparatus was built, consisting of a 2
Dust type Dust layer glow temp Minimum Ignition
m long perspex tube with an internal diameter of 0.3 (C) Temperature (MIT)
m. A vibrating hopper and screw feed arrangement was (C)
designed to feed dust into an air flow in a pipe connected
to the top of the tube. The rate of feed and air flow were Sulphur 250270 280370
both variable so that a wide range of dust concentrations Lycopodium 280 410a
could be produced (see Fig. 1). Woodflour 310320 480500b
Tea (Earl Grey) 300 510a
The burning behaviour of a range of dusts was tested Cornflour (st2) 440450 450500b
by subjecting a line of each dust to a flame ignition Calcium stearate 450 450500b
source. From the results several dusts was selected with Anthraquinone 450 600675b
characteristics covering a range of cloud Minimum
a
Ignition Temperatures and Layer Ignition temperature Measured in BAM oven at HSL
b
values. The ignitability characteristics for the dusts Measured in Godbert Greenwald furnace at Syngenta
c
Other results were taken from (Eckhoff, 1997)
selected are shown in Table 1.
For the main series of tests, coherent smouldering or
burning nests were to be used as the ignition source in
the vertical tube arrangement. In order to obtain sus-
tained combustion some form of airflow either through
or over the smouldering dust sample is usually neces-
sary. Several different methods for producing sustained
combustion were tried and the best was a bank of dust
over which was passed warm air at 50 C. This method
allowed sustainable smouldering nests to be created with
most dusts. Land Cyclops Ti35+ infrared Thermal Imag-

Fig. 2. Baby milk powder burning nest, burning modesmouldering.

ing camera was used to measure the burning temperature

of the various dusts.
The dusts were banked up in a 100 mm diameter tray
as shown in Figs 2 and 3 and ignited using a blow torch.

Fig. 1. Vertical tube apparatus. Fig. 3. Baby milk powder burning nest, burning modeflame.
 J. Gummer, G.A. Lunn / Journal of Loss Prevention in the Process Industries 16 (2003) 2732 31

The burning temperatures of the agglomerates are
listed in Table 2.
A burning nest of dust was positioned inside the verti-
cal tube close to the bottom and its burning mode noted.
A dust cloud was then created within the tube using the
screw feed arrangement.
In some of the tests the burning nest was dispersed
by an air blast from a conical nozzle buried in the dust,
to see if the action of breaking-up the smouldering
deposit would result in ignition of the dust cloud.
4. Results and discussion
An analysis of the results is shown in Table 3, where
the temperatures of the nests are divided into three
bandsapproximately 700, 800900 C; and approxi-
mately 1000 C and above. The type of burning is listed,
along with the Minimum Ignition Temperature of the
dust cloud and whether or not ignition occurred. In the
majority of tests where smouldering was the mode of
combustion, ignitions did not take place even when high
smouldering temperatures were evident and the differ-
ence between this temperature and the cloud MIT was
high. Only when the nest was dispersed and the tempera-
ture difference was high did ignition occur. Sulphur dust
clouds were the only ones which would ignite on
smouldering nests, but even then the temperature differ-
ence between the nest and the cloud MIT was greater
than 500 C. By contrast, if flaming combustion took
place ignition of a dust cloud was practically guaranteed,
even when the flames were small.
The results from the current project are in agreement
with those of Bailey and Walker. In their tests, smould-
ering nests did not ignite dust clouds with MIT values
above 400 C, but flaming nests were able to ignite
clouds of all the dusts used up to the maximum MIT
used, 600675 C. Smouldering nests with a temperature
above approximately 700800 C ignited sulphur clouds.
A review of the wider literature also shows that
smouldering nests are poor sources of ignition. The like-
lihood of ignition of a dust cloud by a hot nest is low
if the nest burns only by smouldering. Ignition depends
crucially on the production of either flame or incan-
descent particles and if flaming does take place then the
risk of an ignition is very high.
When it is considered that in order for a nest to ignite
a dust cloud the following are needed: a dust capable of
holding together as a nest; a means of heating this; a
means of transporting it through the system, a means
of breaking the nest open and producing flames or high
temperatures, and a dust cloud with an explosible con-
centration at the right time, then the risks from glowing
nests are probably overestimated and many events
ascribed to them may have been caused some other way.
A more critical look at actual incidents, including in
particular the nest forming properties of dusts where
nests have been blamed for the incident, might well lead
to a revision of the previously assigned causes of
ignition.
In practical situations, a test similar to the Train Firing
Test is a useful method for determining whether a dust
deposit will either propagate smouldering or produce
flame, although both the temperature and airflow inside
dust-handling plant can have an effect on the burning
behaviour. If the air above the dust deposit is at a tem-
perature higher than normal room temperature, the possi-
bility for ignition of a dust cloud may rise.

Table 2
Burning temperature of dust deposits

Dust tested Type of burning Air applied at 50 C? Temperature range (C)

Wood Smouldering No 690

Wood With flame No 730
Wood Smouldering Yes 850900
Earl Grey fines Smouldering Yes 800940
Lycopodium With flame No 650
Lycopodium With flame Yes 10561173
Lycopodium With flame After air removed 820850
Lycopodium Smouldering Yes 1050
Baby milk powder Smouldering Yes 9501000
Baby milk powder Smouldering No 700
Baby milk powder With flame Yes 960
Cornflour (St2) Smouldering No 800
Cornflour (St2) Small pockets of flame Yes 830
Cornflour (St2) With flame Yes 900
Coal dust Smouldering Yes 1170
Calcium stearate With flame Yes 700
Calcium stearate With flame No 900
Anthraquinone With flame No 860
Hot coil (used on setup tests) 670680
32 J. Gummer, G.A. Lunn / Journal of Loss Prevention in the Process Industries 16 (2003) 2732

Table 3
Ignition conditions

Temperature of Nest C Mode of Combustion MIT of Dust Cloud (C) Ignition Y/N

Wood 690 Smoulder 280370 N

Wood 690 Smoulder 480500 N
Wood 690 Smoulder 410 N
Wood 690 Smoulder 600675 N
Wood 730 Flames 410 Y
Wood 730 Flames 450500 Y
Wood 730 Flames 600675 Y

St 2 Cornflour 830 Flame 450500 Y

Anthraquinone 860 Flame 600675 Y
Tea 800940 Smoulder 510 N
Wood 850900 Smoulder 280370 Y
Wood 850900 Smoulder 410 N
Wood 850900 Smoulder 450500 N
Wood 850900 Smoulder 480500 N
Calcium Stearate 900 Flame 450500 Y
Tea 800940 Smoulder 600675 N

Milk 9501000 Smoulder 410 N

Milk 9501000 Smouldering then dispersed 410 Y
Milk 960 Flame 410 Y
Lycopodium 1050 Smoulder 410 N
Coal1170 Smoulder 410 N
Coal1170 Flame 410 Y
Coal1170 Smoulder 600675 N
Coal1170 Flame 600675 Y

Various tests have been developed to measure the
ignition behaviour of dust deposits in a stream of hot
air. Similar tests could be used to tell whether smoulder-
ing dusts were in danger of flaming. The I.ChemE Guide,
Prevention of fires and explosions in dryers described
tests developed to simulate various conditions and obtain
measurements of the temperature at which exothermic
reaction begins (Abbot, 1990). If the dust deposits and
surrounding conditions properly simulate practical situ-
ations, the temperature at which deposit burning pro-
gresses to flaming combustion could be used as a basis
for safe procedures with an adequate safety margin
incorporated.
References
Abbot, J. (1990). Prevention of fires and explosions in dryersA user
guide (2nd ed.). Rugby, UK: The Institution of Chemical Engin-
eers.
Abbott, J.A. (1985). BMHB Survey of dust fire and explosions in the
UK 197984. British Materials Handling Board, ISBN 0 85624
4554.
Alfert, F., et al. (1988). The ignition capability of nests of smouldering
material and hot objects in industrial plants. VDI-Berichte No 701.
pp. 303319.
Bailey, M., & Walker, N. (2000). Private communication. Syngenta,
Manchester, UK.
Eckhoff, R. K. (1997). Dust explosions in the process industries (2nd
ed.). Butterworth Heinemann.
Griesche, G., & Brandt, D. (1976). Einflussfaktoren auf die Zundtem-
peratur von Staub-Luft-Gemischen. Die Technik, 31, 504.
Harper, D.J., Plain, & Gibson (1997). Use of intrinsically safe circuits
and enclosures to control ignition risk from equipment in powder
handling plant. I.Chem.E. Symposium Series No. 141 p. 463. Haz-
ards XIII. Process Safetythe Future.
ISSA (1998). Determination of the combustion and explosion charac-
teristics of dusts. International Social Security Agency,
Mannheim (1998).
Jaeger, N. (1989). Zundwirksamkeit von Glimmnestern in Staub/Luft-
Gemischem. VDI-Berichte 701, p. 263. Dusseldorf: VDI-Verlag,
GmbH (in German).
Jeske, A., & Beck, H. (1989). Evaluation of dust explosions in the
Federal Republic of Germany, EUROPEX Newsletter No 9, p. 2.
Pinkwasser, T. (1985). On the extinction of smouldering fires in pneu-
matic conveyors. Symposium on Control of the Risks in Handling
and Storage of Granular Foods, Paris.
Pinkwasser, T. (1986). On the Ignition capacity of Free-falling
Smouldering Fires. Euromech Colloquium 208, Explosions in
Industry.
Porter, B. (1989). Industrial incidents. Paper presented at Dust
Explosions: Assessment, Prevention and Protection, 24th Nov-
ember, London.
Testing methods for electrical apparatus installed in a dusty environ-
ment with a potential risk of explosion (2001). EU Contract No.
SMT-CT98-2273. Project EADE. Synthesis report available on
www.safetynet.de.
Zockoll, C. (1989). Ignition effect of smouldering pockets in dust-air
mixtures. VDI-Berichte. 701, p295. Dusseldorf: VDI-Verlag
GmbH.
 Journal of Loss Prevention in the Process Industries 16 (2003) 3340
www.elsevier.com/locate/jlp

A test for electrical ignitions of flammable dust clouds

D. Bennett a, G.A. Lunn a,, J.V. Torrent b, E. Querol b, J.-P. Fritze c
a
Health and Safety Laboratory, Harpur Hill, Buxton SK17 9JN, UK
b
Laboratorio Oficial J.M. Madariaga (LOM), Madrid, Spain
c
Deutsche Montan Technologie (DMT), Dortmund, Germany

Abstract

The Atex Directive specifically includes the explosion hazards arising from the presence of flammable dusts. The European
standards body CENELEC proposed a research project to develop tests for assessing the ignition hazard due to electrical apparatus
used in hazardous dusty environments. This paper describes the work done on developing a test for electrical spark ignitions of
explosive dust atmospheres. A prototype apparatus incorporating the dust explosibility vertical tube and the STA break flash appar-
atus has been developed. Tests using three dusts showed sulphur dust had ignition characteristics close to those of gas Group B,
while other dusts were much less easily ignitable than methane. Round robin tests using a duplicate apparatus and the proposed
test method produced results very close to those obtained using the original apparatus.
Crown copyright 2003 Published by Elsevier Science Ltd. All rights reserved.

Keywords: Dust explosion; Electrical; Ignitions

European Directive 94/9/ECthe ATEX Directive
comes fully into force in June 2003 (DTI, 1994). The
Directive applies to equipment and protective systems,
among other things, that are intended for use in poten-
tially explosive atmospheres. For the first time in regu-
lations, explosion hazards arising from atmospheres con-
taining flammable dusts are specifically included.
Consequently, where appropriate, harmonised European
standards are being prepared to specify methods for
either preventing dust explosions or protecting against
their effects.
The hazards arising from dusty environments are:
ignition and explosion of a dust cloud by electrical or
mechanical means,
ignition and burning of a dust layer or accumulation,
ignition of an explosive dust atmosphere by a burning
deposit or burning nest.
To aid the development of standards CENELEC pro-
posed a research project on electrical ignitions of dusty
environments. The European Union funded this project
(EU, 2001).
The partners in the project were the Health and Safety
Laboratory (HSL), UK, Laboratorio Oficial J.M. Madari-
aga (LOM), Spain and Deutsche Montan Technologie
(DMT), Germany. The objectives of the project were:
To review the current state of the art in testing electri-
cal apparatus in dusty environments through an
enquiry directed at test houses, certification organis-
ations and laboratories in Europe and the USA.
To review current scientific and technical knowledge
on ignitions of dust clouds and deposits.
To develop a test apparatus and test method for
assessing the ignition hazard of electrical apparatus
intended for use in potentially explosive dust atmos-
pheres.
To develop a test apparatus and test method for
assessing the ignition hazard of electrical apparatus
intended for use in dusty environments where layers
and accumulations of flammable dust may be present.
To propose test procedures for standardisation.

This paper reports the work done on developing a test

Corresponding author. Tel.: +44-114-2892005; fax: +44-114- for ignitions of dusty atmospheres by electrical sparks.
2892010. Following a literature review, discussions with inter-
E-mail address: geoff.lunn@hsl.gov.uk (G.A. Lunn). ested organisations and a study of the responses on

0950-4230/03/$ - see front matter. Crown copyright 2003 Published by Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0950-4230(02)00086-4
34 D. Bennett et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 3340
ignition testing from the EU and the USA, a prototype
test apparatus was constructed and a draft test method
developed. A duplicate test apparatus was built and sent
to the partners for a series of round-robin tests. These
tests showed that reproducible results could be obtained
from the test apparatus when the test procedure was fol-
lowed.
1. Literature review
At the start of the project a number of testing and
certifying organisations and laboratories were contacted
throughout the European Union and the USA. Enquiries
were made of these organisations as to the current
methods used for testing for ignition of flammable dusts
by electrical apparatus, both as regards ignition of clouds
by sparking and ignition of dust layers and accumu-
lations by hot surfaces.
None of the eleven responses received described spe-
cific tests for electrical apparatus used in dust clouds.
DMT suggested that gasair mixtures should be used as
substitutes for dust clouds, with the ignition energy of
the gasair mixture matching the ignition energy of the
dusts. This type of substitute test is used in the testing
of hand-held electrostatic spray guns according to EN
50 050: 1986 (EN, 1986). This standard specifies gas
air mixtures with ignition energies of 0.24 mJ (for paint
spray guns) and 5 mJ (for powder spray guns) for use
in ignition tests.
Underwriters Laboratories Inc. of the United States
responded that their test method for intrinsically safe
apparatus in dusty atmospheres was given in the Stan-
dard for Safety (Underwriters Laboratories, 1997). the
Factory Mutual Researchs Approval Standard Class No.
3610: 1999(Factory Mutual Research, 1999) demands
the same requirements as UL 913:
Intrinsically safe apparatus not enclosed in a dust-
tight enclosure shall comply with the spark ignition
requirements of some gases, e.g ethylene for metal
dusts and propane or methane for coal and grain dusts
and fibres and flyings.
Electric sparks and arcs have long been known to be
capable of igniting dust clouds. The minimum energy
required for ignition varies with the properties of the
dust, characteristics of the spark such as the spatial and
temporal distribution of energy, and the geometry of the
electrode gap, the electrode size and shape and the dis-
charge triggering mechanism.
Electrostatic spark discharges are capacitative, and the
theoretical spark energy, neglecting external circuit
losses, is given by 0.5CV2, where C is the capacitance
and V is the voltage difference.
Sparks or arcs generated when live currents are broken
are transient inductive sparks, and the theoretical energy
is given by 0.5Li2, where L is the circuit inductance and
i is the current in the circuit. Discharges from compo-
nents with significant inductance or resistance generally
produce discharges with a longer duration than purely
capacitative discharges, with the result that less total
energy is necessary to ignite a specific dust. Boyle and
Llewellyn (1950) showed that the minimum capacitor
energy capable of igniting a given powder decreased
substantially when a series resistance was included in
the circuit. Because the series resistance absorbs a large
fraction of the capacitor energy during discharge, the
energy going into the spark gap is only a small part of
the theoretical stored energy, generally about 510%
(Eckhoff, 1997). Thus, in Boyle and Llewellyns experi-
ments, at a series resistance of 104105 ohms, for some
dusts the ignition energy was reduced to approximately
1% of the energy necessary without the resistance. Simi-
lar changes were measured by Line, Rhodes, and Gilmer
(1959). Smielkow and Rutkowski (1971) increased the
spark duration either by adding a large inductance or by
adding a large resistance to the circuit, and again the
theoretical ignition energy from the capacitor decreased
by a factor of ten, approximately.
The optimum spark discharge duration measured by
Matsuda and Naito (1983), and estimated from the data
in Boyle and Llewellyn (1950) and Line et al. (1959)
were 0.11.0 ms, decreasing with the net spark ignition
energy (Eckhoff, 1997).
Measurements by Parker (1985) showed that for some
dusts there was a fairly distinct region where the dis-
charge duration, produced the lowest ignition energies.
With other dusts, however, no such region occurred.
Parker used four dusts, and those with the higher ignition
energies showed the optimum spark duration effect.
The physical length of the spark gap also has an effect
on the ignition energy. Tests by Ballal (1980), using
metal dusts and carbon, demonstrated that there was an
optimum spark gap length for each dust, and that it
increased as the ignition energy at the optimum gap
length increased, although the result for carbon did not
fit in with the results for the metal dusts. The optimum
spark gaps fell in the range 27 mm. Measurements by
Norberg, Zu, and Zhang (1988) showed the optimum
spark gap length to be in the range 68 mm, for short
duration capacitative sparks.
There is little published literature on research into the
ignition of dusts by the type of circuits used in intrinsi-
cally safe apparatus, but two programs of work have
been performed using the make and break sparking
apparatus.
The make and break sparking apparatus is known as
the break-flash apparatus and is described in BS-
EN50020 (BSI, 1995). Harper, Plain, Wilston, and Gib-
son (1997) used the standard disc and wires and fixed
them vertically through the walls of a glass tube that
 D. Bennett et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 3340
resembled the Hartmann tube that is used for dust
explosibility testing. The sparking apparatus was pos-
itioned about 5 cm above the Hartmann dispersion cup.
By repeated dispersion of the dust a continuous cloud
was produced, with a concentration passing through the
explosion range. Following calibration of the apparatus
with methane, ten dusts were used with low ignition
energies, as measured with capacitative sparks.
Of the dusts tested, only sulphur had a minimum ignit-
ing current less than that of methane. In a resistive cur-
rent, at low voltage, the minimum igniting current
measurements for sulphur followed the gas Group IIB
data, and then passed through the gas Group IIA to gas
Group I as the voltage increased. In an inductive circuit,
the results for sulphur fell between gas Groups IIB and
IIA. These measurements show that circuits certified as
intrinsically safe for gases and vapours would not
present an ignition risk for all but the most sensitive
dusts.
A similar apparatus to that used by Harper et al.
(1997) was devised by Tolson (Lunn, Rowland, & Tol-
son, 1999) to measure the ignitability of coal dust clouds
when subjected to discharges produced by the IEC spark
test apparatus. The equipment consisted of a circuit con-
taining a variable power supply attached to an induct-
ance of 1.1 Henry. The current in this circuit was made
and broken using a modified version of the IEC spark
test apparatus described in CENELEC Standard EN
50020. The main modifications were:
the spark test apparatus was mounted horizontally;
the disc electrode was copper instead of cadmium;
the rotating electrodes were inside a Hartmann verti-
cal tube apparatus.
The current in the circuit was set by adjusting the
power supply voltage and the coal dustair mixture pro-
duced by introducing a blast of air through a measured
amount of coal dust placed in the dispersion cup at the
base of the Hartmann apparatus. The concentration of
coal dust in air could be changed by altering either the
amount of the dust placed in the dispersion cup, or the
air introduced by the blast. The current was progress-
ively increased to give spark energies of from 50 to 300
mJ but there were no ignitions.
The test equipment was then validated using lyco-
podiumair mixtures. These are known to have ignition
energies of about 8 mJ and were readily ignited with
spark energies of 11 mJ which indicated that the appar-
atus was functioning correctly.
Coal dustair mixtures were eventually ignited with a
spark energy of 400 mJ; this was the maximum possible
with the circuit used. Ignitions at this energy level were
repeatable. This energy level is approximately 1000
times the level necessary for ignitions of methaneair
mixtures.
35
Both the standard test for assessing dust explosibility
and for measuring the Minimum Ignition Energy (MIE)
utilise very similar designs of apparatus. The Hartmann
tubeor vertical tubeconsists of a 1.2 l tube, fitted
with an ignition source that is either a high voltage con-
tinuous spark or a heating coil, with a means for introd-
ucing a blast of air to disperse a dust sample. The con-
centration of the dispersed dust is varied by changing
the amount of dust placed in the tube prior to the test.
By this means the potential explosion range of the dust
is covered. If an explosion occurs the dust is considered
to be flammable at ambient conditions.
The MIE, defined as the lowest capacititively stored
electrical energy that just ignites the most ignitable
dust/air following discharge across a spark gap, is meas-
ured in an apparatus similar to the vertical tube. A
capacitor of known capacitance is charged to a specified
high voltage and then discharged across a gap through
a dust cloud dispersed by an air blast. The energy of the
spark discharge is lowered in steps, and at every step
tests are performed across the explosible range. This
continues until no ignition occurs at any dust concen-
tration.
2. Development of the proto-type test apparatus
The literature survey and discussions with users and
manufacturers indicated that an acceptable way forward
was to combine the vertical tube apparatus familiar from
dust explosibility testing and measurement of MIEs with
the spark test apparatus and its associated test method.
The vertical tube is an accepted way of producing an
explosive dust cloud for testing purposes, for example
the MIKE 3 apparatus for MIE measurements (Adolf
Kuhner, 1997). The spark test apparatus (STA) is an
accepted means of testing the igniting capabilities of
electrical circuits for gases. This combination has the
advantages that there is expertise throughout the world
in using both types of test and the apparatus and test
procedure would fit well into an established series of
tests for dust ignitability and explosibility. A research
apparatus on similar lines has been used successfully by
two groups of workers in the United Kingdom and this
work had demonstrated that the proposed test was tech-
nically feasible.
In addition, a new version of the spark test apparatus
for circuits up to 20 A has been developed under an EU
funded contract SMT4-CT98-2217 (EU, 1999).
A prototype apparatus was constructed using the
Spark Test Apparatus (STA) as described in EN50 020
(BSI, 1995) as modified in project SMT4-CT98-2217.
The dust dispersion chamber is based on the principles,
and where practicable, the dimensions of the MIKE 3
minimum ignition apparatus. Photographs of the appar-
atus are shown in Fig. 13.
36 D. Bennett et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 3340
Fig. 1. HSL Test apparatus for ignition testing of electrical circuits in dust atmospheres.
Fig. 2. STA apparatus in the vertical tube.
The apparatus thus consists of the STA disc and trai-
ling electrode system mounted horizontally inside a ver-
tical Hartmann-type tube with a mushroom shaped dis-
persion system in the base. Successive blasts of air at
timed intervals disperse dust placed in the base to pro-
duce a cloud around the STA disc and the location of
the sparks. Provision is made to pass a gas/air mix
through the chamber for calibration of the STA. The
number of turns of the STA motor are counted and dis-
played with a provision to automatically end the test
when a preset count is reached. An ignition detector
mechanism based on detection of the negative pressure
following ignition has been developed.
Tests have been carried out with three dusts, lyco-
podium, calcium stearate and sulphur. The research has
been aimed at devising a workable test method that takes
into account the variables that may affect the ignition
result. Ignition of dusts by electrical means is a complex
Fig. 3. Dust dispersion system.
mechanism involving the potential for many variables to
effect the outcome. In view of the need to produce a test
system that is reproducible and can be used universally
the procedure has to be kept as simple as possible, and
some of the variables involved need to be confined to
standardised settings. Otherwise the operation becomes
too complicated and unwieldy and not workable in prac-
tice. Following the development research work a test
procedure has been produced that gives reproducible
results.
3. Test procedure
A maximum particle size and drying procedure have
been defined (63 m and 105 oC for 1 h are values com-
monly used in the past), but care has to be taken as
experience with the MIKE 3 apparatus has shown that
 D. Bennett et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 3340 37

drying can cause some dusts to be desensitised. Testing 4. Add 2.5 g of the prepared test dust.
with both the dried and as found dusts may, therefore, 5. Connect the electrical equipment to be tested.
be necessary. 6. Run the test for 400 counts of the STA or until
It is especially important that a sufficiently wide range ignition occurs.
of dust concentrations is used to ensure that the optimum 7. Blow out the chamber with an airline (dry air).
concentration for ignition for a particular dust is passed 8. Repeat 6) and 7) using 5, 7.5, 10, 12.5, 15, 17.5 g
through during the test. During a test, however, the dust of dust.
concentration reduces with time as the dust falls out of
suspension and, provided sufficient dust is used initially,
this ensures that the optimum concentration is passed 4. Results
through at some point although it is likely to be present
for a limited time only. A study of the relationship The results from a validation test using an 8.3% meth-
between the explosibility characteristics and concen- ane mixture are shown in Fig. 4, where they are com-
tration of a number of different dusts shows a fairly flat pared to the ignition curves for resistive circuits pub-
profile between a concentration of about 250 g/m3 and lished in EN 50 020 (BSI, 1995). The present
about 1000 g/m3 for a lot of the dusts with only a few measurements are close to, but slightly above, the Group
having a significant peak.
The air blasts used to disperse the dust are dry (bottled
air is suitable) and the reservoir pressure is defined (the
MIKE 3 uses 7 bar but 3 bar proved satisfactory for
dusts tested here).
A calibration procedure needs to be carried out before
the equipment can be used. The STA part is calibrated
using the same procedure defined in EN 50 020:1995
(BSI, 1995) for methane (24 V d.c., 110 ma, 95 mH
inductance). Once calibration using methane has proved
satisfactory then a fixed weight of a standard dust can
be introduced and the STA electrical parameters set to
a defined value. The tests have shown that lycopodium
dust with a voltage of 24 V and a current of 10 A is
a suitable combination. Ignition should occur within a
stipulated number of STA motor turns displayed on a
counter.
Once the apparatus is fully calibrated, a thorough
clean to remove all traces of lycopodium is required
before the test dust is introduced. The electrical equip-
ment to be tested is connected to the STA. A relay that
operates upon commencements of the test is used as an
interface to prevent heat damage to the tungsten wires
when stationary. The test is run for 400 turns of the STA
or until ignition occurs and is repeated a number of times
with different initial quantities of dust so that a wide
range of concentrations is covered. During the research
work 400 turns proved successful on the three dusts
repeated for seven different starting weights of dust.
Blowing out of the chamber with an air-line is sufficient
after a test before fresh dust is added. This procedure is
repeated until the defined count is reached or an
ignition occurs.
In summary the test procedure is:

1. Calibrate the STA with a methane/air gas mixture

using the procedure in EN 50 020.
2. Calibrate the full apparatus using lycopodium (10 A,
24 V).
3. Dismantle and clean the chamber. Fig. 4. Calibration of HSL spark ignition apparatus: 8.3% methane.
38 D. Bennett et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 3340
1 (methane) curve and indicate that the apparatus is giv-
ing reliable results.
Three dusts have been tested using a resistive circuit
calcium stearate, lycopodium and sulphur. The ignition
results based on the proposed draft test method are
shown in Fig. 5 in a comparison with the published cur-
rent voltage ignition curves from EN 50 020 (BSI, 1995).
Both lycopodium and calcium stearate show ignition
conditions higher than the methane curves. Sulphur,
however, is much easier to ignite, and falls close to the
ignition curve for IIB gasestypically ethylene. This
result supports previous work that shows ignitions of sul-
phur falling on or close to the IIB ignition curve.
Fig. 5. Ignition results for lycopodium, calcium stearate and sulphur
dusts compared to published ignition curves.
5. Round robin tests
5.1. Tests at LOM
In the tests performed by LOM, around 25 V were
used for the motor supply, about 28 air pulses were
obtained for 400 revolutions and various resistor combi-
nations were used to obtain the desired spark intensities.
A calibration check with methane at 8.3% was perfor-
med, with ignition obtained at 0.120 A. Three dust
samples delivered by HSL were tested: lycopodium, cal-
cium stearate and sulphur. Dust weights ranged from 2.5
to 20 g.
5.2. Tests at DMT
The tests were carried out with lycopodium, calcium
stearate and sulphur, as provided by HSL. The revol-
utions of the wire holder of the wolfram wire were set
to 80 min1. The tests were stopped in case of non-
ignition after 400 revolutions. The ignition current was
varied at a given voltage by altering the corresponding
resistance. By this means, the power was reduced for
each following test until an ignition of the dustair mix-
ture did not occur. The current that just did not lead to
an ignition of the dust-air mixture is recorded as well as
the current that just led to an ignition.
The weight of the dust was increased step by step and
the powder was reduced if necessary. The tests were car-
ried out with lycopodium and calcium stearate with a
weight of 1.25, 2.5 and 5 g. Larger weights of dusts
did not seem to be practical because they could not be
dispersed completely. For 7.5 g, for instance, a major
part of the dust remained in the dispersing device.
Weights of up to 15 g were used for the tests with sul-
phur dust. The sulphur dust adhered to all the surfaces
in the interior of the test apparatus following initial dis-
persal. The sulphur was dried up to remaining humidity
of 0.08%. The lycopodium and the calcium stearate were
kept in the same status as they were delivered.
The calibration of the test apparatus was carried out
with a 8.3 vol% methaneair mixture. The ignition cur-
rent was set to 110 mA with a voltage of 24 V. The
circuit has an air coil with an inductance of 95 mH.
Wolfram wires with a diameter of 0.2 mm were used
for the calibration. Problems arose when calibrating the
test apparatus had to be sealed particularly during cali-
bration to have an exact gas concentration.
The DMT measurements plus those of HSL and LOM
are shown in Fig. 6 and listed in Table 1.
6. Discussion
Because of the more complex nature of dust clouds
and the greater number of variables when compared to
 D. Bennett et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 3340 39

Table 1
Voltages and currents at the ignition of dusts

HSL LOM DMT

Voltage V Current A Current A Current A

Lycopodium
23 9.67 8.995 No ignition at
9.71 A
30 4.15 4.49 3.13
40 1.43 1.33 1.32
50 1.05 0.508 0.73
60 0.6 0.496 0.59
Calcium stearate
25 9.03 8.995 8.81
30 4.59 5.5 4.04
40 1.43 1.82 1.2
50 1.05 1.044 1.04
60 0.6 0.521 0.59
Sulphur
25 0.52 0.523 0.52
30 0.45 0.44 0.44
40 0.24 0.171 0.24
50 0.15 0.129 0.2

The attrition was so intensive that the electrode had to

Fig. 6. Comparisons of measurements using the test apparatus.
gases and vapors there was a concern that applying the
STA test to dust clouds would give less consistent
results. The results obtained and the practical experience
gained whilst developing the test method show this not
to be the case. The dusts proved to behave in a repeat-
able manner during development of the test and the part-
ners measured very similar ignition conditions during the
round robin tests. Contamination of the cadmium disc
by the test dust did not seem to have any effect on the
sensitivity of the test.
DMT found that the cadmium electrode wore aware
rapidly and the abrasion was so extreme after 40 single
tests that the cadmium electrode could not be used any-
more. The abrasion was mainly visible by splintering.
Due to this high cadmium emission, corresponding work
protection measurements were necessary for the tests.
While executing the tests and cleaning test apparatus the
testing officers had to wear a respirator with a P 3-filter
and gloves. All three electrodes delivered by HSL were
too soft for the wolfram wire of 0.38 mm diameter used.
be exchanged after each test series.
HSLs experience was, however, different. The wear
rate and subsequent amount of cadmium deposited was
not excessive during the HSL tests and the same disc
was used for the verification tests using methane and all
three dusts before being regarded as worn out.
Excessive wear problems experienced by DMT of the
cadmium discs may be due to the material being too soft.
The discs used in the apparatus supplied by HSL were
the last remaining stock of undetermined composition.
It is possible they could be softer than the one used by
HSL. It is important that a reduced amount of overlap
between wire and disc (0.5 mm rather than the 1 mm
for the standard wires) is used to compensate for the
increased stiffness of the thicker wire. Endurance tests
of 20,000 turns were carried out during the development
of the high current STA for project SMT4-CT98-2217
(EU, 1999) with negligible erosion of disc or wires. Test-
ing bodies such as the UK Electrical Equipment Certifi-
cation Service (EECS) have used this higher current ver-
sion successfully. The use of cadmium as a disc material
poses a potential health risk but the use of similar extract
system and precautions as are required for the standard
STA equipment should be adequate.
Although the composition of the dust cloud at the
moment of ignition is unknown this is true of other dust
ignition tests that use the same method of dust disper-
sion. The procedure has been developed to ensure that
there is a very high probability that the most easily ignit-
able mixture is traversed as dust concentrations change
and that it will be subjected to the spark. The highly
satisfactory agreement obtained by the partners using
40 D. Bennett et al. / Journal of Loss Prevention in the Process Industries 16 (2003) 3340
this apparatus and method indicates that the procedure
is satisfactory in this regard.
Based on the dusts tested and the results of the round
robin tests the equipment can be used as a viable test
for individual pieces of electrical equipment. Based on
a comparison of the results obtained with the curves for
gases and vapors in the European Standard EN 50 020
(BSI, 1995), then there seems to be a good possibility
of using the equipment to produce similar curves for
various dust groups. The results with sulphur dust fall
very close to the curve for ethylene and show that some
dusts are very sensitive and equipment would have to
be certified accordingly. The results obtained using a
similar experimental apparatus by Harper et al. (1997),
further verify that the HSL test produces reproducible
results.
On the other hand the lycopodium and calcium stear-
ate results indicate that a less stringent classification for
these and probably a great many more dusts would be
appropriate. The fact that their MIKE 3 MIE values are
fairly different when the STA apparatus produces almost
identical results suggests that dusts may lend themselves
to groupings.
References
Adolf Kuhner, A. G. (1997). MIKE 3 technical and operations manual.
Switzerland: Adolf Kuhner AG.
Ballal, D. R. (1980). Ignition and flame quenching of quiescent dust
clouds of solid fuels. Proc. Roy. Soc. Lond. A, July, 479500.
Boyle, A. R., & Llewellyn, F. J. (1950). The electrostatic ignitability
of dust clouds and powders. J. Soc. Chem. Ind. Trans, 69, 73181.
BS EN 50 020:1995 (1995) Electrical apparatus for potentially explos-
ive atmospheres. Instrinsic safety (i). British Standards Insti-
tution. London.
DTI (1994). Guidelines on the application of Directive 94/9/EC of 23
March 1994 on the approximation of the laws of the Member States
concerning equipment and protective systems intended for use in
potentially explosive atmospheres. Available from DTI. The docu-
ment contains the text of the Directive as an Appendix.
Eckhoff, R. K. (1997). Dust explosions in the process industries.
Oxford: Butterworth Heinemann.
EN 50 050:1986 (1986). Electrostatic painting and finishing equipment
using flammable materials. Part 1. Specification for hand-held
spray guns and associated apparatus.
EU (2001). Testing methods for electrical apparatus installed in a
dusty environment with a potential risk of explosion (EADE). EU
funded project contract no. SMT-CT98-2273. Synthesis report
available on the SAFETYNET web site: www.safetynet.de.
EU (1999). EU funded project SMT4-CT98-2217gas ignition test for
heavy current low voltage circuits. Final report.
Factory Mutual Research (1999). Approved Standard. Intrinsically safe
apparatus and associated apparatus for use in Class 1, 11, and
111, Division 1 hazardous (classified) locations. Class number
3610.
Harper, D. J., Plain, K. E., Wilton, J. S., & Gibson, N. (1997). Use
of intrinsically safety circuits and enclosures to control ignition risk
from equipment in powder handling plant. I.Chem.E. Symposium
Series No. 141. p463. Hazards XIII. Process safety. The future.
Line, L. E., Rhodes, H. A., & Gilmer, T. E. (1959). The spark ignition
of dust clouds. Journal of Physical Chemistry, 63, 290.
Lunn, G. A., Rowland, D. B., & Tolson, P. (1999). Electrical ignitions
and use of flameproof enclosures in coal dust and methane atmos-
pheres. Transactions of the Institution of Mining and Metallurgy
A, 108, 71.
Matsuda, T., & Naito, M. (1983). Effects of spark discharge duration
on ignition energy for dust/air suspensions. In J. K. Beddow (Ed.),
Particulate systems, technology and fundamentals (p. 189). New
York: Hemisphere Publishing Corp/McGraw-Hill.
Norberg, A., Xu, D., & Zhang, D. (1988). Powder Ignition energy
measured utilizing a new fluidized bed ignition chamber. Report
ISSN 0349-83 52 (September), Institute for High Voltage Research,
Uppsala, Sweden.
Parker, S. J. (1985). Electric Spark ignition of gases and dust. Ph.D.
thesis, (August) City University, London. Dust part also published
as report no. 8 53351-4 (December) Chr. Michelsen Institute,
Bergen, Norway.
Smielkow, G. I., & Rutkowski, J. D. (1971). Badania Zjawiska
Zaplonu Mieszanin Pylowopowietrznych wywolanego wyladonani-
ami iskrowymi. Chemia Stosowana, XV3, 283.
Underwriters Laboratory Inc. (1997). Standard for Safety. Intrinsically
safe apparatus and associated apparatus for use in Class 1, 11 and
111, Division 1, hazardous (classified) locations. UL 913. 5th edi-
tion.
 Journal of Loss Prevention in the Process Industries 15 (2002) 373383
www.elsevier.com/locate/jlp

Explosion venting of bucket elevators

P. Holbrow a, G.A. Lunn a,, A. Tyldesley b
a
Health and Safety Laboratory, Harpur Hill, Buxton, Derbyshire SK17 9JN, UK
b
Health and Safety Executive, TD5, Magdalen House, Bootle, Merseyside, UK

Abstract

This paper is a report of an experimental programme on the explosion protection of bucket elevators by venting. The project
was a collaborative effort with funding by the Health and Safety Executive and manufacturers and users of bucket elevators through
the British Materials Handling Board. Two bucket elevators were used in the projecta single leg elevator and a twin-leg elevator.
Four dusts were used with KSt values up to 211 bar m s1 and dust clouds were produced by dust injection and by normal operation.
Reduced explosion pressures were measured and guidance has been derived from the results. This guidance is in terms of vent
spacing as a function of the KSt value of the dust. Crown Copyright 2002 Published by Elsevier Science Ltd. All rights reserved.

Keywords: Dust explosion; Venting; Bucket elevators

Bucket elevators are widely used in the handling of
large quantities of bulk powders and are normally the
preferred means of conveying where long vertical lifts
are required for free flowing powders/granular products.
As such they are to be found in nearly all animal feed
mills, bulk grain stores and many of the larger instal-
lations handling powders in the food industry. Powder
or granular products inevitably spill from the buckets
during operation and fall down the upleg of the elevator.
The finer material is likely to remain in suspension while
the coarser material falls back to the boot. At the top of
the elevator most of the powder will discharge down the
off-take chute, but some will normally be carried over
into the downleg of the elevator. Thus both legs are
likely to contain a dust cloud of unknown concentration,
constantly agitated by the moving buckets, all the time
the elevator is in operation. Various sources of ignition
are foreseeable in such units and explosion incidents
have been reported.
Explosion venting is one method for explosion protec-
tion of bucket elevators. The current Institution of
Chemical Engineers guidance requires that the vents
equal in cross sectional area to the limbare positioned
according to the guidance for ducting. Alternatively, a
Corresponding author. Tel.: +44-114-2892005; fax: +44-114-
2892010.
E-mail address: geoff.lunn@hsl.gov.uk (G.A. Lunn).
spacing of 6 m between vents is used. The guidance also
requires that the top casing and the boot must be
explosion relieved (Lunn, 1992).
There appears to be no guidance from other European
organisations such as the German VDI, but the USA
National Fire Protection Association document NFPA-
91: Fires and dust explosions in agricultural and food
products facilities gives guidance in somewhat more
detail than Lunn (1992) but still recommends a vent
spacing of approximately 6 m.
There is, however, no evidence that any current guid-
ance spells out the optimum venting requirements of
elevators, and there is little published work on elevator
explosion tests. Gillis and Fishlock (1982) carried out
venting and suppression experiments on a twin leg elev-
ator and some guidance was given.
This paper describes an experimental programme, a
collaborative effort by the Health and Safety Executive
and manufacturers and users of bucket elevators under
the British Materials Handling Board, in which
explosion venting of bucket elevators was studied using
a single leg elevator and a twin leg elevator with a num-
ber of flammable dusts. Reduced explosion pressures
have been measured and the results used by the Health
and Safety Executive to develop guidance for practical
use.

0950-4230/02/$ - see front matter. Crown Copyright 2002 Published by Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 0 2 ) 0 0 0 2 1 - 9
374 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383
1. Experimental
Two bucket elevators have been used in this pro-
gramme: a single leg elevator and a double leg elevator.
Both elevators were mounted in a tower with access lev-
els at 2.7 m intervals.
1.1. Single leg elevator
A schematic diagram of the elevator is shown in Fig.
1. The single leg steel casing is rectangular in shape,
with a cross section of 1.22 m0.945 m, in which the
chain linked buckets of nominal dimensions 540 mm
wide280 mm390 mm, with a spacing of approxi-
mately 450 mm, run up and down. It has a fixed speed
drive mounted at the head of the elevator powered by
an 11 kW motor and gearbox that drives the buckets at
a speed of approximately 35 m/min. The drive pulley
and a deflector pulley are mounted within the head of
the elevator and a return pulley is mounted at the boot
of the elevator.
Explosion relief vents were installed at each level,
including the top face of the elevator, with dimensions
equal to the nominal cross section of the elevator casing
(1.22 mm0.945 m), apart from level 8 where, because
Fig. 1. Single leg bucket elevator.
of the restriction of supporting steel members, a slightly
smaller vent was installed0.945 m0.7 m. Plastic vent
panel closures were used for the majority of the tests.
Stainless steel vent panels were also used in some of
the tests.
Dust injection cylinders were located at each of the
nine levels at intervals of 2.7 m. Their position at each
level alternated from side to side. Distribution of the dust
injection points along the elevator enabled a well-dis-
persed dust cloud to be produced throughout the elevator
casing. An ignition source could be fitted at level 1, level
5 or level 9 (see Fig. 1). Pressures were measured at up
to nine positions along the full height of the elevator.
1.2. Twin leg elevator
The twin leg elevator (Fig. 2) was supplied by Carier
Bulk Materials Handling Ltd and represents a typical
elevator used in the bulk handling industry. The casing
was designed to a stronger specification than normal to
Fig. 2. Twin leg bucket elevator.
 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383 375

enable it to withstand the explosion pressures developed measured at 14 points, located in both legs at 3 m inter-
in the tests. The elevator head, boot and a 1.5 m long vals.
leg section were each hydrostatically tested to 1.5 bar g.
The overall height from the base of the boot to the top 1.3. The dusts
of the head section was 17.75 m.
The maximum dimensions of the steel buckets were Four dusts were used in the tests:
308 mm wide175 mm deep130 mm high and were
bolted to the 320 mm wide rubber belt. The belt was milk powder: KSt=86 bar m s1, Pmax=7.4 bar g;
driven by the 0.6 m diameter crowned head pulley at a cornflour A: KSt=147 bar m s1, Pmax=7.9 bar g;
speed of 3 m/s. Typically, when full to capacity, each cornflour B: KSt=211 bar m s1, Pmax=8.0 bar g;
bucket would carry approximately 1.7 kg of cornflour or cornflour C: KSt=180 bar m s1, Pmax=8.7 bar g.
1.3 kg of milk powder. Discharge takes place by cen-
trifugal action as the buckets pass around the head pul-
ley. 2. Experimental results
The design clearances are, approximately: between the
tip of the buckets and the front of the casing: 70 mm, 2.1. Single leg elevator
between the sides of the buckets and casing: 41 mm, and
between the rear of the belt and the casing: 55 mm. A series of tests was performed to determine the opti-
Explosion relief vent openings were installed at mum conditions of injection pressure, dust concentration
approximately 3 m intervals on both legs and measured and ignition delay that produced the highest explosion
305 mm wide457 mm high (0.139 m2). The bottom pressures. The optimum conditions were used through-
edge of the first relief panel was 2.875 m from the base. out the main test programme.
A single explosion vent was located at the side of the
head. Either aluminium foil or proprietary stainless steel 2.2. Effect of ignition position
panels with a nominal burst pressure of 100 mbar were
used as the vent closures. The cleaning door at the boot Three ignition positions have been used in the com-
was modified to incorporate a safety panel designed to plete series of explosion teststop (level 9), middle
protect the boot in the event of excessive pressure. This (level 5) and bottom (level 1) of the elevator. The results
was covered with a strong burst panel with a bursting show that any one position is not significantly more haz-
pressure in excess of 400 mbar. ardous than the others. There was a tendency, where the
Dust could be dispersed into the elevator using a ignition source was located at level 1 or level 9, for the
pressure injection system or by a recirculation system. peak pressure to be measured at level 9. The likely cause
Dust was injected into each leg at each level simul- of this is the congestion in the elevator head, with buck-
taneously via nozzles located flush with the inside of the ets, drive and deflection pulley wheels all mounted in
casing. Pairs of nozzles were positioned at each level. close proximitythese obstacles would tend to produce
Seven injector assemblies were fitted to the elevator, one enhanced turbulence and act to restrict the venting of
at each floor level. the explosion.
In the recirculation system, dust is initially loaded into When the igniter was located at level 5 the explosion
the elevator via a chute at the bottom of the up-leg and propagated towards the head and the boot and resulted
conveyed to the head where it is discharged into a recy- in peak pressures at a range of locations. Although there
cle leg. The discharged dust falls under gravity through was no definite pattern to the location of the peak press-
the leg to the elevator inlet and is reconveyed back up ure, its most frequent location was at level 9.
the elevator. The recycle leg has a square cross section
measuring 250 mm250 mm and incorporates an inter- 2.3. Effect of moving buckets
mediate 2 m3 capacity holding bin. The bin and the leg
are protected by explosion relief panels. The bin is fitted The results showed that operation of the buckets had
with two explosion relief panels on the top face and the no significant effect on the reduced explosion pressure
recycle leg has four explosion panels. Removal of dust compared to when the buckets were stationary.
from the elevator is achieved by directing the dust, as it
flows from the bin, to a discharge duct by the operation 2.4. Measurements of reduced explosion pressure
of a diverter valve. The diverter valve is located 3 m
below the bin. A series of experiments was performed with milk
The ignition source was installed in the elevator cas- powder, cornflour A and cornflour B, using vent clos-
ing at either level 1 upleg (close to the boot), level 7 ures with various opening pressures. The opening press-
(close to the head of the elevator) or at an intermediate ure of a vent closure was measured during each test so
point in the leg (see Fig. 2). Explosion pressures were that an indication could be obtained of the overpressure
376 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383

at which the venting process began. Initially, tests were

done with all vents acting as explosion relief; as the pro-
gramme proceeded vents were progressively blanked off
to decrease the available venting area.
The overpressure that develops at different points in
the bucket elevator depends on the ignition position and
the number of vent openings. Some examples of the
pressure variation along the elevator during an explosion
are given in Table 1. Numbers 19 in Table 1 are corn-
flour A tests; 1013 are cornflour B tests. Variations
in the dust dispersion, explosion propagation and press-
ures at which vents open lead to a spread of maximum
reduced explosion pressures for nominally identical
experiments.
The results have been analysed by plotting the
maximum value of reduced explosion pressure measured
in a particular test, Pred, against the vent opening press-
ure measured during the test, Pstat.

2.4.1. Cornflour B
Fig. 3 shows all relevant test results using Cornflour
B, for various explosion-venting arrangements.

Fig. 3. Reduced explosion pressure vs vent bursting pressure.

2.4.2. Cornflour A
Fig. 4 shows all the relevant test results using corn-
flour A. 2.5. Analysis of the results

In order to produce a worst-case analysis of the

2.4.3. Milk powder
Explosions of milk powder generated very low press-
ures, and, often, pressures were not sufficient to burst
any of the vent covers. In the explosion tests that did
burst the vent covers, pressures did not rise beyond the
bursting pressure of the cover.
results, the points for different total vent areas have been
enveloped by straight lines. From each of these lines, an
upper value of the reduced explosion pressure at Pstat
values of 0.1 bar and 0.05 bar has been estimated. These
pressures are plotted against the total vent area in Figs.
5 and 6, respectively. The total vent areas necessary to

Table 1
Explosion pressures in single-leg elevator tests

Vent covers blanked off Ignition location Pressures

Level 1 (mbar) Level 3 (mbar) Level 5 (mbar) Level 7 (mbar) Level 9 (mbar)

1 levels 1, 9 level 1 150 220 100 120 130

2 levels 1, 4, 7 level 5 220 180 200 60 230
3 levels 1, 4, 7 level 1 80 100 300 260 270
4 levels 2, 4, 6, 8 level 5 240 300 180 170 280
5 levels 4, 8 level 1 60 50 180 120 140
6 levels 4, 8 level 5 150 240 90 110 110
7 level 9 level 9 60 90 130 120 170
8 levels 2, 3, 4, 5, 6, 7, 8 level 5 440 690 680 840 350
9 levels 2, 3, 4, 5, 6, 7, 8 level 5 609 1410 1360 1370 770
10 levels 1, 9 level 9 230 100 280 140 530
11 levels 1, 4, 7 level 5 350 300 550 160 530
12 levels 1, 4, 7 level 9 450 250 440 310 160
13 level 9 level 9 140 160 160 550 610
 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383 377

Fig. 6. Explosion pressure vs total vent area at a vent opening press-

ure of 0.05 bar.

limit the reduced explosion pressure to either 0.5 bar g

or 1 bar g have been marked on Figs. 5 and 6. The results
Fig. 4. Reduced explosion pressures vs vent bursting pressure. for milk powder have been included on Fig. 5. The lines
for KSt=175 bar m s1 have been obtained by linear
interpolation.

2.6. Twin leg elevator

Because there is a relatively large space around the

buckets in a single leg elevator, it is generally easy to
propagate a flame through the entire casing. In a twin
leg elevator, however, the space around and between the
buckets is limited and it is unclear, at first sight, whether
the buckets either act as turbulence inducers in the flow
ahead of the flame and thus cause the explosion to accel-
erate, or act as obstacles to flame propagation and so
decrease the explosion velocity or prevent its propa-
gation altogether.
In order to answer this question, explosion tests were
done in which all buckets were removed from the elev-
ator and then replaced in stages until a full complement
was re-fitted. The guidance derived from these results is
based only on the tests with a full complement of buck-
ets.

2.7. Effect of ignition location

With vents at 3 m intervals (fully vented) and no

Fig. 5. Explosion pressure vs total vent area at a vent opening press-
ure of 0.1 bar.
buckets installed, the most effective location of the
ignition source for explosions of cornflour A was at the
head; with cornflour B explosions the most effective
location was at the boot. However, when buckets were
378 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383
installed, the explosion pressure tended to increase when
the igniter was located at level 7. Therefore, in the
majority of tests the ignition source was located in the
elevator head at level 7. This was at a point in the head
where the free volume was greater than elsewhere in the
elevator and so maximum development of the primary
explosion would occur prior to the expanding flames
making contact with the elevator walls and buckets.
2.8. Effect of bucket spacingfully vented elevator
Tests were carried out, initially, without buckets
installed, followed by tests with a range of bucket spac-
ings with the buckets running. In principle, the presence
of the buckets could produce two effects: (a) inhibit
flame propagation, (b) increase turbulence of the flame.
The elevator was fully vented, with vents at 3 m intervals
and with a vent at the head. A range of ignition locations
was used.
Without the buckets installed, flames of both cornflour
A and cornflour B propagated through the elevator.
The more reactive dust, cornflour B, produced a
slightly higher peak pressure (211 mbar) compared to
cornflour A (191 mbar).
To test the flame blocking ability of the buckets, they
were installed at 3 m spacing and were positioned
between the vents in a stationary position. In the station-
ary mode the buckets prevented propagation of the corn-
flour A flame and the pressure did not exceed the burst
pressure of the explosion panels; cornflour B flames
propagated through the elevator and the explosion press-
ure increased to 275 mbar. In the running mode, the
buckets still inhibited flame propagation with cornflour
A. However, cornflour B flames still propagated
through the elevator, with the explosion pressure further
increased, to 314 mbar. This provides evidence that the
presence of the buckets increased the turbulence in the
case of cornflour B explosions but the buckets inhibited
flame propagation with cornflour A, although this was
not always the case. In one test with 1 m spacing of the
buckets, the elevator running and ignition at the head, a
cornflour A flame propagated past the buckets from the
head to the boot after which it propagated up the down-
leg, producing 273 mbar at the boota pressure compa-
rable with a cornflour B explosion that produced 265
mbar, in a nominally identical test.
The results in Table 2 demonstrate the progressive
increase in explosion pressure with cornflour B as the
number of buckets in the elevator increases. With corn-
flour A, however, the buckets tended to inhibit flame
propagation, producing accompanying low pressures.
2.9. Explosion tests with varied vent configurations
The peak explosion pressures were measured for a
range of vent configurations using the four dusts. The
Table 2
Peak explosion pressuresfully vented twin leg elevator
Peak explosion Peak explosion
pressure pressure
cornflour A cornflour B
No buckets 191 211
Buckets at 3 m spacing 110 314
Buckets at 1 m spacing 273 265
Buckets at 0.28 m spacing 117 519
Buckets at 0.14 m spacing 110 659
buckets were running in all the tests. Some typical
results are given in Table 3, demonstrating the variation
of pressure along the elevator and the behaviour of the
flame.
Pressure data from the tests with the buckets spaced
at 280 and 140 mm have been plotted and are presented
in Figs. 7 and 8, respectively.
Vent spacing was set at 3, 6 and 12 m. Generally,
flame propagation was rare with cornflour A and peak
pressures were measured usually close to the ignition
in the head. Cornflour B explosions propagated into
the elevator legs and to the boot, with peak pressures
measured either in the boot or the upleg. Explosions of
cornflour C also propagated into the upleg to the boot
and into the downleg. In one test, with a vent spacing
of 6 m, the primary explosion in the head propagated to
the boot via the downleg in the direction of the bucket
travel and propagated to level 3 in the upleg. A second-
ary flame then re-emerged at level 5 in the downleg and
persisted for approximately 4 s at the vent after which
the flame re-emerged at level 5 in the upleg and at the
head, thus demonstrating how unpredictable flame
propagation can sometimes be. No flame propagation
took place in any of the milk powder tests.
2.10. Tests with the recirculation system
These tests were performed to check that worst case
conditions were being tested by dust injection tests and
that the test programme adequately covered explosions
experienced during actual running.
The elevator was cleaned out and the appropriate vent
configuration was installed. Cornflour was manually
loaded into the elevator boot and the elevator was run
for approximately 34 min to recycle the dust before the
igniter was fired. The test conditions were:
Cornflour A recycle tests
Dust: cornflour A
Bucket spacing: 280 mm
Igniter positions: level 7 (hood)
Vent configurations: vent spacings 12 m
Dust loading: l75200 kg
Cornflour B recycle tests
 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383 379

Table 3
Pressures in some twin-leg elevator tests

Dust Pitch of the Vent covers Peak pressures (mbar) Comments

buckets (m) Blanked off

Boot Upleg Downleg Head

Cornflour B 0.28 none 519 190 306 176 Flame propagated down both legs to the boot.
Cornflour B 0.28 levels 2, 4, 6 563 484 224 215 Flame propagated down the upleg to the boot and up the
downleg.
Cornflour B 0.28 levels 2, 4, 6 610 650 350 280 Flame propagated down the downleg to the boot and up the
upleg.
Cornflour B 0.28 levels 2, 4, 6 847 402 157 123 Flame propagated from the head down the upleg and up the
downleg
Cornflour B 0.28 levels 2, 4, 5, 6 640 332 293 260 Flame propagated from the head to the vents at level 3
Cornflour B 0.28 levels 2, 4, 5, 6 300 3031 921 194 Flame propagated from the head to the vents at level 3
Cornflour B 0.14 none 659 300 402 117 Flame propagated down both legs
Cornflour B 0.14 levels 2, 4, 6 1320 1303 420 875 Flame propagated down the downleg and then into the upleg
with flame from level 5
Cornflour C 0.14 none 221 143 169 118 Flame propagated from the head down the downleg and up
the upleg to the head
Cornflour C 0.14 levels 2, 4, 6 356 319 233 129 Flame propagated from the head down the downleg and up
the upleg to the head
Cornflour C 0.14 levels 2, 4, 5, 6 570 517 330 329 Flame propagated from the head down the downleg and up
the upleg to the head
Cornflour C 0.28 levels 2, 4, 6 780 777 418 175 Repeat of test 95, flame propagated down the downleg and up
the upleg with further secondary flame
Cornflour C 0.28 levels 2, 4, 5, 6 108 135 140 155 No flame propagation
Cornflour C 0.28 levels 2, 4, 5, 6 306 200 951 138 Flame propagated in the downleg to level 3

Fig. 7. Explosion pressure vs vent spacing. Buckets at 280 mm spac- Fig. 8. Explosion pressure vs vent spacing. Buckets at 140 mm spac-
ing; vent opening pressure=0.1 bar. Twin leg elevator. ing; vent opening pressure=0.1 bar. Twin leg elevator.

Dust: cornflour B
Bucket spacing: 280 mm The results are summarised in Table 4.
Igniter positions: level 7 (hood) and part way down In the tests with the recirculation system, the peak
the elevator explosion pressures were significantly less than those
Vent configurations: vent spacings of 3, 6 and 12 m developed by similar tests using the dust injection sys-
Dust loading: 100 kg tem. The lower pressures are likely to be the result of a
380 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383

Table 4
Pressures from explosions using recycling system: twin leg elevator

Dust Vent covers Ignition location Peak pressures (mbar) Comments

blanked off

Boot Upleg Downleg Head

Cornflour A levels 2, 4, 5, 6 Head 88 100 94 97 No flame seen

Cornflour B none Head 216 129 174 154 Flame propagated into downleg to the
boot, then to level 2 and 6 upleg.
Cornflour B none Downleg, 0.5 m from the top 194 85 91 111 Flame propagated into downleg to the
boot.
Cornflour B none Head 60 90 140 90 Flame propagated into the downleg to
level 3.
Cornflour B levels 2, 4, 6 Head 136 135 145 152 Flame propagated into the downleg to
level 3 and then to level 3 upleg.
Cornflour B levels 2, 4, 6 Downleg, 0.5 from top 142 118 246 92 Flame propagated into the downleg.
Cornflour B levels 2, 4, 5, 6 Head 92 356 172 317 Flame propagated into downleg then
into upleg

reduction in turbulence and differences in the dust con- 3. Discussion

centration. The comparative data are shown in Table 5.
Generally, the direction of explosion propagation was 3.1. Single leg elevator
into the downleg following the direction of the bucket
movement and occasionally into the upleg via the boot. Figs. 5 and 6 provide the information from which the
These tests show that continuing the operation of the vent spacing for dusts with different KSt values can be
elevator after the explosion can extend the duration of estimated.
the explosion compared to when the dust is injected. In Fig. 9 shows how the total vent area required limiting
one test, secondary explosions and external explosions reduced explosion pressures to 1.0 and 0.5 bar varies
continued until the operation of the elevator was with the KSt value when the value of Pstat is 0.1 and
switched off after approximately 1.5 min. Until the 0.05 bar.
buckets were shut down, their movement continued to
feed cornflour to the external flames, perpetuating com-
bustion outside the elevator. Large sustained fireballs,
typically 5 m in diameter, were produced in the tests and
dust deposits that had settled out on the platforms under
the vent openings were ignited.

Table 5
Peak Pressures in the twin leg elevator with different dust cloud forma-
tions

Vent Ignition Peak pressure (mbar)

spacing location

Injection system Recirculation system

3m head 519 216

3m downleg 194
6m head 650 152
6m downleg 246
12 m head 3031 356
12 m downleg no ignition of cornflour
Fig. 9. Total vent area vs KSt value. Single leg elevator.
 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383 381

Table 6 a vent spacing of 14 m will limit reduced explosion

Vent spacing: single leg elevator pressures to the vent bursting pressure if this is no
greater than 0.10 bar.
KSt bar m s1 Pstat bar g Pred bar g Vent spacing (m)

150 0.05 1.0 19 3.2. Twin leg elevator

0.5 10
0.10 1.0 14 The reduced explosion pressure data for bucket spac-
0.5 7 ing of 140 or 280 mm are combined in Fig. 11. This
175 0.05 1.0 7 diagram may be used to estimate vent spacing providing:
0.5 4
0.10 1.0 5
0.5 4 (i) the vents open at a pressure not exceeding 100 mbar;
200 0.05 1.0 5 (ii) the area of the vent is not less than the cross-sec-
0.5 3 tional area of the elevator leg;
0.10 1.0 4 (iii) a vent is positioned at the head and a vent is located
0.5 3
as close as possible to the boot.

The data suggest that a vent spacing of 10 m will limit

The vent spacing is calculated by assuming that one
vent is positioned in the boot and one in the head of the
elevator, and the remaining total vent area is distributed
along the elevator assuming each vent has an area equal
to the cross-sectional area of the elevator. The vent spac-
ings for several combination values of KSt, Pred and Pstat
taken from Fig. 9 are listed in Table 6. The spacing read
from Fig. 10 is rounded down to the nearest metre.
The data from the milk powder tests are shown in Fig.
5. In neither of the tests in which venting occurred did
the reduced explosion pressure exceed the vent opening
pressure, which was 125135 mbar. In the two tests
where venting occurred, the vent nearest the ignition
position opened, along with vents approximately 1012
m from the ignition position. These results indicate that
the reduced explosion pressure to 1 bar for dusts with
KSt values between 150 and 175 bar m s1 and a spacing
of 5 m is required for dusts with KSt values between 175
and 200 bar m s1. For dusts with KSt values between
100 and 150 bar m s1, a spacing of 14 m will limit the
pressure to 1 bar. For dusts with KSt values below 100
bar m s1, the reduced explosion pressure does not
exceed the bursting pressure of the vent cover even at
very high vent spacing.
In the early stages of the project, a different twin leg
elevator, 17.6 m in height and with buckets spaced at
approximately 0.32 m, had been used for some explosion
tests with cornflours A and B. Because of its age and
atypical installation of buckets, the results from the tests
using this elevator have not been used in formulating the
guidance. The peak pressures measured in the tests
where the explosion propagated through the elevator,

Fig. 10. Vent spacing as a function of total vent area. Single leg elev- Fig. 11. Explosion pressure vs vent spacing for twin leg elevator.
ator. Vent opening pressure=0.1 bar.
382 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383

however, are listed in Table 7 along with the vent spac-
ing information. A comparison of these data with the
information in Fig. 8 shows that the lines for guidance
drawn in Fig. 8 envelop the results from this twin leg
elevator.
4. Guidance
4.1. Single leg elevators
Vent openings should have an area equal to the cross-
sectional area of the elevator leg and the minimum
requirement is that vents should be fitted in the head and
as close as is practicable to the boot. This generally
means a vent within 6 m of the boot or within the rec-
ommended spacing, whichever is the lesser. The spacing
between vents along the elevator is listed as a function
of the dust KSt value, the vent burst pressure and the
reduced explosion pressure in Table 6.
For dusts with KSt values of 150 bar m s1 or less, a
vent spacing of 6 m will limit the reduced explosion
pressure to 300 mbar, when the vent static burst pressure
is 0.1 bar.
For dusts with KSt values of 100 bar m s1 or less,
vents installed in the head and boot of the elevator, with
none intervening, will limit the reduced explosion press-
ure to 0.5 bar. For dusts with KSt values of 80 bar m s1
or less, a vent spacing of 14 m will limit the reduced
explosion pressure to the vent bursting pressure if this
is no greater than 0.1 bar.
For dusts with a KSt value of 80 bar m s1, a vent
spacing of 20 m will limit the reduced explosion press-
ure to 250 mbar.
4.2. Twin leg elevators
Vent openings should have area equal to the cross-
section of the elevator leg and the least requirement is
that vents should be fitted in the head and as close as is
practicable to the boot. This generally means within 6
m of the boot or within the recommended vent spacing,
whichever is the lesser. The static burst pressure of the
vent closure should not exceed 0.1 bar.
The spacing of additional vents depends on the KSt
value of the dust.
(a) Although explosions are possible with dusts of low
KSt, generally the pressures developed by dusts with
KSt values below 100 bar m s1 are not significant,
and no additional vents are required.
(b) Dusts with a KSt value of 150 bar m s1 are able to
develop significant pressures, although the likeli-
hood of explosion propagation through the elevator
is low. Vents additional to those at the head and boot
may be required on long elevators if the casing is
comparatively weak. The graphs in Figs. 7, 8 and
11 should be used to estimate the reduced explosion
pressure for a given KSt value and vent spacing.
(c) Dusts with KSt values above 150 bar m s1 will
propagate explosions, and vents additional to those
in the head and boot are required on elevators taller
than 6 m. The graphs in Figs. 7, 8 and 11 should be
used to estimate the reduced explosion pressure for
a given KSt value and vent spacing. The strength of
the elevator should then be designed appropriately.
(d) No data are available for dusts with KSt values
greater than 210 bar m s1.
The guidance given in this section differs to that given
in NFPA-91 in some important respects. In this guidance
the area of a vent is assumed to be equal in cross-sec-
tional area to the leg in which it is installed; in NFPA-
91 a vent is assumed to have an area equal to 4/3 of the
cross-sectional area of the leg. NFPA-91 suggests the
vent area at a given position be divided into two equal
vents installed in opposite sides of the leg; the present
tests did not, however, show reaction forces to be

Table 7
Peak pressures in early tests with a twin leg elevator

Dust Vent spacing (m) Explosion pressure (mbar) Location of pressure peak

Cornflour A average of 2.5 m 231 Boot

Cornflour B average of 2.5 m 122 Level 5 upleg
Cornflour B average of 2.5 m 202 Level 4 upleg
Cornflour B average of 2.5 m 212 Level 6 upleg
Cornflour B average of 2.5 m 333 Level 3 upleg
Cornflour B average of 2.5 m 430 Level 6 upleg
Cornflour B average of 2.5 m 439 Level 4 upleg
Cornflour B 3.8 m in middle of leg; 2.5 m elsewhere 599 Level 4 downleg
Cornflour B 3.8 m in middle of leg; 2.5 m elsewhere 797 Level 3 upleg
Cornflour B 4.41 m at bottom; 2.7 m in middle of leg; 5.29 m at top 581 Level 5 downleg
Cornflour B 4.41 m at bottom; 2.7 m in middle of leg; 5.29 m at top 466 Boot
Cornflour B 4.41 m at bottom; 5.43 m in middle of leg; 5.29 m at top 566 Level 4 upleg
Cornflour B 4.41 m at bottom; 5.43 m in middle of leg; 5.29 m at top 589 Level 5 downleg
 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 15 (2002) 373383
important. NFPA-91 suggests that when the belt speed
is below 2.5 m s1 and the capacity is less than 106 m3
h1 explosion venting is not required; the present results
show, however, that belt speed has no noticeable effect
on the reduced explosion pressure.
It is essential that the elevator stop quickly in the
event of an explosion and this may be achieved by trip
switches on vent panels, but because of uncertainty as
to which panels may open, a trip on a single panel is
not likely to be reliable. A sensitive pressure switch, or
switches, or trips fitted to more than one panel are rec-
ommended.
Vents should not open into regularly occupied areas,
and wherever possible should be either ducted to the out-
side or fitted with a flameless venting device.
383
Acknowledgements
The assistance of Steve Andrews, Des Brookes and
Andrew Nicol during this project is gratefully acknowl-
edged.
References
Lunn, G. (1992). Guide to dust explosion prevention and protection.
Part 1venting (2nd ed.). Rugby: Institute of Chemical Engineers.
NFPA-91: Fires and dust explosions in agricultural and food products
facilities. Quincy, MD: National Fire Protection Association
(1999 edition).
Gillis, J. P., & Fishlock, F. H. (1982). Explosion venting and sup-
pression of bucket elevators. Journal of Powder and Bulk Solids
Technology, 6(2), 516.
 Journal of Loss Prevention in the Process Industries 15 (2002) 305310
www.elsevier.com/locate/jlp

Minimum ignition energy (MIE) a basic ignition sensitivity

parameter in design of intrinsically safe electrical apparatus for
explosive dust clouds
R.K. Eckhoff
University of Bergen, Process Technology Programme, Physics Dept., Allegaten 55, N-5007BergenNorway

Abstract

In general terms, the purpose of any safety standard is to define borderlines between safe and unsafe conditions, with reasonable
safety margins. The electrical spark ignition sensitivity of dust clouds (MIE) varies over at least eight orders of magnitude. Therefore,
in the case of intrinsically safe electrical apparatus to be used in the presence of explosive dust clouds, substantial differentiation
of the minimum requirements to prevent ignition by electrical sparks is needed. The present paper proposes a method by which
adequate differentiation of required maximum permissible currents and/or voltages in intrinsically safe electrical circuits to be used
in explosive dust clouds can be achieved. In essence, the concept is to use conservative first-order ignition curves, calculated or
estimated from the experimental MIE value of clouds in air of the actual dust. Charts to be used for design purposes are given in
the paper. Internationally standardised test methods allow MIE for clouds of any dust to be determined, at least down to the range
of a few mJ. There is, however, a need for a supplementary method covering the range of lower energies, down to 0.01 mJ.
2002 Elsevier Science Ltd. All rights reserved.

Keywords: Dust explosions; Minimum ignition energy; Ignition curves; Intrinsic safety; Electrical apparatus

1. Introduction use in areas containing combustible dusts constitutes part

Gas and dust explosions can be intiated by a range of
different ignition sources. An overview for gases was
given by Eckhoff and Thomassen (1994), and for dusts
by Eckhoff (1997). Various electrical apparatuses can
give rise to electrical sparks that may cause ignition.
Therefore, careful standardisation is required for electri-
cal apparatuses intended for use in explosive atmos-
pheres. Such standards are produced by the International
Electrotechnical Commission (IEC), and the correspond-
ing regional organisations in America, Australia and Eur-
ope.
Within IEC there is an ongoing effort to harmonise
electrical equipment standards for explosive dust clouds
with the corresponding standards for explosive gas
atmospheres. Development of a new separate standard
for design of intrinsically safe electrical apparatuses for
University of Bergen, Process Technology Programme, Physics
Dept., Allegaten 55, N-500, Bergen, Norway.
E-mail address: rolf.eckhoff@fi.uib.no (R.K. Eckhoff).
of this work (IEC, 2001).
It is true that explosive gas mixtures and explosive
dust clouds, once existing, exhibit similar ignition and
combustion properties. However, as discussed in detail
by Eckhoff (2000), the ways in which the two kinds of
clouds are generated and sustained in practice are very
different. In the case of explosive gas atmospheres,
intrinsically safe apparatus is widely used. In the case
of explosive dust clouds, however, electrical circuitry,
switches, relays, etc. to be used in areas containing such
clouds, are, for several good reasons, protected against
ingress of dust by means of dust-protected or dust-tight
enclosures (IP 5X or 6X).
However, there are some special applications where
there is a genuine need for intrinsically safe apparatus
even in environments containing combustible dusts. One
example is capacitive-level indicators for silos, bins, etc.
for storage of solid bulk materials. In this case, a live
capacitor plate, in the form of a bare metal rod carry-
ing a voltage with respect to earth, is directly exposed
to combustible powder/dust inside the silo or bin.
According to Klotz-Engmann (2001) an ignition risk

0950-4230/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 0 2 ) 0 0 0 0 3 - 7
306 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 15 (2002) 305310
could arise from electrical sparks generated by direct
contact between the energised bare rod and any
grounded metal part of the silo. This possibility has to
be considered in Zones 20 and 21, at least when taking
into account possible faults.
The intrinsic safety requirements currently adopted in
such cases by many producers and users of electrical
apparatus are those of gas group IIB. This has, in fact,
also been proposed in the current draft ExDi standard
for dusts, produced by IEC (2001). One argument put
forward in favour of this approach is that the same types
of electrical apparatuses can be used both for practically
all kinds of combustible gases and vapours, and for all
dusts, even the most spark-ignition-sensitive ones, such
as very fine sulphur. This obviously simplifies life for
the producers of electrical apparatuses, who can stan-
dardise on only one type.
However, the purpose of the kind of standards in ques-
tion should be to draw objective borderlines between
safe and unsafe conditions, with reasonable safety mar-
gins. Therefore the approach currently adopted by IEC
is unreasonable if the electrical spark ignition sensitivity
of dusts vary substantially. In that case, an intrinsic
safety standard for dusts should exhibit a corresponding
marked differentiation of the maximum permissible cur-
rents and voltages for preventing incendiary sparks in
electrical apparatus. The main objective of the present
paper is to suggest that this is necessary, and can be
achieved by basing the specification of maximum per-
missible circuit parameters on the experimentally
determined MIE value of the dust in question.
2. Range of MIEs for dust clouds
In the past it was thought the absolute lower limit for
MIEs of dust clouds in air was of the order of 10 mJ,
i.e., about two orders of magnitude above the values of
IIA gases. For example, Dorsett and Nagy (1968) pub-
lished a figure of 15 mJ for a very fine sulphur powder
of average particle size 4 um. However, Eckhoff (1975)
showed that the true MIE for very sensitive dusts was
considerably lower, by at least one order of magnitude,
than the values indicated by USBM in the 1960s. More
recently, Bartknecht (1993) indicated an extreme lower
limit, by reporting a MIE value of very fine sulphur pow-
der of only 0.01 mJ, i.e., practically identical with the
MIE of carbon disulphide.
As documented by Bartknecht (1993) extensively and
by Beck et al. (1997), MIEs of clouds of various dusts
in air span from perhaps as low as 0.01 mJ at the lower
end, to beyond the 1KJ range at the upper. Hence, there
is an obvious need to establish a differentiated system
for specifying the maximum permissible spark energy
requirements to be satisfied by ExiD equipment for
dusts.
3. Experimental assessment of MIE of dust clouds
The strong influence of the spark discharge time on
MIE for dust clouds was first indicated by the results
of Boyle and Llewellyn (1950) and Line et al. (1959).
Experiments by Eckhoff and Enstad (1976) gave further
support to the validity of the observation by Line et al.
that the shock wave produced by sparks of very short
duration can, in fact, prohibit ignition by blowing the
dust away from the spark before the ignition process gets
underway. Section 5.3 of Eckhoff (1997) provides a
more recent summary of the state-of-the-art on electric
spark ignition of dust clouds.
One important conclusion is that conservative experi-
mental determination of MIE of dust clouds requires the
use of electric sparks of sufficiently long discharge times
to prevent significant disturbance of the dust cloud by
the shock wave generated by the spark discharge. This
is achieved by introducing a 12 mH inductance in the
discharge circuit, and this feature has been included as
a compulsory option in the current international (IEC,
1994) and new European (CEN, 2001) standard methods
for MIE determination for dust clouds. (This inductance
is to be removed when the purpose of the test is to assess
the sensitivity to ignition by electrostatic discharges.)
A major limitation of the test apparatuses commonly
used so far is that they cannot generate sparks of ener-
gies significantly below 13 mJ. Efforts should be made,
therefore, to develop generators for synchronised sparks
of energies down to 0.1 mJ, or even down to 0.01 mJ.
Research to meet this need is currently being initiated
in Norway.
4. MIE and ignition curves for gases
In the current Exi standards for gases CENELEC
(2000) and IEC (1999), the ignition curves play a central
role. These are empirical borderlines separating the com-
binations of circuit parameters that cannot produce
incendiary electric spark discharges from those that can.
The curves are determined experimentally by means of
a special ignition apparatus specified in the same stan-
dards. In this apparatus the explosive gas atmosphere is
exposed to electrical sparks generated by breaking or
closing live electrical circuits. Three different types of
ignition curves are determined, viz. capacitive, inductive
and resistive. The dependence of the ignition curves on
the ignition sensitivity of the gas is taken care of by
having complete sets of such curves for the four different
standard gases: methane, propane, ethylene and hydro-
gen. All other gases and vapours are then grouped in
four corresponding categories according to their standard
minimum ignition current, determined in a closely speci-
fied way.
 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 15 (2002) 305310 307

Figure 1 shows the four generally accepted capacitive

ignition curves currently used for the four standard gas
groups. However, the figure also shows the ultimate
conservative ignition curves, based on the equation
0.5 CU2 MIE, where MIE is determined experimen-
tally under optimal conditions for ignition. As would be
expected, the latter curves are more conservative than
the standard experimental ones over the entire ranges of
voltages and capacitances.
Figure 2 shows the four generally accepted corre-
sponding inductive ignition curves currently used for the
four standard gas groups. The figure also shows the ulti-
mate conservative ignition curves, based on the equ-
ation 0.5 LI2 MIE, where MIE is measured under
standardised, optimal conditions for ignition. In this
case, the latter curves are more restrictive than the stan-
dard experimental ones for all combinations of currents
and inductances, only as long as the latter is at least 1

Fig. 2. Experimentally determined standard inductive ignition curves

(solid lines) for the four gas groups I (methane), IIA (propane), IIB
(ethylene), IIC (hydrogen), taken from IEC (1999) and CENELEC
(2000). The dotted straight lines are the corresponding theoretical
lower limit ignition curves based on the equation 0.5LI2 MIE.

mH. In the range of L1 mH, the electrical circuits used

Fig. 1. Experimentally determined standard capacitive ignition
curves (solid lines) for the four gas groups I (methane, R = 0), IIA
(propane), IIB (ethylene), and IIC (hydrogen), taken from IEC (1999)
and CENELEC (2000). The dotted straight lines are the corresponding
theoretical lower limit ignition curves based on the equation 0.5 CU2
MIE.
to determine the ignition curves behave essentially as
purely resistive circuits. In agreement with this, it can
be seen that the maximum current values of the curves
in Fig. 2, and the corresponding voltage values, fall on
the respective corresponding resistive ignition curves in
Fig. 3.
Figure 3 shows the corresponding four generally
accepted resistive ignition curves for the four standard
gas groups. The figure also shows a unique resistive
ignition curve for clouds in air of lycopodium powder,
determined very recently by Lunn (2001). However, in
the case of resistive ignition curves, there is no simple
theory available for constructing conservative curves
corresponding to those in Figs. 1 and 2 (dotted lines).
Therefore, an empirical substitute had to be found, and
the dotted lines in Fig. 3 represent the simple equation
I 15000 MIE0.45 1 / U2 (1)
where I is in A, MIE in J, and U in V. As can be
seen, this equation works for all the three gas groups, as
well as for the recent unique data for lycopodium
308 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 15 (2002) 305310
Fig. 3. Experimentally determined standard resistive ignition curves
(solid lines) for the four gas groups I (methane), IIA (propane), IIB
(ethylene), IIC (hydrogen), taken from IEC (1999) and CENELEC
(2000) and for clouds of lycopodium in air, taken from Lunn (2001).
The dotted straight lines are corresponding linear ignition curves based
on the empirical correlation equation (Eq. (1)), where I is in A, MIE
in J, and U in V.
obtained by Lunn (2001). Lunn also gives an experi-
mental resistive ignition curve for a sulphur powder,
showing close resemblance with that of ethylene (II B
gases). However, since MIE for this dust is not known,
it is not possible to confirm whether Eq. (1) works even
for this dust. It is known that MIEs of sulphur dusts are
generally exceptionally low, and for a very fine sulphur
powder, Bartknecht (1993) reports a value as low as 0.01
mJ. Therefore, a MIE value of 0.07 mJ, as for ethylene,
would not seem unreasonable for a somewhat coarser
sulphur dust than that tested by Bartknecht.
There is an evident need for more experimental
ignition data for dusts, in particular for dusts of consider-
ably higher MIEs than that of lycopodium (7 mJ), for
further substantiation of Eq. (1). If new data should show
that a modification of the equation is required, adequate
fit to new data can most probably be obtained by just
changing the constant 15000 and the MIE exponent 0.45
to more appropriate values.
5. Conservative ignition curves for dust clouds
based on MIE
For capacitive circuits, conservative ignition curves
can be obtained directly from the equation 0.5 CU2
MIE. Charts showing this graphically are given in
Fig. 4.
In the case of inductive circuits, conservative ignition
curves can be obtained directly from the equation
0.5 LI2 MIE for L 1 mH. For smaller L, the
ignition current will be independent of L, and the circuit
will essentially be resistive. Charts showing conservative
ignition curves for inductive circuits using MIE as a
parameter are given in Fig. 5.
For resistive ignition circuits, the situation is less
straightforward. Recent research (Lunn, 2001) showed
that the resistive ignition curve for clouds in air of a
sulphur dust was very close to that of ethylene in air (IIb
gases). This is most interesting in view of the indication
given by Bartknecht (1993) that MIE of clouds of
ultrafine sulphur dust may be as low as 0.01 mJ, i.e., of
the same order as MIE of carbon disulphide. This means
that a somewhat coarser sulphur powder, which may
have been used by Lunn, could well have had a MIE of
about 0.07 mJ, corresponding to that of ethylene (IIb
standard gas). This could indicate that Eq. (1) may be
used for first-order conservative estimates of resistive
ignition curves even for dust clouds.
A graph representing this equation is given in Fig. 6.
It is clear, however, that there is considerable uncertainty
involved in extensive extrapolation of this equation into
the regions of higher MIE. Therefore, there is a need for
more research to establish more reliable resistive circuit
ignition data for dust clouds in the range of higher MIEs.
6. Conclusions
1. Electric spark ignition sensitivities of dust clouds,
expressed in terms of their minimum ignition energies
(MIE), vary substantially, across at least eight orders
of magnitude. Therefore, substantial differentiation of
current and voltage limitation requirements for intrin-
sically safe equipment to be used in explosive dust
clouds (ExiD), is essential in intrinsic safety standards
for dusts.
2. The experimentally determined MIE value for a given
dust, determined by current international standard
 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 15 (2002) 305310 309

Fig. 4. Theoretical conservative capacitive ignition curves for dust clouds based on the equation 0.5 CU2 MIE.

Fig. 5. Theoretical conservative inductive ignition curves for dust clouds based on the equation 0.5LI2 MIE, valid for L 1 mH

methods, using a 12 mH inductance in the capacitive
spark-generating circuit, is an obvious, conservative
basic criterion for such differentiation.
3. For capacitive circuits, conservative safe combi-
nations of circuit parameters are obtained by
satisfying the criterion 0.5 CU2 MIE.
4. For inductive circuits, conservative safe combinations
of circuit parameters are obtained by satisfying the
criterion 0.5 LI2 MIE as long as L 1 mH. For
L 1 mH, the circuit is essentially resistive.
5. For resistive circuits, the use of a simple empirical
relation between current and voltage, using MIE as a
sensitivity parameter, may be a solution. The empiri-
cal relation in Eq. (1), where I is in A, MIE in J and
U in V, was shown to apply to the four standard gases
methane, propane, ethylene and hydrogen, as well as
to clouds in air of lycopodium powder. This powder
is the only one for which both the resistive ignition
curve and the MIE have been reported. Whether the
proposed relation is also valid for other dust, in parti-
cular those of significantly higher MIE values than
that of lycopodium, remains to be investigated.
310 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 15 (2002) 305310
Fig. 6. Theoretical resistive ignition curves based on the empirical
equation (Eq. (1)) obtained from correlating ignition data for gases.
6. There is a need for a standard method for experi-
mental determination of MIE for dust clouds in the
important energy range below 1 mJ, which is not
covered by typical standard test apparatuses in current
use. A research programme aimed at developing such
a method is currently being set up in Norway.
References
Bartknecht, W. (1993). Explosionsschutz. Berlin: Springer-Verlag.
Beck, H., Glienke, N., & Mohlmann, C. (1997). BIA-Report 13/97:
Combustion and explosion characteristics of dusts. HVBG, Sankt
Augustin, Germany, ISBN: 3 88383 469 6
Boyle, A. R., & Llewellyn, F. J. (1950). The electrostatic ignitability
of dust clouds and powders. J Soc Chem Ind Trans, 69, 173181.
CEN (2001). Determination of minimum ignition energy of dust/air
mixtures Draft prEN 13821, 2001-04-27. CEN, Central Secretariat,
rue de Stassart 36, B-1050 Brussels, Belgium.
CENELEC (2000). Electrical apparatus for potentially explosive
atmospheres Intrinsic safety i. European Standard EN 50020,
CENELEC, Central Secretariat, rue de Stassart 35, B-1050 Brus-
sels, Belgium.
Dorsett, H.G., & Nagy, J. (1968). Dust explosibility of chemicals,
drugs, dyes, and pesticides. Report of Investigations 7132, US
Department of the Interior, US Bureau of Mines.
Eckhoff, R. K. (1975). Towards absolute minimum ignition energies
for dust clouds. Combustion & Flame, 24, 5364.
Eckhoff, R. K., & Enstad, G. G. (1976). Why are long electric sparks
more effective dust explosion initiators than short ones? Combus-
tion & Flame, 27, 129130.
Eckhoff, R. K. (1997). Dust explosions in the process industries.
Oxford: Butterworth-Heinemann.
Eckhoff, R. K. (2000). Design of electrical equipment for areas con-
taining combustible dust. Why dust standards cannot be extensively
harmonised with gas standards. Journ Loss Prev Process Ind, 13,
201208.
Eckhoff, R. K., & Thomassen, O. (1994). Possible sources of ignition
of potential explosive gas atmospheres on offshore process instal-
lations. Journ Loss Prev Process Ind, 7, 281294.
IEC (1994). Method for determining minimum ignition energy of
dust/air mixtures, International standard IEC 1241-2-3. Inter-
national Electrotechnical Commission, Central Office, rue de Var-
embe, Geneva, Switzerland.
IEC (1999). Electrical apparatus for explosive gas atmospheres
Part 11: Intrinsic safety i, International standard IEC 60079-
11. International Electrotechnical Commission, Central Office, rue
de Varembe, Geneva, Switzerland.
IEC (2001). Electrical apparatus for use in the presence of combust-
ible dust Part 11: Protection by intrinsic safety iD. Inter-
national draft standard IEC 61241-11. International Electrotechn-
ical Commission, Central Office, rue de Varembe, Geneva,
Switzerland.
Klotz-Engmann, G. (2001). Private communication between Dr. Klotz-
Engmann and R.K. Eckhoff, July 2001
Line, L. E., Rhodes, H. A., & Gilmer, T. E. (1959). The spark ignition
of dust clouds. Journal of Physical Chemistry, 63, 290294.
Lunn G.A. (2001). Testing methods for electrical apparatus installed
in a dusty environment with a potential risk of explosion. Synthesis
Report No. SMT-CT98-2273 from a research project funded by the
European Union under the Competitive and Sustainable Growth
programme 19982002. Available: http://www.safetynet.de/.
06 Electrical power supply and utilization (scientific, technical)

electrode. This is then baked under controlled conditions. In case of needed for the leaf meal product. This concept provides a means for
usage as anodes in steel electric furnaces (or as other carbon and rural economic development with a sustainable approach to production
graphite products), the electrodes could undergo further processing agriculture.
like pitch impregnation or graphitization. During heat treatment, some
of the organics are destructively distillated, vaporized or decomposed, 02/00475 Natural coal as precursor for anode of Li-ion cell
resulting in carbon deposition in the electrode. As the vaporized Frackowiak, E. et al. Mol. Phys. Rep., 2000, 27, 26-30.
materials exit the body of the electrode they cause porosity in the walls, Electrochemical insertion of lithium into low temperature cokes
which results in reduction in density, current carrying capacity and obtained from pyrolysis of lignite at 700, 900 and 1000C has been
flexural strength. The paper present studies to improve some correlated with their elemental composition, microtexture and surface
physicochemical characteristics of these electrodes (such as coefficient functionality. The kind of surface groups strictly affects the specific
of thermal expansion, mechanical strengths, density, pore volume, reversible capacity of lithium storage which reaches values in the range
porosity, etc.), obtained in different manufacture steps, by addition of from 310 to 370 mAh/g. The microtextural properties of carbon have
varieties of coal tar pitch. These include attempts to improve the been determined from nitrogen adsorption data and it has been shown
chemical compatibility of the coke-pitch system in the mixture and that a small amount of mesopores present together with basic groups in
establish the method and the point of introduction of additive, the the coke 700C contribute to a large hysteresis between insertion and
concentration required and appropriate analysis control during the deinsertion processes. The potentiodynamic investigation proved that
entire manufacture. Methods of analysis used include TG and the process of lithium insertiomdeinsertion is kinetically controlled by
porosimetry. The microstructure of the electrodes was studied through the diffusion of lithium ions. The coke prepared at 1000C is a cheap
a wide range and the data obtained include pore size and pore volume material which gives quite interesting performance for lithium storage.
distribution, surface area, porosity, particle size distribution and type
of pores. The overall results clearly indicate better characteristics and
performance for electrodes with additives as against electrodes without 02/00476 Novel techniques for evaluation of explosion-safe
them, such as lower porosity, lower thermal expansion coefficients and characteristics of dust for power plant fuels and temperature
greater mechanical strength. These data are analysed with respect to of dust-air-gas mixture downstream the grinder (separator)
the process step and electrode type. Dudorov, Y.D. Elektr. Sm., 2000, 3, 24-28. (In Russian)
Pulverized coals, oil shales, and peat from a number of deposits and
basins are classified into four categories based on their explosion
02lOO472 Improvement of the efficiency of flue gas propensities. Maximum temperatures of pulverized fuel-air-gas mix-
desulfurization plants - use of adipic acid in the tures downstream the pulverizer (separator) which do not increase
Frimmersdorf Brown Coal Power Plant of RWE Energie AG their explosibility are also given.
Hoffmann, W. et al. VGB Tech. Vu. Grosskrafr,~erksherr.. [Tagmgs-
her.] VGB-TB, 1999, 435, PaperR8/i, PaperR8/1_PaperR8/12. (In 02/00477 Oa structure L12/3[Ni1,3Mn2/3]02: a new layered
German) cathode material for rechargeable lithium batteries. I.
Addition of 300 ppm adipic acid to the limestone suspension improved Electra-chemical properties
the efficiencv of flue aas desulphurization bv 6% (absolute) in full-scale Paulsen, J.M. er al. J. Eleclrochem. Sot., 2000, 147, (3), 861-868.
tests. This enabled energy saving owing io switching off additional Layered Li-Mn oxide-based materials (Liz,s[Ni,Mn, _.]-Oz) with the
pumps of the absorber circuit. unconventional Oz structure were prepared by ion exchanging Na for
Li in the corresponding P2 structure sodium bronzes, NaUs
(Ni,Mnt .)Oz. Powder X-ray diffraction (XRD) was used to confirm
02/00473 Integrated fuel cell electric power generation the purity and structure of the phase. The Bragg peaks from
system for submarine applications Lizis[Nil,sMt&,s]Oz can be indexed based on a hexagonal unit cell
Epp, D.G., Baumert, R.M. U.S. US 6,063,515 (Cl. 429-17; HOlM8/04), with dimensions a = 2.8609 A and c = 10.061 A, and the diffraction
16 May 2000, Appl. 996,101, 22 Dee 1997. 10. pattern can be approximately matched assuming that the atoms are in
An integrated electrochemical fuel cell power plant, including a fuel 2a and 2b positions of space group P63mc. However, the initial
processing subsystem is suitable for use in a submarine. Selection and structure is more complex, because some observed diffraction peaks
control of operating temperatures and pressures in the various (i.e. 111) are forbidden in this space group. In .tl/u XRD and
subsystems and components of the fuel cell power plant are important electrochemical testing were used to investigate the behaviour of 02-
aspects of submarine power plant design. A catalytic burner is used to type Liz,?[Ni,Mn, _x]O, as a cathode for rechargeable lithium batteries.
heat a heat transfer fluid, which in turn is used to heat a vapourizer and It exhibits a large reversible capacity of approximately 180 mAh/g
reformer in the fuel processing subsystem. A hydrogen separator is divided into two plateaus centred near 2.8 and 3.9 V. In situ XRD
used to obtain a substantially pure hydrogen fuel stream from the proves that after an initial phase transformation, during which all
reformate stream, which is directed to and recirculated through the forbidden diffraction peaks vanish, the extraction and insertion of
fuel cell stack in a closed-loop. The fuel processor is operated at high lithium occurs without any first order phase transitions. The material
pressure which facilitates the discharge of waste exhaust streams remains layered and does not transform to the spine1 structure during
overboard, and the fuel cell is operated at a substantially lower cycling. It shows good capacity retention versus cycle no. at both 30
pressure. and 55. Differential scanning calorimetry measurements on charged
electrodes show exotherms that are about one order of magnitude less
intense than those from charged Li,CoOz electrodes.
02iOO474 Integrated protein production and electricity
generation using renewable alfalfa feedstock in a 02lOO476 Operation status of Krupp Uhde GmbH
combination advanced IGCC and feed processing Puertollano IGCC plant
arrangement Schellberg, W. Pe/rotech (Tokyo), 2000, 23, (2), 169-172. (In
Delong, M.M. et al. Proc. Renewhie Adv. Energy SJ.SI. 21st Cenlury, Japanese)
1999, 91-100. The design and operation conditions of the title coal-fired power plant
A feasibility study was conducted to determine the viability of a co- in Spain are described, including IGCC plant scheme, initial operating
production concept of alfalfa leaf meal as a concentrated protein results, and problems encountered in the plant operations.
animal feed and the generation of electricity from the remaining stem
material. Alfalfa is a well-known and widely-planted crop that offers
02/00479 Preparation of carbon based material for electric
environmental and soil conservation advantages when grown as a 4-
and special uses from degassed petroleum coke and other
year segment in a T-year rotation with corn and soybeans. Alfalfa fixes
nitrogen from the air, thereby enhancing soil nitrogen and decreasing raw materials
Duinea, N. et al. Rom. RO 108,549 (Cl. COlB31/04), 30 Jun 1994,
the need for manufactured nitrogen fertilizer. With alfalfa yields of
Appl. 142,488, 14 Nov 1989. 5. (In Romanian)
8.96 metric tonnes/ha (4 dry tons per acre) per yr and with separated
Carbon based material for electric (electrodes, electric brushes, etc.)
alfalfa leaves being sold as a high-value animal feed, separated alfalfa
and special uses is perpared by mixing degassed coke 52-60%, natural
stems can be economically viable fuel feedstock for a gasifiericombined
graphite lo-14%, purified coal tar 26-38% at 100. The cooled product
cycle power plant. This paper reports on a feasibility study for an
is pressed to prepared specimens mainly in form of plates.
integrated biomass power system, where an energy crop (alfalfa) is
coupled to a processing plant and a power plant (integrated
gasification combined cycle with hot gas cleanup) in a way that benefits 02/00460 Process simulation of integrated systems of
the joint venture of an alfalfa producers cooperative and a utility entity. smelting reduction ironmaking-gasoline synthesis-
The sale of a mid-level protein animal feed co-product and electricity electricity generation
both support the production cost of alfalfa. The co-product/fuel Zhang, X. el cl. Huagong Yejin, 2000, 21, (2), 164170. (In Chinese)
processing operation uses a common train of equipment, thereby The integrated system of smelting reduction ironmaking-gasoline
requiring neither product to carry the total cost. The power plant synthesis-electricity generation, in which the coal gas produced in the
provides an important continuous demand for the feedstock and results ironmaking module is first cleaned and then used as the fuel for the
in continuous supply of leaf product to provide a reliable supply electricity generation module or raw material of the gasoline synthesis

46 Fuel and Energy Abstracts January 2002

 Journal of Loss Prevention in the Process Industries 14 (2001) 443447
www.elsevier.com/locate/jlp

Characteristics of the vibrating-mesh minimum ignition energy

testing apparatus for dust clouds
a,*
K.S. Choi , M. Yamaguma b, T. Kodama b, J.H. Joung c, M. Takeuchi d

a
Graduate School of Science and Engineering, Ibaraki University, 12-1, Nakanarusawa-cho 4-chome, Hitachi-shi, Ibaraki 316-8511, Japan
b
National Institute of Industrial Safety, 1-4-6, Umezono Kiyose-shi, Tokyo 204-0024, Japan
c
Department of Safety Engineering, Seoul National University of Technology, 172, Gongreung-dong, Nowon-gu, Seoul 139, South Korea
d
Department of Electrical and Electronic Engineering, Ibaraki University, 12-1, Nakanarusawa-cho 4-chome, Hitachi-shi, Ibaraki 316-8511,
Japan

Abstract

A novel apparatus for testing the minimum ignition energies of flammable dusts is introduced. Unlike the conventional apparatus
(the Hartmann tube), this new apparatus employs a vibrating mesh to produce a dust cloud. Using three kinds of powders, namely
lycopodium, anthraquinone and polyacrylonitrile, which are designated as the samples for calibration by the International Electro-
technical Commission (IEC, 1994) standards, fundamental characteristics were experimentally investigated. Concerning the mini-
mum ignition energies (MIEs), the new testing apparatus worked well for two samples, lycopodium and polyacrylonitrile. The MIE
for anthraquinone, however, was by far larger than the expected value. We concluded that the aggregation of anthraquinone particles
is the main cause of the difference and is attributable to both the tribo-charges acquired by the particles when passing through the
mesh and the filamentary shape of the crystal. Other essential factors for characterizing the testing apparatus, such as the concen-
tration of dust, the shape and spacing of the sparking electrode system, and the impedance of the sparking circuit are discussed.
2001 Elsevier Science Ltd. All rights reserved.

Keywords: Dust explosion; Electrostatic discharge; Minimum ignition energy; Vibrating-mesh

1. Introduction prior to handling them. For long, the measurement of

Dust explosions caused by electrostatic sparks have
always been a concern in industries handling a large
amount of flammable dust like flour or fine plastic pow-
ders. Nowadays, as powder technologies have been rap-
idly progressing, a variety of new flammable powders
have been produced and used in various industrial pro-
cesses. Some of those powders are so sensitive that even
a spark with 1mJ or less energy can ignite them
(Matsuda & Yamaguma, 2000).
Therefore, dust explosions have become a common
concern in companies, which handle flammable pow-
ders, regardless of their amount. In planning protective
measures for dust explosions, it is inevitable to estimate
the minimum ignition energy (MIE) for relevant dusts
* Corresponding author. Fax: +81-424-91-7846.
E-mail addresses: choiks@anken.go.jp (K.S. Choi), takeuchi@
ee.ibaraki.ac.jp (M. Takeuchi).
MIEs has been dominantly done with the Hartmann tube
testing apparatus although it is costly and requires oper-
ational skill to obtain reliable data. To make the
measurements drastically easier, quicker, and more
economical, a simplified testing apparatus that makes use
of a vibrating mesh as a dust cloud generator was pro-
posed by some of the authors, and many interesting
results were obtained (Cheung, Kodama, & Yamaguma
1997, 1998). Nevertheless, it still remains to be determ-
ined whether the data taken with the new apparatus show
good agreement with those with the traditional Hartmann
tube. This should be urgently clarified to materialize a
practical version of the new method. Therefore, after
making a few modifications, we evaluated the current
version using several powders specified by the Inter-
national Electrotechnical Commission (IEC) standard. In
addition to obtaining the MIE data, other factors that are
imperative with regard to the performance of this kind
of device have been investigated.

0950-4230/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 0 1 ) 0 0 0 3 2 - 8
444 K.S. Choi et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 443447
Fig. 1. Structure of the powder hopper of the vibrating-type testing
device.
2. Experimental
2.1. General idea of the new apparatus
Unlike the conventional Hartmann tube, in which the
dust cloud to be ignited is produced by blowing off the
settled powder with compressed air, the new apparatus
produces it by sieving with a fine mesh. The actual
structure including the dimensions of the powder hopper
is shown in Fig. 1. In the hopper, a U-shaped stainless
mesh holder is set, and a certain amount of powder is
loaded into it. Then, this holder is placed at the top of
a cylindrical combustion case with sparking electrodes
fixed at its side and, flame arresters made of coarse cop-
per mesh were placed under the hopper and on its bot-
tom. Since the combustion case is rigidly fixed on a
vibrator as in Fig. 2, the powder passes though the mesh
downward, and a streaming dust cloud is formed in the
middle of the case. The vibration strength, which affects
the dust concentration, is adjusted with the volume selec-
tor at eleven different levels, from 0 (minimum) to 10
(maximum). If sufficient spark energy is given, the dust
cloud is ignited to a blaze. The flame, however, will be
quenched and extinguished immediately by the cooling
effect of the flame arresters and will not leave any harm-
ful effects except a small and often negligible amount
Fig. 2. Vibrating minimum ignition energy measuring apparatus.
of cinders and ashes. The operation for MIE measure-
ment can usually be continued even after combustion has
occurred as long as the powder remains in the hopper.
Furthermore, unburned powder settled on the base of the
case may be reused for another testing if impurities such
as ashes are carefully removed. The size of the mesh
should be determined so that all the powder in the hop-
per can pass through without changing its size distri-
bution.
Thus, reliable MIE testing could be carried out more
easily, rapidly, and economically with this new appar-
atus provided that it gives the same data as the conven-
tional apparatus does.
2.2. Setup and method
Besides the vibrating mesh system described above,
the experimental setup was made up of a power source
equipped with a high voltage dc generator and a set of
capacitors ranging from 100 to 5000 pF, a digital storage
oscilloscope (Textronix TDS520), a current amplifier
(Textronix AM503), a personal computer (PC) for con-
trolling and data processing, and other auxiliary devices
as shown in Fig. 2. The discharging or sparking elec-
trodes had either sharp or round tips made of stainless
steel rods with a diameter of 2 mm. Either an inductor
of 0.94 mH (slightly smaller than the 1.0 mH specified
in the IEC standard) or a resistance ranging from 100
k to 3 M was serially connected to the discharging
circuit to vary the duration of the spark. In this experi-
ment, three kinds of powderlycopodium, anthraqui-
none, and polyacrylonitrilewere selected and put to the
test as samples in accordance with the calibration
method for this kind of equipment described in the IEC
standard. The typical testing procedures were as follows:
1. Putting the sample powder in the hopper and firmly
mounting it on the combustion case.
2. Setting the level of voltage of the power source at the
initial level and switching on the vibrator.
3. Manually triggering the ignition circuit to make the
electric spark. If no ignition was obtained, the trig-
gering operation was repeated every one second up
to ten times.
4. Changing the level of voltage or the magnitude of
vibration, if necessary, and continuing testing until the
least sparking energy level (the MIE) was obtained.
In this experimentation, ignition was acknowledged to
have taken place when the dust cloud caught fire by elec-
tric sparking and the flame reached either the upper or
the lower flame arrester.
The sparking energy W (J) was calculated in either of
the following ways:
1. When an inductor was connected, the formula below
was used.
 K.S. Choi et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 443447
W1/2C(V21V22), (1)
where C (F) is the capacitance of the capacitor and
V1 (V) and V2 (V) are the voltage of the power source
before and after sparking, respectively.
2. When a resistor was connected, the calculation was
done with the PC using the formula below.
t all round. Neither lycopodium nor polyacrylonitrile par-
W n(t)i(t)dt (2)
0
where n(t) (V) and i(t) (A) are the voltage and the
current of the spark, respectively, and t (s) is the dur-
ation of the spark.
In order to evaluate the MIEs obtained by the new
method, the data were also taken using the same samples
with the conventional Hartmann tube apparatus
(MIKE-3).
3. Results and discussion
3.1. Evaluation of MIE data obtained by the new
method
The MIEs obtained by both the new and conventional
apparatuses together with the target values stated in the
IEC standard are given in Table 1. In this case, we
believe that the data taken with MIKE-3 should be
regarded as the benchmarks rather than those found in
the IEC standard (rightmost column) because our
samples might have not been exactly the same as those
expected by the standard. The MIEs taken with the
vibrating mesh and the Hartmann tube were similar in
the case of lycopodium and polyacrylonitrile. The MIEs
for polyacrylonitrile were slightly different, yet this
problem can be solved by tuning-up carefully the new
apparatus. In the case of anthraquinone, however, the
difference seems to be too large to be made up by mere
adjustment, and the reason of the difference should be
clarified.
We finally noticed that the shape and size distribution
of the anthraquinone dust would be attributable. It is well
Table 1
MIEs measured by the new (vibrating mesh) and the conventional
(Hartmann Tube) method (L=0.94 mH)
Substance MIE (mJ)
Vibrating mesh Hartmann tube Target value
(MIKE-3) recommended
by IEC
Lycopodium 2329 1030 510
Anthraquinone 2934 13 26
Polyacrylonitrile 3441 1030 26
445
known that the greater the particle size, the more energy
that is needed to ignite the dust. If the size distribution
changed, it could affect the flammability of the anthra-
quinone dust.
As shown in Fig. 3(a), all lycopodium particles had
almost the same size and round shape. Polyacrylonitrile
(b) particles come in various sizes, but their shapes were
ticles had any other sub-components. Anthraquinone
particles (c), however, consisted of numerous small slen-
der crystals of irregular shape. It was also found that the
median diameter of particles passing through the vibrat-
ing mesh, as measured by the laser diffraction method,
was 178 m, as shown in Table 2. Since the anthraqui-
none powder was sieved with a 100-mesh (150 m in
bore) prior to the test, this value of 178 m is much
larger than expected. There are several reasons for the
enlargement of the particles. First, there are attractive
forces among particles due to the electrostatic charge on
particles passing through the mesh. All materials are
more or less electrified when they are in contact with
different materials. The charge amounts of the samples
measured by the Faraday cage method are shown in
Table 3.
It is obvious that anthraquinone obtained a conspicu-
ously large amount of charges in comparison with the
others. Taking the electrical attractive forces exerted
among particles into consideration, aggregation through
this process would be quite possible. Second, the slen-
der, filamentary structure of the crystal could also be
responsible for the aggregation. Filamentary substances
naturally tend to become intertwined. A visual evidence
of the aggregation of anthraquinone particles is given in
Fig. 4. The picture depicts how anthraquinone particles
stick together on the mesh.
These facts suggest that a method to enhance the dis-
persion of particles should be devised for making the
new apparatus more practical.
3.2. Estimation of dust concentration
One of the weak points of our apparatus is that the
dust concentration is not easily estimable. Generally
speaking, as the vibration strength increases, the concen-
tration becomes higher. Therefore, it will be possible to
obtain the MIE for a certain dust without knowing the
exact concentration. In a real situation, however, it is
often required to determine the ignition energy at a cer-
tain concentration level. Consequently, the relationship
between the vibration level and the dust concentration
needs to be clarified.
As a trial, we employed a rather simplified model in
which each particle flowed at a constant velocity,
obtaining an even concentration at any point in the dust
cloud. Thus, we obtained the following formula:
NM/(rS), (3)
446 K.S. Choi et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 443447

Fig. 3. SEM images of powders used in the experiments. (a) Lycopodium; (b) polyacrylonitrile; (a) anthraquinone.

Table 2 second). With regard to lycopodium powder, the concen-

Comparison of median diameter of powders between before and after tration and the least ignition energy at each vibration
passing through mesh
level were obtained as shown in Fig. 5. In this case, the
Substance Median diameter (m) spacing of the sparking electrodes was 4mm, and a
After passing Before passing resistor of 100 k was used in place of the inductor.
through mesh through mesh The MIE in this condition was 4.6 mJ at 0.71 kg/m3.
Lycopodium 38 31
Anthraquinone 178 18 3.3. Influence of discharge conditions on the MIE of
Polyacrylonitrile 73 27 lycopodium

The influence of electrode spacing, electrode shape,

and circuit resistance on the MIE of lycopodium was
Table 3
Charge of powders after passing through the vibrating mesh
examined at a constant dust concentration. The electrode
spacing of 4 mm gave the lowest ignition energy for
Substance Charge amount (C/kg) lycopodium, where the circuit resistance was 100 k
Minimum Maximum Mean and both the electrodes were round. When the electrode
spacing was 2 mm or less, the cooling effect by heat
Lycopodium 1.5 2.4 1.9
Anthraquinone 2.3 11 6.1
conduction at a discharge electrode was greater than in
Polyacrylonitrile 0.1 0.54 0.23 the other two cases (4 and 6 mm), bringing about a larger
MIE. When the electrode spacing was 6 mm or more,
convection and radiation in the electrode space promoted
loss of heat, which made the MIE higher.
As to the shape of discharge electrode, electrodes with
sharp tips gave smaller MIE values than those with
round tips in an LC circuit. This can be attributable to
the shorter discharge duration for sharper tips because a
larger radius of curvature promotes heat dissipation. In
an RC circuit, however, the sharp tips and the round tips

Fig. 4. Aggregation of anthraquinone particles on the mesh (CCD

camera image).

where N (kg/m3) is the concentration, M (kg/s) is the

weight of powder per second fed from the hopper, r
(m/s) is the average velocity of the particles, and S (m2)
is the horizontal cross-sectional area of the dust stream.
The velocity r was determined by tracing particles Fig. 5. Dust concentration and MIE of lycopodium with respect to
with a high-speed video camera (up to 600 frames per the vibration step.
 K.S. Choi et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 443447
gave similar MIE values. This can be attributable to the
relatively long discharge duration and the impartial dissi-
pation of electric power during that time.
The circuit resistance of 100 k gave the lowest
energy for ignition, 4.6 mJ, where the electrode spacing
was 4 mm and both the electrodes were round. These
results suggest that the ignition of dust depends on the
electric power and the discharge duration.
4. Conclusion
The new apparatuses for testing dust MIEs were put
to the test to evaluate their data in comparison with those
by the conventional method. The data for powders with
round shapes, i.e. lycopodium and polyacrylonitrile,
agreed with the benchmarks on the whole. The MIE for
anthraquinone was much larger than its reference. It is
concluded that the high-level tribo-charge and the fila-
mentary crystal sub-component of the particle enhance
the aggregation of the particles, leading to insensi-
tiveness to electric sparks. If this shortcoming is over-
come, this method will be another reliable testing
method.
Other factors essential to the device are as follows:
1. When a 100 k resistance was connected to the
447
sparking circuit and the spacing of the sparking elec-
trodes was set at 4 mm, the lowest level of MIE, 4.6
mJ, was attained for lycopodium powder.
2. Electrodes with sharp tips gave smaller MIEs than
those with round tips in an LC sparking circuit, while
shape made no difference in an RC circuit.
Acknowledgements
This work was supported financially by the Korea
Research Foundation in 1999.
References
Matsuda, T., & Yamaguma, M. (2000). Tantalum dust deflagration in
a filter bag collecting device. Journal of Hazardous Materials A,
77, 3342.
Cheung, W. L., Kodama, T., & Yamaguma, M. (1997). Vibrating mini-
mum ignition energy measurement system. The lecture paper
presented at the Institute of Static Electricity of Japan (pp. 181
186).
Cheung, W. L., Kodama, T., & Yamaguma, M. (1998). Development
of vibrating minimum ignition energy for powder. In ESAIJ Joint
Symposium on Electrostatics Proceedings (pp. 143150).
IEC, International Standard (1994). 1241-2-3.
 Journal of Loss Prevention in the Process Industries 14 (2001) 495502
www.elsevier.com/locate/jlp

Cornstarch explosion experiments and modeling in vessels ranged

by height/diameter ratios
a,*
S. Radandt , Jianye Shi a, A. Vogl a, X.F. Deng b, S.J. Zhong b

a
Forschungsgesellschaft fur angewandte Systemsicherheit und Arbeitsmedizin (FSA), 68165 Mannheim, Germany
b
Industrial Explosion Protection Institute (IEPI), Northeastern University, 110006 Shenyang, PR China

Abstract

Height to diameter (H/D) ratio is one of the important parameters affecting premixed particleair combustion characteristics.
This paper focuses on the behavior of cornstarch combustion in closed vessels with changed H/D ratios and fixed volumes; and a
combustion model is employed to simulate the experiments. An EulerianLagrangian approach for two-phase flows was used in
the model and conservation equations of unsteady turbulent two-phase reacting flows were solved in two-dimensional domains.
Heat loss to the vessel walls was taken into consideration in the model. The simulation results have a good agreement with those
of experiments. Further simulations were carried out for higher H/D ratios from 8 to 15. These results show that H/D=8 is a
changing point. When H/D8, the maximum pressure and the rate of maximum pressure rise decrease with increasing H/D ratios.
While H/D8, the both have an increasing tendency with increasing H/D ratios. 2001 Published by Elsevier Science Ltd.

Keywords: Cornstarch; Dust explosion experiment; Closed vessel; Modeling; CFD

1. Introduction experiments. Our projects aimed to get more knowledge

Numerous experiments in elongated vessels show that
the height to diameter H/D ratio has an effect on the
behavior of dust explosions (Bartknecht, 1993; Radandt
and Shi, 1997). NFPA 68 (NFPA68, 1988) also suggests
that explosion venting of vessels of H/D5 should be
considered specially. Pu (1988) had made laboratory
scaled tests with different H/D ratios and volumes. Bart-
knecht (1998) had done large-scale explosion tests in
vessels of which volumes and H/D ratios were changed
but surface to volume ratio was fixed. Radandt (1983)
had done venting experiments in silos with both horizon-
tal and vertical vessels (V=20 m3, H/D=6.25). Vogl
(1996) investigated flame propagation in pipelines with
different lengths (40, 48, and 60 m) and diameters (0.10,
0.15 and 0.20 m). Gardner, Winter, and Moore (1986)
had studied explosion development and the transition
from deflagration to detonation in coal dust/air suspen-
sions in a long duct (H/D=70, D=1.2 m).
Tests in vessels with changed H/D ratios and fixed
volumes are left unknown in the former researches and
* Corresponding author.
E-mail address: fsa@radandt.de (S. Radandt).
on dust explosion behaviors and phenomena on this
aspect, and to lay a basis for dust combustion modeling.
By using the combustion modeling, once the initial con-
ditions are given, the explosion behaviors can be pre-
dicted. This paper focuses on the maize starch combus-
tion in closed vessels with changed H/D ratios and fixed
volumes. That is, the volume is kept to be 1 m3 and H/D
ratios are changed from 1, 2, 4 and 6, respectively.
Generally, dust combustion models considering local
flow variables can be cataloged into two classes accord-
ing to multi-phase sub-models. One model named Eurl-
erianEulerian (EE) considers particle phases as some
special kind of fluid. Particles are grouped according to
the particle size and each group is considered as a kind
of fluid species. Another model is the EulerianLagrang-
ian (EL) model. In this model particles are treated as
non-continuous phases.
A lot of work had been done by former researchers
in the field of dust flame propagation and combustion
modeling. Clark and Smoot (1985) had modeled the coal
dust flame acceleration using three-zone model. Smirnov
and Nikitin (1996) had modeled dust explosion in a 1.25
m3 confined vessel. In their papers EL multi-phase
model was used, and volatilization and combustion of
char were taken into consideration. Korobeinikov, Mar-

0950-4230/01/$ - see front matter 2001 Published by Elsevier Science Ltd.

PII: S 0 9 5 0 - 4 2 3 0 ( 0 1 ) 0 0 0 4 4 - 4
496 S. Radandt et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 495502

Nomenclature
A interface area of heat convection, m2
As pre-exponential factor, unit is mechanism dependent
Bs temperature exponential factor
c dust concentration, kg/m3
Cp specific heat capacity, J/(kg K)
Cd drag force coefficient
cs Rosseland mean absorption coefficient
D diameter of the test vessel, m
Ds diffusion coefficient of specie s, m2/s
dp diameter of particle, m
E specific mixture energy (including chemical energy), J/kg
Ep energy flux from particulate phase, J/(m3s)


fd drag force, N


G gravity acceleration constant, m/s2
Gs mass flux of a single particle combustion, kg/s
H height of the vessel, m
h heat transfer coefficient, W/(m2K)


I unit tensor of the second order
Is mass flux of diffusion in gas phase, kg/(m2s)
Iq heat flux of thermal conduction in gas phase, J/(m2s)
Is heat flux of thermal radiation, J/(m2s)
k turbulent kinetic energy, J/kg
Mp momentum flux from particulate phase, kg/(m2s2)
mp mass of a particle, kg
Nu Nusselt number
P, p pressure, 105 Pa
Pr Prandt number, 0.7
Q rate of heat transfer, J/(m3s)
q specific heat, J/kg
Re Reynolds number
RMS Root Mean Square velocity, m/s
Sp net mass flux from particle phase, kg/(m3s)
T temperature, K
t time, s


u velocity vector of gas phase, m/s

up velocity vector of a particle, m/s
V volume, m3
W molecular weight, kg/mol
ws net reaction rate of species sth in gas phase, kg/(m3s)


x position vector of a particle, m
Ys mass fraction of sth species in gas phase

Greek

a volume fraction of gas phase

as mass fraction of specie s in Sp
r density of gas, kg/m3
 S. Radandt et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 495502 497

l thermal conductivity coefficient, J/(s m K)

t stress tensor of second order, N/m2
ss Schmit number, 1.0
sB Stefan Boltzmann Constant, 5.67108 J/(m2 k 4)
n kinetic viscosity, m2/s
e radiation emissivity; decay rate of turbulent kinetic energy, J/(kg s)

Superscripts

g gas
l laminar properties
t turbulent properties
w wall

kov, Klemer, Klemens, and Wolanski (1996) had used 2. Experimental configuration
an EE method to model dust explosions and detonations
A test method with a ring-pipe-nozzle to disperse dust
and radiative heat transfer was taken into consideration.
is recommended by ISO6184 guideline (1985). When the
A simple dust flame propagation model on turbulent
volume of the vessels are small and kept to be constant,
fluctuation velocity (RMS velocity) and laminar burning
however, and H/D ratios are changed bigger the shapes
velocity was integrated in a gas explosion code FLACS
of the vessels will be changed from cubic one to pipe.
(FLame ACceleration Simulation). Explosions in a 20
In this case, it is difficult to get homogeneous dustair
m3 chamber were simulated with this modified compu-
mixture by using only one ISO recommended ring
tational fluid dynamics (CFD) code by Van Wingerden
nozzle, and more ring-nozzles in a vessel will influence
(1996). Krause and Kasch (1998) had modeled the flame
on the pre-ignition turbulence. To reduce those short-
velocity and laminar burning velocities using code FLU-
comings, a new mushroom nozzle system was designed
ENT 4.
for these specific tests. Four mushroom nozzles are
Of these methods EE one is relative simple because
mounted on the surface of the wall of a vessel. The tur-
it only needs minor modification of the gas combustion
bulence induced by mushroom nozzles is reduced to a
code to incorporate the function to deal with dust com-
lower level, and more homogeneous dustair mixture
bustion. This method, however, is time consuming when
can be generated.
particle size groups increase, and there are unnecessary
Fig. 1(a) shows the experimental layout of ignition
diffusions for the particle phases.
and dispersion system in a H/D=1 vessel. Four small
EL method adopts an Eulerian approach for the gas
dust chambers connected with a air reservoir are linked
phase and a Lagrangian approach for the particle phase.
to the nozzles and the compressed air can be discharged
The trajectories, mass changes, motions and combustion
at the same time by electromagnetic valves. Ignition is
of particles in the calculation domain are traced in Lag-
triggered after a pre-defined time (ignition delay time)
rangian coordinates. Because of the large amount of the
of the dispersion. One setup of the test vessels (H/D=2)
particles, some particles (typically several thousands) are
is shown as Fig. 1(b).
represented as one computational particle (model
The pre-existing turbulence is a function of ignition
particle).
delay time. In the current experiment setup, when the
The EL model is adopted in present modeling. In
ignition delay time is 0.38 s, the RMS turbulent velocity
this model, dust combustion process is considered as a
is similar to that of 0.60 s in a standard 1 m3 vessel test
transient two-phase turbulent reacting flow. The flow and
chamber. The sample dust is cornstarch and some of the
combustion phenomena are governed by a set of conser-
characteristics are listed in Table 1.
vation equations. The following sub-models are employ-
ed:
3. Physical model
Multi-phase model; 3.1. Gas phase equations
Turbulence model;
(arYs)
Combustion model of gas phase and of particle phase; (arYs

u )IsasSpws (1)
Model of heat transfer to the vessel wall. t
498 S. Radandt et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 495502

Table 1
Characteristics of cornstarch

Medium diameter (m) 15

Moisture (weight) 3.5 (dried)
Minimum ignition energy (J) 0.005
Minimum ignition temperature (K) 673
Lower explosible limit (g/m3) ~40
Maximum pressure (105 Pa) 9.4
Dust explosion index Kst (105 Pa m/s) 217

( t u )Ep
where

nt
Isar DI Y
ss
(4)

2 2
t ar(vIvt)
u )T

u (


u I ark I (5)
3 3

Iq lI
arCpnt
Pr
T (6)

16 sB 3
Is T T. (7)
3 cs
The kinetic turbulent viscosity is got by:
ntcmk 2/e (8)
and k, e are determined by ke model.

3.2. Particle phase equations

d

up

mp f dmp

G (9)
dt

d
x

up (10)
dt
dTp
mpCp Qsurface reactionQconvectiveQrediation (11)
dt
where
Fig. 1. Experiment layout of H/D=1 and H/D=2 vessels, V=1 m3. (a)
1






H/D=1, (b) H/D=2. f d Cdarpd 2p( u u u p)| u u u p|. (12)
8

(ar u )


(ar u u )(ap) t Mp (2)
t 3.3. Combustion mechanism

(arE)
(ar

u E)(ap

u )(IqIs) (3) The combustion mechanism had been described in
t detail by Zhong, Deng, and Li (1998). The volatilization
 S. Radandt et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 495502 499

of particles, combustion in gas phase and heterogeneous

combustion of fix carbon were taken into consideration.

3.4. Heat loss to the wall of a vessel

It is assumed that before the fireball reach the wall of

a vessel, the radiation from the fireball to the wall can
be ignored. In the calculation domain, only the cells
adjacent to the wall are considered.
The heat transfer to the wall includes heat convection
and heat radiation. The convection transfer from a cell
to the wall can be expressed as:
qchA(TgTw) (13)
where the convection heat transfer coefficient is
obtained by:
Nul
h (14)
D
where Nu the Nusselt number, l the thermal conduc-
tivity coefficient and D the diameter of the vessel. Nu
can be obtained by Chen (1990):
Nu0.023 Re0.8 Pr0.33 (15)
where

|

u |D
Re (16)
nI
Pr is assumed to be 0.7. The radiation from a cell to the
wall is:
qrAesB(T 4gT 4w) (17)
where sB, Stefan Boltzmann Constant sB=5.67108 Fig. 2. Typical temperature fields in the H/D=1 vessel. (a) At 25 ms
W/(m2 k 4); e, the absorption factor of a wall, ranged after ignition, (b) at 65 ms after ignition.
from 0.37 to 0.9 according to the lubricity and bright
degree of the wall surface. The wall temperature is
unknown, but it can be obtained by solving the heat con- Table 2
Pmax and (dp/dt)max under dust concentration in the H/D=1 vessel
duction equation inside the vessel wall, or taken as a
constant for simplification. Dust Experiment Simulation Experiment Simulation
concentration
(g/m3)
4. Comparison of the experiment and modeling Pmax (105 Pa) (dp/dt)max (105 Pa/s)
results
250 8.2 8.3 145 185
Typical temperature profiles of the simulation are 500 10.7 11.0 242 250
750 10.6 10.7 248 261
shown as Fig. 2(a) and (b).
1000 10.5 9.5 210 193
1250 9.4 8.7 182 167
4.1. Explosion pressure and maximum rate of pressure
rise as the function of dust concentration

The maximum pressures and maximum rates of press- experiment and simulation increase at first with increas-
ure rise of different dust concentrations for H/D=1 are ing dust concentration (250500 g/m3), keep nearly con-
listed in Table 2. stant in the middle (500750 g/m3) and then fall down
Fig. 3 demonstrates that both Pmax and (dp/dt)max of (750 g/m3). From Table 2 and Fig. 3, it can be seen
500 S. Radandt et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 495502

Fig. 3. Pmax and (dp/dt)max vs. dust concentrations in the H/D=1 ves- Fig. 4. Pmax and (dp/dt)max in different vessels ranged by H/D ratios.
sel. (a) Pmax vs. concentration, (b) (dp/dt)max vs. concentration. (a) Pmax to different H/D ratios, (b) (dp/dt)max to different H/D ratios.

that the simulation results have a good agreement with for a H/D=6 one. Higher heat loss results in lower press-
experiment ones. ure.
The compressing effect, on the other hand, plays an
4.2. The effect of H/D ratio increasingly role with increasing H/D ratio. This effect
can be described as follows: In a long duct, compared
It can be found in Table 3 and Fig. 4 that simulation with a cubic enclosure with the same volume, there is
results and experiment ones have a good agreement narrow space for the expansion of hot combustion gases.
when H/D ratio is changed from 1 to 6. Pmax and The gases can only expand in two directions (sometimes
(dp/dt)max decrease with increasing H/D ratio. The rea- one direction), unlike those in the cubic vessel they can
son of this phenomenon is that the heat loss to the wall expand in every direction. The unburned mixtures in the
in vessels with smaller H/D is smaller, because the inner duct will be more pre-compressed than in the cubic ves-
surface areas of vessels are smaller. For instance, the sel. Pre-compressing will make higher temperature and
surface area is 3.69 m2 for a H/D=1 vessel, but 6.71 m2 velocity. Higher velocity will then generate higher turbu-
lence, and therefore the deflagration becomes stronger
Table 3 and stronger.
Pmax and (dp/dt)max in vessels ranged by H/D ratios When H/D8, both Pmax and (dp/dt)max increase with
increasing H/D ratio. That means when H/D8 the com-
H/D Experiment Simulation Experiment Simulation pression effect plays dominant role when compared the
role of heat loss (when H/D8). In a H/D=15 vessel,
Pmax (105 Pa) (dp/dt)max (105 Pa/s)
Pmax can achieve 17.5105 Pa and (dp/dt)max up to
1 10.73 11.0 248 261 1070105 Pa/s, which is a very sharp increase when
2 10.51 10.8 220 240 compared with the results of the H/D=8 vessel (Pmax
4 10.47 9.8 161 184 =9.2105 Pa and (dp/dt)max=156105 Pa/s). This result
6 9.89 9.4 120 158 is reasonable when referring to Gardners experiments
8 N/A 9.2 N/A 156
10 N/A 10.7 N/A 605 (1986). It had been found in his experiments that in a
15 N/A 17.5 N/A 1070 42 m long duct, Pmax could be (3080)105 Pa, and det-
onation had been observed. Pressuretime curves in dif-
 S. Radandt et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 495502 501

ferent positions (distance from the center) from the

H/D=1 vessel to H/D=15 one are shown from Figs. 58.
The pressuretime histories at different positions in
vessels of H/D=16 are nearly the same. That means a
uniform distribution of pressure in the cubic and cubic-
like vessels. On the contrary, in vessels of H/D=1015,
different positions have different pressuretime histories.
The pressure and rate of pressure rise at the end position,
which are far away from the ignition source, are much
higher than those at the center. After about 0.120.15 s,
the pressures at different positions in the vessel will
become the same.

5. Conclusions and discussion

Experiments in vessels ranged by H/D ratio show that

when H/D increases from 1 to 6, the maximum press-
ure Pmax decreases a little, and the maximum rate of
pressure rise decreases apparently;
Simulation and experiment results have a good agree-
ment in vessels when H/D is between 16;
Simulation shows that when the H/D ratio increases
from 8 to 15, Pmax and (dp/dt)max have different tend-
ency than those when H/D ratio increases from 1 to
6. Both Pmax and (dp/dt)max increase with increasing Fig. 6. Explosion pressuretime histories in H/D=4 and 6 vessels. (a)
H/D=4 vessels, (b) H/D=6 vessels.
H/D ratios when H/D are changed from 8 to 15.
According to the simulation H/D=8 is the changing

Fig. 5. Explosion pressuretime histories in H/D=1 and 2 vessels. (a) Fig. 7. Explosion pressuretime histories in H/D=8 and 10 vessels.
H/D=1 vessel, (b) H/D=2 vessels. (a) H/D=8 vessels, (b) H/D=10 vessels.
502 S. Radandt et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 495502
Fig. 8. Explosion pressuretime history in a H/D=15 vessel.
point of the decreasing tendency (H/D=18) and
increasing one (H/D=815);
The heat loss has the most important effect in H/D6
vessels. While the compressing effect may play a
dominant role in H/D8 vessels.
Acknowledgements
This work of modeling part was supported by the
National Natural Science Foundation of China (Project
No. 29376228) and the experiment part by Forschung-
sgesellschaft fur angewandte Systemsicherheit und
Arbeitsmedizin (Project No. F-05-9302). The authors
acknowledge Dr. F. Hauert for providing turbulence
measurement data.
References
Bartknecht, W. (1993). Explosionsschutz Grundlagen und Anwendung.
Berlin: Springer pp. 537570.
Bartknecht, W. (1998). Einflu des Hohen/DurchmesserVerhalt-
nisses Duckentlasteter Behalter auf den Flachenbedarf. Basel:
Ciba-Geigy Bericht D 15/88.
Clark, D. P., & Smoot, L. D. (1985). Model of accelerating coal dust
flames. Combustion and Flame, 62, 255269.
Chen, S. (1990). Heat transfer. Beijing: High Education Press of
China.
Gardner, B. R., Winter, R. J., & Moore, M. J. (1986). Explosion devel-
opment and deflagration to detonation transition in coal dust/air
suspensions. In: The 21st Symposium (International) on Combus-
tion (pp. 335343). The Combustion Institute.
ISO 6184/1 Explosion Protection SystemPart 1 (1985). Determination
of Explosion Indexes of Combustible Dusts in Air.
Korobeinikov, V. P., Markov, V. V., Klemer, J., Klemens, R., & Wol-
anski, P. (1996). The unsteady flows of dusty gases with chemical
reactions. In: Proceedings of International Symposium on Hazards,
Prevention and Mitigation of Industrial Explosions and 7th ICDE
(pp. 6.526.59). Bergen, Norway.
Krause, U., & Kasch, T. (1998). The influence of flow and turbulence
on flame propagation through dustair mixtures. In: International
Symposium on Hazards, Prevention and Mitigation of Industrial
Explosions and The 8th International Colloquium on Dust
Explosions, IL.
NFPA 68 Guide for Venting of Deflagrations, Nation Fire Protection
Association (1988).
Pu, Y. K. (1988). Fundamental characteristic of laminar and turbulent
flames in cornstarch dust/air mixtures. Ph.D. thesis, Department of
Mechanic Engineering, McGill University, Canada.
Radandt, S. (1983). Staubexplosionen in Silos. Symposium Heft 12 (pp.
1851), Berufsgenossenschaft Nahrungsmittel und Gaststatten.
Radandt, S., & Shi, J. Y. (1997). Dust explosion in different vessels
ranged by height to diameter ratios. FSA/BGN Internal Report.
Smirnov, N. N., & Nikitin, V. F. (1996). Dustair mixtures evolution
and combustion in confined and turbulent flows. In: Proceedings
of International Symposium on Hazards, Prevention and Mitigation
of Industrial Explosions and 7th ICDE (pp. 5.52669), Bergen,
Norway.
Vogl, A. (1996). Flame propagation in tubes of pneumatic conveying
systems and exhaust equipment. In: The 29th Loss Prevention Sym-
posium, New Orleans.
Van Wingerden, K. (1996). Simulation of dust explosion using a CFD
code. In: Proceedings of International Symposium on Hazards,
Prevention and Mitigation of Industrial Explosions and 7th ICDE
(pp. 6.426.51), Bergen, Norway.
Zhong, S. J., Deng, X. F., & Li, G. (1998). Modeling maize starch
explosions in the 12 m3 silo. In: Proceedings of the International
Symposium on Hazards, Prevention and Mitigation of Industrial
Explosions, Schaumburg, IL.
1
2
3
4
5 Journal of Electrostatics 51}52 (2001) 435}439
6
7
8
9
10 Investigation of electrostatics during sulphur
11
12 crushing operations
13
14 Sun Keping *, Zhao Hangyu , Liu Yong

15
The Research Section of Electrostatics, Shanghai Maritime University, Box 119, 1550 Pudong Dadao Road,
16 Shanghai 200135, China
17
Promax Technology Ltd., Xi+an 710068, Shanxi Province, China
18
19
20 Abstract
21
22 Electrostatic discharge caused "re and explosion accidents at a sulphur factory in Shanghai.
This work describes the test result of potential at the factory workshop, and presents several
23
protective measures in order to prevent dangerous discharge.  2001 Elsevier Science B.V. All
24
rights reserved.
25
26 Keywords: Electrostatic discharge; Dust explosion; Powder handling; Hazards
27
28
29
30
31 1. Introduction
32
33 High insulation powder in pneumatic pipe can be highly charged. The mixture of
34 powder}air in a collection bag or storage may be ignited by electrostatic discharge
35 (ESD). A "re or explosion may be caused. Accidents of this kind are often reported
36 [1}3]. But sulphur powder explosion accidents and protective measures are not
37 always mentioned.
38 Several times "re disaster and three times explosion accidents have occurred in the
39 last few years in the Shanghai Sulphur Factory.
40 In order to investigate the reason for the accidents, we measured the potential inside
41 the storage and at the edge of the collection bag, and found that the potential was very
42 high. We think that it may be ESD which caused the accidents.
43
44
45
46 * Corresponding author. Fax: #86-21-58795940.
47 E-mail address: sun-keping@greathorse.com.cn (S. Keping).

0304-3886/01/$ - see front matter  2001 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 4 - 3 8 8 6 ( 0 1 ) 0 0 0 3 7 - 7
 436 S. Keping et al. / Journal of Electrostatics 51}52 (2001) 435}439

1 We installed a new kind of eliminator [4] at the entrance of the storage, and
2 mounted a corona wire which is earthed at the centre of the collection bag. The static
3 energy can be safely released by the wire corona discharge. These measures have
4 a good e!ect for nearly one year. Fire and explosion accidents have not yet occurred.
5
6
7 2. Test on-site
8
9 2.1. Flow chart of the workshop
10
11 The #ow chart of the workshop is shown in Fig. 1. Lump or piece sulphur is fed into
12 the crusher (shown as 3 in Fig. 1) through the hopper (as 2 in Fig. 1) in order to get
13 sulphur powder. The powder will be selected by two classes winnow machine. The
14 bigger particles will "rst deposit into metal silo no. 1 (as 7 in Fig. 1) through winnower
15 no. 2 (as 4 in Fig. 1) and the smaller particles will deposit into metal silo no. 2 (as 8 in
16 Fig. 1) through winnower no. 2 (as 5 in Fig. 1). A few particles with very smaller
17 diameter will #ow out with the air#ow. They will be collected by collection bag (as 6 in
18 Fig. 1) in order to control pollution and recovery the sulphur powder.
19 The "re disaster and explosion accidents took place at the silo no. 1 (two times) and
20 the collection bag (6 times). The diameter of silo no. 1 is 4 m. The collection bag is
21 made of polymeric material. Its diameter is about 0.4 m, length is 1.2 m, respectively.
22
23
24
25
26
27
28
29
30
31
32 Fig. 1. Flow chart of the sulphur workshop. 1: blower; 2: hopper; 3: crusher; 4: winnower no. 1; 5: winnower
33 no. 2; 6: collection bag; 7: silo no.1; 8: silo no. 2.
34
35
36
37
38
39
40
41
42
43
44
45 Fig. 2. Test position of the collection bag. 1}3 are test points.
 S. Keping et al. / Journal of Electrostatics 51}52 (2001) 435}439 437

1 2.2. Test on site

2
3 First, we tested the potential at points 1}3 of the collection bag, respectively. The
4 2nd point is just at the deposit dust surface, as shown in Fig. 2. The 1st point is about
5 15 cm above the dust surface. The 3rd point is about 15 cm under the 2nd point. The
6 test result is shown in Table 1.
7 The test result indicates that the static potential is high. The accumulated static
8 energy is very easy to release. Any light strike or impact may cause "re or explosion.
9 Second, we tested the potential in silo no. 1. Test position is near the pressure-relief
10 valve of the silo. The test result is shown in Table 2.
11 At the same time, we also test the speci"c charge of the sulphur powder by Faraday
12 pail at the entrance of silo no. 1. The test result is shown in Table 3.
13 Table 3 indicates that the speci"c charge of the sulphur powder is very high. We
14 think that it may be the `Maurera discharge or spark discharge inside the silo that
15 causes the accidents [5,1].
16
17 Table 1
18 The potential at the edge of collection bag (temperature: 193C, humidity: 78%)
19 1st point (kV) 2nd point (kV) 3rd point (kV)
20
21 !10 !20 !30
22 !11 !20 !30
23 !10 !21 !30
!11 !20 !31
24 !10 !20 !30
25 Average value !10 !20 !30
26
27 Note: The thickness of deposit dust is about 0.4 m while testing.
28
29
30 Table 2
31 The powder level surface potential at the edge of silo no. 1 (temperature: 193C, humidity: 78%)
32 Sulphur powder level (m) Potential (kV)
33
34 1.0 25
35 1.2 30
36 1.6 50
1.8 60
37
38
39
Table 3
40 The speci"c charge of the powder (temperature: 193C, humidity: 78%)
41
42 Speci"c charge (
C/kg)
43
First time test 1.7
44
Second time test 1.8
45
 438 S. Keping et al. / Journal of Electrostatics 51}52 (2001) 435}439

1 3. Protective measure
2
3 3.1. Eliminator
4
5 We have designed a new kind of eliminator [4]. This eliminator is specially used in
6 the pneumatic pipe. It has 12}32 spray nozzles which can spray strong ion wind
7 around the pneumatic pipe in order to neutralize the sulphur particle charge. We
8 installed one of the eliminators at the entrance of silo no. 1. It can decrease the charge
9 density of the powder. The potential at the edge of silo no. 1 was reduced considerably
10 after using this kind eliminator (as shown in Table 4).
11 The speci"c charge of the powder decreased to 0.8
C/kg after using the eliminator.
12 By the way, the eliminator has a certi"cate of explosion-proof by authorities in
13 China [4]. Its basic circuits are safe and the whole eliminator is with proof-explosive
14 special measures.
15
16 3.2. Corona wire
17
18 Lindbauer presented that earthed conductive dividing wires can considerably
19 reduce the potential by corona current inside the tank or storage [6]. This considera-
20 tion led to our work to install an earthed wire at the centre of the collection bag. We
21 expect that corona discharge will occur around the earthed wire in a su$ciently high
22 static "eld. The static energy can be safely released.
23 We experimented with three wires (H0.5, H1, H1.5 mm, respectively) installed at
24 the centre of the bag. We found that H1 mm wire has the best e!ect. The potential at
25 the deposit powder surface inside the bag obviously decreased, as shown in Table 5.
26 We connected an amperemeter between the wire and earth. We found that there is
27 a steady corona current through the wire to earth of about 0.9}2.1
A.
28
29
30 Table 4
31 Contrast data of the potential at the edge of silo no. 1 (temperature: 193C, humidity: 78%)
32 Sulphur powder level (m) Potential (kV) potential (kV)
33 (not using the eliminator) (using the eliminator)
34
35 1.0 25 12
36 1.2 30 15
1.6 50 23
37 1.8 60 28
38
39
Table 5
40 Contrast data of the potential at the powder surface inside the bag
41
42 Installed the earthed wire (kV) Without the earthed wire (kV)
43
9.2 21
44
9.0 20
45
 S. Keping et al. / Journal of Electrostatics 51}52 (2001) 435}439 439

1 Presently, we are doing an experiment with 12 corona wires at a circle inside the
2 collection bag in order to reduce the particle speci"c charge e!ectively.
3 We will report the results when new data has been listed.
4
5
6 4. Conclusion
7
8 The test results on site indicated that the speci"c charge of the sulphur powder is
9 about 1.7}1.8
C/kg and the potential inside the silo is about 25}60 kV. It may be
10 Maurer discharge or spark discharge causing the accidents at the workshop. A new
11 kind of special pipe eliminator installed at the entrance of the sulphur powder storage
12 can considerably reduce the speci"c charge of the particle and the potential inside the
13 silo.
14 The corona wire at the centre of the collection bag can safely release the accumulat-
15 ive static energy and can obviously reduce the possibility of ESD hazards.
16
17
18 References
19
20 [1] T.B. Jones, J.L. King, Powder Handling and Electrostatics, LEWIS Publishers, 1991.
21 [2] B. Maurer, Discharge due to electrostatic charging of particles in large storage silos, Ger. Chem. Eng.
2 (1974) 189}195.
22 [3] M. Glor, Hazards due to electrostatic charging powder, J. Electrostat. 16 (1985) 175}181.
23 [4] Keping Sun, Investigation on static explosion-proof of aldehyde resin powder in pneumatic pipe,
24 Proceedings of the sixth International Colloquium on Dust Explosion, Shenyang, China, Northeastern
25 University Press, 1994, pp. 420}425.
26 [5] B. Maurer, M. Glor, G. Luttgens, L. Post, Institute of Physics Conference Series No. 85: Electrostatics
87, Bristol, 1987.
27 [6] R.L. Lindbauer, Reducing hazardous space potential in tankers: a theoretical and experimental
28 approach, J. Electrostat. 1 (1975) 274}294.
29
30
31
32
 Journal of Loss Prevention in the Process Industries 14 (2001) 111122
www.elsevier.com/locate/jlp

Ignition of dust clouds by brush discharges in oxygen enriched

atmospheres
ystein Larsen *, Janicke H. Hagen, Kees van Wingerden
Christian Michelsen Research, Postboks 6031 Postterminalen, 5892 Bergen, Norway

Abstract

Brush discharge ignitions of sulphur dust in oxygen-enriched atmospheres have been established. From a total of approximately
300 trials, brush discharges were able to start explosions of sulphur dust in three trials. In these trials the atmospheres contained
55 vol%, 60 vol% and 70 vol% oxygen, respectively. The work is motivated by the fact that no brush discharge ignition of dust
air mixtures has been observed in laboratory trials, despite an equivalent energy of the discharge above the minimum ignition
energy of some dusts. By adding oxygen to the atmosphere in which the brush discharge is generated, the dustair mixture will
be more prone to ignition. If the critical oxygen concentration to establish ignition of very sensitive dusts such as sulphur is
considerably higher than the oxygen concentration in air, it may be stated that a brush discharge cannot ignite dustair mixtures
at atmospheric conditions. High-speed video recordings show that all three ignitions occurred at the top edge of an evolving dust
cloud. The research is still ongoing. 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Brush discharge; Equivalent energy; Charge transfer; Dust cloud; Ignition

1. Introduction and objective
An increased use of non-conductive materials in the
process industry may lead to a higher probability of
charge accumulation. Common earthing techniques are
often not suitable and the frequency of electrostatic dis-
charges may increase. Brush discharges are among the
most usual forms of electrostatic discharges occurring
when a charged isolator comes into the neighbourhood
of a blunt earthed conducting object. Work has shown
that common hydrocarbons are easily ignited by brush
discharges (Gibson & Lloyd, 1965; Heidelberg, 1967).
An equivalent energy of brush discharges in gasair
atmospheres was found to be in the range 14 mJ (Glor,
1981). This energy should be sufficient to ignite dust
with low minimum ignition energy (MIE). However
such ignitions have never been reported in spite of com-
prehensive laboratory tests.
Potential reasons have been discussed. Electrical
forces may attract dust towards the electrode and the
non-conducting plate. A dust deposit on the electrode
divides the discharge into many weaker discharges
* Corresponding author. Tel.: +47-55-57-4040; fax: +47-55-57-
4041.
(Luttgens & Glor, 1989). A dust cloud in between the
non-conducting plate and the electrode may hamper the
charge transfer by reducing the electrical field. Some
theories focus on the differences in the spark discharge
characteristics between a brush discharge and a capaci-
tive spark between two electrodes. The character of these
two types of discharges is rather different in time and
space.
Various critical parameters for ignition of gas by
brush discharges have been investigated. These are: the
non-conducting plate surface humidity and charge trans-
fer in the brush discharge (Gibson & Lloyd, 1965), the
charge density on the non-conducting plate and the non-
conducting plate surface polarity (Lovstrand, 1981), the
size of the electrode (Heidelberg, 1967) and the non-
conducting plate surface potential and the time charac-
teristics of the brush discharge (Johnson, 1977/78;
Kramer & Asano, 1979).
It is well known from literature (Eckhoff, 1991), that
adding oxygen to an atmosphere where a dust cloud is
generated will make the dust more sensitive to ignition
compared to a dust cloud in an atmosphere of pure air.
By generating brush discharges in oxygen enriched
atmospheres where dust is whirled up, critical oxygen
concentrations just giving ignition of the dust can be
determined. If this critical oxygen concentration is far

0950-4230/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 0 0 ) 0 0 0 3 4 - 6
112 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122
beyond the oxygen concentration in air, it would be
reasonable to believe that brush discharges are not able
to ignite dust clouds in pure air.
2. Experimental apparatus and procedure
2.1. The brush discharge generator
A brush discharge generator, based on the apparatus
used by Heidelberg (1970) was developed. The appar-
atus is shown in Fig. 1 and consists of a gas tight cubic
chamber of 0.550.550.55 m. The left side wall and
the bottom as seen in the figure are made of steel and
PVC. The front wall is made of plexigass, allowing for
visual inspection during the experiments. All conductive
parts are shielded to avoid charge dissipating to ground.
The right wall consists of a 6 mm thick PVC plate. On
to this, non-conducting plates can be mounted on the
inside by plastic screws. A metal plate with more than
200 sharp-pointed corona needles is used to charge the
non-conducting plate. The metal plate was supplied with
high voltage from an adjustable high voltage generator
(030 kV). When this plate was placed near the non-
conductive plate, charges were transferred to the non-
conducting surface. Discharges between ground and
negatively charged non-conducting surfaces are tra-
ditionally named positive surface discharges. These are
found to be far more incendive than discharges between
ground and positive charged surfaces (Lovstrand, 1981).
The charging voltage was therefore 30 kV.
In order to obtain high charge density on the non-
conducting plate and avoid charge dissipation to the sur-
roundings before brush discharges were generated, an
earthed circular metal plate could be placed against the
non-conducting plate from the outside of the chamber
through a circular opening in the centre of the PVC wall.
Brush discharges were generated by moving a spherical
electrode pneumatically towards the non-conducting
Fig. 1. Brush discharge generator.
plate until a discharge occurred. For reasons of safety,
movement of the various parts was done pneumatically
yielding remote control.
Gasair mixtures introduced into the chamber are pre-
pared using a gas-mixing panel where the desired gas
concentration is easily mixed. To reduce any pressure
build up in case of an explosion, the backside and the
top of the rig work as explosion vents. A thin plastic
sheet covers these openings during tests where explosive
mixtures were attempted to ignite. From a dispersion
nozzle through a hole in the bottom of the rig nitrogen
can be introduced in case of fire. The brush discharge
generator is placed in a fireproof tent. Two extraction
ducts, one above and one below the generator, carry the
excess gas in the tent outdoors via a ventilation system.
2.2. Dust dispersion and control system
A circular dust dispersion cup with a mushroom
nozzle is mounted underneath the bottom plate of the
generator near the non-conducting plate. Through a cir-
cular hole dust can then be dispersed up into the area
where the brush discharges are generated. The air blast
used to disperse the dust is delivered from a 725 ml
pressurised container via a magnetic valve. The dust dis-
persion and the electrode movement are controlled from
a PC and a control unit. Various delay times between
the dust dispersion and start of the electrode can be
selected in order to generate brush discharges within the
dust cloud and not before or after the dust cloud is
present. Various perforated plastic discs can be mounted
on top of the dispersion opening in order to avoid
agglomeration of the dust sample.
2.3. Charge measurement circuit
An RC circuit in series with the electrode measures
the charge transfer during the brush discharge. The volt-
age build-up on a 10 nF capacitor is measured with a
Tektronix TDS 360 digital oscilloscope. This voltage
build-up was discharged to ground through a resistance
of 15 M. This gives a time constant: t=RC=150 ms
which is well beyond the duration of a brush discharge
(t106). The measuring circuit is shown in Fig. 2.
The measured voltage V, is easily converted to charge
q, by the following relation:
qCV
where C is the capacitance of the capacitor.
Through the front window a high-speed video camera
with a maximum recording rate of 12 000 frames per
second monitors the experiments. An electrical signal
generated by the brush discharge triggers the camera.
The exact electrode position during charge transfer can
be determined from the video recordings.
 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122
Fig. 2. RC-circuit used for measuring the charge transfer during
brush discharges.
2.4. Test procedure
The following procedure is applied during the tests.
First, all remaining charge on the non-conducting plate
is discharged to earth. The plate is cleaned with isopro-
panol and distilled water. A weighed dust sample is then
evenly distributed into the dispersion cup. Next the dis-
persion container is pressurised using a gas mixture with
the same composition as present in the rest of the rig.
The vent openings are covered with plastic sheets and
the generator is flushed with the desired gas mixture until
the gas composition at the inlet and the outlet is equal.
The gas flow is then shut off, and the gas mixture is
allowed to stabilise. The charging plate is put in the
charging position, about 10 mm from the non-conducting
plate, and the metal plate is placed against the back of
the non-conducting plate. The high voltage is turned on
and the non-conducting plate is charged during 30
seconds. When the charging is completed the charging
plate is moved to its rest position on the other side of
the chamber and the high voltage is turned off. The metal
back plate is then removed and the electrode is moved
towards the non-conducting plate when dust is dispersed.
Ignition was detected by sight and hearing.
By using dry air and placing a heater lamp close to
the back of the non-conducting plate, all the experiments
were performed in atmospheres where the humidity is
below 30%.
The time intervals between the various actions are
held constant. Depending on which kind of tests are per-
formed some of the actions mentioned above are
excluded.
113
3. Results and discussion
3.1. Optimising the apparatus and the test procedure
Many of the parameters influencing the incendivity of
a brush discharge are related to the apparatus and the
experimental procedure. In order to obtain high charge
transfer a systematic study of the effect of various
experimental parameters on the charge transfer in the
brush discharge was performed. In all tests every para-
meter was held constant except for the one being investi-
gated.
3.1.1. Diameter non-conducting plate
The charge transfer from three non-conducting plates
made of PVC with diameters of 200 mm, 300 mm and
400 mm is presented in Fig. 3. The average charge trans-
fer in the brush discharges from the three non-con-
ducting plates was found to be 71.4 nC, 436.8 nC and
422.8 nC respectively.
3.1.2. Type of non-conductive material
The charge transfer from two types of non-conducting
material (polycarbonate and PVC) was compared. The
results presented in Fig. 4 show that the mean charge
transfer from the polycarbonate and the PVC plate are
370.4 and 422.8 nC respectively.
3.1.3. Non-conducting plate thickness
Three different plates of PVC with a thickness of 5
mm, 8 mm and 11 mm respectively and a diameter of
300 mm, were tested. The charge transfer from the vari-
ous plates is presented in Fig. 5. The highest amount of
charge transfer was obtained with 8 mm non-conducting
plate where the average charge transfer was 436.2 nC.
Choosing a thinner or a thicker non-conducting plate
gave a significant decrease in charge transfer. Relatively
the charge transfer for the 8 mm non-conducting plate
is 56% and 20% higher than the 5 mm and the 11 mm
insulators respectively.
3.1.4. Effect of diameter and speed of the grounded
electrode
To see the effect of different electrode diameters and
electrode speeds on the charge transfer in the brush dis-
charge, three electrodes with different diameters were
tested. Each electrode was tested considering five differ-
ent velocities towards the insulator. When the electrode
approached the non-conducting plate and the electrode
speed was relatively slow, two and three discharges
could be observed in some of the tests. This phenomenon
was observed for all the velocities with the 20 mm elec-
trode diameter but only for the lowest electrode velocity
with the 40 mm electrode diameter. In the tests with the
60 mm electrode diameter all charge was released in one
single brush discharge. During the tests with the 20 mm
114 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122

Fig. 3. The effect of various non-conducting plates of on the charge transfer in brush discharges. (8 mm PVC, electrode diameter=60 mm and
electrode=0.8 m/s. Black dots represent an average of 10 trials).

Fig. 4. The effect of various types of non-conducting plates on charge transfer in brush discharges. (Non-conducting plate thickness=8 mm,
diameter=400 mm, electrode diameter=60 mm and electrode speed=0.8 m/s. Black dots represent an average of 10 trials).

electrode several sparks could be observed regardless of
the electrode speed. In almost all tests where several
sparks were initiated, the first brush discharge was the
discharge with the highest charge transfer. An oscillo-
scope signal showing three successive brush discharges
in one test is seen in Fig. 6.
In Fig. 7 the average charge transfer in these tests is
plotted. A dashed horizontal line divides the plot area.
Above this line all the charge was released in one single
brush discharge. Below the dashed line the charge trans-
fer from the non-conducting plate was divided into sev-
eral discharges. The average value of the charge transfer
is calculated from the largest of the discharges. The
charge average is therefore rather low for the 20 mm
electrode for all velocities and the 40 mm electrode for
the lowest velocity.
The results in Fig. 7 show a marked increase in the
charge transfer when the speed is increased from 0.15
m/s to 0.6 m/s for all three diameters. For the 40 mm
which shows the most distinct tendency, this can be
explained by the division of the charge transfer into more
brush discharges and the fact that the average charge is
calculated only from the largest brush discharge. The
same explanation can also be given for the fact that the
20 mm electrode curve is well below the 40 mm and the
60 mm electrode curves. If the electrode speed is further
increased the curves flatten for the 20 mm and the 40
mm electrode around 0.8 m/s. For the largest 60 mm
electrode, further increase in the speed gave no increase
but rather a slight decrease in the charge releases.
If the total charge transfer (the sum of all charge trans-
fer in one test) is plotted the situation is rather different.
 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122
Fig. 5. The effect of non-conducting plate thickness the on charge transfer in brush discharges. (Non-conducting plate diameter=300 mm, electrode
diameter=60 mm and electrode speed=0.8 m/s. Black dots represent an average of 10 trials).
Fig. 6. Example of oscilloscope measurement of three successive dis-
charges in one test. The corresponding charge transfer in these three
discharges is 320 nC, 38 nC and 144 nC. (8 mm polycarbonate, elec-
trode diameter=20 mm and electrode speed=1.6 m/s).
Total charge transfer is now significantly higher for
the smallest electrode compared to the electrode with the
largest diameter, see Fig. 8.
The charge transfer in the brush discharges as a func-
tion of brush discharge spark length is presented in Fig.
9. As can be seen from the figure only a few of the
sparks occurred closer than a distance of 56 cm from
the non-conducting plate. The figure indicates also that
short sparks initiate more charge transfer from the non-
conducting plate compared to longer sparks.
115
3.1.5. Humidity
Tests were performed in air atmospheres with differ-
ent levels of humidity.
The results presented in Fig. 10 show a significant
drop in the charge transfer from brush discharges when
the air humidity was increased.
3.1.6. Discussion and conclusion optimisation
apparatus and test procedure
When the non-conducting plate diameter is enlarged
above 200 mm the higher charge transfer can be under-
stood by the fact that more charge is available for the
brush discharge due to the larger surface area. However
as pointed out by Gibson and Lloyd (1965) an upper
limit can be reached for the surface where the flow of
charge from the non-conducting plate to the electrode is
limited by the surface resistance. This may explain why
there is no increase in the charge transfer when the diam-
eter of the non-conducting plate is increased from 300
mm to 400 mm.
The higher surface resistivity of polycarbonate com-
pared to that of PVC (1016 and 1013 respectively)
may explain the higher charge transfer for the PVC plate
(10% higher). This may be explained by the fact that
more charge can flow to the point of discharge due to
the lower resistivity of PVC (assuming a similar charge
density at the moment of discharge).
From the results in Fig. 5 it is evident that the highest
charge transfer in the brush discharges is obtained with
a non-conducting plate thickness of 8 mm. During the
charging a potential is generated across the non-con-
ducting plate. The electrons on the inner surface of the
non-conducting plate are restrained by charges of
opposite polarity at the outside generated by the
116 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122

Fig. 7. The effect of various electrode diameters and electrode speed on the charge transfer in brush discharges. (8 mm polycarbonate with 400
mm diameter). Each mark is an average of 10 trials. The average charge for points below the dashed line is calculated from the largest brush
discharge when more than one brush discharge occurred in the same test.

Fig. 8. The effect of various electrode diameters and electrode speed on the total charge transfer in brush discharges. (8 m polycarbonate with
400 mm diameter). Each mark is an average of 10 trials. If more discharges occurred in a test the total charge transfer was taken into account.

grounded metal plate. If the non-conducting plate is too
thin the electrical forces may hold on to more of the
electrons impeding them for being released to the elec-
trode during the brush discharge. However, if the insu-
lator is too thick, the electrical forces set up by the
grounded metal plate is lower due to higher volume res-
istivity. This results in a lower charge density on the
inner surface of the non-conducting plate and a lower
charge transfer to the brush discharges. A certain optimal
non-conducting plate thickness therefore exists.
Varying the speed and the size of the electrode shows
that a certain electrode size and speed is necessary to
collect all the charge transfer into one single brush dis-
charge. When the electrode diameter is small and the
electrode speed low, several weaker discharges may
occur instead of one powerful brush discharge, see Fig.
7. To increase the incendivity of the brush discharge it
is preferable that all the available energy is discharged
in one single breakdown. It should therefore be favour-
able to use relatively large electrodes and a fast move-
ment of the electrode towards the non-conducting plate.
When dust is introduced into the chamber high electrode
velocity will also make the timing between the start of
electrode and the dispersion of the dust easier. This is
because of the limited period of time the dust is present
in the area where the brush discharge occurs. The disad-
vantage may be generation of turbulence in the dust
cloud by the electrode, which may hamper ignition. The
minimum ignition energy of dust may increase by sev-
eral orders of magnitude if initial turbulence is present.
From Fig. 9 it can be seen that for the 60 mm elec-
trode the distance from the non-conducting plate where
the brush discharge most often occurred was 56 cm.
This electrode seemed to generate rather reproducible
 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122 117

Fig. 9. Spark length and charge transfer in brush discharges generated in pure air. (8 mm PVC plate with 400 mm diameter, electrode diameter=60
mm and electrode speed=1.2 m/s. Black dots represent the average).

Fig. 10. The effect of air on humidity on charge release in the brush discharges. Each mark is an average of 10 trials. (8 mm PVC, with 400
mm diameter, electrode diameter=40 mm and electrode speed=1.0 m/s).

discharges due to the spark length. Knowing this dis-
tance will make the timing of dust dispersion easier.
Due to the results in Fig. 10 the humidity in the
chamber should be kept as low as possible. High air
humidity leads to condensation at the surface, which
reduces the surface resistance.
In addition to the parameters presented above an opti-
mal charging period was found to be approximately 30
seconds. It was also seen that it was important to gener-
ate the brush discharge within a relative short time inter-
val after charging was completed. The conclusions of
the optimisation of the procedures and apparatus are
summarised in Table 1.
3.2. Determination of the equivalent energy in brush
discharges
Brush discharges were generated in various mixtures
of propane, air and nitrogen. The propane content of the
mixture was constant at 4.2 vol% while the nitrogen con-
tent in the air was increased in steps of 0.2 vol% from
24 vol% up to 32 vol%. The results presented in Fig. 11
show the frequency of ignition for various mixtures of
propane, air and nitrogen compared with results obtained
by Glor (1981) and Schwenzfeuer and Glor (1993). The
tendency is clear. For both of the tested electrode speeds,
0.4 m/s and 1.5 m/s, the frequency of ignition drops sys-
tematically as the content of nitrogen content in the
atmosphere increases above 22 vol% and 24 vol%
respectively. The highest nitrogen content giving
ignition is 30 vol% for both electrode velocities.
The result of the present study are in between those
of Glor (1981) and Schwenzfeuer and Glor (1993) indi-
cating a maximum equivalent energy of 3.6 mJ in the
brush discharges (see Fig. 12). A slightly higher prob-
ability of ignition can be seen for the highest electrode
velocity. Flow velocity and turbulence intensity is
known to increase the minimum ignition energy of a
118 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122

Table 1
Favourable experimental parameters for high charge transfer in the brush discharge

Parameter Result

Non-conducting plate diameter 300 mm

Non-conducting plate thickness 8 mm
Electrode diameter 40 mm
Electrode velocity 0.6 mm
Charging period 30 s
Charging distance 10 mm
Humidity 40%

Fig. 11. The brush discharges ignition frequency of 4.2 vol.% propane in air with various amounts of added nitrogen. Results from tests with
two different electrode speeds are compared with results reported by Glor (1981) and Schwenzfeuer and Glor (1993). (8 mm PVC with 400 mm
diameter, electrode diameter=40 mm).

Fig. 12. Minimum ignition energy of 4.2 vol% propane in air as a function of added nitrogen (Glor, 1981).

fuel. The opposite effect would therefore be expected as m/s of the gas mixture prior to the ignition, such flow
a fast moving electrode may initiate more disturbances effects are only minor and can be neglected, see Ballal
in the gas atmosphere compared to a slow moving one. and Lefebvre (1975). Fig. 13 shows the ignition of a
However, if we assume a maximum flow speed of 1.5 mixture of 8.5 vol% methane in air by a brush discharge.
 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122
Fig. 13. Ignition of a mixture of 8.5 vol% methane in air by a brush
discharge. (8 mm polycarbonate with 400 mm diameter, electrode
diameter=60 mm and electrode speed=0.8 m/s).
3.3. The influence of presence of dust on the brush
discharge
Tests were executed in order to see the effect of the
presence of dust on the charge transfer and spark length
of a brush discharge. When dust is introduced into a
region with a strong electrical field, electrical forces may
act on the dust, giving rise to dust deposits both on the
spherical electrode and on the surface of the non-con-
ducting plate. How such dust layers affect the charge
transfer in brush discharges was therefore investigated.
The procedure was to blow dust by air towards the elec-
trode or the non-conducting plate prior to the brush dis-
charge generation. Brush discharges were also produced
and characterised within a dust cloud whirled up
between the electrode and the non-conducting plate. The
average charge transfer and discharge length from a ser-
ies of ten trials for each set-up is presented in Fig. 14
and summarised in Table 2.
The relative decrease of charge transfer when dust is
introduced as a dust layer on the electrode, a layer at
the non-conducting plate or as a dust cloud is nearly
30% in all cases.
Fig. 14. The influence of dust on the spark length and charge transfer in a brush discharge. (8 mm PVC with 400 mm diameter, electrode
diameter=60 mm and electrode speed=1.2 m/s).
119
3.4. Brush discharge appearances
Pictures of brush discharges were taken by a Nikon
F1 camera with a Kodak TMZ P3200 film. The brush
discharges were generated in complete darkness with the
camera aperture fully opened. From the pictures the
rather varying behaviour of a brush discharge was evi-
dent. Most of the discharges appeared as a single stem
near the electrode, which branched into weaker dis-
charge channels towards the non-conducting plate. But
also several intense discharge channels between the elec-
trode and the non-conducting plate could be seen, see
Fig. 15.
Fig. 16 shows two photographs of brush discharges,
one generated with and one without dust deposits on the
electrode. Due to the darkness the length of the spark
could not be determined but the proportion between the
discharges is correct. To the left a typical brush dis-
charge is shown. A high-energy plasma stem branches
into weaker discharges near the non-conducting plate.
The charge transfer in this discharge is 442 nC. As can
be seen from the right picture a weaker and longer brush
discharge is generated when dust is present on the elec-
trode. The charge transfer when dust is present is 218
nC which is a decrease in charge transfer of more than
50% compared to the clean electrode discharge.
3.5. Ignitions of sulphur in an oxygen enriched
atmosphere.
Several tests were performed in oxygen enriched
atmospheres where the amount of dispersed dust and the
dispersion pressure were varied from 1.0 g to 5.9 g and
from 1.25 barg to 7.0 barg, respectively. From a number
of approximately 300 tests with sulphur, three ignitions
120 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122
Table 2
Spark length and charge transfer in brush discharges generated in the presence of dust
Amount of dust
No dust
Dust layer on the electrode 44.2 g/m2
Dust layer on the non-conducting 4.0 g/m2
plate
Dust cloud g/l
Fig. 15. Pictures of brush discharge with several discharge channels.
(Charge transfer=506 nC. Electrode diameter=60 mm).
have been obtained. The oxygen concentration, amount
of dispersed dust, dispersion pressure, spark length and
charge transfer in these tests are summarised in Table 3.
A high-speed video sequence of one of the ignitions
is shown in Fig. 17. The sequence shows the sulphur
dust cloud as it moves up between the entering electrode
to the left and the non-conducting plate to the right. The
flow speed of the dust is 78 m/s. A numbered scale
shows the distance (in cm) between the electrode and
the non-conducting plate.
3.6. Discussion ignition tests
The tests in Fig. 14 clearly demonstrate that the pres-
ence of dust both as a cloud or as a dust layer on the
non-conducting plate or electrode affects the charge
transfer. A similar conclusion can be drawn from Fig.
16. A decrease of the charge transfer of 30% in brush
discharges when dust is present is significant and reduces
the incendivity of a brush discharge. It is however, not
the only explanation for not being able to ignite dust air
mixtures by brush discharges (if this is the case).
From Fig. 17 it can be seen that the edge of the dust
cloud had just reached the height of the electrode when
Average spark length (cm) Average charge transfer % of max charge transfer
(nC) without dust
5 449
14.1 304 67.7
5 330 73.5
6.2 308 68.6
Fig. 16. Pictures of brush discharges showing the difference in brush
discharge appearance when dust is presented on the electrode (right,
218 nC) compared to a brush discharge generated with a clean elec-
trode (left, 442 nC). (Electrode diameter=60 mm).
the ignition occurs. This was also the case for the two
other tests where ignition was obtained. As only the top
of the dust cloud has reached the brush discharge region,
it may be assumed that the influence by the dust on the
brush discharges in these tests is marginal. If so, the
incendivity of a brush discharge may be approximately
equal to brush discharges generated in gaseous atmos-
pheres. This observation may give rise to a new way
of approaching the problem of dust ignition by brush
discharges. Dust present all over the area where the
brush discharge occurs may hamper the discharge and
eliminate the possibility of ignition. But, in situations
comparable to the one present here, the small inter-
ference by the dust may give rise to an increase of the
ignition probability. In cases where dust enters a region
where charge already has been generated such a scenario
may occur. Work has shown that dust particles may
reduce the electrical breakdown voltage of the atmos-
 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122
Table 3
Condition during tests with suIphur resulting in ignition by brush discharges
Test no. Vol% oxygen in air Dust weight (g)
1 70 5.83
2 60 5.83
3 55 5.86
Fig. 17. High-speed video recordings of ignition of a dust cloud of
sulphur in oxygen enriched air (oxygen content 55 vol%). The spark
occurs 9 cm from the non-conducting plate surface.
phere significantly (Eckhoff, 1970). Results by Al-
Arainy (1997) show that the breakdown voltage of pol-
luted rod gaps is mainly controlled by the dust film coat-
ing on the electrodes and not the airborne particles. A
brush discharge emanates from the grounded electrode
when the surface density of the induced charge reaches
a certain level. This may explain the longer spark lengths
obtained with dust deposits on the electrode. It can there-
fore be assumed that the dust itself actuates the discharge
between the grounded electrode and the non-conducting
plate. Tests have shown that with the experimental set-
up to obtain ignition in gasair mixtures, brush dis-
charges normally occur at a distance 5 cm from the non-
conducting plate. The fact that the brush discharge
occurs at a longer distance from the non-conducting
plate in the dust ignition tests supports this theory. If so,
the distance between the electrode and the non-con-
ducting plate may not be the only decisive parameter
for an electrical breakdown to occur. The probability of
similar occurrences may thus increase if dust enters a
121
Dispersion pressure Spark length (cm) Charge transfer (nC)
(barg)
6.5 10 124
6.5 7 268
6.5 9 278
region where optimal conditions (high potential differ-
ence and optimal electrode shape) are present.
4. Conclusion
An apparatus generating brush discharges has been
developed. Tests have been performed to optimise the
apparatus. Parameters influencing the charge transfer in
a brush discharge are: non-conducting plate diameter,
non-conducting plate thickness, electrode diameter and
electrode velocity towards the non-conducting plate. The
effect of charging time, charge dissipation to the sur-
roundings and humidity on the charge transfer were
investigated and an optimal test procedure could be
established.
It has been demonstrated that the presence of dust
influences the charge transfer in a brush discharge. A
dust layer on the charged non-conducting plate or at the
earthed electrode reduces the charge transfer by approxi-
mately 30 percent. A similar decrease in charge transfer
was observed when brush discharges were generated
within a dust cloud of sulphur.
Photographs show that the appearances of brush dis-
charges are more diffuse and less intense when dust is
introduced as a layer on the electrode.
Brush discharges were produced in dust clouds of sul-
phur dispersed in atmospheres enriched with various
amounts of additional oxygen. At the moment of writing
this paper, three ignitions have been obtained. The oxy-
gen concentrations in the air in these cases were 70
vol%, 60 vol% and 55 vol% and the charge transfer in
the brush discharges 124 nC, 268 nC and 278 nC
respectively.
More work needs to be done to verify whether ignition
of dust by brush discharges in pure air is possible or
is unlikely.
5. Further work
Further work will be carried out with different dust
concentrations and lower oxygen concentrations in order
to find the critical oxygen concentration just giving
ignition of the dust by the brush discharge. Other dust
types will also be tested. How the presence of dust
122 . Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122
affects the brush discharge duration and appearances will
be investigated.
Acknowledgements
The authors acknowledge the support given by the fol-
lowing companies: Alpharma A/S, Borregaard Industries
Limited, Brandforsk, DSM Research, Elkem ASA, FK
Group, Schweizerisches Institut zur Forderung der Sich-
erheit, Norgesmllene DA, Norsk Hydro ASA, Forestia
AS, Tine Norwegian Dairies BA, Pharmacia & Upjohn,
SCA Hygiene Products AB, RhonePoulence Industrial-
isation, Forsikring & Pension. They also want to thank
Dr. Martin Glor and Dr. Klaus Schwenzfeuer
(Schweizerisches Institut zur Forderung der Sicherheit)
and Prof. Rolf K. Eckhoff for several fruitful dis-
cussions.
References
Al-Arainy, A. A. (1997). Influence of dust particles on DC discharges
in symmetrical air gaps. European Transactions on Electrical
Power, 7, 275.
Ballal, D. R., & Lefebvre, A. H. (1975). The influence of flow para-
meters on minimum ignition energy and quenching distance. In:
15th Symposium (Int) on Combustion, The Combustion Institute.
Eckhoff, R. K. (1970). The energy required for the initiation of
explosions in dust clouds by electrical sparks. Thesis for the degree
of Master of Philosophy. The University of London.
Eckhoff, R. K. (1991). Dust explosions in the process industry.
England: Butterworth-Heinemann.
Gibson, N., & Lloyd, F. C. (1965). Incendivity of discharges from
electrostatically charged plastics. British Journal of Applied Phys-
ics, 16, 1619.
Glor, M. (1981). Ignition of gas/air mixtures by discharges between
electrostatically charged plastic surfaces and metallic electrodes.
Journal of Electrostatics, 10, 327.
Heidelberg, H. (1967). Generation of igniting brush discharges by lay-
ers on earthed conductors. In: Static Electrification Conference,
1967, p. 147.
Heidelberg, H. (1970). Entladnungen an electrostatisch aufgeladenen
nichtleitfahigen Metallbeschichtungen. PTB-Mitteilungen, 6, 440.
Johnson, J. K. (1977/78) The ignition of vapour and droplets by liquid-
to-metal sparks. Journal of Electrostatics, 4, 53.
Kramer, H., & Asano, K. (1979). Incendivity of sparks from surfaces
of electrostatically charged liquids. Journal of Electrostatics, 6,
361.
Luttgens, G., & Glor, M. (1989). Understanding and controlling static
electricity. Germany: Expert Verlag.
Lovstrand, K. L. (1981). The ignition power of brush discharges
experimental work on the critical charge density. Journal of Elec-
trostatics, 10, 161.
Schwenzfeuer, K., & Glor, M. (1993). Test to determine the ignition
frequency of dust by brush discharges. Journal of Electrostatics,
30, 115.
 Journal of Loss Prevention in the Process Industries 14 (2001) 139151
www.elsevier.com/locate/jlp

Safety strategy against potential hazards due to the handling of

powders in a blending unit
*
Norbert Jaeger
Safety Testing Laboratory, Ciba Specialty Chemicals, Additives Division, McIntosh, AL, USA

Abstract

To ensure and maintain process safety in the chemical industry a systematic hazard search and evaluation, i.e. Risk Analysis
is indispensable. The knowledge of the ignition behavior of dust-air is important for such a risk assessment in a chemical production
plant. This paper describes, based on a practical example, the strategy of performing a Risk Analysis and the use of the minimum
ignition energy and minimum auto ignition temperature as very important safety indexes in practice. 2001 Elsevier Science Ltd.
All rights reserved.

Keywords: Minimum ignition energy; Dust explosion hazards; Hazard evaluations; Electrostatic discharges

1. Introduction a granulation unit. The blend components are manufac-

The knowledge of the ignition behavior of dustair
mixtures is important for risk assessments in chemical
production plants.
This paper describes the systematic approach in
searching for dust explosion hazards, the appropriate use
of technical measures to eliminate the previous identified
hazards or their reduction of the severity to an accept-
able level.
Preventive measures against the occurrence of electro-
static discharges and mechanically generated sparks dur-
ing powder handling operations will be also described
by using practical examples from a blending unit for
powders.
The aim of this paper is to assist people dealing with
powders. It reflects the present state of the art and current
knowledge of the assessment and measures associated
with powder handling.
2. Process description
The blend unit is designed to produce two or three
component blends for sale as powder and as a feed for
* Current address: Home and Fabric Care, Ciba Specialty Chemi-
cals, Consumer Care Division, Grenzach, Germany. Tel.: +49-0-7624-
12-2235.
E-mail address: norbert1.jaeger@cibasc.com (N. Jaeger).
tured in separate process units, adjacent to the granu-
lation unit. All the components are conveyed with N2
through pneumatic conveying systems into storage silos.
Blend components are screw feed from surge bins into
weigh hoppers. From there, the material can be trans-
fened into a blender. Once the blend has been completed,
the batch is discharged to a surge hopper and a new
blend can be initiated. From the surge hopper the final
blend can be transfered into the packaging facility for
packaging in drums, FIBCs and bulk containers. Fig. 1
shows a simplified schematic drawing of the unit.
3. Risk analysis
Safety guidelines, regulations and technical instruc-
tions facilitate control of the more commonly occurring
safety problems. However, in practice, every chemical
process constitutes a particular combination of chemi-
cals, plant and process conditions, To ensure and main-
tain process safety in the chemical industry, therefore, a
systematic predictive hazard search and evaluation, i.e.
Risk Analysis, is indispensable. Its objective is to
obtain synthetic experience before the technical
implementation of processes or plant start-up so that
from the outset planned, appropriate measures can be
taken Ciba-Geigy, 1992.
In evaluating Risk, consideration must be given to

0950-4230/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 0 0 ) 0 0 0 3 6 - X
140 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151

Fig. 1. Blending unit.

both the Probability of occurrence of an incident and the

Severity of the consequences.

3.1. Systematic approach

To obtain useful results, a risk analysis must incorpor-

ate the stages shown in Fig. 2.
A Risk Analysis for the blending unit was performed
in accordance to the above described procedure, modi-
fied for the search of possible explosion hazards (see
Fig. 3)
The necessary information to carry out a Risk Analy-
sis in its full depth is not yet available in the fast devel-
opment stages of a process. However, to acquire the
desired level of integrated safety of the processes and
facility, it is important that safety considerations of the
process accompany the project from its earliest stages of
development. As the project increases through the stages
of development, these considerations will become more Fig. 2. Analysis steps Ciba-Geigy, 1992.
 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151
Fig. 3. Analysis steps (example).
detailed and will finally lead into the detailed risk analy-
sis described below.
3.2. The steps in risk analysis
Deviations from safe working conditions can lead to
hazards. A systematic search for hazards is therefore
only possible if safe working conditions and their limits
are defined and the consequences of deviations from
them are known. In the forefront, therefore, are the
chemicals involved in the process, the intended reactions
and the criteria, which must be maintained for safe pro-
cessing.
3.3. Compilation of basic data
The basic data give fundamental information on the
inherent hazards of the chemicals and of the process.
They form the basis of the whole risk analysis. They
should have been compiled primarily in the course of
process development from the knowledge available
worldwide from practice and literature. The type of
required safety data are based on various specific internal
and external safety standards (NFPA 1997a,b, 1998;
NFPA 1997c, 1993).
3.4. Systematic search for hazards
This is the most creative step of risk analysis. The
chemical process is considered in the context of plant,
personnel and operating practice. In addition to operating
instructions and plant instructions therefore, the plant
141
specific data and plant drawings must be available for
the analysis. This analysis can only be successful if the
drawings correspond to the actual situation. Furthermore,
a site inspection is essential, since critical details and
hazards such as, for example, cables carried over sharp
edges, leaks, scale formation, dismantled fittings and
also influence from neighboring plants, will not be found
on any drawing.
In the systematic search for hazards, all possibilities
are sought in the critical areas. For the specific powder
handling operations one must focus on:
chemicals
physical processes
plants
energy
personnel
external influences
which could lead to deviations from the safe process
conditions defined.
These deviations should be set out in the list of haz-
ards together with statements of cause and effect,
initially without evaluation or correlation. Various
methods are available for a systematic search for hazards
CCPS, 1995.
4. Dust explosion hazards in the blending unit
An explosion hazard can exist when dusts are pro-
duced, stored or processed in a plant and these materials
are present as a mixture in air. An explosible mixture
is present, when combustible dusts are present in such
quantities in air that an explosion occurs after an
ignition.
An explosion requires three conditions to exist simul-
taneously (see Fig. 4):
Fig. 4. Requirements for the occurrence of explosions ISSA, 1996.
142 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151
fuel or flammable material in sufficient quantity and
effectively mixed with
air and
an effective ignition source
Using preventive measures against explosions require
the reliable exclusion of one of the conditions necessary
to generate an explosion as shown in Fig. 5.
An explosion can thus be excluded with certainty
by either
avoiding the development of explosible mixtures
(combustible dusts, flammable gases), or
replacing the atmospheric oxygen with an inert gas,
working in a vacuum or using inert dust, or
preventing the occurrence of effective ignition sourc-
es.
All three measures are summarized under the heading
preventive explosion protection. It should be ensured
that at least one of the three conditions is eliminated or
so strongly reduced, so that an explosion is no longer
possible or at least very rare. Their appropriate elimin-
ation prevents an explosion from starting.
To determine the necessary explosion protection
safety measures, the knowledge of the safety data of the
materials being used in the production unit and the effec-
tiveness of the identified ignition sources must be
known.
4.1. Basic safety data information
In order to sufficiently survey the danger potential of
a dust, one is normally forced to carry out a multitude
of laboratory tests that are based on standardized
methods. The test results are described in the form of
parameters known as safety characteristics. Thorough
knowledge of the ignition behavior of dustair mixtures
towards mechanically generated sparks or electrostatic
discharges (Minimum Ignition Energy, MIE) and hot
surfaces (Minimum Autoignition Temperature, MAIT) is
extremely important for assessing the hazards in dust-
carrying plants. The ignition behavior determines essen-
Fig. 5. Hazard triangle.
Fig. 6. Combustibility test at room temperature.
tially the extent, and hence the cost of the protective
measures, to be used. This is especially true for the use
of the protective measure prevention of ignition sour-
ces and also for the understanding of the ignition
phenomena regarding static electricity, e.g. brush dis-
charges, bulk surface discharges, spark discharges, pro-
pagating brush discharges. Knowledge of a powders
minimum ignition energy, particle size distribution and
specific resistivity allows one to define the necessary
protective measures for different operations based on the
vessel size and the conductivity of the material of con-
struction Luettgens & Glor, 1989.
4.1.1. Burning class
The test is designed to determine the ignitibility and
the combustibility of a dried product according to the
following criteria:
Can the product be ignited at all?
Is the ignition propagated?
Is there a fire or a flameless reaction?
The tests are carried out in a ventilated laboratory hood.
The air velocity at the test location is approx. 0.2 m/s.
The superimposed light air stream is necessary to vent
the inert gas which are generated upon ignition of the
product and which may inhibit the combustion behavior
of the dust (see Fig. 6).
If the combustibility of a product is of interest at elev-
ated temperatures, the combustibility test can be done
at elevated temperatures (e.g. at the anticipated drying
temperature, see Fig. 7).
Fig. 7. Test at increased temperature in a glass tube.
 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151
Table 1
Definition burning classes
Test result
No ignition
Brief ignition, rapid extinction
Localized combustion or glowing with practically no spreading
Glowing without sparks (smoldering) or slow de-composition without flames
Burning fireworks or slow quiet burning with flames
Very rapid combustion with flame propagation or rapid decomposition without flame
The combustibility of the product is rated in accord-
ance with the course of reaction and characterized as a
class number (see Table 1).
4.1.2. Dust explosion characteristics
The Twenty-Liter Apparatus (see Fig. 8) is used in
the determination of explosion indices of combustible
dusts. Dust explosivity, lower explosion limit (LEL),
maximum explosion overpressure (Pmax), maximum
explosion constant (Kmax) and the limiting oxygen con-
centration (LOC) can be measured with this device. The
maximum explosion pressure and the maximum
explosion constant are necessary to design explosion
venting or explosion suppression system.
4.1.3. Minimum autoignition temperature
The Minimum Autoignition temperature, MAIT, is
defined as the lowest temperature of a heated surface at
which the most readily ignitable mixture of a dust with
air just ignites. It provides information on the ignition
behavior of a dust suspension when quickly passing over
a hot surface (see Fig. 9).
Fig. 8. Twenty-liter-test apparatus.
143
Class Reference
No spreading of fire 1 Table salt
2 Tartaric acid
3 D+Lactose
Fire spreads 4 H-acid
5 Sulfur
6 Black powder
Fig. 9. BAM furnace.
4.1.4. Minimum ignition energy (MIE)
The minimum ignition energy, MIE, of a combustible
substance is the lowest value of the electrical energy
stored in a capacitor, which on discharge just suffices
to ignite the most readily ignitable fuelair mixture at
atmospheric pressure and room temperature. To help
assure a standardized test procedure, a test apparatus
known as MIKE 3 (Fig. 10) of the third generation has
been specially developed by Kuhner AG, Switzerland
and has been made commercial available. Other test
apparatus are also available to determine the MIE in
accordance to an American Society for Testing Materials
(ASTM) standard on MIE of dusts ASTM, 1999.
The MIE is usually quoted as a range: The lower value
represents the highest energy at which no ignition is
found in at least 10 experiments. The higher value, on
the other hand, is the lowest energy at which the dust
air mixture is just ignited.
The method for the determination of MIE is described
in the International Standard of the International Electro-
technical Commission (IEC) and in an ASTM standard.
The MIE is generally obtained with an inductance in the
discharge circuit. However, in order to assess the incen-
divity of electrostatic discharges in industrial operations
towards dustair mixtures, the MIE must also be determ-
ined without an inductance in the discharge circuit. With
flammable gases and easily ignitable dusts, the influence
of the inductance is generally not detectable.
144 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151

related to a dust explosion hazard of the handled material

in the blending unit.

4.2. Ignition sources

The use of Avoidance of Ignition Sources as a pro-

Fig. 10. MIKE 3 apparatus for determination of the minimum
ignition energy MIE of dusts.
4.1.5. Powder volume resistivity
To characterize the static dissipative properties of a
material, its powder volume resistivity, rR, has to be
determined. It must be kept in mind that resistivity is
not an absolute property of a powder and depends very
strongly on moisture content and on the method used for
measurement. From an electrostatic point of view, dusts
are considered to be conductors, e.g. incapable of storing
charge, if the resistivity is less than or equal 109 m.
Table 2 summarizes the relevant generated safety data
Table 2
Summary of dust explosion data
Product Product
A B
tective measure requires a comprehensive hazard evalu-
ation as part of a detailed risk analysis, to determine all
possible ignition sources that may occur during pro-
duction. There are a large number of different ignition
sources that one must consider in industrial operations
(see Fig. 11). Not every ignition source has sufficient
energy to ignite all types of explosible atmospheres.
Therefore, it is necessary to investigate the ignition
sources in detail in order to determine the ignition hazard
in conjunction with the expected explosible mixtures.
Trivial ignition sources (welding, smoking etc.) must
be excluded by using organizational measures. Ignition
sources, which could result from the process itself intro-
ducing energy into the product being handled (e.g. mech-
anical, friction energies must be consider during a risk
analysis. This holds particular true for products that have
a tendency to form glowing particles and, if in the course
of the process, glowing particles could be formed or
entrained. Electrostatic discharges are ignition sources
that are often underestimated in industrial operations.
These discharges occur frequently in most products
handling processes.
4.2.1. Mechanically generated sparks
Mechanically generated sparks and resultant hot sur-
faces together are regarded as one of the more important
causes of ignition in industrial practice. With mechan-
ically generated sparks, a distinction is made between
grinding, impact and Friction sparks which are formed
by brief contact (5 s) between materials. MechanicaIly
generated hot surfaces, on the other hand, are formed
by relatively long rubbing (5 s) against steel. The hot
Product surfaces show considerably better incendivity in com-
C

Burning class @ 20C 2 2 2

Burning class @ 100C -(Melted) 2 2
Impact sensitivity Negative Negative Negative
DSC
Exothermic onset [C] 400 400 400
Delta H [kcal/kg]
Exothermic onset in air (grewer 300 300 300
oven)
Pmax [bar] 8.0 8.3 9.1
Kmax [barm/s] 158 244 301
Minimum auto ignition temp. 340 440 400
(BAM) [C]
Minimum ignition energy (with 1,3 1,3 1,3
inductance) [mJ]
Powder resistivity [m] 201012 201012 201012
Particle size distribution (median 46 56 16
value) [microns]
Fig. 11. Examples of possible ignition sources ISSA, 1996.
 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151 145

Table 3 4.2.2. Mechanically generated hot surfaces

Influence of the relative circumferential speeds vc on the danger of
ignition for combustible dusts
vc1 m/s There is no danger for ignition
vc1 10 m/s Every case has to be judged separately considering
the product and material-specific characteristics
vc10 m/s In every case there is danger for ignition
Mechanically generated hot surfaces represent an
ignition hazard if, irrespective of the MIT and of the
MIE, the surface temperature is 1100C or higher and
the hot surface area by itself is large enough (see Fig.
13) Bartknecht, 1993; Eckhoff, 1997.
Higher surface temperatures and larger surfaces have
a better incendivity, lower temperatures and smaller sur-
faces a poorer incendivity.

4.2.3. Smoldering pockets

parison with the short-lived mechanically generated
Lumps of smoldering material always represent a haz-
sparks. Neither ignition source appears in industrial
ard when the dust can be classed as capable of forming
practice from the normal metallic materials of construc-
such lumps, i.e. its burning behavior class at 100C is
tion rubbing against each other or against stone if the
greater than 3. A smoldering lump surface of a cube
relative circumferential speeds vc are less than or equal
Ao=9600 mm2 and a surface temperature of To=900C is
to 1 m/s (see Table 3). This is not valid for cerium iron,
sufficient to ignite the mixtures of dusts with a minimum
titanium and zirconium Anon, 1991.
ignition temperature MIT600C (see Fig. 13) Jaeger,
The ignition behavior of mechanically generated
1989; Bartknecht, 1993. Higher surface temperatures
sparks in dust air mixtures depends on the minimum
and larger surfaces have a better incendivity, lower tem-
ignition energy MIE and the minimum autoignition tem-
peratures and smaller surfaces a poorer incendivity.
perature MAIT of the dust in question. The ignition
effectiveness of mechanically generated is dependent
4.2.4. Nature and origin of static electricity in
from the combination of the materials. According to Fig.
production plants
12, it can be stated that the type of spark-producing
An electrostatic charge by itself does not necessarily
material, together with the MIT and the MIE require-
represent an ignition hazard. Such a hazard exists only
ment, determines whether an ignition of dust air mixture
when the charge is so high that discharges occur owing
has to be anticipated from friction, grinding or impact
to the high electrical field (Fig. 14) Britton, 1999.
sparks Anon, 1991.
The individual steps, which lead to the occurrence of
The mechanically generated sparks can thus be
charge build up and discharge, are always the same
assigned different equivalent ignition energies towards
(Fig. 15):
dust air mixtures with a MIT500C. For example, if
Charge Separation process (usually between
the MIT of a dust is 300C, steel-friction sparks can
separation: product and plant units) lead to
ignite this dust only with a MIE (equivalent energy) up
charging of the surfaces in contact.
to 3000 mJ. The equivalent energy, also known as the
Charge Charges can accumulate on products,
equivalent ignition energy EIE, is the amount of energy
accumulation: plant units, packaging containers and
which when transformed into an electrical spark dis-
operators, etc.
charge has the same incendivity as the sparks shown in
Charge As soon as a connection of sufficient
Fig. 12.
dissipation: conductivity is established between the
ground and the sites of the
accumulated charge, the charge can

Fig. 13. Equivalent energies of mechanically generated hot surfaces

Fig. 12. Equivalent energies of mechanically generated sparks. and lumps of smoldering material.
146 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151
Fig. 14. Charge accumulation ISSA, 1996.
Fig. 15. Basic scheme of electrostatics ISSA, 1996.
dissipate to the ground.
Discharge: If the charge continues to accumulate
because the charges formed processes
an not flow to ground quickly enough,
a discharge will occur when the
breakdown field strength is reached.
In a powder handling production plant, static charges
can occur at the surface of solids and powdery sub-
stances. At this point, it is important to make a distinc-
tion between conductors of electricity (e.g. metals) and
non-conductors or insulators (e.g. plastics). Electrically
charged particles (electrons) can move freely on a con-
ductor, whereas on an insulator they arc fixed in one
place. If a substance has excess or a deficiency of
charged particles, it is said to be charged.
If two uncharged (neutral) objects, of which at least
one is an insulator, are brought into close contact with
each other and then rapidly separated, both of the objects
become charged. This occurs because charged particles
will first of all pass from one object to the other but
then, during separation, can not return fast enough.
An electrostatic discharge can be incentive when the
energy released is equal to or greater than the minimum
ignition energy (see Section 4.1.4 for definition) of a
mixture. The energy released depends, among other
things, on the type of discharge. This in turn depends on
the geometry and material of the participating surfaces as
well as on certain other conditions. Table 4 summarizes
the ignition behavior of several types of electrostatic dis-
charges.
5. Protective measures
When combustible dusts are handled, avoiding an
explosive atmosphere by keeping the dust concentration
Table 4
a,b
Ignition sources: Incendivity for
Brush discharge MIE 3 mJ1
Conicle pile discharge MIE 1 J2
Spark discharge MIE 1 J
Propagating brush discharge MIE 10 J
a
For brush discharge incendivity it is assumed that dusts are not
ignitable from a brush discharge. The initial assumption that a dust
cannot be ignited from a brush discharge is justified by considerable
industrial experience with handling powders with low MIEs and exten-
sive laboratory testing.
b
Estimated vale of conical pile discharge incendivity from Glor
and Maurer (1993).
 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151
outside the explosive range is rarely possible due to sedi-
mentation or whirling up of the material being handled.
Thus, as a matter of principle, an explosive atmosphere
can only be avoided with certainty by reducing the oxy-
gen concentration, e.g. inerting. In practice, however, the
possibilities to apply inerting are also limited. For such
situations, avoidance of effective ignition sources and/or
explosion-proof design are the only measures available.
In the following, protective measures are outlined for
typical powder handling operations in the blending
facility. As previously stated, the following product and
plant properties are important for an accurate hazard
assessment:
Minimum ignition energy, MIE, of the bulk material
(measured without inductance in the discharge
circuit),
Minimum Autoignition Temperature, MAIT, of the
bulk material,
Volume resistivity rR of the powder,
Particle size distribution of the bulk material and
mean value, M,
Volume and shape of the silo or container (volume
and shape of the product heap and of the dust cloud).
Unless otherwise stated, the following sections are based
on the assumption that the bulk materials are handled
without flammable gases or vapors being present.
5.1. Pneumatic conveying system
With pneumatic transport, potential ignition sources
include: glowing particles from the product, mechanical
sparks from connecting equipment and, in the case of
highly insulating materials, electrostatic discharges (such
as Propagating Brush Discharges, Fig. 16). Inside the
Fig. 16. Examples of propagating brush discharges ISSA, 1996.
147
transport duct, there is however, no need to consider
ignition hazards due to electrostatic discharges if the fol-
lowing conditions are observed:
The diameter of the duct is less than 1 m;
The MIE of the bulk goods is 1 mJ,
The duct is electrically conducting and is grounded;
The duct is not lined with an insulating material of
2 mm or with a breakdown voltage of greater than
4 kV. In our experience, deposits of insulating pro-
ducts in the duct are not critical if they are not burnt
or melted on.
Normally, pneumatic transport is installed for fill-
ing silos or large vessels; consequently, the same
requirements and considerations must be observed as
for normal filling and emptying (see Section 5.2).
5.2. Filling and emptying operations
In filling and emptying operations, electrostatic
ignition hazards are of prime importance due to the elec-
trostatic charging occurring during separation process.
The hazard comprises possible charge accumulation not
only on plant units, incl. drum and container, but also
in the case of insulating bulk materialon the bulk
material itself which is shown in Fig. 17.
Assuming the insulating bulk material carries a
charge, filling represents the more hazardous of the two
operations for the following reasons. In a filling oper-
ation, the bulk material undergoes dispersion (e.g. grav-
ity feed, pneumatic transport, etc.) and can therefore be
charged in the separation processes occurring in trans-
port. The bulk material and hence its associated charge
is then packed in a small space in which the charge is
not able to flow to ground sufficiently quick, even with
a conductive and grounded receiver. This generates a
high space charge density and hence a high electric field.
In addition, consideration must also be given to the prob-
lems associated with heat accumulation and the possi-
bility of entrainment of smoldering lumps.
Based on the product and plant properties mentioned
Fig. 17. Filling (charging) and emptying (discharging) operations.
148 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151
Table 5
MIEa [mJ] 1 13
MAIT [C] Do not process 530
a
The MIE values must be determined with an addition inductance in the discharge circuit.
above, a decision tree and the matrices are available pro-
viding guidance for the required safety measures
Jaeger & Siwek, 1999.
5.3. Mixing
Even when handling highly sensitive dusts, the occur-
rence of effective ignition sources in mixing is unlikely,
provided the following conditions are met:
In the filling and emptying of the mixer, the measures
applied are the same as those in the filling and empty-
ing of containers.
In the filling and emptying of the mixer, the mixing
elements must be off or run at a circumferential speed
(tip speed) which does not exceed 1 m/s. This restric-
tion must be assured by technical safeguards.
If the mixer is closed and is filled to fill level of 70
vol.% or more, the circumferential speed of the mix-
ing elements is no longer restricted (see Fig. 14).
Any insulating coating must have a breakdown volt-
age of less than 4 kV. Product build-up must be
checked if a homogeneous layer can be formed.
Circumferential speeds up to 10 m/s can be tolerated
in a mixer with a product fill height of less than 70
vol.%, provided the combination of material values
as listed in Table 5 are present.
Nauta mixers with bottom support of the helical screw
can heat up during operation and care must be exer-
cised with substances capable of spontaneous
decomposition. (Fig. 18)
Fig. 18. Mixing operation.
310 1030 30100 100300 3001000 1000
500 465 430 395 360 325
5.4. Dust separation
In the case of dust separators, especially in filters, the
dust explosion hazard must not be underestimated. The
probability of occurrence of a fine dust atmosphere
sensitive to ignition is large. In addition, the entrainment
of ignition sources (e.g. formation of smoldering lumps)
and the danger of ignition through electrostatic charging
are of prime importance (see Fig. 19).
Electrostatic charging must be prevented by the fol-
lowing measures:
Grounding of all conductive apparatus parts. Parti-
cular attention must be paid to the grounding of all
conductive parts which could possibly be insulated
from ground if a filter cloth made of insulating
material is used (e.g. filter supports, clamps). This
must be specially checked after repair and mainte-
nance work.
With a MIE3 mJ or in the presence of flammable
gases or vapors in the air being cleaned, electrically
conducting filter materials must be used. An excep-
tion to this rule can be applied when the protective
measure inerting is employed. Continuity of the
conductivity and safe grounding must be checked.
Multiple washing can have an adverse effect on the
continuity of the filter material conductivity and thus
require repeat checking.
All inner walls on which dust can impact at high
speed must not have any insulating inner coatings
with a high electrical breakdown strength (breakdown
voltage must be less than 4 kV; periodic checks are
required).
In general, with dusts with a MIE10 mJ it is advisable
Fig. 19. Electrostatic ignition sources (dust collector).
 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151
Table 6
Environment bulk No explosible Explosible dust Flammable
material atmosphere atmosphere gases or vapors
MIEa1 J A B C and the inliner is not made of plastic.
3 mJMIEa1 J B B C
MIEa3 mJ C C C
a
MIE measured without inductance in the electrical circuit.
to implement explosion protection measures which go
beyond the avoidance of effective ignition sources or to
consult the responsible specialist departments. It should
further be noted that the fan must be installed on the
clean airside of the filter and that dust deposits must be
avoided in the pipe and fan housing (periodic check or
install a dust control unit).
5.5. Use of flexible intermediate bulk container FIBC
Flexible intermediate bulk containers are used on ever
increasing scale in the powder handling industry.
Depending the hazard situation at the location where
they are used, they must meet different requirements (see
Table 6) in order to avoid ignition hazards caused by
electrostatic charging (Fig. 20).
A: No special requirements.
B: Breakdown voltage of the FIBC wall material must
not exceed 4 kV in order to prevent propagating
brush discharges.
C: The bag material including the slings must be elec-
trostatic conductive. The resistance to the ground
measured at any bag location (inside and outside)
must be fess than 100 M (108 ). The flexible bulk
bag must have a grounding strap. The conductivity
and the necessity for grounding must be clearly mark-
ed.
Most FIBCs on the market today are made of poly-
Fig. 20. FIBC requirements.
149
propylene ribbon fabric. To pass the type B classification
the following requirements are recommended:
any inner PE coating/liner present is not thicker than
2030 m
FIBCs on the market that meet the requirement of a
type C are constructed as follows:
The basic fabric consists of conductive material (e.g.
plastic with sufficient admixture of carbon)
The basic fabric consists of non-conductive material,
but web contains interwoven threads of conductive
plastic material, which are interconnected.
The basic fabric consists of non-conductive material,
but the web contains interwoven metal threads, which
are interconnected.
The basic fabric consists of non-conductive material,
but the FIBC has an internal conductive coating.
To meet the specification as a type C bag, the following
requirements are recommended:
The FIBC must be clearly labeled indicating that it
is conductive and that grounding is required during
charging and discharging (see Fig. 21).
The FIBC must have a clearly marked area for the
attachment of the grounding clamps.
The lifting naps must also be made of conductive
material and have a leakage resistance of less than
108 to the FIBC body.
It is extremely important to keep in mind that the dis-
charge from an ungrounded bag can occur at single
point. Such a discharge is strong enough to ignite dust
Fig. 21. Type C FIBC.
150 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151

clouds. Using a type C bag requires a permanent ground-

ing of the bag during the entire time that the bag is filled
or discharged (see Fig. 21).
The generated charge in the product pile cannot fully
dissipate to ground. Small discharges can occur along
the surface of pile. These electrostatic discharges are too
weak to ignite dust clouds if the volume of the bag is
less than 2 m3 but strong enough to ignite solvent vapors.

6. Safety concept for the blending unit

Based on the determined hazards related to the differ-

ent process operations in the unit it is obvious that the
formation of an explosible dust atmosphere and the pres-
ence of ignition sources cannot be excluded with cer-
tainty. The following safety measures were used to
reduce the probability and the severity of such an Fig. 23. Extinguishing barrier.
unwanted event:

All electrical design within the confines of the facility
are Class II Division 2 Group G with the exception
of a three foot radius around the packaging machines
which will be Class II Division 1 Group G. The
classification was done in accordance to NFPA 499
NFPA, 1997a.
There are no moving parts at tip speeds of greater
than 1 m/s.
All components are made out of conductive material.
All conductive parts are reliably bonded and
grounded.
Tramp metal is rejected from the system through
metal detectors.
All units with a volume of greater than 1 m3 are pro-
tected by the use of constructional measures against
explosions. This was necessary because the formation
of a conical pile discharge cannot be excluded with
certainty.
Silos with a volume of 500 ft3 or less are equipped
with explosion suppression systems (see Fig. 22) and
extinguishing barriers (see Fig. 23) for each convey-
ing line entering the silo. Silos and the conveying
lines are rated to withstand a reduced explosion press-
ure of 15 psig. Explosion venting was ruled out as a
protective measure in this case due to the indoor
location of these silos.
The three 3500 ft3 silos for the storage of the basic
materials is protected by an explosion venting system
and extinguishing barriers for each conveying line
entering the silo. These three silos are rated to with-
stand a reduced explosion pressure of 50 psig.
Installed vent discharge ducts will ensure a safe
release from the enclosure to the outdoors NFPA,
1997b (see Fig. 24).
Silos that do not receive product through pneumatic
conveyors are inerted with nitrogen.
All dust collectors are protected with explosion sup-
pression systems.
All pneumatic conyeyings are done by nitrogen.

Fig. 22. Installed HRD bottles on top of a silo. Fig. 24. Explosion vent discharge duct.
 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151
References
Anon (1991). Perrys chemical engineers handbook. (7th ed). New
York: McGraw-Hill.
ASTM 2019-99 (1999). Standard test method for minimum ignition
energy of a dust cloud in air.
Bartknecht, W. (1993). Explosion protection, basics and application
(in German). Berlin: Springer-Verlag.
Britton, L. G. (1999). Avoiding static ignition hazards in chemical
operations. CCPS.
CCPS (1995). Guidelines for process safety documentation, ch. 6, Pro-
cess hazard analysis.
Information C-2. Guide to risk analysis, Ciba-Geigy internal Infor-
mation, September.
Eckhoff, R. K. (1997). Dust explosions in the process industries. (2nd
ed). Oxford: Butterworth-Heinemann.
Glor, M., & Maurer, B. (1993). Ignition tests with discharges from
bulked polymeric granules in silos. J. Electrostatic, 30, 123134.
ISSA (1996). Static electricity-ignition hazards and protection meas-
ures. Heidelberg, Germany.
Jaeger, N. (1989). Zuendwirksamkeit von Glimmnestern in Staub/Luft
Gemischen. Dusseldorf: VDI Bericht.
Jaeger, N., & Siwek, R. (1999). Prevent explosions of combustible
dusts. Chemical Engineering Progress, AIChE, June.
151
Luettgens, G., & Glor, M. (1989). Understanding and controlling
static electricity. Ehingen, Germany: Expert Verlag.
NFPA (1993). NFPA 77, Static electricity.
NFPA (1997a). NFPA 499, Recommendation for the classification of
combustible dusts and of hazardous (classified) locations for elec-
trical installations in chemical process areas.
NFPA (1997b). NFPA 69, Standard on explosion prevention systems.
NFPA (1998). NFPA 650, Standard for pneumatic conveying systems
for handling combustible particulate solids.
NFPA (1997c). NFPA 654, Standard for the prevention of fire and
dust explosions from the manufacturing, processing and handling
of combustible particulate solids.
Norbert Jaeger is production manager optical brighteners with Ciba Spe-
cialty Chemicals, Grenzach, Germany. From 1994 until 2000 he was
working for Cibas NAFTA Central Safety Testing Laboratory, where he
was responsible for explosion technology and static electricity prevention.
He joined Ciba in 1987 as a member of its Corporate Safety and Environ-
ment unit in Basle, Switzerland. From 1988 through 1994, he was the
safety and loss prevention manager for Cibas additives production site in
Lampertheim, Germany. Jaeger holds an MS in safety science from the
Universitiy of Wuppertal, Germany. He is a member of ASTM Committee
E-27 on Hazard Potential of Chemicals, Subcommittee 05, Explosivity
and Ignitibility of Dust Clouds, a member of AIChEs Center for Chemical
Process Safetys reactive chemicals subcommittee, and a member of the
working group on dust explosions of the International Social Security
Association.
 09 Combustion (burners, combustion systems)

thermogravimetric analysis, grid heater and Curie-point pyrolyser. The oxidation. A percolation model was used to simulate the char structural
original kinetic parameters for the first two experimental techniques were changes during combustion in regime III and the predicted particle size
obtained by applying the non-isothermal model of decomposition of distributions qualitatively agreed with the experimental measurements.
functional groups in independent parallel reactions and for the Curie-point
pyrolyser an isothermal expression of the model was used. Data obtained 00/02799 Modification method of coal and combustion method
with each technique have been extrapolated to the conditions of the other of the modified coal
two techniques, Effects of differences in heating rate, experimental Hasegawa, S. and Hanzawa, M. Jpn. Kokai Tokkyo Koho JP 11 124,587 [99
environment, analysis and amount of sample are discussed, as well as the 124,587] (Cl. ClOL5/00), 11 May 1999, Appl. 97/291,918, 24 Ott 1997. 7.
influence of assumed reaction order in the models. The results provide (In Japanese)
criteria for the range of validity and possible extrapolation of experimental It is possible to modify the method for the preparation of coal for boilers by
kinetic data. continuously pulverizing coarse coal having a particle size of 0.1-5 mm from
coal lumps by feeding the coal with an incombustible gas to a pulverizer to
form pulverized coal having particle size 0.1-10 pm. Then the pulverized
00/02795 Kinetics of NO reduction by char: effects of coal rank coal is fed to a cyclone with an incombustible carrier gas, the mineral
Guo, F. and Hecker, W. C. Symp. (Inr.) Comhus~., [Proc.J, 1998, 2, 3085- contents in the pulverized coal are removed and the demineralized coal and
3092. the incombustible gas are withdrawn from the cyclone and fed to the boiler
There is potential in the heterogeneous reaction of NO with coal char as the for combustion.
basis for re-burning and post-combustion clean-up processes to control NO,
combustion emissions. The reaction is also important in understanding NO
formation and reduction during coal combustion. The kinetics of NO 00/02800 Optimization of operation and environmental
reduction by chars made from coals ranging in rank from lignite to low- characteristics of a thermal contact pyrolyrer in power blocks
volatile bituminous were examined in a packed-bed reactor at temperatures with a circulating bed
of 723-1173 K, graphite and coconut char were also studied. Low-rank Chernyavskii, N. V. er rrl. Ekotekhnol. Resursoshererhenie, 1998, 3, 17-20.
chars were significantly more reactive than high-rank chars with the Ts, (In Russian)
(temperature required for 50% NO conversion) varying from 870 K for An experimental optimization has been carried out of operating regimes
NDL to 1100 K for graphite for a given set of conditions. For all chars and ecological characteristics of a thermal contact pyrolysis reactor of a
studied, the reaction was first order with respect to NO partial pressure and pilot plant with circulating spouted bed. It is shown how it is possible to
exhibited an activation energy (Ea) shift from loo-160 kJ/mol at low obtain the medium-calorific pyrolysis gas with more than 80% of volatile
temperatures to 190-250 kJ/mol at high temperatures. The shift to distinctly conversion and more than 90% of coal sulfur capturing. In addition an
different and higher E,s at higher temperature is opposite to what would optimal scheme of the thermal contact pyrolysis reactor operating in
be expected if a reaction is shifting from chemical rate control to mass circulating fluidized bed is given.
transfer control and suggests different mechanisms or rate-determining
steps at high and low temperatures. Although all chars exhibited the shift in 00102801 Pilot scale study of trace element vaporization and
E,, the shift temperature and the E, within each temperature regime condensation during combustion of a pulverized sub-
tended to increase with increasing rank. Also, the relative reactivity of char bituminous coal
depended on organic char surface area and on inorganic content. Senior, C. L. Fuel Processing Technology, 2000, 63, (2-3). 149-165.
Trace metal emissions from coal-fired power plants are largely associated
with the fly ash. The work reported here is part of a larger effort to develop
00/02796 Laser diagnostics of combustion processes: from a fundamental model for transformation of trace metals in coal to air toxic
chemical dynamics to technical devices emissions from coal-fired power plants. Because the time-temperature
Ebert, V. Isr. J. C/rem, 1999, 39, (I), l-24. history of the combustion gases determines the condensation behaviour of
Valuable tools in the study of the molecular dynamics of gas-phases and gaseous species, experimental study of these phenomena require a realistic
heterogeneous reactions and complex technical combustion processes, post-combustion environment. Pilot-scale combustion of a Powder River
include laser-based rn siru diagnosttc techniques with high temporal, spatial Basin coal was conducted using realistic post-combustion conditions. Trace
and spectral resolution. Results of recent experiments are presented in element distributions were measured in the submicron fly ash at the inlet to
which laser-induced fluorescence among other techniques were applied to an electrostatic precipitator. Flame temperature had a dramatic impact on
investigate elementary chemical gas-phase and catalytic combustion the amount of certain trace elements such as arsenic and selenium in the
reactions, internal combustion engine processes, as well waste incineration submicron ash, indicating that these elements vaporize during the
and coal combustion. combustion process. The amount of vapourization was not sensitive to coal
grind. There is evidence for the reaction of arsenic and selenium vapours
with the large (supermicron) ash in the post-combustion flue gas via a
00102797 Modeling of toxic metal emissions from solid fuel surface reaction. The correlation between arsenic and calcium in the ash
combustors suggests the formation of calcium arsenate. No such correlation was
Yousif, S. er 01. S_rmp. (Inr.) Comhusr.. [Proc.J, 1998, 2, 1647-1654. observed between selenium and calcium.
It is possible that particulate matter emitted from solid-fuel combustors
contains toxic metals in concentrations sufficient to endanger health. Metal
00102802 Pitch pyrolysis kinetics: isothermal heat treatment
contents are enriched in the sub-micron fraction, the one most likely to
avoid capture by conventional collection devices and the one capable of
experiments and model
Py, X. and Daguerre, E. Fuel, 2000, 79, (6), 591-598.
deep pulmonary ingestion. Mathematical modelling to predict toxic metal
The present study focused on the modelling of the evolution under heat
emissions from solid-fuel combustors and of metals partitioning with
treatment of the so-called y, 4 and a-resins fractions usually used for the
respect to particle size is described. This model accounts for particle
characterization of pitch composition. The model was based on phenom-
formation and growth by the combined effects of coagulation and metal
enologic basic mechanisms such as gas-liquid, liquid-liquid mass transfers
vapour deposition on particle surfaces. Based on experimental evidence,
and chemical reactions with respect to the overall resins assimilated to basic
particle size distribution is sub-divided into two ranges: the sub-micron
equivalent chemical species. It has been confronted with success to
fraction and much larger residual ash particles. Conservation equations are
experimental data investigated at three temperature levels (400, 430,
solved for the total number and mass concentrations of particles. The
450C) and eight soaking times (from 0 to 12 h) for two different materials:
modelling is included in an existing procedure for solid-fuel combustors
petroleum and a coal tar pitch. The identified kinetic parameters were
combustion and applied to predict lead and cadmium emissions from a
found to be in agreement with usual values encountered in literature for
laboratory combustor fired with coal and wastewater treatment sludge.
similar basic phenomena. This led to an accurate predictive tool intended
Predicted emissions of these two metals and the metals enrichment in the
for designing reactors and optimization as well as pitch reactivity
sub-micron class compared rather well with measured values. Discrepancies
comparison.
were attributed to the effect of gas-phase and surface chemical reactions
that, were not modelled, due to the absence of kinetic data.
00102803 Predictive method of pressure venting for dust
explosions in large vessels
00/02798 Modeling the fragmentation of non-uniform porous Pascaud, J. M. and Brossard, J. Comhus~. Sci. Technol., 1998, 138, (l-6),
char particles during pulverized coal combustion 159-177.
Liu, G. Fuel, 2000, 79, (6), 627-633. In this study, a simple model to predict the evolution of the combustion
Char fragmentation during pulverized coal combustion was studied using an characteristics of solid-gas mixtures in a closed vessel was extended to a
Australian bituminous coal. The coal was combusted with air in a drop tube wide range of dust suspensions in the general case of vented explosions.
furnace at a gas temperature of 1300C. The char samples were collected at The model was tested to obtain a risk evaluation and to improve safety
different levels of char burnout and their structure was examined using conditions in industrial plants. A simple representation of the combustion
scanning electron microscopy. Approximately 40% of the char particles phenomena based on energy transfers between particles and the action of
formed after pyrolysis were cenospheres with a highly non-uniform porous specific molecules is presented. The pressure venting due to the vent
structure and a large central void. A large number of fine particles were breaking is calculated from thermodynamic characteristics given by the
also observed in the char samples with burnout levels between 30 and 50 model and taking into account the mass rate of discharge of the different
vol%, which suggests that significant fragmentation occurs during the early products deduced from the standard orifice equations. The application
combustion stage. A mathematical model was developed relating the conditions determine the fuel ratio of the used mixtures, the nature of the
fragmentation of cenospherical char particles with the macropores in the chemical kinetics and the calculation of a universal set of parameters. The
particle shell. The formation of these macropores partially results because proposed model is compared to Bartknechts experimental results and
of the carbon removal from the surface of the thin shell due to surface shows the model representativeness for varied dust suspensions (corn-

310 Fuel and Energy Abstracts September 2000

 09 Combustion (burners, combustion systems)

starch, cellulose and aminophenazone) in very large use conditions such as experiments. The peaks varied in shape, temperature and size, and showed
the concentration in a rich or a lean mixture, the vessel volume or its a marked dependence on coal rank. This can be attributed to the varying
geometry and the choice of the vent area or the venting pressure. amounts of the different functional groups in the coals studied. The nitric
oxide released during pyrolysis, together with its precursor species, received
00/02804 Purification of internal combustion engine exhaust special attention.
gases and catalysts used thereof
lizuka, H. er al. Jpn. Kokai Tokkyo Koho JP I I 114,377 [99 114,377) (Cl.
BOlD53/94), 27 Apr 1999, Appl. 97/284,858, 17 Ott 1997. 9. (In Japanese) 00102809 Sources of unburned carbon in the fly ash produced
The gas purification involves employing an alkali or alkaline earth metal from low-NO, pulverized coal combustion
composite oxide whose standard free energy is less than that for formation Veranth, J. M. e! (I/. S~,mp. (Ini.) Cornbust., /Proc./, 1998. 2, 1737-1744
of its sulfates so as to make a lean exhaust NO, chemical adsorbed or When low-NO, pulvertzed coal combustion takes place, the unburned
absorbed to the oxides. The use of the catalysts gives themselves improved carbon in the fly ash produced is shown to consist of a mixture of soot and
SO,-poisoning protection and the exhaust gases improved NO,-purification. coal char. Soot was identified by the presence of chains or aggregates of IO-
50-nm-diameter primary particles in electron microscope images of both
laboratory samples and a sample of fly ash from a power plant operating
00/02805 Pyrolysis kinetics of lignocellulosic materials-three
low-NO, burners. Laboratory samples showed increasing carbon content
independent reactions model
with decreasing nitrogen oxide (NO,) concentration. Experiments included
Orfao, J. J. M. er al. Fuel, 1999, 78, (3). 3499358
a high-NO, base case and four low-NO, cases consisting of: staged
Thermogravimetric analysis with linear temperature programming in the
combustion with short (0.5 s) residence time; staged combustion with long
presence of nitrogen and air was used to study the behaviour of biomass (1.5 s) residence time; a low-NO, burner with slow mixing; and re-burning
components, such as cellulose, lignin and hemicellulose. The results were using coal as the re-burning fuel. Comparison of the base case that used
compared and the pyrolysis kinetics of cellulose detected. Using simple premixed coal and air with long-residence-time staged combustion case
reactions it was not possible to model the thermal decomposition of xylan
shows a decrease in NO, from over 900 ppm to below 200 ppm and an
and lignin with acceptable errors. Thermograms were detected for pine and
increase jn carbon in the ash from 4% to over 30%. Fly ash from staged
eucalyptus woods and pine bark, under inert (nitrogen) or oxidizing (air) combustion was a mixture of large soot aggregates, porous char, and
conditions. The pyrolysis of these lignocellulosic materials was modelled spherical particles of mineral ash, whereas the ash from re-burning lacked
with good approximation by three first-order independent reactions. One of the large aggregates. For all laboratory conditions, carbon content in the
these reactions is associated with the primary pyrolysis of cellulose, its particle fraction with an aerodynamic diameter over 10 pm was higher than
parameters being previously detected and fixed in the model. The model in the 1-2.5 Frn diameter fraction. Both soot aggregates and char
parameters are the activation energies and pre-exponential factors for the contributed to the high carbon in the large particle fraction. The difference
pyrolysis of the remaining two pseudo-components and two additional in carbon burnout between the two staging conditions was consistent with
parameters related to the biomass composition. This report also proposes a published soot oxidation rates. In studies of carbon content of pulverized
method to detect this composition. Comparisons were made between the coal fly ash both char burnout and soot formation need to be considered.
calculated results and data from the literature, and they were in agreement
with one another.
00102810 Spray vaporization in nonpremixed turbulent
00/02806 Release of biomarkers from sulfur-rich kerogens combustion modeling: a single droplet model
during hydrous pyrolysis Rtveillon, J. and Vervisch, L. Comhusrion arrd Flame, 2000, 121, (l/2). 755
Koopmans, M. P. er al. Prepr. Symp.-Am. Ckem. SW., Div. Fuel Chem., 90.
1999, 44, (2). 364-367. The injection of liquid fuel is a common procedure in turbulent combustion
Hydrous and anhydrous pyrolysis experiments in artificial maturation were devices operating in the non-premixed regime. Various numerical models
performed at 200-360C for 72 hours over claystone and limestone, in order may be found in the literature to calculate such turbulent flames, using
to examine the release of biomarkers from sulfur-rich kerogens when they either Reynolds averaged Navier-Stokes techniques (RANS) or large eddy
mature. Hydrous pyrolysis generated large quantities of n-, isoprenoid, and simulation (LES). The typical inputs of non-premixed turbulent combustion
cyclic alkanes, and Cs7_ss-alkenones; in addition, desulfurization of the modelling are the mean and the fluctuations of the mixture fraction. In
polar fraction of the minerals yielded substantial amounts of biomarkers. ln computational fluid dynamics codes, Euler-Lagrange spray modelling may
contrast, desulfurization of the polar fraction of the minerals heated to provide the mean source of mixture fraction. However, the sources of
200C by anhydrous pyrolysis did not release biomarkers. This strongly fluctuations of mixture fraction due to vaporization require more closures.
suggested that anhydrous pyrolysis is not capable of converting sulfur- Direct numerical simulation (DNS) provides a way of estimating these
bound biomarkers in the polar fraction to free biomarkers. The process of sources and, using DNS of droplets evaporating in a turbulent flow, it is
releasing biomarkers from sulfur-rich kerogens relies on the presence of described how they play an important role in the time evolution of fuel/air
water, especially when clay minerals are involved in the process. mixing in a dilute spray. The statistical properties of the spray and of the
scalar field are analysed to propose a single droplet model (SDM) to
00/02807 Screening of potential mineral additives for use as evaluate these sources. SDM calculates mean values of the Eulerian source
fouling preventatives in Victorian brown coal combustion of fuel conditioned on the mixture fraction.
Kyi, S. and Chadwick, B. L. Fuel, 1999, 78, (7). X45-855.
Laboratory tests have been undertaken on a suite of 12 mineral samples to
ascertain their effectiveness as fouling alleviating additives to coal during 00/02811 Strongly swirling gas-particle flows and coal
combustion. Two samples of Victorian coal from the Loy Yang and combustion in a cyclone combustor
Yallourn coal mines were used in the investigation. The laboratory tests Zhou, L. X. er crl. Symp. (lnt.) Comhusr., [Proc.]. 1998, 2, 3119-3126.
include measuring the total chemisorption capacity of the minerals for A three-dimensional phase Doppler particle anemometer system (PDPA) is
different sodium compounds (NaCI, NaOH and Na2SOJ) using a simple used to measure strongly swirling gas-particle flows with swirl numbers 1.5
thermogravimetric technique. The second method involves ashing mixtures and 1.0 in a cyclone combustor. In the case of swirl number 1.5 it is
of the minerals and coal in air. Chemical analysis of the sodium content simulated using a unified second-order moment (USM) model of
within the mixtures (before and after ashing) is used to determine the anisotropic two-phase turbulence. The prediction results using the USM
effectiveness of the different additives, with a high sodium retention model are then compared with those obtained using the isotropic k-r-kp
indicating good potential for fouling mitigation. Minerals containing both two-phase turbulence model and the PDPA measurements. From the
alumina and silica were found to be the best sorbents for NaCl in the measurements and predictions it is clear to see that for strongly swirling
chemisorption experiments. They are also found to be the best additives for flows with axial and tangential inlets there is no recirculation zone in axial
retaining sodium within the coal ash (for both coal types) as well. In velocity profiles; the gas and particle tangential velocity profiles have the
contrast, minerals exhibiting high NaOH and Na2S04 retention were not Rankine-vortex structure, but the solid-body rotation zone is much larger
necessarily good candidates for use as coal additives. This indicates that the than that for weakly swirling flows; then near to and finally slightly smaller
sodium released from the coal during combustion is primarily in the form of than the tangential ones; both axial and tangential particle time-averaged
NaCI, agreeing with thermodynamic modelling of the combustion process. and fluctuation velocities lag behind the gas one; and the USM model can
The best additive candidates are kaolin, bentonite, miclay (micaceous clay), better predict the Rankine-vortex structure of tangential velocity profiles
diatomite, pumice, pyrophyllite and overburden. On a cost basis kaolin and the anisotropy of two-phase turbulence in strongly swirling flows than
mineral types and overburdens (provided there are sufficient reserves) the k-r-kp model. For the optimum design of the coal-fired cyclone
appear to be the best candidates for use as fouling preventative additives combustor that burns three to five mm coal particles, the research results of
during the combustion of Victorian low-rank coal. the two-phase flows were used. These combustors are beneficial because
they provide higher combustion efficiency and lower NO, formation.
00/02808 Simultaneous thermogravimetric-mass spectrometric
study on the pyrolysis behavior of different rank coals
Arenillas, A. er al. 1. Ana[~sis Appl. Pyrolysis, 1999, 50, (I), 31-46. 00102812 Study of the low temperature pyrolyzed coal by
The pyrolysis behaviour of an anthracite and three bituminous coals of thermogravimetry
different volatile matter content was studied by simultaneous thermo- Zheng, Y. and Dai, Z. Meirm Zhuunhucr, 1998, 21, (4). 48-51. (In Chinese)
gravimetry-mass spectrometry. This system was optimized by using calcium The isothermal and non-isothermal method was combined to study the
oxalate as a reference for calibration. A normalization method that pyrolysis of Yanzhou coal and Datong coal before and after low
permitted a semi-quantifiable comparison between the volatile species of temperature pyrolysis. The variation of pyrolysis properties was found,
the coals was also developed. The instantaneous evolution of the volatile also the value of pyrolytic kinetics. More significant was the difference of
compounds was studied by means of temperature programmed pyrolysis weight loss at different temperatures.

Fuel and Energy Abstracts September 2000 311

 Journal of Loss Prevention in the Process Industries 13 (2000) 183199
www.elsevier.com/locate/jlp

Overview of dust explosibility characteristics

*
Kenneth L. Cashdollar
Pittsburgh Research Laboratory, National Institute for Occupational Safety and Health, Pittsburgh, PA, USA

Abstract

This paper is an overview of and introduction to the subject of dust explosions. The purpose is to provide information on the
explosibility and ignitability properties of dust clouds that can be used to improve safety in industries that generate, process, use,
or transport combustible dusts. The requirements for a dust explosion are: a combustible dust, dispersed in air, a concentration
above the flammable limit, the presence of a sufficiently energetic ignition source, and some confinement. An explosion of a fuel
in air involves the rapid oxidation of combustible material, leading to a rapid increase in temperature and pressure. The violence
of an explosion is related to the rate of energy release due to chemical reactions relative to the degree of confinement and heat losses.
The combustion properties of a dust depend on its chemical and physical characteristics, especially its particle size distribution. In
this paper, the explosion characteristics of combustible dusts will be compared and contrasted with those of flammable gases, using
methane as an example. These characteristics include minimum explosible concentration, maximum explosion pressure, maximum
rate of pressure rise, limiting oxygen concentration, ignition temperature, and amount of inert dust necessary to prevent flame
propagation. The parameters considered include the effects of dust volatility, dust particle size, turbulence, initial pressure, initial
temperature, and oxygen concentration. Both carbonaceous and metal dusts will be used as examples. The goal of this research is
to better understand the fundamental aspects of dust explosions. Published by Elsevier Science Ltd.

Keywords: Dust; Explosibility; Explosion

1. Introduction explosibility. Both carbonaceous and metal dusts are

In industries that manufacture, process, generate, or
use combustible dusts, an accurate knowledge of their
explosion hazards is essential. Various books have been
published since 1980 on the general subject of the
explosion hazards of dusts and powders (Bartknecht
1981, 1989, 1993; Field, 1982; Nagy & Verakis, 1983;
Cashdollar & Hertzberg, 1987; Eckhoff, 1991). The
present paper is an update of a previous Pittsburgh
Research Laboratory1 (PRL) paper (Hertzberg & Cash-
dollar, 1987) on the general topic of the explosion haz-
ards of dusts. The basic variables that influence the
characteristics of a dust explosion will be discussed in
general terms without specific reference to particular
practical systems. One purpose of this paper is to provide
assistance and guidance to the practising safety engineer
at a plant regarding the important variables in dust
* Tel.: +1-412-386-6753; fax: +1-412-386-6595.
E-mail address: kgc0@cdc.gov (K.L. Cashdollar).
1
The Pittsburgh Research Laboratory was part of the U.S. Bureau
of Mines before transferring to the National Institute for Occupational
Safety and Health (NIOSH) in October 1996.
used as examples of combustible dusts. Although many
of the examples in this paper use coal dust, the concepts
are applicable to other dusts as well.
This paper is not meant to be an overview of the many
areas of dust explosion research throughout the world.
Instead, it is only meant to be an overview of some of
the dust explosibility characteristics that are important
for safety engineers to consider at industrial plants.
2. Dust explosion requirements
The three requirements for combustion are a fuel, an
oxidizer (usually air), and an adequate heat or ignition
source. This is often called the fire triangle. The fuel
can be any material capable of reacting rapidly and exo-
thermically with an oxidizing medium. In this case, the
fuel is a combustible dust. For a dust explosion, the dust
must be dispersed in the air at the same time that the
ignition source is present. The resulting rapid oxidation
of the fuel dust leads to a rapid increase in temperature
and therefore pressure. This explosion may be a defla-
gration or a detonation, depending on the rate of reaction

0950-4230/00/$ - see front matter. Published by Elsevier Science Ltd.

PII: S 0 9 5 0 - 4 2 3 0 ( 9 9 ) 0 0 0 3 9 - X
184 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199

and resulting burning velocity. The discussion in this The mechanism of flame propagation for many dusts
paper is mainly confined to deflagrations. The destruc- is combustion of flammable gases emitted by particles
tive pressure forces of an explosion can destroy struc- heated to the point of vaporization or pyrolysis
tures and endanger personnel. The violence of an (Hertzberg, Zlochower & Cashdollar, 1988b; Cashdollar,
explosion is dependent on the rate of energy release due Hertzberg & Zlochower, 1989). Some other dusts can
to chemical reactions relative to the degree of confine- propagate a flame through direct oxidation at the particle
ment and heat losses. The requirements for a dust surface (Hertzberg, Zlochower & Cashdollar, 1992). For
explosion are often called (Stephan, 1990) the either mechanism, a finer size of dust is likely to react
explosion pentagonconsisting of fuel, faster than a larger size of dust of the same material.
dispersion/suspension, oxidizer, heat/ignition source, Particle shape and porosity can also greatly affect the
and confinement. The confinement is usually the walls particle surface area and the reaction rates. Therefore,
of the equipment or building in which the dust is dis- the dust particle size and shape are of primary impor-
persed, but it could also come from self-confinement if tance in regard to dust explosibility characteristics. Dusts
the reaction is fast enough. It is possible to have a are often defined as material that is minus 20 mesh
destructive explosion even in open air if the reaction is (850 m) (Stephan, 1990; Nagy, 1981) or minus 40
so fast that pressure builds up in the dust cloud faster mesh (420 m) (NFPA, 1998). However, the larger
than it can be released at the edge of the cloud. dust particles participate inefficiently in the flame propa-
Whether the reacting material is a gas or a dust, the gation process. It is the finer fraction of the dust particles
combustion products are usually gases so that the that contributes the most to the hazard because the finer
explosion process in a closed system is most simply particles have a greater surface area per mass and there-
understood in terms of the ideal gas law: fore react faster. The finer dust particles are also more
m easily dispersed in air and remain airborne longer.
PVnRT RT (1) An example of a dust particle size distribution is
M shown in Fig. 1. The cumulative distribution is shown
where the absolute pressure, P, times the system volume, in Fig. 1A and the differential distribution in Fig. 1B as
V, is proportional to the temperature, T. The pro- semi-logarithmic plots. The two types of size distri-
portionality constants are the number of moles, n, and butions are shown both as surface area weighted and as
the universal gas constant, R. The number of moles is mass or volume weighted curves. The surface median
equal to the mass of gas, m, divided by the average mol-
ecular weight, M. For a typical accidental explosion, air
is usually the oxidant and the fuel may be a dust, gas,
or hybrid mixture. Because air consists mainly of nitro-
gen, there is usually little change in the number of moles
of gas during combustion. Therefore, to a first approxi-
mation, a rapid combustion reaction in a closed system
results in:
Pmax Tb
(2)
P0 T0
where Pmax is the maximum absolute explosion pressure,
P0 is the initial absolute pressure, Tb is the absolute tem-
perature of the burned gas and T0 is the initial absolute
temperature. The faster the combustion reaction is, the
more adiabatic the system will be, and the more nearly
will the explosion pressure approximate the ideal
relation in Eq. (2). If the number of moles of gas changes
significantly during combustion or if the explosion vents
from the container volume, the maximum explosion
pressure will be significantly changed.
In this paper, the terms flammability and explos-
ibility are used interchangeably to refer to the ability
of an airborne dust cloud and/or gas mixture to propagate
a deflagration after it has been initiated by a sufficiently
strong ignition source. Historically, the term flamm-
ability has been used more often for gases, and explos- Fig. 1. Dust particle size distributions by surface area and mass: (A)
ibility more often for dusts. cumulative distribution; (B) differential distribution.
 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
diameter (23 m) and the mass median diameter (42 m)
can be determined from the 50% points on the cumulat-
ive curves in Fig. 1A. The cumulative curve also shows
that the dust has 82% by mass minus 200 mesh (75
m). The differential curves in Fig. 1B are often more
useful in visualizing the size distribution. Other ways of
identifying a representative particle size for the dust in
Fig. 1 are the surface mean diameter (DS=30 m) and
the mass mean diameter (DW=50 m), calculated from
the data in Fig. 1B. Because the combustion of the dust
cloud is greatly dependent on the surface area of the
dust, a mean particle diameter based on surface area is
perhaps more appropriate than one based on mass. Vari-
ous books on particle size analysis (e.g. Allen, 1975;
Irani & Callis, 1963), may be useful in better under-
standing this aspect of dusts. It should be noted that dif-
ferent particle size analysis instrumentation may give
somewhat different results for the same dust because of
the different particle sizing methods used.
The combustion properties of a dust depend on its
chemical and physical characteristics, especially its par-
ticle size distribution. Published dust explosibility data
can give an indication of the hazards associated with a
particular type of dust. However, it is preferable to deter-
mine the explosibility characteristics of an industrial dust
by test, because published data are for a particular size
distribution that may be different from the dust in ques-
tion. Particle shape and porosity are also important con-
siderations in the explosibility of a dust. In general,
shapes with greater surface area will propagate flame
more readily and therefore be more hazardous.
It should be noted that there is no US standardized
test for whether or not a dust is explosible. There are
tests to determine whether a dust can be ignited by an
electric spark (Dorsett, Jacobson, Nagy & Williams,
1960) or what the maximum explosion pressures
(ASTM, 1999a) or minimum explosible concentrations
(ASTM, 1999b) are using stronger chemical ignitors.
One reason for this lack of an explosibility test is that
the question of whether or not a dust can be ignited and
propagate a flame depends greatly on the ignition source.
However, different industries have different views about
what would be an appropriate or likely ignition source.
For some industries, it could be an electrostatic spark;
for others, it could be a flame; and for the mining indus-
try, it could a very large flame from blown-out explos-
ives. Therefore, each industry has to decide what are the
likely or possible ignition sources and which dusts could
be ignited by them.
3. Laboratory equipment for dust explosibility
evaluation
The explosibility characteristics of dust clouds are
often measured in closed volume chambers. The 1.2-L
185
Hartmann tube (Nagy & Verakis, 1983; Dorsett et al.,
1960) is often used for preliminary screening tests and
for minimum ignition energy (MIE) measurements.
However, it may yield false negatives for dusts that are
difficult to ignite with a spark but that are ignitable by
stronger ignition sources. It is also not recommended
(ASTM, 1999a) for measuring rates of pressure rise. The
20-L chambers are used for explosibility measurements
such as maximum explosion pressures, maximum rates
of pressure rise, minimum explosible concentrations, and
inerting effects. An example of a 20-L laboratory
chamber (Cashdollar & Hertzberg, 1985) is shown in
Fig. 2. This is the standard laboratory test chamber
designed and used at the PRL for studying the explos-
ibility and inerting of combustible dusts. There is another
style of 20-L chamber designed by R. Siwek (Bartknecht
1981, 1989; Siwek 1977, 1985, 1988) that is in wide use
in Europe and elsewhere. There are also 1-m3 (1000-L)
chambers (Bartknecht 1981, 1989; Cashdollar & Chatra-
thi, 1993). The 1-m3 chambers may give more realistic
measurements of minimum explosible concentrations,
maximum explosion pressures, and maximum rates of
pressure rise, but the testing is more time consuming and
requires much larger dust samples than the 20-L cham-
bers.
The PRL 20-L chamber is made of stainless steel, and
has a pressure rating of 21 bar. Two optical dust probes
(Cashdollar, Liebman & Conti, 1981; Conti, Cashdol-
lar & Liebman, 1982) are used to measure the uniformity
of the dust dispersion at the positions shown in Fig. 2.
The optical probes measure the transmission through the
dust cloud, with path lengths of 38 or 95 mm. The strain
gauge pressure transducer measures the explosion press-
ure and rate of pressure rise (dP/dt). The data from the
various instruments are collected by a high speed per-
sonal computer (PC) based data acquisition system. The
experimental dust concentration reported for the 20-L
chamber is the mass of dust divided by the chamber vol-
ume. After the dust and ignitor have been placed in the
chamber, the chamber is partially evacuated to an absol-
ute pressure of 0.14 bar,a. Then a short blast of dry air
(from a reservoir at 9 bar) disperses the dust and raises
the chamber pressure to about 1 bar,a. There is a total
ignition delay of 0.4 s from the start of dispersion until
ignition for the standard test procedure in the PRL
20-L chamber. The standard procedure for the Siwek 20-
L chamber has an ignition delay of 0.06 s and a reser-
voir pressure of 20 bar, resulting in a higher level of
turbulence. The usual ignition sources used for the 20-
L tests are electrically activated, pyrotechnic ignitors
manufactured by Fr. Sobbe2 of Germany. These ignitors
are available in various energies from 250 to 10,000 J.
2
Mention of any company name or product does not constitute
endorsement by NIOSH.
186 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199

Fig. 2. Vertical and horizontal cross sections of PRL 20-L explosibility test chamber.

The 2500-J ignitor is comparable in energy to an entire
book of 20 pocket matches, all ignited at once. The
Sobbe ignitors are much stronger than the electric sparks
used in the 1.2-L Hartmann tests.
4. Explosion characteristics
4.1. Pressures and rates of pressure rise
Examples of the pressure data for a weak and a mod-
erate coal dust explosion are shown in Figs. 3 and 4.
The absolute pressure (Figs. 3A and 4A) and rate of
pressure rise (Figs. 3B and 4B) are plotted versus time.
Fig. 3 shows the data for a 20-L chamber explosion test
of a low volatile bituminous coal at a dust concentration
of 125 g/m3, which is just above the minimum required
for an explosion. The pressure trace in Fig. 3A starts at
the partially evacuated value of 0.14 bar,a. The blast of
air that disperses the dust starts at 0.1 s and ends at 0.4 s
on the pressuretime trace. The ignitor is activated at
0.5 s at a chamber pressure of 1.0 bar,a. The maximum
explosion pressure is about 3 bar,a or a pressure rise of
about 2 bar. In Fig. 3B, the rate of pressure rise,

Fig. 3. Typical pressure data for a weak dust explosion.

 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199 187

Fig. 4. Typical pressure data for a moderate dust explosion.

(dP/dt)ignitor, for the ignitor is larger than (dP/dt)ex for
the dust explosion itself. It is important to determine
(dP/dt)ex rather than the ignitor effects. Fig. 4 shows data
for a larger explosion of the low volatile coal dust at a
higher concentration of 200 g/m3. The maximum
explosion pressure is about 5.5 bar,a or a pressure rise
of 4.5 bar. For this explosion, the (dP/dt)ex for the dust
explosion is greater than (dP/dt)ignitor.
Examples of absolute pressure versus time traces for
typical dust explosions at a concentration of 600 g/m3
in the constant volume 20-L chamber are shown in Fig.
5. The traces are for two carbonaceous dusts and six
metal dusts. The relative reactivity of the dusts can be
estimated from either the peak explosion pressure or the
maximum rate of pressure rise. The aluminum (Al) has
the highest reactivity, in part because it is much finer in
size than any of the other dusts in Fig. 5. Next in order
of reactivity is the magnesium (Mg) dust, followed by
the two carbonaceous dusts (polyethylene and coal). The
polyethylene and the high volatile bituminous coal (hvb)
have similar maximum pressures, but the polyethylene
has a faster rate of pressure rise. The titanium (Ti) dust
has a lower explosion pressure than the carbonaceous
dusts, and the iron (Fe) and zinc (Zn) dusts are even
lower. The dust with the lowest reactivity in Fig. 5 is the
tantalum (Ta) dust, which barely reaches its maximum
pressure by 250 ms. The relative reactivities of these
dusts are dependent not only on the intrinsic reactivities
of the materials but also on the specific particle sizes of
the dusts.
The pressure evolution of an explosion in a constant
volume system is predicted by classical combustion
theory (Lewis & von Elbe, 1961, pp. 367381). For the
ideal case, the absolute pressure as a function of time,

Fig. 5. Explosion pressure traces for carbonaceous and metal dusts.

188 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199

P(t), in a constant volume, spherical explosion is related volume. This may be approximately true in the 1-m3
to the fractional volume, V(t), occupied by the fireball chambers. However, in the 20-L chambers with pyro-
during the time of propagation, t, as follows technic ignitors, it is certainly not true, and the ignition
(Hertzberg & Cashdollar, 1987): and combustion are more volumetric (Zhen &
Leuckel, 1997).
P(t)P0 V(t)
k (3)
PmaxP0 V0 4.2. Dust concentration effects
where P0 is the initial absolute pressure, V0 is the
chamber volume, and k is a correction factor related to In order to study the overall explosibility character-
the difference in compressibility between burned and istics of a dust, tests must be made over a range of con-
unburned gases. For spherical propagation from a centrations to determine the worst case. Explosibility
point source, data for the high volatile Pittsburgh bituminous coal dust

are shown in Fig. 6 as a function of dust concentration.
V(t) r(t) 3 Sbt 3
At the top of the figure, the transmission data measured
(4)
V0 r0 r0 by the optical dust probes are shown. The transmission
is measured over a 0.1-s time interval just before the
where r(t) is the fireball radius, r0 is the chamber radius,
dust is ignited. As described in Cashdollar et al. (1981)
and Sb is the flame speed given by:
and Conti et al. (1982), the transmission t is related to
Sb
dt
dr(t) ru
S
rb u
(5)
the mass concentration Cm by Bouguers law:
t=exp(3QCm/2rDS), where Q is a dimensionless

where ru/rb is the density ratio of unburned to burned

gases (at constant pressure). The burning velocity, Su, is
the rate of flame propagation relative to the unburned
gas ahead of it. The flame speed, Sb, is relative to a fixed
reference point. Note that both Sb and Su are for turbulent
not laminar conditions for dust explosions. For spherical
propagation in a spherical chamber, the maximum press-
ure is reached just as the flame contacts the wall. At that
instant, k=1. Differentiating Eq. (3) with respect to time
and substituting Eqs. (4) and (5) into the results gives:
dP(t) r(t)2dr(t)
3(PmaxP0) 3 (6)
dt r0 dt

3(PmaxP0)
ru r(t)2
S
rb u r30
Eq. (6) shows that the maximum rate of pressure rise
should also occur at the instant the flame front contacts
the wall. Setting r(t)=r0=(3V0/4p)1/3 and letting
ru/rbTb/T0Pmax/P0, gives:

KSt
dP(t)
dt max
0 4.84
V1/3 Pmax
P0
1 PmaxSu (7)

Eq. (7) is the cubic law, and KSt is the size normalized
maximum rate of pressure rise.3 The subscript St refers
to Staub, the German word for dust. Because it is size
normalized, the KSt value is used in the practical design
of venting systems (NFPA, 1998).
This derivation of the cubic law is based on the ide-
alized condition where the vessel size is large compared
with either the dust flame thickness or the ignitor flame

3
Note that this equation has been corrected from the version in eq.
(7) in Hertzberg & Cashdollar (1987). Fig. 6. Explosibility data for high volatile bituminous coal dust.
 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199 189

extinction coefficient, is the path length, r is the den-

sity of a particle, and DS is the surface mean particle
diameter. The data in Fig. 6A generally follow the
expected linear relationship on this semi-logarithmic
plot. At the highest dust concentrations, there is some
upward curvature, probably due to increased agglomer-
ation. The scatter in the data is probably due to variations
in the agglomerated particle size of the air dispersed
dust.
In Fig. 6B, (dP/dt)V1/3 is the volume normalized
maximum rate of pressure rise. Note that the turbulence
level was lower in the PRL 20-L chamber for these tests
than that recommended in ASTM E1226 (ASTM,
1999a). Therefore, the (dP/dt)V1/3 data in Fig. 6B are not
recommended for the sizing of vents according to ISO
Standard 6184/1 (ISO, 1985), NFPA Guide 68 (NFPA,
1998), and VDI Standard 3673 (VDI, 1983). These con-
sensus standards are based on the higher turbulence level
of the Siwek 20-L chamber and the 1-m3 chamber
(Bartknecht 1981, 1989). These PRL 20-L data are, how-
ever, useful as a relative measure of explosion hazard.
At the higher turbulence level recommended in ASTM
Standard E1226, the maximum (dP/dt)V1/3 data for this
Fig. 7. Explosion pressure data for high volatile bituminous (hvb)
Pittsburgh coal would be roughly three times higher. The coal and polyethylene dusts, compared with those of methane gas.
maximum absolute explosion pressures (with the press-
ure rise of the ignitor subtracted) are shown in Fig. 6C.
Because there are small variations from test to test in hydrocarbon gases or dusts, the measured LFL or MEC
the chamber pressure at the time of ignition, these data generally corresponds to a calculated adiabatic tempera-
were normalized to a starting pressure of 1.0 bar,a. The ture (Hertzberg et al., 1988b) of 1300 to 1500 K. This
data in Fig. 6 show that below a certain dust concen- is the limit flame temperature, which is the minimum
tration, explosions are not observed. This is the mini- temperature needed to keep a flame propagating. Exper-
mum explosible concentration (MEC) or lean flammable imentally, the LFLs of most hydrocarbon gases are easy
limit (LFL). For this coal, the measured MEC in the to measure because the gases have low ignition energies.
20-L chamber is 80 g/m3. This is the same as the Much stronger ignition energies are needed for dusts
80 g/m3 MEC-value measured for the same coal in a (Cashdollar & Chatrathi, 1993; Hertzberg, Cashdollar &
1-m3 chamber using a 10-kJ ignitor (Cashdollar & Zlochower, 1988a). However, if too strong an ignition
Chatrathi, 1993; Cashdollar, Weiss, Greninger & Chatra- energy is used relative to the test chamber volume, the
thi, 1992). At higher dust concentrations in Fig. 6, the result will be an overdriven ignition (Cashdollar &
maximum pressures and rates of pressure rise level off Chatrathi, 1993). A standard method for measuring the
as all of the oxygen in the chamber is consumed, but MEC of a dust cloud is ASTM E1515 (ASTM, 1999b).
there is no evidence of a rich limit for the coal dust. In contrast to the two dusts in Fig. 7, the methane gas
Typical of dusts, there is more scatter in the rate of shows a rich limit. For the dusts, the maximum pressures
pressure rise data than in the pressure data. level off at concentrations of 200 to 300 g/m3 as all of
A summary of the 20-L chamber pressure versus con- the oxygen in the chamber is consumed. At even higher
centration data for the bituminous coal and polyethylene dust concentrations, although the mixtures are nominally
dusts is shown in Fig. 7, where the data are compared fuel rich, the pressure nevertheless remains constant. The
with those for methane (CH4) gas. The data for the two normal rich limit observed for hydrocarbon gases such
carbonaceous dusts are similar, except that the poly- as CH4 is not observed for the dusts. An explanation of
ethylene has a lower MEC and a slightly higher this effect, at least for many dusts, is that the solid phase
maximum explosion pressure. This is because the poly- fuel must first devolatilize before it can mix with the air
ethylene has a volatility of 100% compared with 37% (Hertzberg et al., 1988b). As soon as sufficient volatiles
volatility for the coal, and it has a higher H:C ratio than are generated to form a stoichiometric concentration of
the coal. The methane gas has a LFL or MEC similar to volatiles in air, the flame front propagates rapidly
that of the polyethylene. This shows that the completely through the mixture before excess fuel volatiles can be
volatilizable polyethylene reacts similarly to the methane generated.
gas at low concentrations (Hertzberg et al., 1988b). For Fig. 8 shows explosibility data from the 20-L chamber
190 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
Fig. 8. Explosibility data at very high dust concentrations for high volatile bituminous coal and polyethylene dusts.
with 2500-J ignitors for the high volatile coal dust and
for polyethylene dust at very high concentrations. This
shows that these dusts explode even at concentrations
beyond 4000 g/m3. There is, of course, an increased
uncertainty in the dust dispersion effectiveness at these
very high concentrations. The decrease in pressure at
higher concentrations may be due to the increased heat
sink of the very large dust concentrations. The decrease
in dP/dt at higher concentrations may be due to the
increased heat sink effect and/or to the possible decrease
in turbulence due to the large mass of dust. Deguigand &
Galant (1981) had previously observed an apparent
upper limit at 4 kg/m3 for coal dust, but this may have
been only an ignitability limit because they used an elec-
tric spark ignition source that was much weaker than the
2500-J Sobbe ignitor used here. Mintz (1993) observed
some upper limits under conditions of reduced oxygen
and at large coal particle sizes. In principle, there are
rich limits for dusts. Eventually, the large mass of excess
fuel will become too much of a heat sink and the flame
temperature will be reduced below its limit value. How-
ever, for most practical purposes, dusts can be con-
sidered to have no rich limit of explosibility. This obser-
vation has also been made by Wolanski (1992).
Examples of scanning electron microscope (SEM)
photomicrographs of coal before and after explosions are
shown in Fig. 9. The dust was a narrow size distribution
of Pittsburgh coal with a mass median diameter,
Dmed23 m. The original unburned particles are shown
at two magnifications on the left side of the figure. They
are compared to the burned post-explosion particles in
the four frames on the right side of the figure. The
burned particles are mainly char residues that are often
larger than the original particles. In the flame, the bitum-
inous coal particles become molten as shown by the
rounded particles on the right. Some particles form
cenospheres. The particles also devolatilize in the flame,
and the volatiles are emitted through the blow holes
seen in the char residues. Additional SEM photomicro-
graphs for various post-explosion dust residues are in
Ng, Cashdollar, Hertzberg & Lazzara (1983).
Metal dusts show similar explosibility data to carbon-
aceous dusts, as shown by the data for two sizes of iron
dust in Fig. 10. The Fe-1 dust was finer in size and had
Dmed4 m; the Fe-2 dust had Dmed45 m. The
explosion pressures, rates of pressure rise, and measured
explosion temperatures are shown as a function of dust
concentration. Fig. 10C shows the measured explosion
pressure (absolute) for each test, corrected for the press-
ure rise due to the ignitor. Fig. 10B shows the size nor-
malized maximum rate of pressure rise, (dP/dt)V1/3, for
each explosion test. As for the coal dust data in Fig. 6,
the iron data in Fig. 10 show that explosions are not
observed below a certain dust concentration. The MECs
for the Fe-1 and Fe-2 dusts are about 220 and 500 g/m3,
respectively, based on the procedures of ASTM E1515
(ASTM, 1999b). However, there is considerable uncer-
tainty in these values, especially for the Fe-2 dust, due
 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
Fig. 9. Scanning electron microscope photographs of bituminous coal particles before and after explosions.
to the scatter in the data. At the higher dust concen-
trations, Pmax and (dP/dt)maxV1/3 level off as all of the
oxygen in the chamber is consumed. Similar to the car-
bonaceous dusts in Fig. 8, the iron metal dusts show no
evidence of a normal rich limit.
The explosion temperatures shown in Fig. 10A were
measured with a six-wavelength infrared pyrometer
(Cashdollar & Hertzberg, 1982). The pyrometer
observed the continuum radiation from the particles, and
temperatures were calculated from the best Planck curve
fit to the infrared radiance data. The maximum measured
particle temperatures for the Fe-1 dust were 1800 K,
well below the maximum calculated adiabatic tempera-
ture, Tad,max=2250 K, for ideal combustion at constant
pressure (Cashdollar, 1994). The maximum measured
particle temperatures for the Fe-2 dust were even lower.
These experimental temperatures are only those of the
particles in the explosion; the gas temperatures may be
different. For all three explosion characteristics shown
in Fig. 10, the Fe-1 dust has higher values than the Fe-
2 dust, showing that it is more reactive, due to its finer
particle size.
4.3. Particle size effects
Most of the previous explosibility data were measured
using rather broad size distributions of the dusts. Fig.
11 shows explosibility data from the 20-L chamber for
Pittsburgh bituminous coal dust as a function of mass
median particle diameter. The data for the narrow size
distributions are shown as the solid circles and solid
curve. These data for narrow distributions are compared
191
with the data crosses for the broad size distributions of
coal dusts. The MEC-values in the bottom section of the
figure are relatively independent of particle size for the
finer sizes. At the larger sizes, above 100 m, the MEC-
values increase with particle size until a size is reached
that can not be ignited. The top two sections of Fig. 11
show that the maximum pressures and rates of pressure
rise are found at the finest sizes tested, as expected. The
pressures decline slowly and the pressure rise rates
decrease faster with increasing particle size. At some
size between 200 and 300 m, the narrow sizes of Pitts-
burgh coal dust can no longer be ignited. These data are
typical for narrow size distributions of carbonaceous fuel
dusts. A broad size distribution is just a combination of
narrow distributions, and these data show that it is the
finer particles in a broad distribution that contribute the
most to its hazard. The MEC data crosses for the broad
size distributions show little difference from the narrow
size distribution data below Dmed100 m. However,
the broad size distributions ignite and propagate at larger
Dmed sizes than the narrow size distributions. The press-
ure and dP/dt data for the broad size distributions are
somewhat higher than those for the narrow size distri-
butions, even in the Dmed range of 20100 m. These
effects are probably due to the tail of fine particles in
the broad size distributions. These fine particles were
removed from the narrow size distributions. The main
conclusion of Fig. 11 is that particle size has an
important effect on the explosibility of coal dusts and
other carbonaceous dusts.
Data showing the effect of particle size for iron dust
are shown in Fig. 12. Because the size distributions were
192 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
Fig. 10. Explosibility data for two sizes of iron dust.
broader and the Dmed values less certain, the data are
shown as bars rather than points. The explosion data are
similar to those for the coal dust. The maximum values
for pressure and rate of pressure rise are found at the
finest particle size. The MEC values are relatively size-
independent at the finer sizes and increase above 30 m
until a size is reached that can not be ignited. Additional
data for size effects of aluminum dusts are in Cashdol-
lar (1994).
4.4. Effects of oxygen concentration
One of the ways to prevent a dust explosion is to inert
the atmosphere so that there is insufficient oxygen for a
flame to propagate. This removes one side of the fire
triangle or explosion pentagon, thereby preventing com-
bustion. One of the most common inerting gases is nitro-
gen, which is the main constituent of air. To determine
the limiting oxygen concentration for coal dust
explosions in the 20-L chamber (with 2500-J ignitors),
the dusts were dispersed with various oxygennitrogen
mixtures instead of normal air at 20.95% O2. Fig. 13 is
an example of the reduced oxygen data for coal dust.
The explosions are denoted by the solid circles and the
nonexplosions by the open circles. The data for coal dust
in air are shown at the top of the figure. In air, the dust
ignites and burns at all coal concentrations above the
MEC of 80 g/m3. At the bottom of the figure,
explosions still occur at 14% down to 11.5% O2. At 11%
O2, the coal dust ignited only in one out of eight tests.
At even lower oxygen concentrations, the dust could not
be ignited. The boundary between oxygen concentrations
that support combustion and those that do not support
combustion is the limiting oxygen concentration, LOC.
As a safe margin, NFPA 69 (NFPA, 1997) recommends
keeping the system oxygen concentration at least 2%
lower than the measured LOC. Gases other than nitrogen
can also be used to reduce the oxygen concentration.
Carbon dioxide is usually more efficient than nitrogen
for inerting carbonaceous dusts, but it is often less effec-
 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
Fig. 11. Effect of particle size on the explosibility of coal dusts: ,
for narrow distributions; , for broad distributions.
tive than nitrogen in inerting metal dusts (Nagy, Dor-
sett & Jacobson, 1964).
The oxidant for a dust explosion is usually the oxygen
in air, although other gases can also be oxidizers. Oxy-
gen concentrations greater than 21% tend to increase the
burning velocity, and concentrations less than 21%
reduce the burning velocity. An example of the effect of
varying oxygen concentration on the explosion pressure
and rate of pressure rise for a carbonaceous dust is
shown in Fig. 14. The data are from DiPalma (1998).
The solid data symbols are at a dust concentration of
500 g/m3 and the open circle data symbols are at a dust
concentration of 375 g/m3. Fig. 14A shows that the rate
of pressure rise varies almost exponentially with oxygen
concentration on this semi-logarithmic plot. The
explosion pressure in Fig. 14B varies roughly linearly
with oxygen concentration, although there is some scat-
ter in the data. If the dust concentration was varied at
each oxygen concentration to obtain the highest Pmax
value, the explosion pressure would be expected to
increase linearly with oxygen concentration, based on
the ideal gas law (Eq. (1)). However, this linear relation-
ship would change as the LOC is approached. Near the
LOC, the pressure would decrease very rapidly with
193
Fig. 12. Effect of particle size on the explosibility of iron dusts.
decreasing oxygen concentration, until the mixture
would no longer be explosible.
Because of the large effect of varying oxygen concen-
tration on the explosion characteristics of dusts, it is
important to test the dust at the appropriate O2 concen-
tration. When determining the explosion characteristics
for a dust in air, it is important to measure the O2 content
of the air cylinders used for the tests. Gas cylinders
that are filled with air that has been compressed and
dried have the normal 20.95% O2. However, many air
cylinders are filled with synthetic or reconstituted air
that has been mixed from liquified oxygen and nitrogen.
The O2 content of these cylinders has been observed to
vary considerablyfrom 19% to 26% O2.
4.5. Effect of temperature
The thermal ignitability of coal dust is shown in Fig.
15, as measured in the PRL 6.8-L furnace (Conti, Cash-
dollar & Thomas, 1993). The tests resulting in ignitions
(solid circles) and non-ignitions (open circles) are plot-
ted on a graph of initial furnace temperature versus dust
cloud concentration. The solid curve is the temperature
boundary between the upper region of the graph where
194 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199

Fig. 13. Effect of reduced oxygen concentration on coal dust explos-

ibility.

the coal dust cloud will thermally autoignite and the

lower region where the dust may be flammable but does
not thermally autoignite. The lowest point of the curve
is the minimum autoignition temperature (MAIT) for the
coal530C. This 6.8-L furnace is one of several listed
in ASTM standard test E1491 for the measurement of
the MAITs of dusts (ASTM, 1999c).
The effect of temperature on the ignitability and
explosibility of coal dust is shown in Fig. 16. The dotted
curve (from Fig. 15) shows the autoignition temperature
for the coal as a function of dust concentration. The dot-
ted curve is the temperature boundary between the upper
region of the graph where the coal dust cloud will ther- Fig. 14. Effect of oxygen concentration on explosion pressures and
mally autoignite and the lower region where the dust rates of pressure rise (data from DiPalma, 1998).
may be flammable but does not thermally autoignite. In
addition to the 6.8-L furnace data, explosibility tests
were also conducted in the 20-L chamber at temperatures
above ambient but below the temperature at which the
CTCT0 273+T0
273+T
[10.00072(TT0)] (8)

dust would autoignite. For these tests, the 20-L chamber
was wrapped with electrical heater tape and insulated to
reach the elevated temperature. A thermocouple meas-
ured the set temperature of the chamber before the test.
The solid circle data points in Fig. 16 show the MEC-
data (Cashdollar, 1996) for the coal dust at near ambient
(60C) and at an elevated temperature of 180C. The
experimental data points are extrapolated to even higher
temperatures (solid curve) using the modified Burgess
Wheeler law (Zabetakis, 1965; Conti, Cashdollar, Hertz-
berg & Liebman, 1983) for hydrocarbons:
where CT is the limit in terms of mass concentration at
temperature T, CT0 is the limit at T0, and the temperatures
are in C. The dust concentrations to the right of the
solid curve are flammable (explosible) and the region to
the left of the curve is nonflammable. For comparison,
the measured lean flammable limit data for methane gas
as a function of temperature (dashed curve, from Cow-
ard & Jones, 1952, p. 43) are also shown. The decrease
in the LFL or MEC with increase in temperature is simi-
lar in form for the dust and gas.
At higher dust concentrations, the maximum
 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
Fig. 15.
Fig. 16. Effect of temperature on the flammability and thermal ignit-
ability of coal, compared with methane gas.
explosion pressure for the bituminous coal was also mea-
sured at elevated temperature in the 20-L chamber
(Cashdollar, 1996). At near ambient temperature, Pmax
for the dust was 6.6 bar,a. At an elevated temperature
of 180C, Pmax was 4.8 bar,a. This observation of lower
explosion pressures at elevated temperature was also
reported previously by Wiemann (1987). The inverse
195
Thermal ignitability of coal dust.
relationship of explosion pressure with initial tempera-
ture is expected from the ideal gas law (Eq. (1)) because
there are fewer oxygen molecules at elevated tempera-
ture to react with the coal. The ratio of measured
maximum explosion pressure (absolute) at elevated tem-
perature to that at ambient temperature is approximately
the same as the ratio of ambient to elevated temperature
in degrees kelvin.
The limiting oxygen concentration for coal dust was
also measured at elevated temperature in the 20-L
chamber. The measured LOC value (Cashdollar, 1996)
for the dust decreased from 11% at ambient tempera-
ture to 10% at 180C. This effect of lower LOC
values at elevated temperature was also observed pre-
viously by Wiemann (1987).
4.6. Effect of pressure
The effect of initial chamber pressure (Hertzberg et
al., 1988a) on the MEC or LFL of gases and dusts is
shown in Fig. 17. When the methane concentration is
expressed in volume percent in Fig. 17A, the LFL is
shown to be constant as the pressure varies from 0.5 to
3 bar. When the CH4 is expressed in mass concentration
in Fig. 17B, the LFL is shown to vary linearly with
pressure. In Fig. 17C, the LFLs of the Pittsburgh coal
and polyethylene dusts also vary linearly with pressure.
A similar relationship was found by Wiemann (1987)
for a brown coal dust.
Bartknecht (1989) and Wiemann (1987) report data
on the effect of initial pressure on the Pmax and KSt
values. Both show that Pmax increases linearly with
increase in initial pressure, over the range of 14 bar.
They also show that KSt increases with initial pressure.
196 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
Fig. 17. Effect of pressure on the lower flammable limits for coal
and polyethylene dusts, compared with methane gas.
4.7. Hybrid mixtures of dusts and gases
Another important factor in the explosibility hazard
of a dust is the possible co-presence of a flammable gas.
Hybrid mixtures of a combustible dust (coal) and a
flammable gas (CH4) were also studied in the 20-L
chamber using 2500-J ignitors (Cashdollar, 1996). Data
for a low volatile bituminous (lvb) coal are shown in
Fig. 18A, and for a high volatile bituminous (hvb) coal
in Fig. 18B. The flammable limits for mixtures of coal
and CH4 are shown by the data points and solid curves.
The areas above and to the right of the curves are explos-
ible (flammable) and the areas below and to the left of
the curves are nonexplosible (nonflammable). The data
for mixtures of Pittsburgh coal and CH4 in Fig. 18B
show a linear or near-linear mixing relationship similar
to Le Chateliers Law for hydrocarbon gases (Zabetakis,
1965; Kuchta, 1985, pp. 4850). All of the solid circle
data symbols are for 2500-J ignitors. The measured LFL
for the pure CH4 with this 2500-J ignitor is 4.4%, but
this is an overdriven system as shown by tests in a larger
120-L chamber (Hertzberg et al., 1988a). The more
appropriate LFL for CH4 is the 4.9% value measured
with a 1000-J ignitor in the 20-L chamber and shown as
Fig. 18. Lower flammable limits for hybrid mixtures of low and high
volatile coal dusts with methane gas.
the symbol in the figure. The data for hybrid mixtures
of the low volatile Pocahontas coal and CH4 in Fig. 18A
show some curvature. This is probably due to the even
greater difference in ignitability between the low volatile
coal and the CH4. That is, the dust becomes more easily
ignited as more CH4 is added. Therefore, the curvature
is more likely an effect of ignitability rather than an
effect of flammability. Ideally, the true mixing relation-
ship would be determined in a much larger chamber,
such as a 1-m3 chamber, where a very strong ignition
source could be used for the dusts without overdriving
the CH4 gas. For most practical situations for mixtures
of hydrocarbon dusts and gases, the linear mixing law
of Le Chatelier would be sufficient. This approximately
linear relationship for the lean limits of coal dust and
CH4 gas mixtures was also observed by Amyotte and
colleagues (Amyotte, Mintz, Pegg, Sun & Wilkie, 1991;
Amyotte, Mintz, Pegg & Sun, 1993) using 5000-J igni-
tors in a 26-L chamber. This linear mixing relationship
is also applicable to mixtures of two carbonaceous dusts
(Hertzberg & Cashdollar, 1987). However, it is not
applicable to mixtures where the two components have
greatly different limit flame temperatures, such as a car-
bonaceous dust and hydrogen gas (Hertzberg & Cashdol-
lar, 1987).
 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
4.8. Effect of added inert dust
The addition of an inert powder to a combustible dust
and air mixture can reduce the explosibility through the
absorption of heat. In the mining industry, coal dust
explosions are prevented by the addition of limestone
rock dust to the deposited coal dust (Nagy, 1981). Since
the limestone is incombustible, it acts as a heat sink to
reduce the flame temperature of the dust mixture below
its limit value. The inerting of coal dust by the addition
of limestone rock dust has been studied in the PRL
20-L laboratory chamber, and the results were compared
to those from full-scale experimental mine tests
(Cashdollar et al., 1992; Cashdollar, 1996; Cashdollar &
Hertzberg, 1989; Greninger et al., 1991). The laboratory
data are shown in Fig. 19. In the top part of the figure,
Fig. 19. Effect of added inert dust on the explosion characteristics
of bituminous coal dust.
197
the LFL or MEC of the coal dust shows almost no effect
with added rock dust until there is over 50% rock dust
in the mixture. At higher rock dust percentages, the LFL
increases, until the mixture can not be ignited at 75%
rock dust. The two lower parts of the figure show the
maximum pressure and rate of pressure rise as a function
of rock dust percentage in the mixture. At each rock dust
percentage, the coal dust concentration was varied over
a series of tests to determine the maximum pressure and
dP/dt. The explosion pressures show only a slight
decrease with added rock dust content up about 70%.
Between 70% and 80% rock dust, the pressures drop
rapidly as the mixture becomes totally inerted and flame
no longer can propagate. The rates of pressure rise
decline almost linearly with increased rock dust content
over the entire range. The 20-L laboratory data for the
rock dust inerting of coals shows relatively good agree-
ment with large-scale data from the PRL experimental
mine (Cashdollar et al., 1992; Greninger et al., 1991;
Weiss et al., 1989). Therefore, the laboratory chamber
can be used for preliminary testing to reduce the number
of large-scale tests. The mining regulations are still
based on the results of the large-scale research.
In addition to the use of inert powders premixed with
the combustible dust in order to prevent ignition and
flame propagation, inert powders are also used in sup-
pression systems to extinguish propagating explosions.
5. Conclusions
The data examples reported in this paper show that
laboratory test chambers are useful in studying a wide
range of explosion characteristics of dusts. For both car-
bonaceous and metal dusts, the finer sized dusts are the
more hazardous. Because of the importance of particle
size, it is critical that representative samples of dusts be
collected for explosibility evaluation. Because of the
possible accumulation of fines at some location in a pro-
cessing system, ASTM E1226 (ASTM, 1999a) and
E1515 (ASTM, 1999b) recommend that the test sample
be less than 200 mesh. It is also important to consider
the effects of the initial system temperature, pressure,
and oxygen concentration on the explosion character-
istics.
Acknowledgements
The author acknowledges the assistance of G.M.
Green in the conduction of the size analyses and the 20-L
chamber explosion tests and C.E. Lucci for the computer
program for data acquisition and analysis. Both are from
the Pittsburgh Research Laboratory of NIOSH. The
author also thanks M. Hertzberg (retired from the Bureau
of Mines) for introducing him to the subject of dust
198 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
explosions and helping him to understand combustion
theory.
References
Allen, T. (1975). Particle size measurement. London: Chapman and
Hall.
Amyotte, P. R., Mintz, K. J., Pegg, M. J., Sun, Y. H., & Wilkie, K.
I. (1991). Laboratory investigation of the dust explosibility charac-
teristics of three Nova Scotia coals. Journal of Loss Prevention in
the Process Industries, 4, 102109.
Amyotte, P. R., Mintz, K. J., Pegg, M. J., & Sun, Y. H. (1993). The
ignitability of coal dustair and methanecoal dustair mixtures.
Fuel, 72, 671679.
ASTM (1999a). Standard test method for pressure and rate of pressure
rise for combustible dusts E1226-94. In ASTM annual book of stan-
dards, vol. 14.02. West Conshohocken, PA: American Society for
Testing and Materials.
ASTM (1999b). Standard test method for minimum explosible concen-
tration of combustible dusts E1515-98. In ASTM annual book of
standards, vol. 14.02. West Conshohocken, PA: American Society
for Testing and Materials.
ASTM (1999c). Standard test method for minimum autoignition tem-
perature of dust clouds E1491-97. In ASTM annual book of stan-
dards, vol. 14.02. West Conshohocken, PA: American Society for
Testing and Materials.
Bartknecht, W. (1981). Explosions: course, prevention, protection.
Berlin: Springer.
Bartknecht, W. (1989). Dust explosions: course, prevention, protec-
tion. Berlin: Springer.
Bartknecht, W. (1993). Explosions-Schutz (explosion protection).
Berlin: Springer.
Cashdollar, K. L. (1994). Flammability of metals and other elemental
dusts. Process Safety Progress, 13, 139145.
Cashdollar, K. L. (1996). Coal dust explosibility. Journal of Loss Pre-
vention in the Process Industries, 9, 6576.
Cashdollar, K. L., & Chatrathi, K. (1993). Minimum explosible dust
concentrations measured in 20-L and 1-m3 chambers. Combustion
and Science Technology, 87, 157171.
Cashdollar, K. L., & Hertzberg, M. (1982). Infrared pyrometers for
measuring dust explosion temperatures. Optical Engineering, 21,
8286.
Cashdollar, K. L., & Hertzberg, M. (1985). 20-L explosibility test
chamber for dusts and gases. Review of Scientific Instruments, 56,
592602.
Cashdollar, K. L., & Hertzberg, M. (1987). In: Cashdollar, K. L., &
Hertzberg, M. (Eds), Industrial Dust Explosions (Proceedings of
the Symposium on Industrial Dust Explosions, Pittsburgh, PA, June
1013, 1986), STP 958 (363 pp). West Conshohocken, PA: Amer-
ican Society for Testing and Materials.
Cashdollar, K. L., & Hertzberg, M. (1989). Laboratory study of rock
dust inerting requirements: effects of coal volatility, particle size,
and methane addition. In Proceedings of the 23rd International
Conference of Safety in Mines Research Institutes (September 11
15, 1989, Washington, DC) (pp. 965977).
Cashdollar, K.L., Liebman, I., & Conti, R.S. (1981). Three bureau of
mines optical dust probes. US Bureau of Mines RI 8542, 26 pp.
Cashdollar, K. L., Hertzberg, M., & Zlochower, I. A. (1989). Effect
of volatility on dust flammability limits for coals, gilsonite, and
polyethylene. In 22nd Symposium (International) on Combustion
(pp. 17571765). Pittsburgh, PA: The Combustion Institute.
Cashdollar, K. L., Weiss, E. S., Greninger, N. B., & Chatrathi, K.
(1992). Laboratory and large-scale dust explosion research.
Plant/Operations Progress, 11, 247255.
Conti, R. S., Cashdollar, K. L., & Liebman, I. (1982). Improved optical
dust probe for monitoring dust explosions. Review of Scientific
Instruments, 53, 311313.
Conti, R.S., Cashdollar, K.L., Hertzberg, M., & Liebman, I. (1983).
Thermal and electrical ignitability of dust clouds. US Bureau of
Mines RI 8798, 40 pp.
Conti, R.S., Cashdollar, K.L., & Thomas, R.A. (1993). Improved 6.8-L
furnace for measuring the autoignition temperatures of dust clouds.
US Bureau of Mines RI 9467, 26 pp.
Coward, H.F., & Jones, G.W. (1952). Limits of flammability of gases
and vapors. US Bureau of Mines Bulletin 503, 155 pp.
Deguigand, B., & Galant, S. (1981). Upper flammability limits of coal
dustair mixtures. In 18th Symposium (International) on Combus-
tion (pp. 705715). Pittsburgh, PA: The Combustion Institute.
DiPalma, J. (1998). Cytec Industries, Inc., Stamford, CN, personal
communication.
Dorsett, H.G., Jr., Jacobson, M., Nagy, J., & Williams, R.P. (1960).
Laboratory equipment and test procedures for evaluating explos-
ibility of dusts. US Bureau of Mines RI 5624, 21 pp.
Eckhoff, R. K. (1991). Dust explosions in the process industries.
Oxford: Butterworth Heinemann.
Field, P. (1982). Dust explosions. Handbook of powder technology,
vol. 4. Amsterdam: Elsevier.
Greninger, N. B., Cashdollar, K. L., Weiss, E. S., & Sapko, M. J.
(1991). Suppression of dust explosions involving fuels of inter-
mediate and high volatile content. In Proceedings of the 4th Inter-
national Colloquium on Dust Explosions (November 49, 1990,
Porabka-Kozubnik, Poland) (pp. 208228).
Hertzberg, M., & Cashdollar, K. L. (1987). Introduction to dust
explosions. In Industrial dust explosions, STP 958 (pp. 532). West
Conshohocken, PA: American Society for Testing and Materials.
Hertzberg, M., Cashdollar, K. L., & Zlochower, I. A. (1988a). Flamm-
ability limit measurements for dusts and gases: ignition energy
requirements and pressure dependences. In 21st Symposium
(International) on Combustion (pp. 303313). Pittsburgh, PA: The
Combustion Institute.
Hertzberg, M., Zlochower, I. A., & Cashdollar, K. L. (1988b). Vola-
tility model for coal dust flame propagation and extinguishment.
In 21st Symposium (International) on Combustion (pp. 325333).
Pittsburgh, PA: The Combustion Institute.
Hertzberg, M., Zlochower, I. A., & Cashdollar, K. L. (1992). Metal
dust combustion. In 24th Symposium (International) on Combus-
tion (pp. 18271835). Pittsburgh, PA: The Combustion Institute.
Irani, R. R., & Callis, C. F. (1963). Particle size: measurement,
interpretation, and application. New York: John Wiley and Sons.
ISO (1985). Explosion protection systemsPart 1: determination of
explosion indices of combustible dusts in air, ISO 6184/1. Switzer-
land: International Organization for Standardization.
Kuchta, J. M. (1985). Investigation of fire and explosion accidents in
the chemical, mining, and fuel-related industriesa manual. US
Bureau of Mines Bulletin 680. pp. 4850.
Lewis, B., & von Elbe, G. (1961). Combustion, flames, and explosions
of gases. New York: Academic Press.
Mintz, K. J. (1993). Upper explosive limit of dusts: experimental evi-
dence for its existence under certain circumstances. Combustion
and Flame, 94, 125130.
Nagy, J. (1981). The explosion hazard in mining. US Mine Safety and
Health Administration IR 1119, 69 pp.
Nagy, J., & Verakis, H. C. (1983). Development and control of dust
explosions. New York: Marcel Dekker.
Nagy, J., Dorsett, H.G., Jr., & Jacobson, M. (1964). Preventing ignition
of dust dispersions by inerting. US Bureau of Mines RI 6543, 29
pp.
NFPA (1997). Standard on explosion prevention systems, NFPA 69.
Quincy, MA: National Fire Protection Association.
NFPA (1998). Guide for venting of deflagrations, NFPA 68. Quincy,
MA: National Fire Protection Association.
Ng, D.L., Cashdollar, K.L., Hertzberg, M. & Lazzara, C.P. (1983).
 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
Electron microscopy studies of explosion and fire residues. US
Bureau of Mines IC 8936, 63 pp.
Siwek, R. (1977). 20-L-Laborapparatur fur die Bestimmung der
Explosionskenngrossen brennbarer Staube (20-L laboratory appar-
atus for the determination of the explosion characteristics of flam-
mable dusts). Thesis at Winterthur Engineering College, Win-
terthur, Switzerland, 109 pp., in German.
Siwek, R. (1985). Development of a 20 ltr laboratory apparatus and
its application for the investigation of combustible dusts. Basel,
Switzerland: Ciba Geigy AG.
Siwek, R. (1988). Reliable determination of the safety characteristics
in 20-L apparatus. In Flammable Dust Explosion (November 24,
1988, St. Louis, MO) (pp. 529573). Holliston, MA: IBC USA.
Stephan, C.R. (1990). Coal dust as a fuel for fires and explosions.
Report No. 01-066-90, US Mine Safety and Health Administration,
Pittsburgh, PA, 10 pp.
VDI (1983). Pressure release of dust explosions, VDI 3673. Dussel-
dorf, Germany: Verein Deutscher Ingeniere.
199
Weiss, E. S., Greninger, N. B., & Sapko, M. J. (1989). Recent results
of dust explosion studies at the Lake Lynn experimental mine. In
Proceedings of the 23rd International Conference of Safety in
Mines Research Institutes, (September 1115, 1989, Washington,
DC) (pp. 843856).
Wiemann, W. (1987). Influence of temperature and pressure on the
explosion characteristics of dust/air and dust/air/inert gas mixtures.
In Industrial Dust Explosions, STP 958 (pp. 3344). West Consho-
hocken, PA: American Society for Testing and Materials.
Wolanski, P. (1992). Dust explosion research in Poland. Powder Tech-
nology, 71, 197206.
Zabetakis, M.G. (1965). Flammability characteristics of combustible
gases and vapors. US Bureau of Mines Bulletin 627, 121 pp.
Zhen, G., & Leuckel, W. (1997). Effects of ignitors and turbulence on
dust explosions. Journal of Loss Prevention in the Process Indus-
tries, 10, 317324.
 Journal of Loss Prevention in the Process Industries 13 (2000) 209219
www.elsevier.com/locate/jlp

Flammability limit measurements for dusts in 20-L and 1-m3

vessels
a,*
John E. Going , Kris Chatrathi a, Kenneth L. Cashdollar b

a
Fike Metal Products, P.O. Box 610, 704 South 10th Street, Blue Springs, MO 64015, USA
b
Pittsburgh Research Laboratory, National Institute for Occupational Safety and Health, Pittsburgh, PA 15236, USA

Abstract

Two types of flammability limits have been measured for various dusts in the Fike 1-m3 (1000-L) chamber and in the Pittsburgh
Research Laboratory (PRL) 20-L chamber. The first limit is the minimum explosible concentration (MEC), which was measured
at several ignition energies. In addition to the three dusts studied previously (bituminous coal, anthracite coal, and gilsonite), this
work continues the effort by adding three additional dusts: RoRo93, lycopodium, and iron powder. These materials were chosen
to extend the testing to non-coal materials as well as to a metallic dust. The new MEC data corroborate the previous observations
that very strong ignitors can overdrive the ignition in the smaller 20-L chamber. Recommendations are given in regard to appropriate
ignition energies to be used in the two chambers. The study also considered the other limiting component, oxygen. Limiting oxygen
concentration (LOC) testing was performed in the same 20-L and 1-m3 vessels for gilsonite, bituminous coal, RoRo93, and aluminum
dusts. The objective was to establish the protocol for testing at different volumes. A limited investigation was made into overdriving
in the 20-L vessel. The LOC results tended to show slightly lower results for the smaller test volume. The results indicated that
overdriving could occur and that ignition energies of 2.5 kJ in the 20-L vessel would yield comparable results to those in the 1-
m3 vessel using 10.0 kJ. The studies also illustrate the importance of dust concentration on LOC determinations. 2000 Elsevier
Science Ltd. All rights reserved.

Keywords: Dust explosions; Minimum explosion concentration; Limiting oxygen concentration

1. Introduction Oxygen Concentration (LOC) or Minimum Oxygen

Many explosibility measurements are needed for
safety or hazard analyses. These include the basic
explosibility parameters, maximum pressure and rate of
pressure rise, as well as explosibility limit parameters
such as fuel concentration, oxidant concentration, and
ignition energy. The fuel concentration limit, often
referred to as Minimum Explosible Concentration, MEC
(or lean flammable limit, LFL) is the lowest concen-
tration of dust dispersed in air that can propagate an
explosion. Today, most MEC measurements are made in
either a 20-L vessel or a 1-m3 (1000-L) vessel. The 20-
L vessel is considerably more convenient to use; the 1-
m3 vessel is expected to produce data that are more rep-
resentative of industrial scale explosions. Another limit
measurement of use in hazard analysis is the Limiting
* Corresponding author. Fax: +1-816-229-5082.
E-mail address: jgoing@fike.com (J.E. Going).
Concentration (MOC). The LOC, which is the term used
in this paper, is the oxygen concentration at the boundary
between propagation and nonpropagation of the dis-
persed dust cloud. LOC data are used, along with an
appropriate safety factor, to establish safe inerting levels
in industrial processes.
Since the MEC and LOC values are experimentally
determined in the laboratory, one of the on-going con-
cerns is that of overdriving the system by a large ignition
source. This is of particular concern with the smaller 20-
L vessel. Dust clouds are inherently more difficult to
ignite than gases and therefore stronger ignition sources
are used in testing. A true limit measurement should
be independent of ignition energy. When the ignition
source is too weak, both the measured MEC and LOC
will be higher than the true value. The system is under-
driven and the results are based more on ignitability than
flammability. In theory, the ignition energy is increased
until the limit measurements are independent of energy.
At some point, however, the energy level is excessive

0950-4230/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 9 9 ) 0 0 0 4 3 - 1
210 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219

for the size of the vessel and the system is overdriven.

In this situation, the energy contributed by the ignitor is
sufficient to combust enough dust so that the result
appears to be an explosion although there is no real
propagation beyond the ignitor flame. Similarly, an
overly strong ignitor can markedly change the initial test
conditions by raising the overall temperature of the sys-
tem, which in turn would lower the apparent limits and
a nonexplosible system would appear to be explosible. A
plot of ignition energy versus the measured limit would
ideally have a vertical asymptote where the limit is inde-
pendent of energy. For most dusts, however, this is not
the case, particularly in the 20-L vessel.
Comparison measurement between the 20-L and 1-m3
vessels can be used to evaluate the overdriving effect.
Overdriving is generally unlikely to occur in the 1-m3
vessel and in principle the 1-m3 vessel can be used to Fig. 1. Vertical cross-section of Pittsburgh Research Laboratory 20-
establish the energy independent limit value. Such com- L chamber.
parisons of vessel size have been made by Hertzberg,
Cashdollar and Zlochower (1988) comparing 20-L vs. Pittsburgh coal, RoRo93, and aluminum powder. Limits
120-L limit data for gases, by Cashdollar and Chatrathi were determined versus energy levels with the goal of
(1992), comparing 20-L vs. 1-m3 MEC data for dusts, establishing the appropriate 20-L ignition energy that
and by Bartknecht (1989) and Siwek (1988), comparing yields data equivalent to 1-m3 data.
20-L vs. 1-m3 data for dusts. Bartknecht (1989) did not
report the effect of different energy sources in the 20-L
vessel, but did note that the LOC results from a 10-kJ 2. Experimental
ignitor in a 20-L vessel were 1.6 times lower than those
found in a 1-m3 vessel with the same ignition energy. The 20-L dust explosibility data were obtained in the
This difference was attributed to the energy of the PRL 20-L laboratory chamber (Cashdollar & Hertzberg,
ignition source affecting the entire vessel volume and 1985) shown in Figs. 1 and 2. The near-spherical
not acting like a point source. Siwek (1988) extended chamber is made of stainless steel and has a pressure
this work by considering the effect of various lower rating of 21 barg. The hinged top is attached with six
energy ignitors in the 20-L vessel. He noted that the 10 19-mm diameter bolts which are not shown. Strain gauge
kJ ignitor gave much higher LOCs in the 1-m3 vessel in pressure transducers measured the explosion pressure.
comparison with those found in the 20-L vessel. Using The data were collected by a high speed personal com-
an ignition energy of 2.5 to 1 kJ in the 20-L vessel
brought its LOC into better agreement with the 1-m3
values. Siwek also compared the MEC values for 16
dusts measured with 10 kJ ignition in both 20-L and 1-
m3 vessels. The issue of overdriving in a 20-L vessel
and its relation to ignition source energy for MEC testing
has also been discussed by Chawla, Amyotte and Pegg
(1996).
The current testing was performed in a 20-L chamber
at the NIOSH Pittsburgh Research Laboratory (PRL),1
located near Pittsburgh, PA and in the 1-m3 vessel
located at Fike Corporation in Blue Springs, MO. Both
vessels have been used extensively for dust and gas test-
ing. This paper reports on the new comparative MEC
tests for RoRo93 (a tetramethylpiperidine derivative
used for a round robin test in 1993), lycopodium, and
iron powder and comparative LOC tests for gilsonite,

1
The Pittsburgh Research Laboratory was part of the U.S. Bureau Fig. 2. Horizontal cross-section of Pittsburgh Research Laboratory
of Mines before its transfer to NIOSH in October, 1996. 20-L chamber.
 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219
puter (PC) based data acquisition system. The dust to be
tested can be placed either in the dust reservoir or on
top of the dispersion nozzle at the bottom of the chamber
(Fig. 1). After the dust and igniter have been placed in
the chamber, the top is bolted on and the chamber is
partially evacuated to an absolute pressure of 0.14 bara.
Then a short blast of dry air (0.3 s duration from a 16-
L reservoir at 9 barg) disperses the dust and raises the
chamber pressure to about 1 bara. The ignitor is acti-
vated after an additional delay of 0.1 s. This results in
a total ignition delay of 0.4 s from the start of dispersion
until ignition. The experimental dust concentration
reported in this paper is the mass of dust divided by the
chamber volume, i.e. the nominal dust loading.
The Fike Corporation 1-m3 chamber (Figs. 3 and 4)
was also used to measure dust explosibilities. The 1-m3
chamber is spherical with an internal diameter of 1.22
m and a wall thickness of 9.5 mm. It has a pressure
rating of 21 barg. The two halves of the sphere are con-
nected by 12 bolts of 51 mm diameter. Two variable
reluctance pressure transducers were used to measure the
explosion pressure. Data from the instruments were col-
lected by a high speed PC based data acquisition system.
The dust injection system for the 1-m3 chamber con-
sists of a 5-L dispersion reservoir, a 19-mm pneumati-
cally activated ball valve, and a rebound nozzle (Fig. 3).
In previous work (Cashdollar & Chatrathi, 1992), a ring
nozzle was used. To create a dust cloud, a weighed sam-
ple of dust is placed in the dispersion reservoir. The res-
ervoir is pressurized with dry air to 20 bara and the
chamber is partially evacuated to 0.88 bara. Activation
Fig. 3. Vertical cross-section of Fike 1-m3 vessel.
211
of the ball valve disperses the dust and air into the 1-
m3 chamber through the rebound nozzle and raises the
chamber pressure to about 1 bara. The ignitor is fired
0.6 s after activation of the ball valve. The reported
experimental dust concentration for the 1-m3 chamber is
the mass of dust divided by the vessel volume.
The dispersion time and measured Kst values (and pre-
sumably the turbulence level) in the Fike 1-m3 chamber
are comparable to those in European 1-m3 chambers
(Bartknecht, 1989). This is the turbulence level in VDI
Standard 3673, ISO Standard 6184/1, and ASTM Stan-
dard E1226 used to determine the maximum rate of
pressure rise of a dust explosion. The Kst and turbulence
levels in the PRL 20-L chamber are lower, but this
should not significantly affect measurements of the MEC
or LOC (Cashdollar & Chatrathi, 1992). The main effect
of increased turbulence at low dust concentrations is to
make the dust cloud more difficult to ignite (Amyotte,
Chippett & Pegg, 1989). However, with the strong igni-
tors used for the tests, the somewhat higher turbulence
level in the 1-m3 chamber should have little effect on
the measurements.
The ignition sources used for the tests were electri-
cally activated chemical ignitors manufactured by Fr.
Sobbe of Germany. The ignitors are composed of 40%
zirconium, 30% barium nitrate, and 30% barium per-
oxide. They are activated electrically with an internal
fuse wire and deliver their energy in about 10 ms. The
Sobbe ignitors are available in energies of 0.25, 0.50,
1.0, 2.5, 5.0 and 10.0 kJ. These are nominal calorimetric
energies based on the mass of pyrotechnic powder in
212 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219

Fig. 4. Photograph of Fike Corporation 1-m3 vessel.

each ignitor. The 5000-J ignitor by itself produces a
pressure rise of about 0.5 bar in the 20-L chamber but
only about 0.01 bar in the 1-m3 chamber. Physical and
chemical properties of the dusts are shown in Table 1.
The gilsonite is an asphaltic material mined in Utah.
The bituminous coal is from the Pittsburgh seam; this
dust has been used for decades as a standard test dust
(Rice & Greenwald, 1929; Cashdollar, Sapko, Weiss &
Hertzberg, 1987). RoRo93 was distributed worldwide by
A. Kuhner AG, of Switzerland, in 1993 as a round robin
test material for Pmax and KSt testing. RoRo93 is a
2,2,6,6-tetramethylpiperidine derivative (light stabilizer).
Lycopodium clavatum (reticulate form) is a plant spore
obtained from the Meer Corporation. The iron is a
minus-325-mesh powder. The aluminum was Alcoa ato-
mized aluminum powder, grade 123. The size distri-
butions were determined from a combination of sonic
sieving data, Coulter counter data, and laser diffraction
particle size data.
3. MEC data and discussion
In a previous report (Cashdollar & Chatrathi, 1992)
in this series, the MECs of three dusts were evaluated
in the PRL 20-L chamber and in the Fike 1-m3 chamber
following the test procedures in ASTM E1515. The
MEC values for gilsonite dust and bituminous coal dust
were measured in each chamber at several ignition ener-
gies. Chemical ignitors with energies from 0.5 to 10 kJ
were used in the tests. These results, given as the top
two dusts in Table 2, indicated that the 20-L chamber
may be overdriven with high energy ignitors. The MEC
values measured in the 20-L chamber with 2.5-kJ igni-
tors were comparable to those measured in the 1-m3
chamber with 10-kJ ignitors. At higher ignition energies
in the 20-L chamber, there was evidence of overdriving.
The explosibility of anthracite coal was also studied in
the two chambers, but the data are not listed in the table.
The anthracite did not ignite at 2.5 kJ, but appeared to

Table 1
Physical and chemical properties of fuels

Parameter Gilsonite Pittsburgh coal RoRo93 Lycopodium Iron Aluminum

Surface mean diameter, Ds

19 30 12 26 16 18
(m)
Mass mean diameter, Dw
37 50 38 28 22 24
(m)
Mass median diameter,
28 44 29 28 23 20
Dmed (m)
75 m (%) 91 81 89 100 97 100
20 m (%) 36 16 37 1 - 42
Moisture (%) 1 1 0 3 0 0
Volatiles (%) 84 37 100 92 NA NA
 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219 213

Table 2
Summary of MEC testing results (g/m3)

Dust 20-L vessel 1 m3 vessel

1 kJ 2.5 kJ 5 kJ 10 kJ 2.5 kJ 5 kJ 10 kJ

Gilsonite 505 355 305 304 393 413 363

Pittsburgh coal 905 8010 6010 5010 905 855 805
RoRo93 363 284 255 344 354 354
Lycopodium 454 305 412 422 422
Iron 25030 20040 21010 1955 1955

ignite at 5 kJ in the 20-L chamber. It did not ignite even

at 30 kJ in the larger 1-m3 chamber.
The previous study (Cashdollar & Chatrathi, 1992)
has now been extended to two additional carbon-based
dusts as well as to a metallic dust with the goal of evalu-
ating the extent of the overdriving phenomenon more
thoroughly. The organic dusts were selected to be non-
coal in order to widen the study. RoRo93 was selected
as it has been widely studied in recent years; lycopodium
was chosen due to its universal acceptance as an explos-
ibility standard (because of its uniform size). Explos-
ibility tests using RoRo93 were done with 2.5, 5, and
10-kJ ignitors in both chambers. The results are shown
in Figs. 5 and 6.

Fig. 6. MEC data for RoRo93 from the 1-m3 chamber.

In these and subsequent figures only the data points

and curves for one or two ignitors will be shown for
purposes of clarity. The top portion of each graph shows
the maximum absolute explosion pressure plotted
against dust concentration. The effect of the ignitor is
partially corrected by subtracting the pressure rise due
to the ignitor itself from the maximum explosion press-
ure. The lower portion shows the maximum rate of
pressure rise, normalized by the cube root of the
chamber volume, (dP/dt)V1/3. When tested at the stan-
dard turbulence level of ASTM E1226, this is known as
the KSt value. The value (dP/dt)V1/3 is proportional to
the maximum flame speed (Amyotte et al., 1989; Hertz-
Fig. 5. MEC data for RoRo93 from the 20-L chamber. berg & Cashdollar, 1987; Hertzberg, Cashdollar & Zlo-
214 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219

chower, 1988). The primary criterion for flame propa-

gation in the 20-L tests was a 1 bar pressure rise,
corrected for the pressure rise of the ignitor itself. How-
ever, an additional criterion of a pressure rate of rise of
1.5 barm/s was also used in evaluating the 20-L results
(Cashdollar & Chatrathi, 1992). Using these criteria, the
MEC for RoRo93 with a 2.5-kJ ignitor in the 20-L
chamber was found to be 36 g/m3. The results shown in
Table 2 show the change in MECs at higher ignition
energies. The criterion for flame propagation in the 1-
m3 chamber is 1 bar pressure rise or an absolute pressure
of 2 bara. Based on this criterion, the MEC for RoRo93
is 35 g/m3 with a 10-kJ ignition source in the 1-m3 ves-
sel. There was no significant pressure rise at lower con-
centrations while the pressure continued to rise at higher
concentrations. The experimental MEC values, given in
Table 2, show little dependency on ignition energy in
the 1-m3 chamber.
Lycopodium testing was performed at 2.5 and 5 kJ in
the 20-L vessel and at 2.5, 5, and 10 kJ in the 1-m3
vessel. The results are shown in Figs. 7 and 8 for one
ignition energy and the summary data are given in Table
2. The 41 g/m3 MEC for the 1-m3 vessel was clearly
independent of ignition energy. The 20-L vessel showed
an energy dependency and gave an MEC of 45 g/m3 at
2.5 kJ and 30 g/m3 at 5 kJ.

Fig. 8. MEC data for lycopodium from the 1-m3 chamber.

For the RoRo93 and lycopodium, as well as for gil-

sonite and bituminous coal, the 2.5-kJ MEC data in the
20-L chamber agreed better with the 10-kJ data in the
1-m3 chamber. At higher ignition energies, there was no
evidence of overdriving in the 20-L chamber.
The iron results are shown in Figs. 9 and 10 and are
listed in Table 2. The MEC was 195 g/m3 at 10 kJ in
the 1-m3 vessel. In the 20-L chamber, the MEC was
200 g/m3 at 5 kJ and 250 g/m3 at 2.5 kJ. In this case,
the 5-kJ ignitor data in the 20-L chamber agreed better
with the 1-m3 data using the 10-kJ ignitor.
The effect of ignition energy on MEC measurement
in the 20-L and 1-m3 chambers was studied in a previous
report (Cashdollar & Chatrathi, 1992). The results of
those tests along with the current results for RoRo93,
lycopodium, and iron powder are summarized in Table
2 and Fig. 11, where the measured or apparent MEC
is plotted versus ignition energy. The pattern observed
previously for gilsonite and Pittsburgh coal in the 1-m3
was again seen with the three new dusts. That is, the
asymptotes are nearly vertical and the measured MECs
from the 1-m3 vessel are essentially independent of
ignition energy over the range studied. The 20-L tests
did not, however, show the same independence of
ignition energy. As the energy increased from 2.5 kJ, the
Fig. 7. MEC data for lycopodium from the 20-L chamber. apparent MEC decreased and was definitely less than the
 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219
Fig. 9. MEC data for iron powder from the 20-L chamber.
1-m3 results for the carbonaceous dusts. This is a result
of overdriving the smaller vessel with too strong an
ignition source. For the carbonaceous dusts, the closest
agreement to the 1-m3 data was found with a 2.5-kJ igni-
tor in the 20-L vessel.
The iron results appear to be somewhat different in
that the 20-L MEC at 2.5-kJ ignition energy was higher
than the 1-m3 MEC. It must be noted that these MEC
values are three to five times higher than those found
with the carbon-based fuels. The data in Fig. 11D show
that for the more difficult-to-ignite dusts (such as this
iron powder) with higher MEC values, the use of a 5-
kJ ignitor in the 20-L chamber may be more appropriate
for agreement with the 1-m3 data at 10 kJ.
Siwek (1988) compared the MEC values measured
with a 10-kJ ignitor in 20-L and 1-m3 vessels. He con-
cluded that there was no significant effect of vessel vol-
ume on the measured MECs, i.e. that the MECs from
the two vessels agreed to within one concentration
increment. However, in his MEC tests, the dust concen-
tration increment was 10 g/m3. A close examination of
the data in Fig. 8 of his paper, however, suggests that
the MECs for six of the 16 dusts studied were between
10 and 30% lower in the 20-L vessel than in the 1-m3
vessel. Two of the dusts were less than 10% higher and
the rest were about the same. This suggests that some
215
Fig. 10. MEC data for iron powder from the 1-m3 chamber.
overdriving was occurring, at least for some of the dusts.
Another difference between Siweks data and the data
presented in this paper is that he reports the MEC as the
highest concentration that does not produce an explosion
rather than as the lowest concentration that just produces
an explosion.
In conclusion, the MEC data presented in this paper
show that overdriving can occur in the 20-L vessel using
5-kJ or 10-kJ ignitors. For the dusts tested, the MEC
data using 2.5-kJ or 5-kJ ignitors in the 20-L vessel
agreed best with the MEC data from the 1-m3 vessel
using a 10-kJ ignitor. The extent of overdriving in the
20-L vessel is dependent on the type of dust. There was
less of an effect (Fig. 11) for the RoRo93, gilsonite, and
lycopodium, all of which have a high volatility. The
overdriving in the 20-L vessel was greatest for the coal
dust, which had a much lower volatility.
4. LOC data and discussion
All of the LOC testing in the 1-m3 vessel was conduc-
ted with the 10-kJ ignition source and the 20-L tests were
conducted mainly with 2.5- or 5-kJ ignition sources. The
criteria for ignition were unchanged from the MEC test-
ing. LOC testing also incorporated the effect of fuel con-
216 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219
Fig. 11. Effect of ignition energy on apparent MEC.
centration on the measurement. In effect, the fuel con-
centration was varied in order to determine the lowest
possible LOC at any concentration, i.e. the worst case.
The oxygen concentration was reduced by adding nitro-
gen to the air. This was done in both test vessels as well
as in the dust injection gas reservoirs for the 20-L and
1-m3 chambers. This approach eliminated any question
about the actual O2 concentration in the initial stages of
dispersion and ignition.
The LOC data for the RoRo93 dust are shown in Fig.
12. Tests were made over a range of dust concentrations
in order to determine the LOC at the worst case. The
ignitions/explosions are shown as the solid data symbols
and the nonignitions/nonexplosions are shown as the
open data symbols. The 1-m3 chamber data using 10-kJ
ignitors are shown in the top part of the figure and the
20-L data using 2.5-kJ ignitors are shown in the bottom
of the figure. In the 1-m3 chamber, the RoRo93 dust
ignited and burned at 11% oxygen but not at 10% oxy-
gen. The reported LOC value is then taken as the average
of these values or 10.5%. In the 20-L chamber, the
RoRo93 also ignited at 11% but not at 10% oxygen, so
the LOC value is also 10.5%. In this case the LOC value
determined in the 20-L chamber using a 2.5-J ignitor
was in agreement with the data from the 1-m3 chamber
using 10-kJ ignitor. All of the LOC data are summarized
in Table 3.
The 20-L data for Pittsburgh coal in Fig. 13 clearly
show that the effect of the ignition source is significant
in this volume. As ignition energy increases, lower quan-
tities of fuel and oxygen are required to create 1 bar
overpressures. A comparison of the 14% oxygen and 150
g/m3 coal coordinates on the three 20-L graphs provides
a good example of the effect of ignition source. With a
1-kJ ignition source, the 14% oxygen and 150 g/m3 point
is outside the flammability envelope. With a 2.5-kJ
ignition source, this point is just inside the flammable
envelope and with a 5-kJ ignition source, this point is
well within the flammability envelope. Increasing the
ignition energy increases the size of the flammability
zone. As listed in Table 3, the LOC in the 20-L chamber
decreased from 13.5% at 1 kJ to 11% at 2.5 kJ and 9.5%
at 5 kJ. In the 1-m3 chamber, the LOC is 13.5% at 10 kJ.
 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219 217

Fig. 12. LOC data for RoRo93.

Table 3 Fig. 13. LOC data for Pittsburgh coal.

Results of LOC testing (% O2)

Dust 20-L vessel 1-m3 vessel LOC value was found at a very high dust concentration
of 1000 g/m3.
1 kJ 2.5 kJ 5 kJ 10 kJ
Comparing the LOCs of aluminum and RoRo93 meas-
RoRo93 10.5 8.5 10.5 ured in the 20-L chamber with the LOCs measured in
Pittsburgh coal 13.5 11 9.5 13.5 the 1-m3 chamber leads to the conclusion that 2.5 kJ is
Gilsonite 10.5 8.5 11.5 an appropriate ignition source for the 20-L chamber. The
Aluminum 9.5 8.5 9.5 2.5-kJ ignition source appears to neither overdrive nor
underdrive the aluminum and RoRo93 systems. How-
ever, comparing the gilsonite and Pittsburgh coal LOCs
The results of the gilsonite tests are shown in Fig. 14 leads to the conclusion that 2.5 kJ does overdrive these
and illustrate the effect of concentration on the LOC. dusts in the 20-L chamber. The 1-m3 LOCs for gilsonite
For example, the 20-L data show an ignition at 11% O2 and Pittsburgh Coal are 11.5% and 13.5%, respectively,
at 100 and 200 g/m3 but failure to ignite and burn at 300 compared with the 20-L LOCs of 10.5% and 11% at
g/m3. The dust did not ignite at 10% O2 at any concen- 2.5 kJ.
tration tested. The data show that the LOC in the 20-L All of the LOC data reported here support the con-
chamber with a 2.5-kJ ignitor was 10.5% compared to clusion reached by Siwek (1988), i.e. that the 10-kJ
11.5% measured in the 1-m3 with a 10-kJ ignitor. ignition source is inappropriate for LOC measurement
As shown in Fig. 15, the LOCs for the aluminum dust in the 20-L vessel. Furthermore, the data also indicate
were found at much higher dust concentrations than for that the 5-kJ ignition source is inappropriate for LOC
the carbonaceous dusts. This shows the importance of measurement in the 20-L chamber, for the dusts studied.
LOC testing over a wide range of dust concentrations. Both the 10-kJ and the 5-kJ sources will overdrive the
The measured LOCs for this aluminum were 9.5% using explosion in the 20-L vessel for most dusts. The results
a 10-kJ ignitor in the 1-m3 and 9.5% at 2.5-kJ and 8.5% reported here (Table 3) and in Siweks paper (figure 20
at 5-kJ in the 20-L chamber. In both cases, the lowest in Siwek, 1988) indicate that the 1-kJ or 2.5-kJ ignition
218 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219
Fig. 14. LOC data for gilsonite.
source may be the most appropriate ignition source to
use for LOC measurement in the 20-L vessel to match
the LOC data from the 1-m3 vessel using a 10-kJ ignitor.
However, there may be some dusts for which overly con-
servative LOC values will be obtained in the 20-L
chamber with the 2.5-kJ ignitor.
Siwek (1988) recommended that a 5-kJ ignition
source should be used for dusts with LOC values below
10%. However, neither the current study nor Siweks
work investigated dusts with LOC values significantly
below 10%. This is an area of LOC measurement that
needs further study. If it is assumed that dusts with LOC
values below 10% are highly reactive and the effect of
the ignition source is to either overdrive or underdrive
these systems, then it might be expected that a 5-kJ
ignition source would more likely overdrive these sys-
tems. Additional data need to be gathered for dusts with
LOC values below 10% to establish a clear recommen-
dation.
To obtain unambiguous and practically usable LOC
values, the 1-m3 chamber is preferred, especially when
LOC values are below 10%. The 20-L LOC values may
be somewhat more uncertain, but the 2.5-kJ data from
the 20-L vessel appear to be conservative, at least for
LOC values above 10%.
In applying the LOC data, one should use a reasonable
Fig. 15. LOC data for aluminum.
safety margin. NFPA 69 recommends keeping the oxy-
gen concentration at least 2% below the measured LOC
value when protecting equipment. It is important to
remember that the LOC values listed in Table 3 are only
for the specific dusts tested and may not be applicable
to other particle sizes of the same materials. Often, finer
sizes of dusts have lower LOC values. It is also
important to recognize that these LOC data are for nitro-
gen inerting of air. Inerting with other gases such as car-
bon dioxide may give different results.
5. Conclusion
The data from this study and the previous work
(Cashdollar & Chatrathi, 1992) demonstrate that over-
driving can occur when using strong chemical ignitors
in the 20-L chamber. The result is that apparent MEC
values are found which are lower than the true values.
For most dusts tested, the best agreement is found
between 20-L chamber data with 2.5-kJ ignitors and 1-
m3 data with 10-kJ ignitors. Overdriving is not a concern
when testing with 10 kJ in the 1-m3 vessel. The advan-
tage of the 20-L chamber is that the explosion tests can
be conducted more quickly and with much smaller dust
samples. In practice, therefore, the 2.5-kJ ignitor is rec-
 J.E. Going et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 209219
ommended for initial testing in the 20-L vessel. For
hard-to-ignite dusts with higher MEC values, a 5-kJ igni-
tor may be more appropriate. If there are significant dif-
ferences in the MEC values obtained at 2.5-kJ and 5-kJ
energies, it may be advisable to go to the 1-m3 vessel
for a final MEC determination. If the dust does not ignite
with a 2.5-kJ ignitor, but does ignite with a 5-kJ or 10-
kJ ignitor in a 20-L vessel, it is necessary to use a 1-m3
vessel with a 10-kJ ignitor for the final determination.
Similar conclusions can be drawn regarding overdriv-
ing the 20-L vessel when making LOC measurements.
These and previous data indicate that both 5- and 10-kJ
energies are too strong for the 20-L chamber. From the
studies carried out to date, the 2.5-kJ ignitor is the most
appropriate energy level for 20-L LOC testing, and in
fact, the 2.5-kJ data may be slightly conservative for
some dusts. Additional LOC studies are needed, parti-
cularly for dusts with LOC values below 10%.
Acknowledgements
The authors wish to acknowledge the assistance of
G.M. Green and PRL in the collection of the 20-L data
and Frankie Terry, Thomas Klug, and Jack Creager of
Fike in the collection of the 1-m3 data.
219
References
Amyotte, P. R., Chippett, S., & Pegg, M. J. (1989). Effect of turbulence
on dust explosions. Progress in Energy and Combustion Science,
14, 293310.
Bartknecht, W. (1989). Dust explosions: course, prevention, protec-
tion. New York: Springer.
Cashdollar, K. L., & Chatrathi, K. (1992). Minimum explosible dust
concentrations measured in 20-L and 1-m3 chambers. Combustion
Science and Technology, 87, 157.
Cashdollar, K. L., & Hertzberg, M. (1985). 20-Liter explosibility test
chamber for dusts and gases. Review of Scientific Instruments, 56,
596602.
Cashdollar, K. L., Sapko, M. J., Weiss, E. S., & Hertzberg, M. (1987).
Laboratory and mine dust explosion research at the Bureau of
Mines. In Industrial dust explosion, STP 958. Philadelphia, PA:
ASTM pp. 107123.
Chawla, N., Amyotte, P. R., & Pegg, N. J. (1996). A comparison of
experimental methods to determine the minimum explosible con-
centration of dusts. Fuel, 75, 654658.
Hertzberg, M., & Cashdollar, K. L. (1987). Introduction to dust
explosions. In Industrial dust explosions, STP 958. Philadelphia,
PA: ASTM pp. 532.
Hertzberg, M., Cashdollar, K. L., & Zlochower, I. A. (1988). Flamm-
ability limit measurement for dust and gases. In 21st Symposium
(International) on Combustion (pp. 303313). Pittsburgh, PA: The
Combustion Institute.
Rice, G. S., & Greenwald, H. P. (1929). Coal dust explosibility factors
indicated by experimental mine investigations 1911 to 1929.
Bureau of Mines Technical Paper 464.
Siwek, R. (1988). Reliability determination of the safety characteristics
in 20-L apparatus. Flammable dust explosion. St. Louis, MO.
 Journal of Loss Prevention in the Process Industries 13 (2000) 221228
www.elsevier.com/locate/jlp

Experimental research on explosibility at high initial pressures of

combustible dusts
E. Conde Lazaro *, J. Garca Torrent
Laboratorio Oficial Madariaga. Universidad Politecnica de Madrid, C/Alenza 1-2. 28003 Madrid, Spain

Abstract

In recent years some explosibility experiments were carried out by LOM (Spain) to look into the behaviour of the combustible
dusts at hyperbaric pressures. In this way, different aspects were studied, such as the effect of initial pressure on the explosion
development, the influence of the effective present turbulence or preventative existing techniques such as inertization.
This paper contains a summary of the different tests performed, which covered a range of initial pressure from 1 to 15 bar.
Several fuels were used within the works, and their main features are described to facilitate understanding the results. 2000
Elsevier Science Ltd. All rights reserved.

Keywords: Dust explosion; Explosibility; Turbulence; Inerting; High initial pressure

1. Introduction normal explosibility studies to high initial pressures, and

The ongoing growth of the world population and the
continuous industrial development is causing a strong
increase in energy consumption, which has to be com-
patible with the preservation of the natural environment
and the clever use of global resources.
There are still significant reserves of fossil fuels to
exploit with inappropriate characteristics for conven-
tional power plants, for example, the Spanish subbitum-
inous coals of low heat value and high sulphur content.
Also, the use of renewable fuels, regarded previously as
low quality combustibles, such as agricultural and forest
residues, is now being promoted by the European Union
for the production of heat and electricity.
However, the new emerging technologies have been
conceived providing for that problem. Technologies like
the Pressurized Fluidized Bed Combustion (PFBC), or
the Integrated Gasification Combined Cycle (IGCC)
operate at elevated pressure to improve the process
efficiency, and incorporate methods to reduce polluting
emissions.
These new processes make necessary the extension of
* Corresponding author. Tel.: +34-1-442-1366; fax: +31-1-441-
9933.
E-mail address: econde@dse.upm.es (E. Conde Lazaro).
to combustibles such as low rank fossil fuels, renewable
fuels and mixtures of them.
It is well known that combustible solids with a fine
particle size can undergo a very rapid combustion, pro-
ducing a pressure wave and a flame front capable of
destroying equipment and installations as they propa-
gate. Fire and explosion risks are less intuitive in solids
handling than in liquids or flammable gases, although
the risk is well known and fought against since the
beginning of the century. Very strong explosions
occurred in the early years (Cybulski, 1975; Bartknecht,
1989; Eckhoff, 1991), leading to specific research and
promotion of the development of prevention and protec-
tion techniques, which have been improving as time
passes.
2. Influence of the initial pressure on the explosion
There are several characteristics defining the nature of
solid substances and their behaviour related to inflam-
mability and explosibility. These properties can be
experimentally determined by measuring some charac-
teristic parameters and establishing certain quantitative
indices.
Explosion severity of a substance can be determined
by recording the variation of pressure against the course

0950-4230/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 9 9 ) 0 0 0 4 7 - 9
222 E. Conde Lazaro, J. Garca Torrent / Journal of Loss Prevention in the Process Industries 13 (2000) 221228
of time when an explosion is generated in a closed test-
ing vessel. Two parameters can be defined upon the
curve, the maximum explosion pressure (MEP) and the
maximum rate of pressure rise (MRPR).
When the initial pressure is elevated, the explosion
severity rises with the initial pressure value, following
a multiplicative law.
2.1. Literature review
Research work about explosions at hyperbaric initial
pressure is limited and, it is normally referred to low
initial pressures. Bodurtha (1980) states that for gases
the relationship between the MEP and the initial pressure
keeps constant, according to the following law:
Pi
MEPMEPa
Pa
where:
Pi is the initial pressure at which the explosion
commences
Pa is the atmospheric pressure
MEPa is the value of the pressure peak at atmos-
pheric pressure
MEP value of the maximum explosion pressure at
the initial pressure Pi.
Palmer (1973) propounds the following model for the
MEP for a coal dust (with a 38% volatile content), based
on experimental studies covering up to 3 bar absolute of
initial pressure:
MEPMEPa 10.75Pi
Pa
where MEP, MEPa, Pi and Pa have the same meaning
as in Eq. (1).
Also Bartknecht (1981) and Wiemann (1987) carried
out experiments at high initial pressures (0.22 bar
absolute with starch for Bartknecht and 14 bar absolute
with coal dust for Wiemmann), finding a proportinal
relation between the MEP and the initial pressure. Later,
Borisov and Gelfand (Eckhoff, 1991) found a similar lin-
ear correlation up to very high initial pressures,
approaching 100 bar.
With regard to the pressure rate, Bodurtha (1980)
states that for gases it rises linearly as initial pressure is
increased, according to the following expression:
dP
dt max
APiB
where: dP/dt is the calculated value of the maximum rate
of pressure rise at Piand A and B are constants that are
different for each kind of material (A=0.63; B=17.2 for
propane. A=1; B=68.9 for hydrogen).
For the pressure gradient Palmer (1973) establishes
the following relation for coal:
dP
AP0.5
i B (4)
dt max
where A and B are constants.
Wiemann (1987) observed a linear relationship
between (dP/dt)max and the initial pressure, although for
high values of initial pressure the gradient started to
level out.
2.2. Experiments and results
In order to study the effect of initial pressure on the
explosion characteristics, a series of tests was carried out
(1)
at four different initial pressures (1, 5, 10 and 15 bar)
over a bituminous coal (35.2% volatiles, 6.4% ashes)
partially inerted with a 36% weight of limestone. The
results obtained are summarized in Table 1.
The test apparatus consists of a 1 m3 spherical vessel
connected to a 25 l dust dispersion vessel through a fast
acting detonator valve. After pressuring the main
chamber to the desired initial testing pressure, and the
dispersion vessel to the desired injection pressure, the
connecting valve is opened to disperse the dust into the
main chamber. The fast acting valve is an elbow shaped
device wich incorporates a mushroom lock, commanded
by an electrical detonator, that hampers the communi-
cation between the recipients. When the detonator is acti-
vated, the lock moves back communicating both recipi-
ents, and then the sample is injected into the main vessel
due to the flow created by the overpressure existing in
(2)
the dust dispersion vessel. A rebound nozzle is coupled
in the inner wall of the sphere to improve the dispersion
process. At a predetermined delay time after the begin-
ning of the dispersion (normally fixed in 600 ms) two
chemical ignitors, with a total energy of 10,000 J,
located at the centre of the sphere, are activated to
initiate the explosion, and then the explosion course is
recorded and processed. Fig. 1 shows a photograph of
the testing facility.
These values of maximum explosion pressure and
pressure gradient were obtained by varying the concen-
Table 1
Bituminous coal (36% limestone) severity parameters at different
pressures
Initial pressure (bar) MEP (bar) MRPR (bar/s)
(3) 1
5
9.7
43.7
155
354
10 84.0 434
15 117.7 521
 E. Conde Lazaro, J. Garca Torrent / Journal of Loss Prevention in the Process Industries 13 (2000) 221228
Fig. 1. Photograph of the 1 m3 sphere.
tration in air of the coallimestone mixture in wide
increasing intervals with the initial pressure rise. Fig. 2
shows the variation of MEP versus concentration for dif-
ferent initial pressures. As it can be observed, the
maximum value corresponds to higher concentrations
and the explosible interval includes higher concen-
trations as initial pressure increases.
Fig. 3 illustrates the tendency of MEP and MRPR with
initial pressure showed by the same material. It can be
noticed in the MRPR trend a slight levelling out at high
initial pressures, that may be explained by the high
Fig. 2. Variation of MEP versus concentration.
223
increase of dust density inside the testing vessel implied
by the achievement of high dust concentrations at such
pressures. Excessive closeness of particles may play a
negative role in the progression of the explosion reac-
tion, lessening the potential value of MRPR.
A simple regression of the data collected in Table 1
provides the following prediction of the explosion press-
ure values as a function of initial pressure, according to
a linear model of the Y=a+bX kind with a correlation
coefficient of 0.998:
MEP3.8277.735Pi (5)
where MEP and Pi have the same meaning as in Eq. (1).
For the pressure rate, a multiplicative model of the
Y=aXb gives a correlation coefficient (0.998) higher than
the linear model (0.959):
MRPR159.43P0.44
i (6)
3. Explosion prevention: inerting
An experimental study was developed over several
subbituminous Spanish coals used in a demonstration
PFBC power plant belonging to the electrical utility
ENDESA, situated in Escatron (Zaragoza, Spain), that
has an operating pressure of 12 bar (Torrent & Menen-
dez, 1993).
The aim of the work was to investigate the influence
of inertization, both adding inert matter to the combust-
ible (in this case the limestone used for the sulphur reten-
tion is also an inhibitor of the explosion) and reducing
the oxygen content of the operating air, and to determine
224 E. Conde Lazaro, J. Garca Torrent / Journal of Loss Prevention in the Process Industries 13 (2000) 221228

Fig. 3. MEP and MRPR increase with initial pressure.

the levels of limestone in the combustible mixture and Table 3

oxygen in air that assured the avoidance of explosion Results obtained in a 21% oxygen atmosphere
risks. Limestone (%) 10 20 30
Table 2 shows the volatile content and the severity Sample MEP Kst MEP Kst MEP Kst
parameters at atmospheric pressure of four samples, that
covered the whole range of variability of the coals. The 001 86.8 378 80.2 320 78.6 146
Kst value, which is called the specific material constant, 002 98.2 524 87.3 375 83.4 284
003 106.4 483 86.1 352 79.2 270
is a mathematical transformation that corrects the effect 004 104.4 753 82.2 297 77.8 257
of the volume of the testing vessel (V) on the MRPR
value:

Kst
dP
dt max
V1/3 (7)
Tables 35 show the results registered for the same
four coals mixed with different percentages of limestone
in atmospheres differing in oxygen content. In all the
tests the initial pressure was 12 bar and the dust concen-
As may be seen, these samples have a high volatile con- tration varied between 5000 and 7000 g/m3 for the
tent, especially if they are refererred to a dry ash free maximum values. The test apparatus was the 1 m3 sphere
(d.a.f.) basis. The coals had a size smaller than 75 m. described before and the inerting gas was nitrogen.
The explosion parameters were obtained in a 20 l sphere It can be observed at a glance that when an explosion
facility (Cesana & Siwek, 1995) and relate to concen- takes place at a high initial pressure such as 12 bar, very
trations ranging from 500 to 750 g/m3 for the high values are registered for explosion pressure (over
maximum values. 100 bar in some cases) and for rate of pressure rise (over
700 bar/s at times). This gives rise, in turn, to Kst read-

Table 2 Table 4
Features of the samples Results obtained in a 19% oxygen atmosphere

Sample Volatile Volatiles MEP (bar) Kst (bar/m/s) Limestone (%) 10 20 30

matter (%) d.a.f.(%) Sample MEP Kst MEP Kst MEP Kst

001 30.3 53.5 6.8 39 001 74.2 194 78.5 185

002 41.4 60.2 7.4 63 002 97.2 416 87.0 311 87.3 339
003 44.5 58.5 7.6 91 003 96.6 596 89.8 390 79.8 203
004 47.9 61.4 7.8 108 004 96.0 390 102.6 567 93.0 436
 E. Conde Lazaro, J. Garca Torrent / Journal of Loss Prevention in the Process Industries 13 (2000) 221228 225

Table 5 Kst variation of sample 002 with different limestone per-

Results obtained in a 17% oxygen atmosphere centages at various concentrations of oxygen in air.
Limestone (%) 10 20 30 From observation of the Tables and Figures, it can be
Sample MEP Kst MEP Kst MEP Kst stated that explosibility decreases, in general, as lime-
stone content increases and the percentage of oxygen in
001 air falls. Although this underlying trend may appear
002 97.4 382 90.3 323 66.1 75
clear, some deviations may be observed among the
003 90.6 309 85.5 264 86.8 239
004 96.0 395 93.9 322 90.3 361 results (see for example, in Fig. 5, for sample 002 with
a 30% limestone content, the Kst increase at 19% oxygen
concentration from the 21% oxygen percentage value),
ings much higher than those corresponding to atmos- probably due to the difficulties entailed by dispersion at
pheric pressures. high initial pressures (see earlier comment).
Fig. 4 represents the trend of MEP for the samples An extended work carried out with seven coals (the
with different limestone contents and Fig. 5 shows the four described and three other subbituminous coals) with

Fig. 4. Variation of the MEP for the samples with different sorbent contents.

Fig. 5. Variation of Kst of sample 002 with different limestone contents at distinct oxygen concentrations in air.
226 E. Conde Lazaro, J. Garca Torrent / Journal of Loss Prevention in the Process Industries 13 (2000) 221228

Table 6
Volatile matter and atmospheric explosion parameters of the mixture and components (20 l sphere chamber)

Sample Volatile matter (%) Humidity (%) Ash (%) MEP (bar) Kst (bar/m/s)

Wood dust 81.87 1.65 0.39 7.8 87

Black lignite 46.10 9.85 3.85 7.7 105
Mixture 53.00 9.5 3.1 8.8 104

different levels of limestone content, percentage of oxy-

gen in air and humidity content indicated that oxygen The tests were performed at 5 and 15 bar of initial press-
was the most influential factor on the occurrence or not ure. Table 7 shows the values of injecting pressure for
of an explosion. Limestone and humidity were the main each initial pressure according to Eq. (8).
influential factors on the explosion evolution once the Table 8 summarizes the MEP and Kst values obtained
explosion had taken place (Garca Torrent, 1995). at different initial pressures, divided in two groups: low
turbulence tests (only one charging deposit used) and
high turbulence ones (both deposits used). The atmos-
4. Influence of turbulence on the explosion pheric values are taken from the determinations pre-
viously performed in a 20 l apparatus (see Table 6).
Among the different studies carried out during a These results, in what refers to these particular experi-
research and development project by five European lab- ments, can be considered as low turbulence tests,
oratories, under the EC Joule Programme, aimed at the because they are equivalent to those measured in the 1
analysis of the behaviour of renewable fuels and mix- m3 vessel with one deposit (Cesana & Siwek, 1995). The
tures of coal and biomass, a study of the turbulence values presented in Table 8 correspond to the maximum
effect on the development of the explosion was included. at each pressure over a wide range of dust concen-
The sample selected for this work was a mixture con- trations.
stituted by black lignite (75%) and wood dust (25%). The data collected in Table 8 can be graphically
Table 6 displays the volatile matter and severity para- depicted to provide an approach to the turbulence effect.
meters at atmospheric pressure of the mixture and the Figs. 6 and 7 show the MEP and Kst plots at different
individual components. All the tests were carried out degrees of turbulence.
with the same sample in order to allow comparison of As can be seen, increased turbulence has much more
results, eliminating dispersing effects caused by the nat- effect on the rate of pressure rise than on the explosion
ure of the dust. pressure, and this should be considered for safety meas-
The explosion parameters are strongly affected by the ures design. An increase in turbulence produces a higher
turbulence immediately before and during the explosion. mobility of particles, the access of oxygen to the surface
Turbulence depends strongly on the injection process.
Therefore, turbulence can be affected by different Table 7
aspects such as the delay time between injection and Injection pressure at each initial pressure
ignition, the magnitude of the injecting pressure and the
rate of mass discharged into the vessel, which can be Initial pressure (bar) Injection pressure (bar)
closely related to the number of charging vessels from
5 27.85
which sample is injected. 15 45
From a practical point of view, different turbulence
levels can be created by using one or two dust deposits,
at different initial pressures. An identical second 25 l
Table 8
dust dispersion vessel was used to disperse the dust into Explosion data of the mixture at low and high turbulence
the 1 m3 sphere at high turbulence conditions. The delay
time was constant for all the shots, and the injecting Initial MEP (bar) Kst (bar/m/s)
pressure was constant for each initial pressure. The pressure
injecting pressure was increased as initial pressure was (bar)
Low High Low High
raised to maintain a similar dispersing effect at each turbulence turbulence turbulence turbulence
initial pressure. The relation between injecting pressure
and initial pressure was calculated according to the fol- 1 8.8 104
lowing relationship: 5 49.2 55 336 784
15 126.6 150 524 1369
Pinjection1.714Pinitial19.285 (8)
 E. Conde Lazaro, J. Garca Torrent / Journal of Loss Prevention in the Process Industries 13 (2000) 221228 227

Fig. 6. MEP data versus initial pressure of the mixture at low and high turbulence.

Fig. 7. Kst values versus initial pressure of the mixture at low and high turbulence.

of the dust particles is promoted and then the burning
rate greatly rises.
Thus, it is remarkable that, in Table 8, there is a great
difference between the Kst value at 15 bar at low turbu-
lence (524 bar/m/s) and at high turbulence (1369
bar/m/s). The high turbulence value is 2.5 times the low
turbulence one. This phenomenon is not shown accu-
rately, because in the high pressure turbulence test at 15
bar, the sphere rupture disks burst out because of the
great severity of the explosion, releasing the high rise of
pressure. The Kst is expected to reach these high values
at elevated pressures, which are normal operating press-
ures at pressurized processes.
5. Conclusions
Every process involving the handling of combustible
dusts is exposed to explosion risks. When the process
operating pressure rises over atmospheric conditions, the
severity of the explosion increases following a multipli-
cative law. The maximum explosion pressure is pro-
portional to the initial pressure. In the case of the
explosion pressure gradient, it has been observed that
there is a tendency to level out at high initial pressures.
Furthermore, this can be aggravated by the appearance
of high turbulence situations. It has been found that
increased turbulence has much more effect on the rate
of pressure rise than on the explosion pressure, and this
should be considered for safety measures design. An
increase in turbulence produces a higher mobility of par-
ticles, the access of oxygen to the surface of the dust
particles is promoted and then the burning rate greatly
rises. A series of experiments carried out at initial press-
ures as high as 15 bar, have shown that explosions at
high turbulence can reach Kst values three or four times
higher than low turbulence ones.
A study carried out at an initial pressure of 12 bar,
with seven coals with different levels of limestone con-
tent, percentage of oxygen in air and humidity content
pointed to oxygen as the most influential factor on the
occurrence or not of an explosion. Limestone and
humidity were the main influential factors on the
explosion evolution once the explosion had taken place.
228 E. Conde Lazaro, J. Garca Torrent / Journal of Loss Prevention in the Process Industries 13 (2000) 221228
Explosion tests based on the addition of an inhibiting
substance to the combustible, the inertization of the test-
ing air by reduction of the oxygen content, and the com-
bination of both procedures, have shown progressive
reduction of explosion severity as the inerting level
increases.
Acknowledgements
The study relating to the influence of turbulence, was
financed by the European Commission, under the RTD
Joule Programme. The study is included in a project car-
ried out by a consortium formed by five European lab-
oratories: VTT from Finland, TNO from Holland, DMT
from Germany, INERIS from France and LOM from
Spain. The other research studies were financed by the
Spanish electrical utility ENDESA.
References
Bartknecht, W. (1981). Explosions, course, prevention, protection.
New York: Springer.
Bartknecht, W. (1989). Dust explosions. Berlin: Springer-Verlag.
Bodurtha, F. T. (1980). Industrial explosion prevention and protection.
New York: McGraw-Hill.
Cesana, C.H., & Siwek, R. (1995). Operating instructions for the 20
litre apparatus (V 5.0). Adolf Kuhner AG, Birsfelden, Switzer-
land, August.
Cybulski, W. (1975). Coal dust explosions and their suppression.
National Science Foundation.
Eckhoff, R. (1991). Dust explosions in the process industry. Oxford:
Butterworth-Heinemann.
Garca Torrent, J. (1995). Coal dust explosions in pressurized pro-
cesses. Final report to ENDESA, Madrid, December.
Palmer, K. N. (1973). Dust explosions and fires. London: Chapman
and Hall.
Torrent, J.G., & Menendez, E. (1993). Explosion tests at elevated
initial pressures. Europex Newsletter no. 22, September.
Wiemann, W. (1987). Industrial dust explosions. In L. Kenneth, & M.
Hertzberg, ASTM STP 958 (pp. 3344). Philadelphia: ASTM.
 Journal of Loss Prevention in the Process Industries 13 (2000) 243251
www.elsevier.com/locate/jlp

Recent development of standardization of testing methods for dust

explosion in Japan
a,*
Masaharu Nifuku , Toei Matsuda b, Heiji Enomoto c

a
National Institute for Resources and Environment, Onogawa 16-3, Tsukuba, Ibaraki, 305-8569 Japan
b
National Research Institute of Industrial Safety, Umezono 1-4-6, Kiyose, Tokyo, 204-0024 Japan
c
Tohoku University, Aoba 1, Aoba-ku, Sendai, Miyagi, 980-8579 Japan

Abstract
The Dust Explosion Committee of the Association of Powder Process Industry and Engineering (APPIE) of Japan has acted to
establish the standard for the member of the Association. Two testing methods have been authorized by the Association. The
methods are established to evaluate sensitivity and severity for dust explosion.
The ignition sensitivity is evaluated by the minimum explosive dust cloud concentration. Tapping sieve equipment and a strictly-
specified Hartmann-type explosion tube have been selected as the testing apparatus. The tapping sieve equipment is applied to
measure the minimum explosive concentration by forming a dust cloud supplied from a sieve on the top of the explosion chamber.
The Hartmann-type tube is used to investigate roughly whether the test powder is explosive or not. A lot of explosion tests had
been carried out to standardize the equipment and evaluate the sensitivity. This APPIE standard is being revised.
Along with the ISO standard for explosion severity, Research Institute of Industrial Safety, the Ministry of Labour published a
Guide to Test Method for Explosion Pressure and Rate of Pressure Rise for Combustible Dusts in 1994. Based on this Guide, The
APPIE Committee has made the draft for the member.
These standards and drafts are aimed to be the Japanese Industrial Standard (JIS). No other similar standard exists in Japan. In
the present paper, technical data and explosion statistics which are the background of the APPIE standard for the ignition sensitivity
are to be described. 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Standardization; Dust; Explosions; Severity; Sensitivity

1. Introduction trial safety point of view, Product Liability Law, and

Tens of million tons of powders are being handled in
the processes of transportation, storage, filtration, etc. in
Japan. There are always potential hazards such as fire
and explosion in these processes, because these powders
are mostly combustible. Although a lot of attention is
paid to preventing industrial hazards, dust explosions
and fires, of a large or small scale, sometimes occur in
the industrial processes. Several annual dust explosion
accidents in Japan have been reported in recent years.
This may be much less than those in many other coun-
tries. However, small accidents and troubles still
accompany the powder handling. Therefore, information
about the explosibility of powders is required even more
in powder processing industries in relation to the indus-
* Corresponding author. Tel.: +81-298-58-8591; fax: +81-298-58-
8558.
E-mail address: nifuku@nire.go.jp (M. Nifuku).
others in Japan.
The Dust Explosion Committee of the Association of
Powder Process Industry and Engineering (APPIE) has
acted to establish the standard for the member of the
Association. The ignition sensitivity is evaluated by the
minimum explosive dust cloud concentration. A lot of
explosion tests have been carried out to standardize the
equipments and evaluate the sensitivity. This APPIE
standard is being revised.
There are many explosion characteristics, such as
minimum explosive dust concentration, minimum
ignition energy, ignition temperature, etc., which are
used to evaluate the explosibility of dusts.
For the evaluation of relative explosibility of dusts,
the U.S. Bureau of Mines proposed to use these three
explosion characteristics, mentioned above, together.
Also, the minimum explosive concentration and the
minimum ignition energy have been independently pro-
posed as an index for evaluating relative explosibility.

0950-4230/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 9 9 ) 0 0 0 4 9 - 2
244 M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251

It has been reported that the minimum explosive concen- Table 1

tration gives an equivalent explosibility rating to that Dust explosion statistics
shown by the explosion sensitivity proposed by U.S. Year Number of
Bureau of Mines (Enomoto, 1988).
Many researchers have experienced that both mini- Explosions Deaths Injuries
mum explosive dust concentration and minimum
ignition energy show a similar explosibility rating. 1952 6 7 26
1953 9 1 16
Therefore, concerning the index for evaluation of rela- 1954 9 1 19
tive explosibility, both of them were taken up. For meas- 1955 4 0 0
uring the minimum explosive dust concentration, uni- 1956 7 7 14
formity of dust cloud is most important and the only 1957 4 2 6
parameter to be concerned about. On the other hand, 1958 8 4 18
1959 7 3 9
measurement of minimum ignition energy requires con- 1960 6 0 1
cern about the dust concentration and how to give the 1961 3 0 6
ignition energy, as well as the uniformity of dust cloud, 1962 8 3 23
because the minimum ignition energy experimentally 1963 11 2 30
obtained varies with the dust concentration and the 1964 7 2 9
1965 12 1 41
method to give ignition energy. Furthermore, a physical 1966 6 3 20
meaning of the minimum explosive dust concentration 1967 8 9 39
is being unveiled (Enomoto, 1993). These discussions 1968 12 4 17
led us to use the minimum explosive dust concentration 1969 8 6 11
as the index for evaluating the relative explosibility of 1970 6 7 5
1971 7 2 12
dusts. 1972 7 6 22
Along with the ISO standard for explosion severity, 1973 12 2 53
Research Institute of Industrial Safety, the Ministry of 1974 7 3 10
Labour published a Guide to Test Method for Explosion 1975 9 3 14
Pressure and Rate of Pressure Rise for Combustible 1976 4 0 3
1977 6 2 2
Dusts in 1994, and the APPIE Committee has made the 1978 8 3 9
draft for the member based on it. 1979 9 2 26
These standards and drafts are expected to be the base 1980 3 0 8
for Japanese Industrial Standard (JIS), because no other 1981 2 0 7
similar standard exists in Japan. In the present paper, 1982 3 3 2
1983 3 0 6
technical data and explosion statistics which are the 1984 3 0 0
background of the APPIE standard for the ignition sensi- 1985 5 3 6
tivity are to be described. 1986 8 3 12
1987 1 0 0
1988 3 0 9
2. Explosion statistics 1989 1 0 0
1990 6 0 11
Powder handling industries are facing troubles and 1991 4 1 7
accidents in many situations. There were many, some- 1992 4 3 3
times serious, accidents long ago, but the accidents have 1993 1 1 1
decreased greatly in recent years. There were about 10 1994 6 7 26
1995 6 0 8
cases or less of dust explosion annually between 1950 Totals 269 106 567
and 1970, and around 5 in the 1980s and 1990s for the
total of all industries in Japan. But the casualty numbers
are not so small. Details of recent statistics on dust
explosion are published elsewhere (Matsuda, 1993) and 3. Activities of APPIE committee
brief summaries are shown in Tables 1 and 2.
Here, it should be mentioned that small scale acci- As discussed so far, there are always potential dangers
dents or troubles, which are not usually reported to auth- of explosion and fire in powder handling processes.
orities and not counted in statistics, are still often experi- Also, physical and chemical information for the
enced in many industries. In fact, there are many explosion is strongly requested for industrial safety with
companies visiting the authors to solve the problems. powders that industries produce and handle because of
Therefore, it could be understood that the powder pro- Product Liability Law. Therefore, the safety aspect of
cess industries are still facing hazardous chances of powder has received great concern and whether the pow-
explosion and fire. ders are sensitive or not for explosion is the important
 M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251 245

Table 2 standards such as ISO and ASTM into consideration.

Dust explosion of various materials
Dust type Number of
Explosions Deaths Injuries
Coal 13 7 41
Inorganic 31 9 28
Metals 64 41
Food and feed 46 17
Chemical synthetic 36 12
materials
Intermediate additives 50 13
Cellulosic materials 28 7 84
Miscellaneous 1 0
Totals 269 106
information for industries. This is because handling of
powders has to be carried out with the greatest attention
if the powders are explosive. Also, the suppliers of pow-
ders have to inform the users of the explosibility.
These backgrounds motivated the establishment of a
committee for the standardization of test methods. This
committee was organized under the auspices of APPIE.
After a few years of investigation by the committee,
APPIE has established a standard for the relative evalu-
ation of explosion sensitivity as the first step. The stan-
dard specifies the minimum explosive dust concentration
as the index, because (a) prevention of explosion is the
rigid standpoint of safety, (b) ISO Standard is not appli-
cable for evaluation of sensitivity, (c) standard for evalu-
ation of sensitivity is strongly desired in industries, and
(d) physical meaning of lower explosion limit is being
unveiled and limitation of its application is being made
clear (Enomoto, 1993). But, there has been no official
standard for the measurement of the minimum explosive
concentration in Japan.
There have been many methods developed for
obtaining the minimum explosive concentration. How-
ever, the published data fluctuates widely, depending on
the researchers. This means that many factors, such as
dust cloud formation method, uniformity of dust cloud
and details of equipment, influence their test results.
The committee had then investigated the details of
testing equipments and methods. Many explosion tests
had been carried out to investigate the universality of
each testing equipment. Finally, a tapping sieve method
(Ishihama, 1961; Enomoto & Matsuda, 1986) was taken
as the standard method for measuring the minimum
explosive concentration. For convenience, a Hartmann-
type explosion tube was also accepted, because experi-
mental results showed that the strictly specified Hart-
mann-type explosion tube gave similar results to those
obtained by the tapping sieve method.
As the second step, the committee also discussed the
standard for testing the explosion severity, taking other
The severity is important to alleviate property damages
caused by the explosion. The committee definitely stands
by coordination with widely accepted standards such as
ISO and ASTM standards. A number of explosion tests
were carried out using many test powders as in the case
of sensitivity tests. The committee has established the
153 draft of standard.
109
79
4. Standard for explosion sensitivity
69
4 As described above, the sensitivity is evaluated by the
minimum explosive dust cloud concentration. Two sets
567 of equipment were applied for the standard. They are the
tapping sieve equipment and the strictly specified Hart-
mann-type explosion tube.
The tapping sieve equipment is shown in Figs. 1 and
2. The dust cloud is formed by tapping the sieve on the
top of the explosion chamber. The test powder is con-
tained in the sieve and it falls into the explosion chamber
by the tapping operation and forms a dust cloud in the
chamber. The powder supply is performed and regulated
by the tapping and friction of tapping rod at the periph-
ery of the sieve. Accordingly, the powder is supplied
through all the face of the sieve. This will help to pro-
duce a more homogeneous dust cloud. After the forma-
tion of a dust cloud an electrical discharge (spark) is
produced in the dust cloud. If flame propagation is pro-
duced by igniting the dust cloud for more than 10 cm,
it is evaluated that the dust cloud is capable of causing
a dust explosion. Then the dust cloud concentration is
measured. The dust sampling is proceeded by tapping
the sieve and closing the shutter (Fig. 2(4)). The shutter
is closed when the top face of the dust cloud reaches the
bottom of the explosion chamber. Thus the powder in
the explosion chamber is sampled.
The amount of test powder in the chamber depends
on the (screen) opening, the tapping force, the frictional
force and the tapping rate. The opening of the sieve has
to be selected to proceed smoothly, but not over supply-
ing the powder and avoiding clogging of the sieve. It is,
therefore, very important to choose the optimum opening
of the sieve and to apply appropriate tapping and fric-
tional forces to the rod. Thus, the homogenous dust
cloud will be formed.
Since the explosibility of dust is influenced by the
powder properties, such as particle size, particle size dis-
tribution, dryness of the particle, etc., the precondition-
ing must be done with great care. The sieving and drying
additives to improve dispersion of powder, etc. are regu-
lated in the test method. The supplied powders are
screened by the sieve (JIS Z 8801 Standard Sieve, open-
ing: 75 m) and the under sieve is used as the test pow-
der. If it is requested to test the supplied powder as it
is, the sample is screened by the sieve with 300 m
246 M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251
Fig. 1.
opening and the under sieve is used as the test powder.
It is recommended to dry the sample which contains a
lot of water content or which is absorptive to moisture,
avoiding chemical change of the powder itself. The test
powder is recommended to have a water content of 1
2% or less. The dried test powder is to be contained in
a dessicater. An additive can be mixed in the sample
with poor dispersibility to improve the dust cloud hom-
ogeneity. Fullers earth, light magnesium oxide or sili-
con dioxide are samples of the additive.
The dust cloud concentration is measured by weighing
the dust dispersed in the explosion chamber, divided by
the chamber volume. The ignition is confirmed by the
flame propagation for 10 cm above the ignition point.
The ignition source is provided by the power source of
15 kV and 20 mA output. In the ignition test, it is neces-
sary to pay attention to avoid oxygen suffocation,
because repetitive ignition tests consume oxygen.
The strictly specified Hartmann-type explosion tube is
shown in Figs. 37. Other details are also described in
the standard. Special attention is paid to the dimensions
Tapping sieve equipment.
of dustcup, air reflector, opening between dustcup and
air reflector, compressed air supply pipes, and timing
of ignition.
Attempts were made to produce dust clouds as hom-
ogenous as possible, because the dust cloud does not
usually have sufficient uniformity. Many explosion tests
were carried out to compare the tapping sieve equipment
with the Hartmann-type explosion tube, and between
experimenters. It was found that the value for the mini-
mum explosive concentration is smaller and that the
reproducibility of the data is better than the case of Hart-
mann-type explosion tube. The major reason for these
differences is due to the difference of homogeneity of
dust cloud between the two. Therefore it is rec-
ommended that the Hartmann-type tube is used for
roughly investigating the degree of sensitivity, not for
obtaining the minimum explosive concentration because
of experimental simplicity. However, it is indicated that
similar results could be obtained if careful attention is
paid to the Hartmann-type tube. For example, the mini-
mum explosive dust concentration of standard lyco-
 M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251
Fig. 2. Explosion chamber.
podium for calibration obtained with the tapping sieve
equipment is 40 g/m3, whereas that obtained with the
specified Hartmann-type explosion tube is 45 g/m3.
In other words, this indicates the possible application
of the Hartmann-type explosion tube even for somewhat
precise evaluation of the explosibility. The published
data so far from the Hartmann tube is not consistent and
it may be possible to mention that the specified Hart-
mann-type explosion tube will produce more reliable
data.
To calibrate the testing equipments, it is recommended
to use both the lycopodium and the 50/50 mixture of the
lycopodium and talc. It is noted that the equipment has
the normal performance if the minimum explosive con-
centration of lycopodium obtained is between 3545
g/m3, and 8595 g/m3 in the case of the mixture of lyco-
podium and talc. The standard lycopodium and talc are
supplied by APPIE.
The lycopodium was selected as the standard powder,
because its dispersibility is excellent, and its chemical
composition and size distribution are known. Also,
another important reason is that the minimum explosive
concentration is adequately low.
It is also necessary to calibrate the testing equipments
at higher dust cloud concentrations, because some pow-
ders are difficult to ignite at such a low concentration.
Since it was hard to find the appropriate sample, an
attempt was made to mix lycopodium with inert sub-
247
Fig. 3. Stricly specified Hartmann-type explosion tube.
stances. The inert materials used and their median diam-
eter are shown in Table 3. The ignition tests were carried
out with both testing equipments.
Fig. 8 shows the relation between the mixing ratio of
inert material and the minimum explosive concentration
obtained with the tapping sieve equipment. Fig. 9 also
shows the same relation, using the strictly specified Hart-
mann-type explosion tube.
It can be understood that the data fluctuation is small
in the case of lycopodium-talc mixture. Also, the lyco-
podium-talc mixture has a larger explosion limit concen-
tration than other cases and the minimum explosive con-
centration obtained with both testing equipments are
roughly the same value.
The 50 wt% mixing ratio of inert material was chosen
because the minimum explosive concentration is appro-
priately high, that is ca. 100 g/m3 and the mixing ratio
is rather simple (the weight ratio is 1:1). It should be
also noted that even if the mixing ratio is not very accur-
ate due to the inadequate mixing operation for example,
the minimum explosive concentration will still stand at
around 100 g/m3.
The results of the explosion test are classified into 3
classes: high explosibility is for 35 g/m3 or less, moder-
ate explosibility between 35 and 100 g/m3, and low
explosibility 100 g/m3 or more.
248 M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251

Fig. 4. Explosion tube.

Fig. 5. Dustcup.

The reason for this classification is somewhat unclear,
because these classes are specified basically by experi-
ences. However, there are some supports for this classi-
fication. For example, coal dust with volatile content of
25% shows a clear difference in the development of
explosion. This implies that the coal dust with less than
25% of volatile content can be ignited only with a rela-
tively large ignition source such as flames of a match and
torch. And the minimum explosive dust concentration of
coal dust with 25% volatile content is around 100 g/m3.
Also powders with smaller ignition energy, such as poly-
ethylene, polypropylene and coal dusts with higher vol-
 M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251 249

Fig. 6. Air reflector, holder for air reflector and pipe connector.

Table 3 5. Standard (draft) for explosion severity

Particle size of inert materials

Material Median diameter (m) The explosion severity is evaluated by the maximum
explosion pressure and explosion index (Kst) using ISO
silica sand 2731 1 m3 equipment, or calibrated 20 or 30 liter explosion
talc 7.29.2
fly ash 1317
spheres. The 30 liter equipment is applied from the oper-
Kanto loam 2731 ational simplicity point of view. Major parts of the test-
ing method specified in the draft are along the ISO stan-
dard. Also, the testing method established by the
Research Institute of Industrial Safety contributed to
Table 4 the draft.
Explosion severity The 20 liter sphere is the one described in ASTM
E1226. The 30 liter sphere is modified from the 20 liter
Explosion class Kst (barm/s) Explosion severity
sphere. Fig. 10 shows the brief description of the 30
St 0 0 0 (non explosive) liter sphere.
St 1 1200 weak The testing method of the 30 liter sphere is based on
St 2 201300 strong the ISO 1 m3 equipment and 20 liter sphere. It is also
St 3 300 extreme requested to calibrate the 30 liter sphere using lyco-
podium.
The explosion index (Kst) is classified into 4 classes
as shown in Table 4.
atile content than 45%, which are regarded as having
strong explosibility, have the minimum explosive con-
centration of about 3040 g/m3 or less and can be ignited
with an electric spark of short duration (Ishihama, 6. Conclusion
Enomoto, Komai, Umezu, & Zhang, 1983). This may
mean that dusts, the minimum explosive dust concen- Since there is no official standard to test dust explos-
tration of which is less than 40 g/m3, can be ignited with ibility in Japan, both the standards (draft) of testing
sparks caused by electrostatic discharge. method for ignition sensitivity and explosion severity are
250 M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251

Fig. 7. Setting of dustcup, air reflector and pipe connector.

Fig. 9. Relation between the mixing ratio of inert material and the
Fig. 8. Minimum explosive concentration of the mixture using the minimum explosive concentration of the mixture by strictly specified
tapping sieve equipment. Hartmann-type explosion tube.
 M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251 251

Fig. 10. 30 liter explosion sphere.

contributing or expected to provide acceptable explos- Safety Engineering, 27 (1), 3136.

ibility information for powders.
However, international coordination for the ignition
sensitivity will be necessary in the future as it is for the
explosion severity. Also some modification of ignition
source for explosion severity test will be necessary,
because the chemical igniter specified in ISO Standard
needs handling permission in Japan.
References
Enomoto, H. (1988). A discussion on the evaluation for the explos-
ibility of dusts (in Japanese). Journal of the Japan Society for
Enomoto, H. (1993). Dust explosions Fundamental Concept. Pro-
ceedings of the Second IUPAC-Workshop on Safety in Chemical
Production, pp. 248255.
Enomoto, H. & Matsuda, T. (1986). Dust explosibility in pneumatic
systems. In Cheremisinoff (Ed.), Encyclopedia of Fluid Mechanics
(Vol. 4), Solids-Gas Flow (pp. 563609). Gulf Pub. Co.
Ishihama, W. (1961). Studies on the lower critical explosive concen-
tration of coal dust cloud. Presented at 11th International Confer-
ence of Directors of Safety in Mines Research, Poland.
Ishihama, W., Enomoto, H., Komai, T., Umezu, M., & Zhang, L. J.
(1983). Studies on the ignition of dust clouds Part 2. Journal of
the Japan Society for Safety Engineering, 22 (1), 2531.
Matsuda, T. (1993). Dust explosion incidents in Japan. Proceedings of
the Second IUPAC-Workshop on Safety in Chemical Production,
pp. 256261.
 Journal of Loss Prevention in the Process Industries 13 (2000) 299309
www.elsevier.com/locate/jlp

Modeling of maize starch explosions in a 12 m3 silo

*
Shengjun Zhong, Xufan Deng
Industrial Explosion Protection Institute, Northeastern University, Box 327, Shenyang, Liaoning Province 110006, PR China

Abstract

This paper presents a 2-dimensional numerical model of EulerianLagrangian multi-phase combustion flow to predict maize
starch explosions in a 12 m3 silo. The flow field after ignition, flame propagation velocity and pressure development histories etc.
during the explosion, are calculated. The data of non-uniform initial conditions including dust concentration, flow velocity and
turbulent RMS velocity in the silo for this model are adopted from Hauert, Vogl and Radandt (1994) [Hauert, F., Vogl, A., Radandt,
S. (1994). Measurement of turbulence and dust concentration in silos and vessels. 6th international colloquium on dust explosions
(pp. 7180), Shenyang, China, August 28September 2, 1994.]. A simple concept of dust granule taking into consideration dust
dispersion efficiency is proposed and introduced. The Lagrangian method is used to trace trajectories and granules, so it is easier
to consider particle size distribution. The ke model is used to simulate the turbulence of the gas phase, and the particles pulsation
is modeled by random vector wind generated by the surrounding gas. In the combustion model, vaporization of water, volatilization
of volatile, gas phase reaction and the particles surface reaction are taken into account. 2000 Elsevier Science Ltd. All rights
reserved.

Keywords: Dust explosions; Modeling; Explosion reaction engineering

1. Introduction shape of particles. For example, the composition of

Dust explosion protection is usually designed accord-
ing to a series of standard experimental dust explosion
characteristcis that describe explosion severity and sensi-
tivity. In real cases however, the intial and boundary
conditions and explosion development in industrial
equipment are very complicated and usually different
from laboratory tests (Deng et al. 1993, 1994; Krause,
Kasch & Gebauer, 1996).
In an industrial application, maximum explosion
pressure Pmax and explosion index Kmax are two
important dust explosion characteristics of pressure
development, and widely used for explosion risk analysis
and explosion protection.
There are many factors that affect explosion pressure
development in an enclosure, such as:
the physical and chemical properties of dusts such as
their chemical composition, size distribution and
* Corresponding author.
E-mail addresses: iepi@mail.neu.edu.cn (S.
dengxf@mail.neu.edu.cn (X. Deng).
organic dusts includes volatile, solid carbon, water
and ash content is very important;
geometry of the enclosure;
initial conditions before ignition including tempera-
ture, pressure, gas and dust flow velocity, turbulence,
dust concentration and dispersion efficiency of the
dust particles in the enclosure.
The effect of turbulence on the rate of explosion
pressure rise has been investigated for many years.
Kauffman et al. (1984), Amoytte and Pegg (1989) and
Pu, Jarosinski and Johnson (1990) have studied the
effects of RMS velocity on the rate of maximum press-
ure rise in 0.95 m3, 1.2 L, and 20 L apparatuses respect-
ively. Hauert et al. (1994) have measured the turbulence
and dust concentration in a 1 m3 vessel and a 12 m3 silo.
Gieras, Klemens and Wolanski (1996) have studied the
relationship between RMS velocity and maximum press-
ure Pmax, maximum rate of pressure rise and flame
propagation velocity in a 1.25 m3 vessel. Smirnov and
Nikitin (1996) have simulated mixture evolution and
Zhong), combustion in confined turbulent flows. However, there
has been more attention to turbulence than dust disper-

0950-4230/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 9 9 ) 0 0 0 5 1 - 0
300 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309

Nomenclature
A pre-exponential factor
B temperature exponential factor
Cd drag force coefficient
CV specific heat of constant volume, kJ/(kgK)
Ds mass transfer diffusion coefficient of specie s, m2/s
E specific mixture energy, J/kg, activation energy, J/mol
Ep energy flux from particulate phase, J/(m3s)

fd gravity acceleration constant, kgm/s2

G gravity acceleration constant, kgm/s2
G mass flux between phase interface of species, kg/s
g mass flux rate per unit area between phase interface, kg/(m2s)
h specific enthalpy, J/kg


I unit tensor of second order
Is mass flux of diffusion, kg/(m2s)
Iq heat flux due to diffusion, J/(m2s)
Is heat flux due to radiation, J/(m2s)
Jq heat flux due to thermal conduction, J/(m2s)
k kinetic turbulent energy, J/kg
k momentum flux from a particle, kgm/s2
Mp momentum flux from particulate phase, kg/(m2s2)
m mass of a particle and its composition, kg
mp mass of a particle
mv mass of vapor decomposed from a particle, kg
mc mass of carbon in a particle, kg
mw mass of water in a particle, kg
p pressure, Pa
Q rate of heat transfer, J/(sm3)
q reaction heat, latent heat or decomposition heat, J/kg
Qsr rate of heat change owing to surface reaction, J/(sm3)
Qcr rate of convective heat transfer, J/(sm3)
Qrd rate of radiation heat transfer, J/(sm3)
Sp net mass flux from particle phase, kg/(m3s)
T temperature, K

u velocity vector of gas phase, m/s
ws net reaction rate of specie s in gas phase, kg/(m3s)
W molecular weight, kg/kmol
Ys mass fraction of sth specie in gas phase

Greek symbols

a volume fraction of gas phase

as mass fraction of specie s in Sp
r density of gas, kg/m3
rp physical density of particle, kg/m3
rp dust concentration, kg/m3


t stress tensor of second order, N/m2
l thermal conductivity coefficient, J/(smK)
s Prandt/Schmidt number; Stefan Boltzmann constant, 5.67108 J/(sm2k 4)
m viscosity, kg/(ms)
n kinetic viscosity, m2/s
e turbulent energy dissipation rate, J/(kgs); radiation emissivity
 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309 301

Superscript

T matrix transverse rate of change

Subscripts

s species in gas phase

p particle properties
r radiation properties
l laminar properties
t turbulent properties, by EBU turbulence combustion model
w water, wall (surface) of particle
c carbon
v volatile
far away from particle surface, gas condition

sion efficiency in both the experimental and numerical 2. Mathematical model of dust explosions
research.
For dust dispersion efficiency, Smoot and Horton 2.1. Governing equations of the gas phase
(1977) found that the incomplete dispersion of dusts gen-
erating the differences of burning velocities, and poor The following equation system is obtained by Favre
dispersion efficiency in the Hartmann Bomb are reasons averaging of the conservation equations of multi-compo-
for the lower rate of explosion rise (Amoytte & Pegg, nent and multi-phase reacting flow (Zhou, 1994, pp.
1989). However, in a numerical simulation of dust 196204, 209210; Smirnov & Nikitin, 1996); the deno-
explosions, dust dispersion efficiency has not been con- tations of Favre averaging are not represented here in
sidered, even though dust dispersion efficiency has a order to simplify the description.
close relationship with explosion characteristics and tur- Mass conservation equation of sth specie:
bulent kinetic energy. (arYs)
In order to scale up characteristics of dust explosions (arYs u )IsasSpws, (1)
t
from laboratory to industrial practice theoretically,
Explosion Reaction Engineering (ERE) was rec- where the last two items are mass flux from the particu-
ommended (Deng et al. 1993, 1994). The merit of ERE late phase and net reaction rate of the sth specie in the
is to use theoretical methods involving numerical mode- gas phase respectively.
ling based on conservation law and transport phenomena By summing all the mass conservation equations of
equations, to predict the severity or sensitivity of species, the mass conservation equation of the gas phase
explosions in equipment, to analyze explosion risk and is as follows:
to optimize the equipment design for explosion preven- (ar)
tion and mitigation. ERE will enable safety engineers to (aru )Sp. (2)
t
more easily use theoretical methods for the prevention
and mitigation of explosions. The model presented in
this paper is part of the research wok of ERE. The momentum equation considering the momentum
In this paper, the maize starch explosions in the 12 transfer between gas and particle phases is:
m3 silo are simulated taking into consideration the fol-
lowing: (aru )
(aru u )(ap) t Mp. (3)
t
dust granule to consider dispersion efficiency;
Eddy Break Up model for volatile combustion; The energy equation is as follows:
vaporization of water from particles;
radiation in the gas phase and between the (arE)
(aru E)(apu )(IqIs) (4)
gas/particle phases. t
302 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309

2.3. Modeling of particles motion in turbulent flow

( t u )Ep,

where E is the specific mixture energy, The drag force f d is expressed as:
1

f d Cdarpd 2p(u u u p)u u u p,
T (13)
u 2
8
E Ys( cV,sdTh0,s) k, (5)
s
2
T0 where u is the turbulent velocity of the surrounding gas,
where k is turbulent kinetic energy; energy flux Iq con- which can be described by the following expression
sists of two parts, conductive heat flux Jq and enthalpy according to the definition of k (Smirnov & Nikitin,
flux S shsIs owing to mass flux: 1996):

ui (u2 1/2
i ) xi, (14)
IqJq hsIslT hsIs. (6) where

s s
1/2
2
i )
(u2 1/2
k , i1,2,3,
3
2.2. Gas phase flux modeling and ke model

and x is a random vector assumed to be with the Gauss

Radiation heat flux Is is modeled by the radiative heat probability distribution function (PDF) for each of its
transfer coefficient lr for the calculation of radiation heat components.
transfer combined with heat conduction (Hjertager, Assume that the probability of a pulsations change
1999): on the interval t to be expressed:

l r
16 s 3
3 Cr
T , (7)
1
Pt1exp t ,
t (15)

where Cr is the Rosseland decay coefficient. k

where t=min(tt,tR), tt= , is the living time of the turbu-
In the gas phase equations, diffusive mass flux Is, e

4d 2prp
stress tensor t and conductive heat flux Jq can be mod- lent eddy, and tR= is the relax time of a particle.
3CdRem
eled as follows:

IsarDsYsar Dl,s vt
ss
Ys, (8)
2.4. Model of water vaporization, volatilization and
surface reaction

2 2 Let mp denote the total mass of a maize starch particle,
t ar(vlvt)(u (u ) u I ) ark I , T
(9) which includes volatile mv, solid carbon mC, water mw
3 3
and ash, ma.

Jq lllr
arvt
st T. (10) 2.4.1. Water vaporization
It is assumed that water diffuses continuously to the
surface of the particle then evaporates from the particle
The turbulent kinetic viscosity vt=cmk 2/e, where e is surface via a vapor film to the surrounding air, so the
the rate of turbulent energy dissipation. model for vaporization of a liquid droplet is adopted
k,e is determined by ke model equations: (Deng et al., 1993):

t

(ark)(aru k) ar
vl+vt
sk
k Gk (11) dmw
dt
l
pdpNu ln[1B],
Cp
(16)

are, where
e


vl+vt
(are)(aru e) ar e (ce1Gk (12) YwYg
t se k B= when (TpTw)
1Yw
ce2are), ,
Cp(TgTw)

B= when (TpTw)
qw
where Gk= t :u is the turbulent energy generation item,
and cm, ce1 and ce2 are constants. YwBwexp(Ew/RTp), (17)
 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309 303

where Yw and Yg are the mass fractions of water at the 1 DsNu

vaporization film and the gas atmosphere respectively. where K= , k d= is the mass transfer coef-
1 1 dp
Tw is the boiling temperature of water, qw the potential +
kd kR

heat of vaporization, Cp and l are the heat capacity and
Es
the conductivity coefficient of water respectively. ficient, kR=AsT Bsexp is the reaction coefficient,
RTw
and Nu*=2.0+0.6Re0.5Pr0.33 is the Nusselt number of
2.4.2. Volatile flux
mass transfer.
The volatilization rate can be expressed as
Then the rate of change of carbon is calculated by:
dmv
mvAvexp(Ev/RTp). (18) mCGO2/bO2GCO2/bCO2 (26)
dt

2.5. Single particle equation system

2.4.3. Surface reaction
The chemical reactions on the surface of a maize par- A single particle has the following characteristics:
ticle are complicated. Near the particle surface, the con-
centration of O2 is relatively low, therefore the oxidation {mp,mv,mc,mw, x p,u p,Tp}, where x p is the position of
is not complete. The main reactions are considered as the particle.
follows: The equation system to model a particles mass,
momentum and energy conservation referring to velo-
C0.5O2CO (19) city, position and temperature is:
CCO22CO (20) dmp
mwmvmc, (27)
The combustion of CO will be considered in the gas dt
phase reaction.
First we consider the mass flux of O2 and CO2 of a du p
mp f dmpG, (28)
single particle due to surface reaction and mass transfer, dt
GO2, GCO2. A coefficient b is introduced to represent the
relation of mass flux: dx
u p, (29)
GO2 0.5WO2 GCO2 WCO2 dt
bO2 , bCO2 , (21)
GC,O2 WC GC,CO2 WC where f d is resistance force to a particle during moving.
where GC,O2, GC,CO2 is the mass flux of solid carbon from dTp
the reaction with O2 and CO2 respectively. mpCp QsrQvQwQcvQrd, (30)
dt
The mass flux of O2 and CO2 due to mass transfer
can be expressed as where

GspdparbsDNuln Ys,Ys,w
bs+Ys,w
. (22)
QsrGC,O2qC,O2GC,CO2qC,CO2,
Qvmvqv,
(31)
(32)
Qwmwqw (33)
In many cases,
QcvpdpNul(TTp), (34)
Ys,Ys,w
bs+Ys,w
1, Ys,w0, ln 1
Ys,Ys,w 1
bs+Ys,w
Ys,.
bs (23) Qrdpd 2pes(T 4T 4p),
where the Nusselt number of heat transfer Nu=2.0+0.6
(35)

Re0.5Pr0.33.
According to surface reaction chemical kinetics,

2.6. Dust granule to consider the dispersion
Es efficiency
Gspd arYs,AsT exp
2
p
Bs
. (24)
RTw
It is known that dusts in equipment are not always
Mass flux of O2 and CO2 can be expressed as con- fully dispersed, and granules consisting of a number of
sidering both reaction and mass transfer processes: undispersed dust particles always exist in a dust/air mix-
ture.
GsKpd 2parYs,, sO2, CO2, (25) A dust granule is a cluster of real dust particles, when
304 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309

-
Total mass, momentum and energy flux are
the stress t of the surrounding gas is strong enough, the expressed as:
granule will be split into many small particles, otherwise
mpmCmvmw, (42)
it can be considered as a big single particle. Parameter
ng is the number of real particles contained in a granule.

The number of real particles comprising a granule, and k mp u F d, (43)

the mass percentage of the granules depend on the level

of turbulent kinetic energy. eQmp(0.5u 2)F du . (44)

In order to reduce the computation and memory
capacity, dust granules are considered as model particles.
Parameter Nm is the number of real particles or granules
to match the model particle.
For the ith model particle representing a group of
where F d is the sum of f d in the cell.
By summing all the fluxes from the particles in a cell
divided by the cell volume, we can obtain the source
term of the gas phase equations:

real particles:
1 1
mtotal,i S p (N )(m ), Mp (N )( k ), E (45)
Nm,i , (36) Vi1,N m i Vi1,N m i p
mreal,i
where mtotal,i is the total mass of a group of real particles
that are represented by the ith model particle, and mreal,i

1
(N )(e )
Vi1,N m i
is the mass of a single real particle.
For the ith particle representing a group of granules:
The volume share is obtained by:

mtotal,i
Nm,i , (37) 1 p
mgranule,i a1 (N ) d 3 (46)
Vi1,N m 6 p
where mtotal,i is the total mass of granules that is rep-
resented by the model particle. mgranule,i is the mass of a where N is the number of model particles in a considered
dust granule in the ith model particle. cell, and V is the volume of the cell.
If the granules represented by the model particle are
split and fully dispersed, Nm will be changed and then 2.8. Combustion model of the gas phase
the diameter of the model particle will be the diameter
of the real single particle instead of a granule. It is assumed that CH4 and CO extracted from volatile
Much work should now be done to give a reasonable and CO generated by surface reaction of solid carbon
criteria stress which is enough to split the dust granule. combusting with oxygen in the gas phase. The heat
A simple method is to assume that the probability of its release will contribute to the temperature rise of the
split is a function of turbulent kinetic energy. gas phase.
For the first step, it is proposed that some percentage The two following reactions without an inverse reac-
of dust is well dispersed and that the rest is considered tion are considered:
as dust granules with a certain diameter, according to
the initial turbulent kinetic energy. 2COO22CO2 (47)

2.7. Modeling flux between phases, closing the source CH42O2CO22H2O (48)
terms of the gas phase equations
The eddy break up (EBU) model is adopted to con-
The total flux of gas components from a single particle sider the turbulence effect on combustion rate. The
is the sum of the fluxes from the surface reaction, volatil- efficient reaction rate can be expressed as (Smirnov &
ization and vaporization of the particle: Nikitin, 1996; Hjertager, 1999)
mCH4av,CH4mv, (38) wsmin(ws,R,ws,t), (49)
mO2GO2av,O2mv, (39) where ws,R is the Arrhenius reaction rate and ws,t, the
mCO2GCO2av,CO2mv, (40) reaction rate determined by turbulent mixing.

mCO
WCO
G
2WCO
G a m .
WO2 O2 WCO2 CO2 v,CO v
(41) wCH4,R(ar)3YCH4Y2O2A3T B3exp E3
RT
(50)

where av,s is the mass fraction of the sth component

in volatile.
wCO,R(ar)3Y2COYO2A4T B4exp E4
RT
(51)
 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309 305

Table 1
Size distribution of the maize starch particles

Diameter, m 30 (20a) 3050 (40a) 60100 (80a)

Volume share 20% 60% 20%

a
Average diameter used in the calculation.

the initial flow field parameters such as dust concen-

tration, flow velocity and RMS velocity are obtained
from Hauert et al. (1994).

3.1. The initial conditions in the 12 m3 silo

The 12 m3 silo is of industrial scale with an L/D ratio

of 3, as shown in Fig. 1. The pneumatic vertical feeding
rate of 3 kg/m3 from the silo top is selected for this
Fig. 1. Schematic diagram of the 12 m3 silo (Hauert, Vogl & Rad- simulation. Initial flow parameters measured by Hauert
andt, 1994).
et al. (1994) are shown in Fig. 2. Unfortunately in Hauert
e et al. (1994), the dust concentration at 3 kg/m3 mass flow
wCH4,tCEBUar min(YCH4,YO2,YCO2,YH2O) (52) rate is not presented, therefore this simulation uses a dust
k
concentration at 7 kg/m3 divided by 1.5, according to
e Hauert et al. (1994). It is evident from Fig. 2 that:
wCO,tCEBUar min(YCO,YO2,YCO2) (53)
k
The maximum mean vertical component velocity
where CEBU is a constant.
and RMS turbulent velocity is near the axis (approx.
700 mm from the wall). The RMS velocity decreases
3. Results of the simulation with decreasing distance to the wall. At the bottom
half of the silo, flow recycling can be observed.
The simulation of maize starch explosions in the 12 The maximum maize starch concentration is also
m3 silo uses the formal equations from (1) to (53) and near the axis. In positions above 3.75 m from the bot-

Fig. 2. The initial conditions in the 12 m3 silo (vertical feeding) measured by Hauert, Vogl and Radandt (1994). (a) The initial gas mean flow
velocity and RMS velocity at 3 kg/m3 dust mass flow rate. (b) The dust concentration at 7 kg/m3 dust mass flow rate.
306 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309
tom, dust concentration decreases with decreasing
distance to the wall and is very low near the wall. In
positions below 3.75 m, dust concentration decreases
from 800 mm to 600 mm and increases from 400 mm
to the wall owing to the gas flow circulation near the
bottom of the silo.
The initial particle size distribution is listed in Table 1.
3.2. The gas temperature and turbulence fields after
ignition
The initial turbulence energy (calculated from RMS
velocity), flow velocity and dust concentration fields
after reprocessing are shown in Fig. 3.
Dust ignition occurs at time zero and then the silo is
closed (no more feeding), and only the cylindrical part
of the silo is considered. The ignition is modeled by an
energy discharge with a specified distribution with space
and time at 1 m from the bottom.
The temperature and turbulent kinetic energy fields at
9.9 ms, 40.0 ms, 80.3 ms and 120.2 ms after ignition
are shown in Figs. 47.
Comparing Fig. 4(b) and Fig. 3(a), it is evident that
turbulence generated by the flame front is greater than
the existing turbulence before ignition. From Fig. 5, it
can be observed that the flame front near the axis of the
cylinder drapes because of the cool effect of a high dust
concentration. Figs. 5, 6 and 7 show that the maize dust
flame propagates along the axis in the 12 m3 silo with
a relatively slower flame propagation velocity near the
axis.
Fig. 3. The initial turbulent kinetic energy, gas mean velocity and dust concentration in the 12 m3 silo at time=0 ms after ignition (dust flow
rate=3 kg/m3).
3.3. The pressure development after ignition
The explosion pressuretime profiles of their different
points along the axis (top, center and bottom) are shown
in Fig. 8, and the time begins from ignition. After that
time, there are some differences because blast waves
make a different fluctuation of pressure in different pos-
itions. After 250 ms the pressure is about 0.85 MPa, and
the maximum rate of pressure rise is 10.6 MPa m1 s1.
3.4. The distribution of temperature along the axis
The temperature distribution along the axis at different
times is shown in Fig. 9. The flame front locates where
the maximum gradient of temperature is found. After
maize starch is ignited at a position 1 m from the bottom,
the flame propagates towards the bottom and upwards
with increasing temperature. Behind the flame front,
there is a slight decrease of temperature.
3.5. The flame trajectories along the axis
A flame trajectory along the axis is obtained by find-
ing the maximum gas temperature gradient of a tempera-
ture profile along the axis. The flame trajectories at dif-
ferent radius positions along the axis in the 12 m3 silo
are shown in Fig. 10.
In general, the flame first accelerates at about 60 ms
after ignition and then the it propagates more slowly
before about 90 ms because the top of the silo is closed.
The flame front at the axis seems to have a jump at about
 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309 307

Fig. 4. The gas temperature (a) and turbulent kinetic energy (b) field in the 12 m3 silo at 9.9 ms after ignition.

Fig. 5. The gas temperature (a) and turbulent kinetic energy (b) field in the 12 m3 silo at 40.0 s after ignition.

75 ms because the flame front at the neighboring pos- 4. Conclusion and discussion
ition drapes and encloses some unburned mixture.
The average flame propagation velocity is about We have simulated maize starch explosions and pre-
2.91 m/s. dict the explosion development in an industrial scale silo
308 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309

Fig. 6. The gas temperature (a) and turbulent kinetic energy (b) field in the 12 m3 silo at 80.3 ms after ignition.

Fig. 7. The gas temperature (a) and turbulent kinetic energy (b) field in the 12 m3 silo at 120.2 ms after ignition.

under non-uniform inital conditions. It will be interesting
to compare the predicted result with further explosion
experiments.
Owing to the complexity of explosions it is difficult
to use empirical measures to predict the explosion devel-
opment in industrial equipment well. However, it will
be possible by using the numerial modeling method:
to verify the model, CFD code and a series relevant
parameters by means of laboratory scale experiments;
to obtain the initial conditions in industrial equipment
by measurement or pre-calculation;
to model the explosion development in the industrial
scale equipment and to verify it with a large scale
experiment.
 S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309
Fig. 8. The pressuretime profile at different positions along the axis
in the 12 m3 silo.
Fig. 9. The temperature profile along the axis in the 12 m3 in different
10 (about 10 40 80, 120 ms).
Fig. 10. The flame trajectory at different radius positions along the
axis in the 12 m3 silo. (PosiR0 0 m from the axis; PosiR1 0.28
m from the axis; PosiR2 0.56 m from the axis; PosiRs wall).
In this simulation it is also found that turbulence gen-
erated by the flame front is much larger than the initial
turbulence. The mechanism of the relationship between
turbulent RMS velocity, dust dispersion effect and Kmax
should be investigated in detail.
In the multi-phase turbulent combustion model, dis-
persion efficiency must be considered. The proposed
309
dust granule concept is a simple approach to handling
this problem. However a lot of further work needs to
be done.
Acknowledgements
The research work of this paper is a part of the contri-
bution to the Explosion Reaction Engineering, which is
involved in the Project No. 29376228 supported by the
National Natural Science Foundation of China. Mr. Shi
Jianye is greatly acknowledged for giving valuable
advice to this paper.
References
Amoytte, P. R., & Pegg, M. J. (1989). Lycopodium dust explosion in
a Hartmann Bomb: effect of turbulence. Journal of Loss Prevention
in the Process Industry, 2.
Deng, X., Xu, R., & Xie, L. (1993). Explosibility and ignitability of
16 types of dust and some opinions on research for dust explosions.
In Proceedings of the 5th ICDE, Warsaw, April 1993 (pp. 219).
Deng, X., & He, J. (1994). The latest development of explosion reac-
tion engineering. In 6th international colloquium on dust
explosions, Shenyang, China, August 28September 2, 1994 (pp.
96115).
Gieras, M., Klemens, R., Wolanski, P. (1996). Evaluation of turbulent
burning velocity for dust mixtures. Proceedings of an international
symposium on hazards, prevention and mitigation of industrial
explosions, & 7th ICDE (pp. 5.35-51), Bergen, Norway, June 23
26, 1996.
Hauert, F., Vogl, A., & Radandt, S. (1994). Measurement of turbulence
and dust concentration in silos and vessels. In 6th international
colloquium on dust explosions, Shenyang, China, August 28Sep-
tember 2, 1994 (pp. 7180).
Hjertager, B. H. (1986). Three dimensional modeling of flow, heat
transfer, and combustion. In Handbook of heat and mass transfer
(pp. 10331350). Houston, TX: Gulf Publishing Company Chap-
ter 41.
Kauffman, C. W., Srinath, S. R., & Tezok, F. I. et al. (1984). Turbulent
and accelerating dust flame. In Proceedings of 20th symposium
(international) combustion, Pittsburgh.
Krause, K., Kasch, T., Gebauer, B. (1996). Velocity and concentration
effects on the laminar burning velocity of dustair mixtures. Pro-
ceedings of the international symposium on hazards, prevention
and mitigation of industrial explosions and 7th ICDE (pp. 5.114),
Bergen, Norway, June 2326, 1996.
Pu, Y., Jarosinski, J., & Johnson, V. G. (1990). Turbulence effect on
dust explosion in the 20-liter spherical vessel. In 23rd symposium
(international) on combustion, the Combustion Institute, Pittsburgh
(pp. 843849).
Smirnov, N.N., Nikitin, V.F. (1996). Dustair mixtures evolution and
combustion in confined and turbulent flows. Proceedings of an
international symposium on hazards, prevention and mitigation of
industrial explosions and 7th ICDE (pp. 5.5266), Bergen, Nor-
way, June 2326, 1996.
Smoot, L. D., Horton, M. D., & Williams, G. A. (1977). Propagation
of laminar pulverized coal-air flames. Proceedings of the 16th sym-
posium (international) on combustion. The Combustion Institute
(pp. 375387).
Zhou, L. (1994). The theory research and numerical modeling of tur-
bulent gas-particle flows and combustion. Beijing: Science Press
of China.
 Journal of Loss Prevention in the Process Industries 12 (1999) 227234

Dust explosion protection in linked vessels: guidance for

containment and venting
P. Holbrow a, G.A. Lunn a,*
, A. Tyldesley b

a
Health and Safety Laboratory, Harpur Hill, Buxton, Derbyshire, SK17 9JN, UK
b
Health and Safety Executive, Division of Science and Technology, Magdalen House, Stanley Precinct, Bootle, Merseyside, L20 3QZ, UK

Abstract

Much industrial dust-handling plant consists of vessels connected by pipelines. If a dust explosion propagates through such a
system, the overall explosion event can be more violent than if a single vessel only is involved, due to a combination of increased
turbulence, pressure piling and a jet flame ignition source in the second vessel. This paper gives guidance, based on an extensive
experimental programme reported previously, on two aspects of the protection of interconnected vessels: containment and venting.
Crown copyright 1999 Published by Elsevier Science Ltd. All rights reserved.

Keywords: Dust; Explosion; Containment; Venting

1. Introduction there is enough evidence to show that in some circum-

The risk of a dust explosion is difficult to avoid in
processes where combustible powders and dusts are
handled. Many fine materials, e.g. coal, wood, flour,
starch, sugar, rubber, plastics, some metals, pharmaceut-
icals, etc., can explode once they are dispersed in air as
a cloud with a suitable concentration and when an effec-
tive ignition source is present.
Dust explosions cause a rapid increase in pressure
when confined in vessels or other process plant. Either
the plant must be built with enough strength to contain
the explosion and to prevent entirely the escape of any
material, even when the explosion is allowed to run its
full course, or some other means of protection, such as
venting or suppression, must be applied.
Guidance on the application of protection techniques
is published in a variety of sources, but usually applies
to isolated vessels. In industry, however, plant items are
connected by pipelines or conveyors and an explosion
may propagate from one vessel to another, causing a sec-
ondary explosion. The conditions under which this sec-
ondary explosion takes place are generally unknown, but
* Corresponding author. Tel.: 44-142-892-000; fax: 44-142-
892-050; e-mail: anne.bailey@hsi.gov.uk
stances there can be a significant increase in the rate of
combustion and violence of the explosion, and a signifi-
cant increase in the explosion pressures.
The British Materials Handling Board (BMHB) has
sponsored a project with the Health and Safety Executive
to investigate the behaviour of dust explosions in sys-
tems of connected vessels. The main purpose of the pro-
ject has been to measure the pressure changes that occur
when a dust explosion is ignited in one vessel, and the
flame transmits into the other. The aim has been to pro-
vide industry with guidance as to the conditions in which
excessive explosion pressures may be generated in con-
tained plants, and when and to what degree additional
venting is required on vented plant. The guidance
derived from the experimental programme is described
in this paper.
2. General
2.1. Containment
Containment is used when hazard quantification indi-
cates that an emission could present an unacceptable
risk, but it may still be a valid option even when this is
not so. For instance, it is often used to protect plants

09504230/99/$ - see front matter Crown copyright 1999 Published by Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 0 - 4 2 3 0 ( 9 8 ) 0 0 0 5 0 - 3
228 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 12 (1999) 227234

operating at sub-atmospheric pressure, e.g. vacuum dry- arrangements that are more than adequate in a single
ers, and is often a suitable option for mills which, when vessel.
of small volume, can be built with enough strength to The reasons for this increase in the rate of combustion
withstand maximum explosion pressures without excess- are the generation of turbulence as the explosion passes
ive cost. Increasingly, however, this technique is being through the interconnecting pipe and the entry into the
used on larger items of plant. second vessel of a relatively large jet flame ignition sour-
Practical application of containment depends on a ce.
knowledge of the maximum explosion pressure gener- The practical result of these effects is that the
ated by the dust explosion. The maximum explosion explosion in the second vessel can be much faster than
pressure, Pmax, can be measured in standard tests, using an explosion in an isolated vessel of the same design,
an apparatus such as the 20 I-sphere. Pmax is usually and consequently, the rate of pressure rise, (dP/dt), will
measured from an initial pressure of 1 bar a. In single be much greater.
vessels, depending on the process conditions, the The rate of pressure rise is the important explosion
maximum pressures are usually in the range 710 times characteristic determining the venting requirements, and
the operating pressure. increased explosion violence will mean an increase in
Industrial plant can then be designed to withstand the these requirements if the plant is to remain safe.
maximum explosion pressure Pmax without rupture.
Because of its insensitivity to vessel size, the maximum
pressure measured in the small-scale tests can be used 3. Guidance
to estimate directly the maximum explosion pressure that
a single vessel must withstand.
It is, then, not difficult to estimate or measure directly 3.1. Containment
the maximum enclosed explosion pressure in a single
vessel, but with more complex configurations this is not 3.1.1. General
possible. With multi-volumes or long ducts, pressure- The guidance is based on a simple mathematical
piling effects can lead to pressures well in excess of model of pressure-piling behaviour described in the
those expected in an explosion at ambient pressure. literature (Lunn, Holbrow, Andrews & Gummer, 1996)
Pressure-piling occurs when an unburned explosible and the results of the experimental programme.
dust cloud is compressed in one part of a plant by an Fig. 1 compares calculations from the model analysis
oncoming explosion in another. When this pre-com-
pressed dust cloud is ignited the explosion begins at an
above ambient pressure and the resulting maximum
pressure is correspondingly higher. The classic case is
transmission of an explosion from a large vessel into a
much smaller one.
If the dust-handling plant does consist of several con-
nected items of equipment, consideration must be given
to isolating the various units to prevent transmission of
flame or other ignition sources, to strengthening the plant
to cope with pressure-piling effects and to increasing the
level of explosion protection.

2.2. Explosion venting

There are a number of accepted methods for calculat-

ing dust explosion vent areas, in single vessels. They are
often easy to use and require only a minimal amount of
informationthe volume of the vessel, the strength of
the vessel, the explosibility rating of the dust, and the
opening pressure of the vent closure.
Often, the calculated value of the vent area errs on
the conservative side: the vent is larger than is strictly
necessary to protect the plant from the real explosions
likely to occur inside it. However, in linked vessels, the
behaviour of the flame and subsequent increases in the
rate of combustion may render ineffective venting Fig. 1. Maximum explosion pressures.
 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 12 (1999) 227234 229

the experimental results are presented in terms of the

maximum explosion pressure measured in the system.
The volume ratio V2/V1 is calculated by taking V2 as the
volume of Vessel 2 and V1, as the volume of Vessel 1
plus the volume of the interconnecting pipe. Vessel 1 is
where the primary ignition takes place.
It is clear from the pressure measurements that the
maximum explosion pressure falls when the diameter of
the interconnecting pipe increases. This is because back-
venting of the secondary explosion is more effective
when the pipe diameter is greater.
The guidance gives estimations of the explosion press-
ures which linked, contained vessels should be designed
to withstand. The highest pressures are worst-case esti-
mations; the analysis assumes slow transmission of the
explosion. In practice, these conditions are not met
because the speed of actual explosions is higher than
assumed, and because of the back-venting effect. The
back-venting effect is, however, taken into account in
the guidance on the basis of the measured pressures. The
Kst value is not a factor in this guidance.

3.1.2. Notes for guidance

Fig. 2. Compression factor (CF).
1. A major problem is that it is rarely possible to predict
with certainty in a system of linked vessels in which
and pressure measurements from the series of experi- vessel the primary ignition will occur. The maximum
mental trials. pressures obtainable depend on the volume ratio of
The analysis produces the calculated maximum poss- the vessels and become very high when this ratio is
ible explosion pressure in the linked vessel system, and less than 0.25, i.e. ignition occurs in the larger vessel.

Fig. 3. Flow chart for design pressure selection in linked contained vessels.
230 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 12 (1999) 227234

2. Volume ratios are taken always to be less than 1 in explosion pressure. Experiments show that the rate of
the advice that follows, i.e. V1 is larger than V2 and pressure rise in the secondary explosion can be very
it is assumed that primary ignition occurs in the larger high, and in practice, this needs to be taken into account
of the two vessels. The volume of the pipework in the design of effective venting.
should be calculated and this guidance is not appli- The following guidance is based on the results of an
cable if the pipe volume is large relative to the ves- experimental programme using various vessel volumes,
sels. The pipe volume should be added to the larger pipe lengths and diameters and explosible dusts, and
vessel when calculating the ratio V2/V1. fully described in reference (Holbrow, Andrews & Lunn,
3. Pmax (bar a) is the maximum explosion pressure for 1996). The guidance has been developed from the results
a given dust as measured in the standard 20 l sphere. of several hundred explosion tests in which the total vent
The pressure compression factor (CF) is estimated area in the linked system was divided so that each vessel
from the graph in Fig. 2. Fig. 2 should not be used had the same effective vent area:
if Pmax exceeds 10 bar a.
4. The flow chart in Fig. 3 describes the relevant con-
1 Av2/V2
Av1/V2/3 2/3
siderations, and depending on the system, the design
pressure may be Pmax, Pmax CF or Pmax (CF
1)/2. Pmax is expressed as bar a. If Pmax is not known where Av is the vent area, V the vessel volume and the
use Pmax 10 bar a. subscripts refer to the vessels. Primary ignition occurs
5. The experimental trials have shown that if the pipe in Vessel 1 and the secondary explosion in Vessel 2.
diameter is less than 0.1 m the probability of trans- As the vent area decreases, the relative effect of link-
mission would seem to be very low, and may be ing the vessels increases. The increase in pressure is gre-
ignored for design purposes. atest when primary ignition occurs in the larger of the
6. This guidance does not apply if the pipe is long linked vessels. Generally, the longer the pipe, the less
enough to allow the flame front to accelerate signifi- the effect is on the explosion pressure, but this is not
cantly (see the I. Chem. E. book on venting of pipe- always so.
lines for the current state of knowledge on this). The higher the explosibility of the dust, the higher the
7. The larger vessel should not exceed 20 m3 in volume. effect on the explosion pressure.
Two worked examples should make this clear:
1. Dust with parameters KST 250 bar m s1, Pmax
8.5 bar a;
vessels of volumes 1.5 and 4.5 m3;
pipeline: diameter 0.3 m, length, 5 m, so volume
0.35 m3;
vessel ratio 0.31;
Fig. 3 gives Pred CF 2.52;
theoretical maximum pressure 8.5 2.52 21.4;
design systems to withstand 21.4 bar a.
2. Same dust, vessels of 8 and 10 m3;
pipe diameter 0.55 m, length 8 m, volume 1.9 m3;
vessel ratio 0.67;
Fig. 3 gives Pred CF 1.9;
theoretical maximum pressure 8.5 1.9 16.2;
design pressure (1.9 1) 8.5/2 12.3 bar a.

3.2. Venting

The progress of the explosion in a system of linked

vented vessels is essentially the same as when the system
is contained: the explosion travels through the intercon-
necting pipe to produce a secondary explosion. The dif-
ference is that the explosion escapes through the vent
openings. The explosion pressures are thus less than
when the system is contained, and the rate of pressure
rise has much more importance in determining the final Fig. 4. Maximum reduced explosion pressure against Av/V2/3.
 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 12 (1999) 227234 231

Fig. 5. Maximum reduced explosion pressure against Av/V2/3. Fig. 7. Maximum reduced explosion pressure against Av/V2/3.

Fig. 6. Maximum reduced explosion pressure against Av/V2/3.

Fig. 8. Maximum reduced explosion pressure against Av/V2/3.
232 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 12 (1999) 227234

Fig. 9. Maximum reduced explosion pressure against Av/V2/3. Fig. 11. Maximum reduced explosion pressure against Av/V2/3.

Fig. 10. Maximum reduced explosion pressure against Av/V2/3. Fig. 12. Maximum reduced explosion pressure against Av/V2/3.
 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 12 (1999) 227234 233

Fig. 13. Maximum reduced explosion pressure against Av/V2/3. Fig. 15. Maximum reduced explosion pressure against Av/V2/3.

Fig. 14. Maximum reduced explosion pressure against Av/V2/3.

Fig. 16. Maximum reduced explosion pressure against Av/V2/3.

234 P. Holbrow et al. / Journal of Loss Prevention in the Process Industries 12 (1999) 227234
3.2.1. Notes for guidance
1. The larger of the vessels should not exceed 20 m3.
2. The length of the interconnecting pipe must not
exceed 15 m, and must not be less than 5 m.
3. The pipe area to volume of the smaller vessel in the
system must not exceed 0.1.
4. The Pmax of the dust should not exceed 10 bar a.
5. The vent area is always divided between the vessels
so that the dimensionless vent area Av/V2/3 where Av
is the vent area and V the vessel volume is equal in
each vessel.
6. The volume ratio V1/V2 is calculated by taking V1 as
the larger volume and V2 as the smaller. The volume
of the connecting pipeline is ignored.
3.2.2. Estimating vent areas for interconnected vented
vessels
This guidance is applicable to compact vessels (L/D
2), Pstat 0.1 bar g and for vessel volumes 20 m3.
Figs. 416 show relationships between the maximum
reduced explosion pressure and the dimensionless vent
area for the conditions listed on each graph. The figures
can be used in two ways: to estimate the maximum
reduced explosion pressure in an interconnected system
when the vent area is known; and to estimate the neces-
sary vent area to limit the maximum reduced explosion
pressure to a given value.
References
Holbrow, P., Andrews, S., & Lunn, G.A. (1996). Dust explosions in
interconnected vented vessels. J. Loss. Prev. Proc. Ind., 9(1), 91.
Lunn, G.A., Holbrow, P., Andrews, S., & Gummer, J. (1996). Dust
explosions in totally enclosed interconnected vessel systems. J.
Loss. Prev. Proc. Ind., 9(1), 45.
 Westfalia adds to centrifuge range
At this years Drinktec Inter- they claim ranks as the potential fields of applica- The two new clarifiers of
brau held in Munich, West- highest performer on the tion are extreme polishing the Hydro-Stop generation
falia Separator AG market with an effective duties such as cold wort complement the existing
unveiled a new Hydro-stop capacity of up to 70,000 I! clarification or beer clarifi- Westfaila series in both the
centrifugal polisher which hr. Westfalia envisage that cation before filtration. upper and lower capacity
ranges. The smallest Hy-
dro-Stop machine type SC
4 with an effective capacity
of up to 900 I/h was specifi-
cally developed for micro-
breweries and pilot plants.
The largest Hydro-Stop
machine type SC 150 is a
further development of the
type SC 120 clarifier. The
effective capacity is up to
70,000 I/hr.
Westfalia report that se-
parators of the Hydro-Stop
generation allow fast and
precise solids ejection. Ad-
ditional features are high
separation efficiency,
smooth, low-vibration opera-
tion, low noise level and high
maintenance-friendliness.
Westfalia Separator AG,
Separation Technology,
Werner-Habig-StraBe 1, D-
59302 Oelde, F.R. Germany.
Tel: +49 25 22 / 77-O; fax:
+49 25 22 / 77-24 88. Inter-
net: http:llwww.westfalia-
separatorzom

Dust collector prevents

explosions risk
Donaldson Torit has in- stringent German guide
stalled a dust collection lines (VD13673).
system at Dunlop Enerka Installed indoors, the
in Drachten, The Nether- Torit unit is fitted with a
lands, thereby avoiding po- bursting disk and a blow-
tential safety hazards off channel on top so any
including the risk of dust explosion would evacuate
explosions. through the roof, ensuring
Dunlop employs over the plant and personnel are
280 people at the Drachten protected. According to
plant which manufactures a the manufacturer, the Torit
range of rubber conveyor Downflo unit uses Ultra-
belts and rubber inner lin- Web filter cartridges, thus -
ings for fire hoses. In order making it 99.999% efficient
to prevent the product from on 0.5 pm particulates.
sticking together it is pow- Donaldson state that the
dered with a zinc stearate. Ultra-Web filter used in the
The dust being generated dust collector has been op-
is sticky and presents a erating successfully since
dust explosion risk. 1981 in a wide range of
To collect the dust and applications, providing a
protect employees, Dunlop low energy alternative com-
installed a 3DF24-R rein- bined with highly efficient
forced Downflo Dust Col- filtration.
lector from Torit. The Donaldson Torit, 65 Mar-
collector handles an air ket Street, Hednesford,
flow of 10,500 m3/hr, and Staffordshire WS12 5AD,
is designed to deal with UK. Tel: +44 1543 425515,
dust explosions by meeting fax: +44 1543 879136
266 Book reviews

regulatory compliance monitoring. In Part III of the book, 11 detailed procedures are
described for determining inorganic substances (N, D, CN, COD, SO,, etc.) in environ-
mental sampling.

Gary F. Bennett

Dust Explosions in the Process Industries by Rolf K. E&off

The first edition of Professor Rolf E&offs book was published in 1991 as a
hardback, and then as a paperback in 1994. It is an up-to-date and comprehensive
publication, and covers in detail the science and technology of dust explosions.
A second edition has now been published. It differs from the first in that an additional
chapter has been included, which is essentially a literature survey of work published
between 1990 and 1996 on research and development in dust explosions.
Chapter One is an extensive overview, which covers the practical aspects of the topic.
Eckhoff reviews why and how dust explosions occur, likely ignition sources, the
consequences of an explosion and the practical measures that can be taken to either
prevent them or minim& their effects. It is an excellent introduction, giving detailed
background on the industrial significance of dust explosions and how they can be dealt
with.
The importance of employing effective prevention and protection techniques is
underlined in Chapter 2, which is devoted to case histories of dust explosion incidents.
The range of materials involved and the variety of incident that occur is demonstrated,
as is how complicated the propagation of an explosion through plant can be.
Chapters 3 to 7 cover in more detail some of the topics in Chapter 1, but with much
emphasis on the formation of dust clouds and the propagation of explosions.
The dust cloud concentration, inhomogeneity and turbulence affect strongly the
propagation of an explosion. Chapter 6 shows how the degree of protection applied to
plant can vary with the state of the cloud. Chapter 3 is a lengthy discussion of dust
dispersion, and reviews the forces of attraction between particles, and how particle
dispersion can be achieved by air flow over dust deposits, and other methods.
Chapter 4 describes the propagation of dust flames. The behaviour of laminar flames
is dealt with comprehensively. The laminar burning velocity of a dust flame is an
important parameter. Various published theories of laminar combustion are discussed,
and experimental techniques for measurement of the laminar burning velocity described.
The importance of turbulence in the propagation of explosions is stressed, and its
influence in compact vessels and elongated enclosures described. Chapter 5 expands on
the ignition of clouds and bulk powders, covering some theories of ignition and
describing in detail experimental work. Chapter 6 is a discussion chapter on dust
explosion venting, and is a good review of the current position, pointing out that
practical venting requirements can differ considerably from prediction.
Chapter 7 discusses laboratory scale testing of the ignitability and explosibility of
dusts.
 Book reviews 267

Chapter 8 is a survey of published work in dust explosions from 1990-1996. This is

as up-to-date as publication allows, and expands on a paper published in the Journal of
Loss Prevention in the Process Industries. The material covers a wide range of topics,
and as a literature survey of the latest findings it is a good place to start. I hope in the
next edition of this very valuable book, much of the information in Chapter 8 will have
been expanded and added to the relevant parts of the earlier chapters.
In summary, this is an excellent book on dust explosions, and should interest anyone
concerned with the topic.

G.A. Lunn
 Journal of
ELECTROSTATICS
ELSEVIER Journal of Electrostatics 40&41 (1997) 511-516

Occurrence of Cone Discharges in Production Silos

M.Glor a and K.Schwenzfeuer a

aExplosiontechnology/Electrostatics Department, Ciba-Geigy AG, CH-4002 Basel,

Switzerland

ABSTRACT

Cone discharges which may occur during the filling of silos or containers with highly
insulating charged product have so far only been observed for granules of several
millimeters diameter or for very coarse product with a median of 700 p.m to 800 ~tm.
After a dust explosion within a production silo which was protected with explosion
relief venting against the effects of a dust explosion, measurements with respect to
static electricity could be performed in case of fine products (median of the mass
distribution 100 to 200 I~m) over a period of more than one year. The results clearly
demonstrate that cone discharges may also occur in case of highly insulating fine
powders, however, in this case the equivalent energy is much lower than for coarse
product.

1. INTRODUCTION

During the last five years several explosions occurred in silos of fine chemicals pro-
duction plants. The silos have a diameter of typically 2 m to 3 m and a volume of 10
ms to about 50 ms. Normally they are filled by pneumatic or pulse flow transport.
Since the products processed in these plants have been known to be very sensitive
with respect to discharges due to the build-up of static electricity the silos have been
protected from the very beginning against the effects of dust explosions by means of
explosion relief venting or explosion suppression combined with pressure shock re-
sistant construction and isolation. The constructional protection worked very well in
all cases. The ignition source, however, could not be identified until recently.
The occurrence and incendivity of cone, discharges has been investigated exten-
sively in the past [1-4]. However all results which have been reported so far on this
phenomenon have been obtained with granules or at least with very coarse product
with particle sizes typically from about 0.7 mm to several millimeters. It was therefore
an important open question whether these discharges may also occur when filling
fine particles into silos and containers and if so, whether the tendency of a
decreasing equivalent energy of ignition with decreasing particle size observed in the
range mentioned above would continue to smaller particle sizes. Based on the wall
known experimental fact that the minimum ignition energy (MIE) of powders
decreases with decreasing particle size [5], the questions raised above become
even more important.

0304-3886/97/$17.00 Elsevier Science B.V. All fights reserved.

PI1 S0304-3886(97)O0095-8
512 M. Glor, K. Schwenzfeuer /Journal of Electrostatics 40&41 (1997) 511-516

2. MEASURING TECHNIQUES

After several explosions have occurred in stainless steel silos of volumes 10 m3 to

50 m3 during the filling with highly insulating fine product of low minimum ignition
energy by pneumatic transport and no ignition source could has been identified, a
project was started to perform measurements with respect to the build-up of static
electricity and the resulting occurrence of discharges. Basically the following
measurement techniques would have been available:
visual, photographic or video registration of the events within the silo viewed
from the top of the silo
probes for the measurement of the electric field mounted flush with the silo wall
or introduced from top of the silo into the space above the powder heap
probes for the measurement of the high frequency electrical signals mounted
within the silo
installation of a ring shaped capacitor at the internal wall of the silo to monitor
charge transfer signals jumping from the charged product to the silo wall
charge to mass ratio of the product entering the silo
calculation of the average input current into the silo based on the mass flow
rate and the charge to mass ratio
Based on the comprehensive experience with measuring electrostatic phenomena in
a test silo [1,2,4] the following measuring techniques have been installed or not
installed: After inspection of the situation in the production silo the visual,
photographic and video registration from the top of the silo was not installed be-
cause there would have been no chance to see any signals through the dense dust
cloud present in the silo during normal production. For the same reason no field
probes have been mounted. Experience with the test rig has shown that these
probes only see the local field as soon as only a small amount of dust is present in
the free space and thus cannot be used to monitor the field resulting from the pow-
der heap. Probes for the measurement of the high frequency electrical signals could
have been mounted rather easily in the silo. Again, tests in the test rig have,
however, shown that not many information is obtained from these signals, since they
cannot be correlated with the strength nor with the location of the discharge events
causing the signals. However the ring shaped capacitor has proven to be very
informative in characterizing the cone discharges in the silo of the test rig. Therefor
such a ring shaped capacitor was mounted in the production silo at a position which
corresponded to the usual maximum filling level of the silo.
For this purpose the interior surface of the stainless steel silo (volume 12 m3,
diameter 2.1 m) was coated with a highly insulating raisin over a height of about 0.8
m. On top of this ring shaped internal coating two separate ring shaped conductive
coatings (aluminum) were deposited each of a height of about 0.25 m. These
coatings were very thin (ca. 100 I~m thickness) and did in no way disturb the filling
and emptying cycles of production. From each of these ring shaped capacitors a
shielded electrical connection to the outside of the silo was installed and connected
to a signal processing unit as shown in Figure 1.
 M. Glor, K. Schwenzfeuer/Journal of Electrostatics 40&41 (1997) 511-516 513

Figure 1. Schematic drawing

of the measuring technique
used in the producUon silo.
vtr~h the ring shaped capaci-
tors the charge transfer sig-
nals from the cone dis-
charges are collected and
stored on a PC after signal

? processing in a RC-circu#
and a transient recorder.

The capacitance of the ring shaped capacitor in the silo including the connections to
the signal processing unit was 140 nF. For an optimal shaping of the transient
charge transfer signals from the cone discharges to the signal processing unit an
additional capacitor of 160 nF and a resistor of R 106 Ohm in parallel were added
resulting in a relaxation time of 300 ms. The signal processing and storage was
controlled by a PC type computer which registered all signals corresponding to a
charge transfer of more than 30 nC. Details can be taken from ref. [4,6].

3. EXPERIMENTAL RESULTS

During normal production the silo was used as an intermediate storage silo between
production and filling of the final product into different sizes of packages. All products
involved were blends from different types of additives used 12 in polyethylene produc-
tion. The resistivity of all products was higher then 5.10 Ohm.m and the median of
the mass distribution of the particle sizes was about 100 to 200 i~m. The average
filling rate was typically about 4 tons per hour. The length of the pneumatic transport
pipe was about 150 m and its nominal diameter 100 ram. The actual transport oc-
curred in the so called pulse flow mode with the transport of about 400 kg within a
few minutes The filling procedure could be changed to a certain extent. Charge
transfer measurements could be performed with central and tangential pneumatic
filling as well as with manual filling by gravity through a rotary valve on top of the
silo.
During a period of 8 months data have Lleen collected on 128 days with tangential
pneumatic input of the product into the storage silo. During this time a total of 428
charge transfer signals above the threshold value of 30 nC have been registered.
The frequency distribution of the discharges of different strength is represented in
Figure 2. Figures 4a and 4b show two of the most energetic discharges observed
during this period with a charge transfer in the range between 500 and 700 nC.
514 M. Glor, K. Schwenzfeuer /Journal of Electrostatics 40&41 (1997) 511-516

50
Figure 2. Frequency
45
distribution of charge
40 transfer signals from
35 cone discharges ob-
30
served in a 12 rn~ stain-
less steel silo during
25 tangential filling with
! 20 highly insulating fine
powder by pneumatic
i: 15
transport,
10
5
0
10 40 70 10 13 16 19 22 25 28 31 34 37 40 43
0 0 0 0 0 0 0 0 0 0 0 0
charge transfer in nC

200 Figure 3. Frequency

distribution of charge
transfer signals from
150
cone discharges ob-
.=_
"o
served in a 12 m 3 stain-
100
less steel silo during
! central filling with highly
E
insulating fine powder by
50 pneumatic transport.

10 40 70 10 13 18 19 22 25 28 31 34 37 40 43
0 0 0 0 0 0 0 0 0 0 0 0
charge transfer in nC

During a period of about 7 months data have been collected on 163 days with cen-
tral pneumatic input of the product into the storage silo. During this time a total of
428 charge transfer signals above the threshold value of 30 nC have been regis-
tered. The frequency distribution of the discharges of different strength is repre-
sented in Figure 3.
During a short period of about 1 week data have been collected on 4 days whilst the
storage silo was being filled manually by gravity through the rotary valve on top of
the silo. During this time only one charge transfer signal above the threshold value of
30 nC could be registered. Figure 5 shows the transient recorder, sequence for this
discharge with a charge transfer of 58 nC.
 M. Glor, K. Schwenzfeuer /Journal of Electrostatics 40&41 (1997) 511-516 515

10"3 12.12.95 20:1i6:30 Hr.: 21 10^3 2q.11.0S 23;~:1l Hr.: 27

1,68 i.r+e
1.SO rdmaZ A [nO] I..~e ganal ~ [n]
l.qO Kanal B [nC] Kanel B [nC]
r. qO 556 nC
1.30 I . 36
1,20 1.20
1.10 1.10
1.00 I .OO
0.90 ).08
e. 80 |.M
8. "tO t . 76
o.6e |.(W
O. SO |,50
e.~e ) . qO
O. 30 D. 3e
O. 20 ~. 20
0.10 D.IO
e.ee =.de
O.O 0.1 0 . 2 0 . 3 O.q 0 . 5 6 . 6 9 . T O.S 0 . 9 1,0 O.O 0.1 0 . 2 0 . 3 e . q O.S 0 , 6 0 . 7 0 . 8 0,9 1.0
ZeLtachee [ o ] 10~86 ZeLtachoe [ e l 18^0

a) b)

Figures 4a and 4b. Voltage versus time curve of the transient recorder sequence in the
case of two very strong discharges registered during tangential pneumatic filling of the silo.
"Zeitachse" means time axis. The ordinate axis shows the charge transfer in pC.

1@'3 Z7.10.95 10:15:55 Nr.: 7G

1 . G8
1 .50 Kanal A (nC] Figure 5. Voltage versus time
1 .qG Kanal B [nC]
curve of the transient recorder
1 . 38
sequence in the case of the only
1 . ZO
1 .10 discharge registered during filling
1 . 00 of the silo by gravity through the
8.98 5 8 nC rotary valve. "Zeitachse" means
0.88
O. "tO
time axis. The ordinate axis
!e. so shows the charge transfer in pC.
. 58

iO . 3 0
Io. 2 e
!8.18
o. oo , i = t ~ i , = i =
e.e 0.1 o.z 0 . 3 o,At 0 . 5 6.8 O.T 0.8 0.9 1.0
ze~ t achoe (el 1e'o

4. DISCUSSION AND INTERPRETATION OF RESULTS

Several charge transfer signals recorded during the pneumatic filling of the storage
silo showing values between 400 and 670 nC lie definitively above the range of 50
to 300 nC which is typical for the charge transfer in brush discharges [7]. There is
therefore a clear evidence that much stronger discharges than brush discharges oc-
curred within the silo. Based on the experience from the test silo these signals can
be attributed to cone discharges. In an earlier paper on cone discharges a relation-
516 M. Glor, K. Schwenzfeuer /Journal of Electrostatics 40&41 (1997) 511-516

ship between the equivalent energy of ignition and the charge transferred in the dis-
charges has been set up (see Figure 9 of ref. [3]). If the strongest discharges regis-
tered in the present measurements with a charge transfer of 400 to 670 nC are
marked on this graph the equivalent energy of the discharges lies within a range
from 3 to 5 mJ. In addition, these results are in good agreement with the estimated
upper value of the equivalent energy of cone discharges obtained from the numeric
formula [4]

W = 5 . 2 2 . D 336 d 1"462 (1)

with W: equivalent energy of cone discharge in mJ, D: silo diameter in m, and d:

median in mm of the particle size distribution of the product forming the powder
heap. In case of a silo with a diameter of 2.1 m, this formula gives an equivalent en-
ergy of 2 to 6 mJ fora median of 0.1 to 0.2 mm respectively.

5. CONCLUSIONS

The results show that the discharges which occurred within the silo had sufficient
energy to ignite very sensitive powders. The minimum ignition energies of the pow-
ders have been determined with the Mike 3 apparatus [8] and lie between 1 and 3
mJ. Thus, even based on a rough statistical estimation, the silo explosions can be
explained rather well: In order to ignite a 2 mJ powder a cone discharge with a
charge transfer above about 300 nC is required [3]. Such a discharge occurred
about every second to fourth week during normal production. Taking into account
that only a coincidence in space and time of a strong discharge and a dust cloud of
optimal concentration results into ignition the occurrence of explosions every few
years can be explained.
The most important conclusion from the results presented in this paper is the fact
that cone discharges do not only occur with highly insulating granules but also with
rather fine highly insulating powders. The upper limit of the equivalent energy of
these discharges can be estimated by a numeric formula, it decreases with the size
of the particles forming the powder heap.

6. REFERENCES
[1] BMaurer et. al. Journal of Electrostatics 23 (1989) 25
[2] M GIor et al. Journal of Electrostatics 23 (1989) 35
[3] MGIor et al. Journal of Electrostatics 30 (1993) 123
[4] M.Glor and B.Maurer VDI Fortschrittberichte 389 (1995)
[5] R.K.Eckhoff, =Dust Explosions in the Process Industries', Butterworth-
Heinemann Ltd, Oxford, 1991
[6] K.Schwenzfeuer, Report D 2/96, Fachbereich ExplosionstechniWElektrostatik,
Konzembereich Sicherheit, Cibe-Geigy AG
[7] N.Gibson and F.C. Lloyd, Brit. J. of Appl. Phys. 16 (1965) 1619
[8] R.Siwek and Ch.Cesena, Process Safety Progress 14 (1995) 107
 Journal of

ELECTROSTATICS
ELSEVIER Journal of Elect rostatics 37 t 1996~ 151 172

An electrostatic suspension method for testing spark

breakdown, ignition, and quenching of powders
G . M . C o l v e r * , S . W . K i m t, T a e - U Y u ~
Department of'Mechanical Engineering, Iowa State Uni~,ersiO'. Ames. IA 50011. USA
Received 23 October 1995: accepted 26 February 1996

Abstract

An innovative electrostatic suspension method, referred to as EPS (electrostatic particulate

suspension), is used for testing the spark breakdown, ignition, and quenching characteristics of
small volume powder suspensions. A highly uniform steady-state suspension is generated
between high voltage dc parallel plate electrodes that can be ignited with a spark discharge from
a high-speed, moving electrode or by pulse discharge from a Ruby or YAG laser. The design
methodology and characteristic suspension times are reviewed. Bulk powder resistivities
< 109 f~ m are recommended for generating a suspension. With cohesive powders such as fine
coal it may be necessary to augment the EPS method with acoustic vibration.
To demonstrate the utility of the EPS method, representative data for coal, aluminum, and
copper powders are presented. Specific results are discussed in the following studies: (I1 Spark
breakdown of a (inert) copper powder in air, (1I) Minimum iqnition enerqy ~MIE) q[ alumi-
num--oxygen-nitrogen/carbon dioxide mixtures at near-quenching conditions, (III) Wall quench-
ing of coal-air and aluminum-air mixtures, and (IV) Inert particle quenching u irh copper particle
propane-air mixtures.

Kew'ords: Ignition conditions; Powder ignition; Minimum ignition energy: Dust explosion:
Electrostatic particle suspension

1. Introduction

A c o m m o n difficulty e n c o u n t e r e d when testing the e x p l o s i o n characteristics of

p o w d e r s is the n o n - u n i f o r m i t y of the c l o u d of particles. T h e w e l l - k n o w n H a r t m a n n
b o m b - t y p e a p p a r a t u s relies on the d i s p e r s i o n of a charge of dust with a short burst of
injected gas followed by ignition of the m i x t u r e with a suitable energy source: whereas,

* Corresponding author.
1Present address: Korea Academy of Industrial Technology, 371-36, Kasan-Dong,Kumchon-Gu, Seoul.
South Korea.

0304-3886/96/$15.00 Copyright (c 1996 Published by Elsevier Science B.V. All rights reserved
PII S 0 3 0 4 - 3 8 8 6 ( 9 6 ) 0 0 0 0 8 - 3
152 G.M. Colver et al./Journal of Electrostatics 3 7 (1996) 151 172

tumbling and open tube methods depend on gravity-fed dispersions [1 3]. The
resulting transient dust clouds are either measured optically or assumed to be of
uniform particulate distribution over the test volume so that particle concentration
can be determined [3 6].
The EPS method produces steady-state clouds of high uniformity while providing
up to four methods for measuring and calibrating particle concentration. 2 The
suspension is sustained by accelerating particles between parallel plate electrodes
using dc electric field strengths of the order kV/cm. If necessary, the test section can be
vibrated to facilitate the breakup of clusters of cohesive particles such as finely ground
coal [7]. When testing flammability, the EPS method is restricted mainly to the lean
and near-rich branches of the flammability curve because of limitations on particle
concentration from charge shielding [8]. Unlike combustible gases, the attainability
of a rich flammability limit for powders remains questionable [3, 9].
The EPS method can be applied to ordinary insulating powders exhibiting a finite
surface conductivity (coal, magnesium, glass) and to high conductivity metals such as
copper and aluminum [10]. As a practical matter we recommend bulk powder
resistivities < 10 9 f~m to maintain bulk powder charge relaxation time less than
about I0 ms at normal temperatures (there is no lower limit for bulk resistivity). Oxide
films on aluminum or copper do not interfere with the method [10]. Because of its
versatility, our EPS facilities have proven to be useful for investigating a variety of
thermo-physical phenomena including electrostatic precipitation [11], particle-wall
heat transfer [12], pneumatic transport [13, 14], and particle cloud dynamics
[15-18].

2. Suspension uniformity

A high degree of suspension uniformity is attainable as well as steady-state opera-

tion using the EPS method. Excellent reproducibility of powder concentration is
possible by monitoring the attenuated signal from a laser beam passed through the
suspension. The spatial uniformity of steady-state suspensions has been confirmed
by (a) scanning the test section in horizontal and vertical directions 3 over extended
periods of time with laser beam attenuation [7,11,12,16,18], (b) collapsing the
suspension (by shorting the plates) and measuring the uniformity of the distribution
of the fallen particles [14], (c) capturing particles leaked from an EPS "pinhole" on
a sticky surface and measuring (by count) the radial distribution of particles about
the axis of the pinhole [15], and (d) visual observation against a lighted back-
ground.

2The methods include:direct particle count after collapsingthe suspension, total mass weight(similar
to direct particle count), laser beam attenuation [7, 11, 12, 16, 18], and external circuit current measure-
ment [10, 24].
3Stratificationin the verticaldirectiondue to gravitationaland charge shieldingeffectsdepends on the
concentration of the suspension [10, 16].
 G.M. Colver et al./Journal of Electrostatics 37 (1996) 151 172 153

Table 1
Characteristic EPS times for spark, ignition, collision, suspension collapse, and flame traverse

Statistical First pulse Second pulse Ignition Particle- Particle Test cell
time lag (spark) delay particle a wall" traverse

Spark 0.1-1 lares 0.1-10 ItS 1 Its

delay
Spark 11~100 ns 5 lOOns
duration
Particle - 6 ms 8 ms
collisions
Suspension - < 120 ms -
collapse
Flame - 1 ms 100 ms
front

A s s u m e s a 1 cm gap at 10 kV, E P S o f 96 [am copper spheres. N = 8 x 103/cm 3 coefficient of restitution

0.62 (copper-copper).

The high uniformity attained is attributed to the dynamic state of the suspension in
which particles, similar to molecules, are in a constant state of collision with each
other as well as with the walls [16, 17] (see collision times, Table 1). The speed
distribution of an EPS of copper spheres has been well represented as M axwellian [15].
For cohesive powders, it should be noted that cloud uniformity can be sharply
reduced, and acoustic vibration may be necessary to help form the suspension. Under
such circumstances, clustering of the particles can also occur depending on the
powder.

2. Suspension design

2.1. Moving needle electrode

Fig. 1 shows the spark kernel for a 136 lam copper sphere suspension with a 1.02 cm
plate electrode spacing. A high-speed moving needle electrode (tip) for triggering the
spark is visible at the upper plate after entry into the test cell. A recoil spring action
ejects the needle from the cell as part of its decelerating-accelerating motion following
entry. The needle electrode was constructed of steel to withstand the high forces
(~40009s max) with the parallel electrodes fabricated from copper [173.
The spark is triggered spontaneously between the high voltage plates by the moving
needle electrode upon entry into the test cell [17, 19]. The needle electrode is initially
accelerated through a vertical tube at speeds of 10 m/s to minimize any disturbance on
cloud uniformity prior to spark breakdown. 4 The particles are retained by a Pyrex

4 Boyle and Llewellyn [3] utilized a gravity fed moving electrode to trigger a powder discharge while
retaining a second stationary electrode in the suspension.
154 G.M. Coh'er et al./Joutvlal o/'Electrostatics 37 (1996) 151 172

Fig. 1. Parallel plate EPS spark test for 136 mm copper spheres: 1.02 cm electrode spacing (moving needle
electrode visible at upper plate with spark reflections above and below).

glass cylinder (2.63 cm i.d.) at a concentration of 1.68 103 particles/cm 3. Larger test
cells to 14.2 cm diameter and 3.5 cm plate separation will be discussed later in the
combustion studies.

2.2. Experimental design Jor spark initiated combustion

Fig. 2 shows the detailed construction of one type of EPS test cell for studying
breakdown and ignition with optional auger feeding of the powder [17].
Alternatively, the particles can be batch-fed to a confined EPS by inserting a
Pyrex cylindrical wall between the parallel plates (Fig. 1). The auxiliary equip-
ment includes a 50 kV dc supply, a 50 kV electrostatic voltmeter, electrometer,
and a high-speed oscilloscope with ns response. In the EPS system of Fig. 2 the
external capacitance (not shown) can be adjusted from 13 to 310 pF to vary the energy
delivered to the spark.
The particles themselves move at speeds of about ! m/s (i.e. 10 times slower than
the needle electrode) and are sustained by the same high voltage source that
supplies the spark energy. A typical field strength of 20 kV/cm is adjusted so as not
to trigger a premature spark in the suspension. Once breakdown has occurred,
the suspension collapses from the short circuit of the spark. The collapse of the
suspension is generally of little consequence since the events of interest (spark
 G.M. Colver et al./Journal of Electrostatics 37 (1996) 151 172 155

VACUUM VI ~ '~ , , I~ ]~3 I

..... "~ __~ "7 COMPRESSED

I 1~
~-~ I AIR RESERVOIR COMPRESSED]]
ELECTROMAGNET ~ AIR TANK [_J

~ S3j
V: VALVE J PNEUMATIC ~ o-.----
SV: SOLENOIDVALVE _~GUN
S: SWITCH ~
] TRANSFORMER~L_
~
_I--L
~--

I IN SPARK
U ~JQ~"~"-MONITORING PARTICLE DC J J
[~,~~CIRCUIT RESERVOIR MOTORL.~.
~

{ n,N ___]Z
BOX

(0-50 kV) ~ ~ ....~ , ~, GEAR BED

Fig. 2. Details of EPS showing suspension, pneumatic gun (moving needle electrode) and (optional) auger
feed of powder [17].

discharge, ignition, flame quenching) will have been completed over times
required for a particle to traverse the electrode gap ~0.01 s [17],
An important measurement for which the collapse of the suspension is critical
is that of the burning velocity of a powder so that the passage of the flame front
is essentially complete before particles can fall through a significant fraction of
the height of the test section. The theoretical time required for a particle to free-
fall a 1 cm gap is about 0.045 s for a particle starting from rest and 0.008 s for a
field-driven particle moving at 1 m/s. Multiple particle bounces with walls
will prolong the motion and extend the total collapse time of the suspension to
about 0.12 s.
By comparison, a radially propagating flame front starting from the spark kernel
and moving at 30 cm/s for aluminum powder requires a time of about 0.10 s to reach
the walls of a 6 cm Pyrex retainer. As the flame propagating time exceeds both the
free-fall and field-driven times of the particle, but not the total collapse of the
suspension, it becomes desirable to measure burning velocity at early stages of the
burn. A high-speed camera is used for this purpose [7].
156 G.h/L Colver et al./Journal ~?fElectrostatics 37 (1996) 151 172

2.3. Characteristic suspension times

Table l gives the characteristic event times associated with the spark, ignition,
burning velocity, and collision processes in the suspension. The statistical time lag is
a random event that can be reduced by irradiating the spark gap and by using clean
(no grease or oxide films) electrodes of low work function such as magnesium
or aluminum 1-20]. Irradiation of the gap was used in our studies. A design require-
ment is that the particle-particle and particle-wall collision times must be longer
than the residence time of the moving needle electrode ( ~ 8 x 10-4 s) prior to sparking
so as to minimize any disturbance to the suspension [17].

3. Spark breakdown in powder suspensions

3.1. The suspension sparking parameter - ND 2

From our EPS spark breakdown studies a useful parameter N O 2 was identified for
characterizing the effects of particle number density (N) and particle diameter (D) on
the breakdown process [17, 19, 21]. This parameter can be interpreted as a measure of
the surface area of a particle per unit volume of the mixture. Correlations for spark
breakdown using N D 2 are reported in [19].

3.2. Sparking probability

In utilizing a moving electrode, it is noteworthy that the sparking process is

a random event with the probability of breakdown (for given voltage) depending on
the penetration distance (L - Lx,) of the needle electrode into the suspension where
L is the parallel plate electrode distance and Lx, the distance from the needle tip to the
opposing parallel plate electrode.
An example of the probabilistic behavior of sparking is shown in Fig. 3 for 96 ~tm
copper particles and a positive needle electrode. A dimensionless needle electrode
penetration distance is defined (1 - xi), where xi = Lx,/L and (1 - x3 is normalized to
the needle electrode position at 50% probability of breakdown as (1 - xg)/(1 - x). The
probability of breakdown with electrode distance is seen to be "'narrowed" with an
increase in the sparking parameter N D 2 but is relatively independent of the applied
breakdown voltage. The curves in Fig. 3 can be well represented by a log-normal
distribution.

3.3. Spark breakdown voltage with particles

The presence of conducting particles can dramatically influence the breakdown

voltage of a point-to-plane gap [17, 19]. Fig. 4 shows the relationship between
particle diameter, particle concentration, needle electrode penetration distance into
the cell, and parallel plate voltage difference at breakdown. For a given needle
penetration distance (1 - x 3 , the average breakdown voltage is reduced with an
 G.M. Colver et al./Journal o/Electrostatics 37 (1996) 151 172 157

1.0

7~=.=___.
L=I.O2cm
I-.0-_ ~ [ ~
.9
D=96um , I f - - -

ND2=O.17#/cm
.8
v : _i-
10.OKV
O 13.0KV
16.0KV
,,-~ - :
O 19.0KV

U')
c-

O
.7 -
~ 0 . 1 7
....
....
ND 2

0.53
0.0
(#/cm)

0.30 []
[i i

RD
.v 6 -
L
(D
(D I"
I..
rm

O .5 - -O
c-
O 12
, B

%)
.A-
i,
(D O
> I

.3

E _!
I
(.9
.2 : ~i5__[-_,
~7

.S --' 1

.I

V
#--
0 I [ I I I
.7 .8 .~ o 1.0 I.I 1.2 1.3

I- Xi
l-x

Fig. 3. C u m u l a t i v e p o s i t i o n a l b r e a k d o w n d i s t r i b u t i o n for parallel plate electrode s p a c i n g of L = 1.02 cm.

(positive needle).

increase in the N D : parameter, that is, either an increase in particle concentration or

particle diameter squared.
Alternatively, for fixed N D 2 the required breakdown voltage decreases as the needle
penetration distance increases. A decrease in breakdown voltage is most easily
explained as a result of the decrease in the spark path length with penetration distance
158 G.3/L Colver et al. /Journal qf Electrostatics 37 (1996) 151 172

10
8 I I t I
]',1oo

\ \ \
10 " I \ \
8

~ ,, o\ \
\ \
7
-- t, \ \
~ '~ o
%

0 961am
\
In 1"1 rl 3561am
10 -2 _
I I 1 I I
.i .2 .3 .t~ .5
(I-x)

Fig. 4. Spark breakdown of EPS-effect of particle diameter and particle concentration (ND 2 parameter) on
needle penetration distance, (constant voltage).

o f t h e n e e d l e e l e c t r o d e . 5 F o l l o w i n g a fixed v o l t a g e c u r v e in Fig. 4, the l a t t e r t w o

m e c h a n i s m s (i.e. d e c r e a s i n g n e e d l e p o s i t i o n a n d i n c r e a s i n g N D 2) a r e seen to be
c o m p e t i n g effects. As w i t h a p a r t i c l e - f r e e b r e a k d o w n , for v o l t a g e s m a i n t a i n e d b e l o w
t h e p a r a l l e l p l a t e limit o f b r e a k d o w n , t h e n e e d l e m u s t a l w a y s t r a v e l a finite d i s t a n c e

5 The enhancement of the field strength at one or both poles of either a charged or neutral particle in an
external field will also likely influence the breakdown.
 G.M. Colver et al./Journal of Electrostatics 37 (1996) 151-172 159

into the test cell before the field at the tip becomes sufficiently large for breakdown to
occur [19].
Interestingly, above a limiting value NDcrl, 2 ~, 0.2 cm- 1, the breakdown voltage
becomes dependent only on the penetration distance of the needle into the cell but not
on the particle concentration or diameter (Fig. 4). The explanation for this limiting
breakdown value NDZ~rit is speculative but suggests that compensating or dominating
effects are present. For example, at the critical value NDc,lt 2 an increase in particle
concentration may be compensated for by a decrease in the ionizing irradiation from
the light source reaching the breakdown region of the needle electrode (with irradia-
tion loss becoming the dominating factor).
Alternatively, local breakdown at the needle tip can reasonably be assumed
to depend on the probability of a particle being in the vicinity of the needle tip as the
local field strength approaches its breakdown value. It is then postulated that with
increasing particle concentration (or diameter squared) the positional probability of
the particle-tip becomes essentially unity at the limiting value ND2~rit and remains
unity for all larger values of the parameter. 6 Here, the competing effects of smaller
needle penetration distance and increased breakdown path length mentioned pre-
viously for the smaller values of N D 2 (<NDZrit) are dominated by the positional
probability of breakdown. These explanations remain speculative.

4. Yag/Ruby laser ignition

A second method that was investigated for powder ignition was pulsed YAG and
Ruby lasers. Laser ignition has advantages in that it can supply the needed energy for
ignition of a powder (~0.1-10 J) [22] and is non-intrusive. The maximum ratings of
our YAG and Ruby lasers were 2 and 5 J, respectively. A near IR window was placed
at the top center of the upper plate electrode. Igniting the powder with the laser beam
entering at the perimeter (outer boundaries) of the suspension is simpler but would
produce a non-symmetric burn pattern.
A problem found with laser ignition of coal powder was that the IR window became
coated thereby reducing the ignition radiation intensity. Over several trials it was
also observed that the window became pitted. To eliminate these problems would
have required some form of gas stream cleansing or a cavity design. In contrast
sparking did not pose any such difficulties and was subsequently used in all of
our studies. Permanently mounted devices within the EPS such as chemical ignitors
or resistance heaters I-1, 2] cannot be tolerated since (as with a fixed needle electrode)
these devices will disturb the electric field pattern and the uniformity of the
suspension.

6Other factors are the breakdown over-voltagerequirementfor a movingneedleelectrode[19] and the

induction charge acquired on particles [-10,16].
160 G.M. Colver et al./Journal o["Electrostatics 37 (1996) 15l 172

5. Minimum spark ignition energy of powders

5.1. Near-quenching limit f o r ignition energy

It has been established experimentally that the minimum spark ignition energy of
a powder depends on such parameters as spark length, spark duration, turbulence
intensity, and test cell volume [2]. For example, a combustible mixture cannot be
ignited in a parallel plate test volume with the plate separation less than the quenching
distance whatever the energy input. It follows that the experimental results reported
here for an EPS cell measuring 2.1 cm height by 6.58 cm diameter should be inter-
preted as the near-quenchin 9 minimum ignition energies (MIE) and can be expected to
be generally larger than those reported for larger volume test cells at the same fuel
equivalence ratio (i.e. same fuel-oxidant-diluent proportions).
In our previous studies minimum ignition energy curves were obtained in various
admixtures of oxygen, nitrogen, and carbon dioxide at ambient conditions of temper-
ature and pressure [23, 24]. Also, flammability curves for aluminum dust mixed with
oxygen and a diluent gas of either nitrogen or carbon dioxide were developed for lean
mixtures. These studies confirm that carbon dioxide is a more effective diluent than
nitrogen in suppressing the flammability of aluminum powder.
The maximum external capacitance for this particular EPS ignition system (not
shown, [23, 24]) was 21 300 pF, which was adjustable down to 302 pF to control the
energy delivered to the spark. For 100% delivery, the capacitance stored energy Ei is

Qv cv
E, ~-- 2' (1)

where Q, v, and C are the stored charge, initial voltage across the parallel plate
electrodes, and the total (external) capacitance, respectively. 7 High-speed oscillo-
graphs of spark current and voltage during breakdown indicate that virtually all of
the stored energy is delivered to the spark at least on smaller capacitance systems
[20]; however, not all of the spark energy goes to heating of the combustible mixture
due to shock and radiation losses. The longer duration diffuse sparks (20-50 ns) have
been shown to be more efficient ignitors than are the shorter filamentary sparks
( ~ 5 ns) [2, 25].

5.2. Typical experimental results-correlations

Fig. 5 shows sample data of minimum ignition energy for an EPS mixture of
aluminum (25-30 gm), oxygen, and carbon dioxide at ambient conditions. The oxy-
gen/total gas fraction (moles/moles) was held at ~b = 0.21 while the parallel plate
electrode spacing was 2.1 cm at a voltage of 20 kV. The stored energy was increased
by trial until ignition occurred. The aluminum concentration (stoichiometry) was then

The external capacitance is placed directly in parallel with the electrodesafter any ballast resistor.
 G.M. Colver et al./Journal o/'Electrostatics 3 7 (1996) l 51 172 161

P=I arm
T = Room Temp.
D = 2 5 - 3 0 ]jm
d = 6.58 cm
L = 2.1 cm
V= 20 KV
= 0.21 (mole/mole)
da- I~nition
X - No ignition

4-

>~

ku

CD
D

x x
CD
x
ixl011

02 + C02

Fig. 5. Spark ignition energy vs. aluminum concentration for 27.5 mm spherical aluminum particle concen-
tration in oxygen-carbon dioxide mixtures showing best fit curve (positive upper electrode) - E],,~ is curve
minimum.

changed and the procedure repeated. As noted above, it was not possible to test the
rich branch of the curve due to the concentration limitations of the E S P method,
although a well-defined curve m i n i m u m could usually be found, called here the lowest
m i n i m u m iynition energy (LMIE).
Table 2 gives flammability data for various particle sizes, diluents, and gas dilution
ratios. Eq. (2) is a regression fit for the constants A, B, and F for the m i n i m u m iynition
eneryy (MIE) at specified oxygen-diluent concentration

Emi e (J) = A _i_ B (log1 o Ca0 + F (log10 Ca1)2, C~1 = ~g. (21
162 G.M. Colver et al. /Journal of Electrostatics 37 (1996) 151-172

Table 2
Minimum ignition energy curves for AI O2-N2 (CO2) mixtures - Eq. (2)

AI-O2 N 2

Run q5 Dia. A B F Elmie Cal,mie Cal

(tool/tool) (lam) Eq. (2) Eq. (2) (test range)
(J) (g/m 3) (g/m 3)

1 0.21 10 15 0.93 3.57 - 1.55 na 30-300

2 0.21 20-25 18.77 - 12.29 2.03 0.169 1064 80-800
3 0.21 25-30 19.6 -- 12.05 1.91 0.594 1427 100-4000
4 0.50 10 15 12.97 - 7.74 1.05 na 40-300
5 0.80 10-15 20.76 - 16.94 3.54 0.494 247 35-240
6 0.15 20-25 23.57 - 14.72 2.32 0.221 1487 10(V900
7 0.50 20-25 21.67 - 15.02 2.61 0.061 754 100-1000
8 1.0 20-25 19.91 - 13.78 2.4 0.130 743 70-900
9 0.5 25-30 19.39 - 12.61 2.08 0.278 1075 90-3500
10 0.8 25-30 18.14 - 11.84 1.96 0.259 1048 80-3000
A1-O2 CO2
11 0.21 10-15 11.12 - 6.84 1.09 0.389 1373 40-300
12 0.50 10 15 11.97 - 7.76 1.35 0.819 748 40-300
13 0.21 20-25 16.98 - 10.24 1.59 0.493 1660 100-1500
14 0.50 20-25 16.47 - 9.99 1.55 0.373 1669 100-2500
15 0.21 25-30 23.45 - 14.42 2.27 0.550 1500 150-3300
16 0.50 25 30 20.51 - 12.68 1.99 0.311 1534 90-2300

= moles O2/[moles 02 + moles N2 (COz)]; Dia. = particle diameter (range); Ca1 = concentration of
aluminum, g/m3; E~m~,= lowest value of minimum ignition energy (C,L~mi~) based on Eq. (2); Italics
= extrapolated value based on Eq. (2), (outside data range); na = sufficiently high concentrations could not
be formed to evaluate E~mie.

T a b l e 2 also gives t h e a l u m i n u m c o n c e n t r a t i o n test r a n g e (Cal test r a n g e ) o v e r w h i c h

d a t a w e r e t a k e n as well as t h e l o w e s t m i n i m u m i g n i t i o n e n e r g y E~mie as g i v e n by Eq. (2)
a n d t h e c o r r e s p o n d i n g c o n c e n t r a t i o n (Cal,lmie). F o r e x a m p l e , the c u r v e in Fig. 5 c o r r e -
s p o n d i n g to r u n # 15 in T a b l e 2 for 2 5 - 3 0 p,m a l u m i n u m p a r t i c l e s in a d m i x t u r e s o f
o x y g e n a n d c a r b o n d i o x i d e (05 = 0.21) has b e e n tested o v e r a l u m i n u m p o w d e r c o n c e n -
t r a t i o n s 1 5 0 - 3 3 0 0 g / m 3 ( = o z / f t 3 10-3), w i t h glmie = 0 . 5 5 0 J at t h e c o n c e n t r a t i o n
Cal,lmie = 1500 g/m 3.
T a b l e 2 i n d i c a t e s t r e n d s for ( n e a r - q u e n c h i n g ) m i n i m u m i g n i t i o n e n e r g y s i m i l a r to
t h o s e f o u n d for l a r g e r v o l u m e t e s t i n g [2], b u t w i t h s o m e e x c e p t i o n s . N o t e t h a t v a l u e s
of Elmi falling s i g n i f i c a n t l y o u t s i d e the t e s t i n g r a n g e o f Cal c a n n o t be a c c u r a t e l y
i n t e r p r e t e d w i t h Eq. (2) since s u c h v a l u e s a r e e x t r a p o l a t i o n s . As such, t h e s e v a l u e s
o f E~mie a r e i n d i c a t e d in italics a n d s h o u l d n o t be c o n s i d e r e d reliable. A l s o a c t u a l
d a t a v a l u e s for E~mie will g e n e r a l l y be s m a l l e r t h a n v a l u e s f o u n d w i t h Eq. (2) since
the s e c o n d - o r d e r fit m a y n o t c o n f o r m to t h e r a p i d c u r v a t u r e o f the data. F o r e x a m p l e ,
for r u n # 15, t h e d a t a of Fig. 5 g i v e Etmi~ ~" 0.35 J r a t h e r t h a n 0.550 J as c a l c u l a t e d
by Eq. (2).
 G.M. Colver et al./Journal of Electrostatics 3 7 (1996) 151- 172 163

A comparison of runs 2 with 3 reveals that for the same oxygen concentration ~b an
increase in particle diameter leads to an increased value of the lowest minimum ignition
energy Elmie. Also, a comparison of runs 2 with 6 or 3 with 9 and 10 give the generally
expected result that for the same particle diameter an increase in the concentration of
oxygen decreases the value of E~mi~. However, in comparing runs 7 with 8 and 14 with
16, opposite trends to those noted above are encountered. The reasons for these
results are not entirely understood. Close examination of the data over the range of
concentrations indicates that in the case of runs 14 and 16, the larger particles in fact
have a lower value of Emie over most of the lean branch of the curve and only cross at
values approaching Elmi~. This can be readily demonstrated by plotting Eq. (2) for the
two runs (14 and 16).
In the case of runs 7 and 8, these curves are observed to be nearly identical over the
ranges tested as can be seen from the similar values of A, B, and F in Table 2. This
would seems to imply that the excess oxygen over a value of ~b = 0.5 acts as a diluent
gas in the presence of carbon dioxide. This is probably simplistic and other possibili-
ties can also be considered. It must be emphasized that our data is for near-quenching
conditions and cannot necessarily be compared with large-volume test data in the
literature [2]. As with gases, we expect (and observe) that the ignition energy will be
greater for the near-quenching conditions.
We must also consider the statistical nature of breakdown (Fig. 3) and that both the
sparking process and ignition are subject to random variation, particularly when
conditions are near-quenching. We have attempted to compensate for such variation
by testing over a range of concentrations. Also, it cannot be overlooked that the
dynamical processes of the EPS method may influence the data since the particles are
moving at speeds of 1-10 m/s. Such speeds are comparable to typical laminar burning
velocities of powders and may induce local turbulence which is known to increase the
ignition energy [2].

6. Quenching distance of powders

The EPS method is well suited for small-volume testing of powders and conse-
quently can be used to measure quenching distance [7]. For parallel walls, the
quenching distance is the smallest characteristic separation distance that will permit
a flame to just propagate against a zero velocity flow. As such the quenching distance
is a safety criteria that can be used to arrest flame propagation.

6.1. Suspension design f o r fixed-needle electrode

Fig. 6(a) shows a coal suspension in a modified EPS system electrically isolated on
3 Teflon supports which in turn are mounted on a vibrating table to aid in the
suspension of cohesive powders. 8 For powder-quenching tests the EPS system

8Free flowing aluminum powder did not require vibration.

164 G.M. Colver et al. /Journal of Electrostatics 37 (1996) 151-172

(a)

(b)

Fig. 6. (a) EPS facility mounted on vibrating table for testing cohesive powders coal powder required
vibration; (b) coal powder ignition (high speed camera}; (c) aluminum powder ignition did not require
vibration.
 G.M. Colver et al./Journal of Electrostatics 37 (1996) 151 172 165

tc)
Fig. 6(c).

capacitance could be varied from 1400 to 53 400 pF [24]. Figs. 6(b) and (cl show,
respectively, the ignition of coal dust (vibration required) and the high-intensity
ignition of aluminum powder (vibration not required). This system was further
modified in that the pneumatically driven needle electrode was replaced with a small
permanent electrode for initiating the spark. The use of a permanent needle electrode,
although undesirable from the point of view of suspension uniformity, is expected to
have a minimal effect on quenching distance measurement since the quenching
process remains dependent on wall-quenching effects at distances removed from the
needle electrode.
To confirm our EPS experiment, we measured the quenching distance of meth-
ane air mixtures at various fuel-to-air ratios, and the results were compared with the
data of Lewis and von Elbe [26]. Fig. 7 shows that the quenching data obtained with
our facility are consistent with their data.

6.2. Typical experimental results'

Figs. 8 and 9 show quenching distance for 17.5 jam spherical aluminum particles
and 16.7 gm Illinois No. 6 coal, respectively. The trends in the two figures are similar
with increasing particle concentration, tending toward a minimum as was observed
for the minimum ignition curves (Fig. 5}. The lowest minimum quenchim] distances for
powders (aluminum ~ 5 mm) are considerably greater (2 - > 3 times) when compared
to gaseous methane in air, Fig. 7.
166 G.lgZ Colver et al./Journal of Electrostatics 37 (1996) 151 172

15
0 NO FLAME
FLAME
------LEWIS & VON ELBE [26]
- - P R E S E N T STUDY
E
E
10 0 0 ~ ~ ~ A ,~ )1oo o
I

r~

n.

o
l! o ./o
i
g
\'~ _ t P I arm.
0 OOX ~ T= 290 K

V = 5 - 15 k~
C 55,000pF
0
0 5 I0 15 20
METHANE CONCENTRATION, % vol

Fig. 7. Quenching distance of methane in air using modified EPS apparatus - comparison with [7, 26].

20 0 NO FLAME
t~ FLAME
SPHERICAL A1
D = 15-20 p m
P'- I arm.
E 15
E " T = 298 K
R. H. = 5 6 ~
u3 V = 7 - 18 kV
.< C =25,000:F
I-
10
(.9
z
-r"
L)
z

5 - oo o o 0 O

I I I
0.0 0.5 1.0 1.5

PARTICLE CONCENTRATION (x 10 ~ ), g
m3

Fig. 8. Quenching distance versus particle concentration for 17.5 pm spherical aluminum particles.
 G.M. Colver et al./Journal of Electrostatics 37 (1996) 151 172 167

30- a, NOFLAME
O FLAME

It1. #6 COAL
v.c. = 27.9%
E 25~O O D = 16.7 pm
E I \ P= I a~.
I \ r = 300 K
1 ~ R.H. = 41%

15~ A ~ ~ t ~ OO
A

&

I0 I I I I
160 210 260 310 360 410
PARTICLE CONCENTRATION, g
m 3

Fig. 9. Quenching distance versus particle concentration for Illinois No. 6 coal.

Our quenching tests indicate that the quenchin 9 distance, the lowest minimum
quenching distance (curve minimum), and the lean flammability limit all increase with
particle size. 9 The quenching distance of coal was observed to decrease with increas-
ing volatile content (the low reactivity qualitatively explains the large quenching
distance ~ 13 m m observed for coal in Fig. 9). Quenching data for powders remains
sparse [2, 6, 27].

7. Gas quenching with inert particles

Inert particles can extract thermal energy and repress the production of flame
radicals. The influence of spherical copper particles on quenching of a combustible gas
mixture of propane and air was investigated using the EPS pneumatic gun of Fig. 2

9The lean (rich) flammabilitylimit is the smallest (largest) concentration of coal that fails to propagate
a flame.
168 G.M. Colver et al./Journal of Electrostatics 37 (1996) 151 172

Fig. 10. Time exposure of sparking followed by burning in copper particle propane air mixture: traces of
particles (path lines) are visible at the outer regions of the cell.

[19]. The particles were sieved in the ranges 30 38, 44-53, and 88 104 lain over
particle concentrations of 1.5 103 to 3 x 104 cm -3.
Fig. 10 is a time exposure following sparking and ignition of a copper particle
p r o p a n e - a i r mixture. The particles are not visible although individual path lines can
be observed in the outer regions of the test cell. The tip of the moving needle electrode
(Fig. 2) is visible at the upper plate (spark reflections top and b o t t o m are also visible).

7.1. Correlations with ND 2 parameter

Fig. 11 gives the results of ignition energy tests performed without particles
( N D 2 = 0) using the parameter N D 2 versus the fuel air r a t i o f ( m o l e - C 3 H s / m o l e - a i r )
for p r o p a n e - a i r mixtures. The particle-free flammability curve has the conventional
U-shape that is characteristic of gases and powders [7, 27]. A lowest minimum ignition
energy ( L M I E ) is seen to occur at a fuel-air ratio somewhat greater than stoichiomet-
ric conditions, JClmie >fstoich, where Jstoich = 0.042. The lower and upper ,flammability
limits can be estimated a s f j . . . . ~< 0.03 andJ~pper ~> 0.07 respectively. F o r comparison
Lewis and von Elbe [26] give fl . . . . = 0.022 and JuVVer = 0.105.
F o r N > 0, Fig. 11 also shows the effect of the variable N D 2 for different values of
ignition energy Ei/Eio where Eio is the energy required to ignite a particle-free
p r o p a n e air mixture at the s a m e f v a l u e . Again these curves exhibit the characteristic
(inverted) U-shape with lower and upper.flammability limits. Increasing either N or
 G.M. Colver e t a / . / ' J o u r n a l o/'Electrostatics 37 (1996) 151 172 169

L = l. O O c m
N = 0.0
L = 0.70mO.B5cm L = 0.70%0.80em
x x
Ei/Eio
.3 -- ~7-2
--0-i0
----A--IO0
/ \
---0 -i000

\ O
//

//
A

"~"
"0

.2
\ I
l
IA/
t
.,,

.,.,rl .~
\
\ \
\ ~'
\
-2
-0
\ I i
II
// ..'"
s/ "~
" 0 \\0\
\ \
\ ~

I .-z

o
x
v
2
111
% / .~2,~
z

//11 ~ ~ \ ",
.i / I /, iS \\ i'1
I I I ~ / ~
/I / O~.. _o/~" ,.
Oi I / ~7 O ~ . % O"""- " "7
A / //
I I
/
V

0 I I I I I
.03 .04 .05 .06 .07

f - FUEL/AIR (mole/mole)

Fig, 1 I. Limiting particle number density for ignition versus fuel air ratio, 96 m c o p p e r p a r t i c l e s .

D to values of ND 2 exceeding a given dashed curve of value E]Eio means that the
mixture cannot be ignited for that energy andfvalue. Thus, higher values of ignition
energy are required as either the diameter or concentration of the copper particles is
increased.
The experiments of Ei/Eio versus ND 2 in Fig. 12 suggest that limiting values of igni-
tion energy exist (for givenfvalues) below which the initial flame will not develop in the
presence of solid particles. This limiting behavior is analogous to that of the quenching
distance. The ignition data were correlated according to the empirical equation [19]
Ei ]-ND2~ m
=exp,~7-. ! , (3)
kn3A
170 G.M. Colver et al./Journal of Electrostatics 37 (1996) 151 172

I o
#0 11"! / ' I
:' i !2 '~ /o/ /
' :""" 4 ? /
' ;i : o " /
I " ' : / I /
I f a- l/ o

20 ' ;/ ~
h" o/ o

,I ~n ! %:. , a / iio o/
"
I

l?"^
I : : ~
I~
! ] I
I
10 i
I &.
" . I 0
I
' I
I
' "
11
/0/
0
/
- I ~II ~,, .,
,~ I,~ I
6
i :!" // I
o
I /io/ //o/
L~
d el/ / / /
i J/ ;/ ""A ./ , ~_1/
g
:

/ /~/ - ~ / 1]~? O Equations(3)&(4)

; ./;~: / ,'~/
v: ip ~ 11 I f I11:3"0 H2(O/em)
! /i : /,* /vo.o3z ........ o.o49
I t; , ~ . " " /~ / -
/ -
0.067 ....... 0.098
/ // / ,.,// a o.o3e . . . . . . . o.zo8
J m: " // / 0 062 ............
: r/. /// ~ 6:o,~2 0.zs7"122
!' P/.'"
...- / // / o ose
/ #..../,y ~ ~.o~o o.1~2'1~
.~:.../~/"
1 I I I f
0 .1 .2
NO2 (#/cm)

Fig. 12. Nondimensional spark ignition energy versus ND 2 parameter for propane-air mixtures;
q~ = fuel/air ratio (mole/mole).

where

H3 = H 2 [In (10)] - 1/u2 (4)

Table values for Ha and H 2 a r e given in Fig. 12 for propane-air mixtures

( f = fuel/air ratio in moles/moles).
 G.M. Colver et al./Journal of Electrostatics 3 7 (1996) 151-172 171

8. Summary
The EPS design methodology and characteristic suspension times associated with
spark breakdown are reviewed. For particles moving at --~1 m/s a moving needle
electrode of 10 m/s is adequate for triggering the spark so as to not to disturb the
suspension. Ignition delays of 1 ms fall well within the suspension collapse time of
120 ms making the system suitable for ignition testing. Similarly, spark durations of
10 Its (1 ms without irradiation) are fast compared to particle collision times of 6-8 ms
and much less than suspension collapse time. The use of YAG and Ruby lasers proved
difficult for igniting powders due to window contamination and pitting. Sparking is
reliable but subject to a breakdown probability curve and may pose an accountability
question on the magnitude of ignition energy depending on whether the spark is
diffuse or filamentary and its duration, etc.
The EPS method has proven to be particularly useful for testing small-volume
suspensions. Its utility has been demonstrated in four studies.
(I) Spark breakdown: A suspension sparking parameter N D 2 is identified together
with a log-normal sparking probability curve for a moving needle electrode with
(inert) copper powder in air.
(II) Ignition energy: Correlation curves for suspension mixtures of alumi-
num oxygen-nitrogen/carbon dioxide are obtained for near-quenching minimum igni-
tion energy (MIE); a lowest minimum ignition energy (LMIE) is identified by varying the
stoichiometry (proportions) of the mixture.
(III) Wall quenching: Data are presented for wall-quenching distance for coal
dust-in-air and aluminum dust-in-air mixtures; a minimum quenching distance is
found to be about 13 mm for coal and 5 mm for aluminum but will vary depending on
the powder concentrations (stoichiometry).
(IV) Particle quenching: An inert (copper) powder was suspended in various admix-
tures of propane, oxygen and nitrogen and correlated with ignition energy and ND 2
for flame quenching conditions. Correlations of spark breakdown for a copper
powder have been published elsewhere [19].
In general, the EPS technique can be applied to any powder exhibiting a finite
surface or volume conductivity including normal insulating powders. Bulk powder
resistivities < 109 m are recommended. For cohesive powders such as coal, acoustic
vibration of the powder may be required to form a suspension.

Acknowledgments
This work was supported by a grant from the ALCOA Foundation. We also
acknowledge constructive comments by the reviewers.

References

[1] M.K. Hertzberg, K.L. Cashdollar and J.J. Opferman, The flammability of coal dust-air mixtures, Bur.
Mines, RI 8360, 1979.
172 G.M. Colver et aL /Journal o f Electrostatics 37 (1996) 151 ~-172

[2] R.K. Eckhoff, Dust Explosions, Butterworth-Heinemann, Oxford, 1991.

E3] A.R. Boyle and F.J. Llewellyn, The electrostatic ignitiblity of dust clouds and powders, J. Soc. Chem.
Ind., LXIX (1950) 173--181.
[4] D.R. Ballal, Further studies on the ignition and flame quenching of quiescent dust clouds, Proc. Roy.
Soc. Lond., A385 [1983a) 1-19.
[5] D.R. Ballal, Flame propagation through dust clouds of carbon, coal, aluminum and magnesium in an
environment of zero Gravity, Proc. Roy. Soc. Lond., A385 (1983b) 21 25.
[6] J. Jarosinski, J.H. Lee, R. Knystautas and J.D. Crowley, Quenching of dust air flames, 21st Syrup. (Int)
on Combustion, Vol. 21 (1986), pp. 1917 1924.
[7] S.-W. Kim, Theoretical and experimental studies on flame propagation and quenching of powdered
fuels, Ph.D. Thesis, Iowa State University, Dept. of Mechanical Engineering, 1989.
[8] G.M. Colver, Electric suspensions above fixed, fluidized and acoustically excited beds, J. Powder Bulk
Solids Tech., 4 (1980) 21 31.
E9] S.E. Slezak, R.O. Buckius and H. Krier, Evidence of the rich flammability limit for pulverized
Pittsburgh seam coal air mixtures, Combustion Flame, 63 (1986) 209 215.
[10] G.M. Colver, Dynamic and stationary charging of heavy metallic and dielectric particles against
a conducting wall in the presence of a DC applied electric field, J. Appl. Phys., 47 (1976) 4839-4849.
[11] X. Liu and G.M. Colver, Capture of fine particles on charged moving spheres: a new electrostatic
precipitator, IEEE Trans. Ind. Appl., 27 (1991) 807 815.
[12] H.-C. Tsai, Particle~wall heat transfer, MS Thesis, Iowa State University, Dept. of Mechanical
Engineering, 1990.
[13] A. Sarhan, Effect of electrically driven particles on air flow in a rectangular duct, Ph.D. Thesis, Dept.
of Mechanical Eng., Iowa State Univ., Ames, 1989.
[14] J.A. Cotroneo and G. M. Colver, Electrically augmented pneumatic transport of copper spheres at
low particle and duct Reynolds numbers, J. Electrostatics, 5 (1978) 205 223.
[15] G.M. Colver and L.J. Ehlinger, Particle speed distribution measurement in an electric particulate
suspension, IEEE Trans. Ind. Appl., 24 (1988) 732 739.
[16] G.M. Colver, Dynamics of an electric (particulate) suspension, in: M.S. Shahinpoor (Ed.) Advances in
the Mechanics and the Flow of Granular Materials, Vol. I, Gulf Pub. Co., Houston, 1983, pp.
355-373.
[17] T.-U. Yu and G.M. Colver, Spark breakdown of particulate clouds: a new testing device,' IEEE Trans.
Ind. Appl., 1A-23 (1987) 127-133.
[18] G. M. Colver and D. L. Howell, Particle diffusion in an electric suspension, Conf. Record. IAS Annual
Meeting, Cincinnati, Sept. 28 Oct. 3, 1980, pp. 1056 1061.
[19] T.-U. Yu, Electrical breakdown and ignition of an electrostatic particulate suspension, Ph.D. Thesis,
Iowa State Univ., Dept. of Mech. Eng., 1983.
E20] J.D. Cobine, Gaseous Conductors, Dover, New York, 1958, p. 188.
[21] G.M. Colver~ S.W. Kim and T.U. Yu, Powder combustion testing using electrostatics, Proc. 2nd Int.
Conf. on Applied Electrostatics, Nov. 4 7, Beijing, China, 1993.
[22] T.B. Jones and J.L. King, Powder Handling and Electrostatics, Lewis, Chelsea, MI, 1991.
[23] S.-W. Kim and G.M. Colver, Spark ignition of aluminum powder in mixtures of oxygen, nitrogen and
carbon dioxide, a new testing method~ in: W.L. Grosshandler and H.G. Semerjian (Eds), Heat and
Mass Transfer in Fires and Combustion Systems, HTD-Vol 148, ASME Winter Annual Meeting,
Dallas, TX, Nov. 1990, pp. 97 104.
[24] S.-W. Kim, Spark ignition of aluminum powder, MS Thesis, Iowa State University, Dept. of
Mechanical Engineering, 1986.
[25] E. Barreto, Ignition by electric sparks, Institutional Physics Conf., Ser. No. 48, 1979, pp. 135 143.
[26] B. Lewis and G. von Elbe, Combustion, Flames and Explosions of Gases, 2nd ed.. Academic Press,
New York, 1961.
[27] E.L. Litchfield, Private communication, in: F.T. Green and J.E. O'Donnell (Eds.), The Quenching
Behavior of Coal Dust-Air Mixtures, Vol. 9, Final Technical Report, US Bur. of Mines Contract No.
J0166076, Kansas City, Mo., Midwest Research Institute, Aug. 1981.
 FuelVol. 75, No. 6, pp. 654-658, 1996
Copyright 1996ElsevierScienceLtd
Printed in Great Britain. All rights reserved
ELSEVIER SO016-2361(96)00006-3 0016-2361/96 $15.00+0.00

A comparison of experimental methods to

determine the minimum explosible
concentration of dusts

Nagesh Chawla, Paul R. Amyotte and Michael J. Pegg

Technical University of Nova Scotia, Department of Chemical Engineering, P.O. Box 1000,
Halifax, Nova Scotia, Canada B3J 2X4
(Received 6 March 1995; revised 21 December 1995)

An experimental investigation of dust explosions was conducted using a 20 L Siwek explosion chamber, to
examine test methods for determination of the minimum explosible concentration (MEC) of combustible
dusts. Two methods, one proposed by the American Society for Testing and Materials (ASTM) and the
other by the International Electrotechnical Commission (IEC), were compared. The MECs of gilsonite,
Pittsburgh coal, oil shale and Pocahontas coal dusts were determined in the 20 L chamber. The test results
were interpreted according to the ASTM criterion and the IEC criterion and compared with data available
for the four dusts from the US Bureau of Mines 20 L chamber and large-scale testing (1 m 3 chamber and
mine-scale). MEC values obtained from the IEC test method were generally lower than the values from the
ASTM test method. The primary reason for this is the difference in recommended ignition energy between
the two methods. Also, an ignition energy of 2500 J was most suitable for determination of minimum
explosible concentrations for most of the dusts studied in the 20 L Siwek chamber. This ignition energy gave
values that were comparable with data from large-scale experiments. Copyright 1996 Elsevier Science Ltd.
(Keywords: explosion limits; dusts; test methods)

This paper describes experimental techniques for deter-
mining the minimum explosible concentration (MEC),
also termed the lower or lean flammability limit, of
combustible dusts. The MEC is the concentration
boundary above which a dust-oxidant mixture will
propagate a flame in the presence of an adequate ignition
source. Knowledge of the MEC is therefore of practical
concern in the safe handling of flammable materials.
The MEC has been measured in different types (sizes
and shapes) of laboratory chambers for many years. The
most common laboratory chamber used has a volume of
20L. Among the test methodologies available for
determining the MEC are those proposed by the
American Society for Testing and Materials (ASTM)
and the International Electrotechnical Commission
(IEC). In this work, tests were conducted to allow a
comparison between the two methods.
The minimum explosible concentration and the upper
or rich flammability limit form the boundaries of the
flammability envelope of a combustible dust. Figure 1
shows the flammability envelope (explosive range) at
ambient temperature and atmospheric pressure for maize
starch, which is representative of a typical natural
organic dust. The MEC and upper limit are shown in
relation to typical dust concentrations permitted in the
context of industrial hygiene and to a typical density of
deposits of natural organic dusts. As noted by Eckhoff 1,
MECs of dusts correspond to concentrations of
airborne particles that are unlikely to occur regularly in
workplaces; primary dust explosions normally occur in
process equipment. The simple representation in Figure 1
also illustrates the potential hazard of secondary dust
explosions occurring by layered dust being raised into
suspension and ignited from a primary explosion.
S T A N D A R D I Z A T I O N OF MEC
MEASUREMENTS
Measurement of the MEC of a heterogeneous fuel-
oxidant mixture requires consideration of factors that
may not prevail in similar measurements for homo-
geneous mixtures. For example, with gaseous mixtures
the fuel and oxidant are separated by only molecular
distances; further, the attainment of a quiescent state
prior to ignition is usually not difficult experimentally.
Solid fuel-gaseous oxidant mixtures present different
challenges. The heterogeneous nature of the combustion
process poses the particular problem of ensuring that a
flammability limit, not an ignitability limit, is being
measured.
Assuming that the conditions of suitable test apparatus
geometry and size are met, the prerequisites for accurate
determination of limit data for combustible dusts are:
(1) careful control of the particle size distribution of the
dust;
(2) physical and chemical homogeneity of the dust
particles;

654 Fuel 1996 Volume 75 Number 6

 Determination of minimum explosible concentration of dusts." N. Chawla et al.

I
DUST
(3) uniform dispersion of the dust throughout the test
11UAL I
i P-.A1~B IIEPosrr apparatus;
I
I~ffl- ! (4) selection of suitable limit criteria; and
X I "." " . : " " ".",: " I
I .'.' :'.'-
I..'~-,. ". : -'..t
"-;I (5) use of an optimum ignition source.
L ' . " . . " , ' . ~" , ' : I
I,'." ~ Y- " ":, - : .I
I ";'~- ~', .; *. "" ,'1
I : : , . , " ~- .,I
Clearly, the measured minimum explosible concentra-
tion is not a fundamental parameter of a particular
I " ..'"'/: '4
I "v" " " . .'.1 material, but is rather an apparatus-dependent quantity.
I I 1 u.i . ':. -...., I t I Any attempt at standardizing MEC measurements in
lO s 10-z 10"1 1 10 10z 103 104 lOs 10' laboratory-scale vessels is essentially an exercise in
manipulating the last three factors in the above list.
MASS OF POWDER/DUST PER UNIT VOLUME [g/m3]
Table 1 gives a comparison of various test parameters
pertaining to the ASTM method 2 and the IEC method 1
Figure 1 Typical flammability envelope of maize starch dust (ref. 1) for MEC measurement. These test conditions, particu-
larly those for the Siwek chamber, are further explained
in the Results and Discussion section. A detailed
description of the Siwek chamber is given in the
Table 1 Comparison of ASTM and IEC test methods
Experimental section. It should be noted that in this
ASTM IEC work the MEC has been defined as the lowest explosible
concentration; strictly speaking, the IEC definition of the
Chamber type USBM a Siwek
Explosion chamber volume (L) 20 20
MEC corresponds to the highest non-explosible concen-
Dispersion reservoir volume (L) 17 0.6 tration. It is also worth noting that both MEC criteria
Dispersion air pressure (bar(g)) 9 20 shown in Table I are explosion overpressures; this is due
Ignition delay time (ms) 400 60 to the nature of the test procedures in which closed vessel
Ignition energy (k J) 2.5, 5 10 combustion occurs.
MEC criterion Explosion Explosion
overpressure overpressure
of 1 bar (g) of 1.5 bar(g)
excluding including
EXPERIMENTAL
igniter effects igniter effects, Tests were made with gilsonite, Pittsburgh bituminous
or 0.4 bar(g)
excluding
coal, oil shale and Pocahontas bituminous coal, encom-
igniter effects passing a wide range of volatility. The volatile matter
and particle size of each dust are given in Table 2.
a United States Bureau of Mines Replicate testing was carried out over a wide range of
dust concentrations and ignition energies.
The experiments were performed using a stainless steel
spherical vessel (Siwek chamber) having a volume of 20 L
Table 2 Dust analyses (Figure 2). A preweighed amount of dust was first placed
Particle size Volatile matter
in the dust storage chamber, the centrally mounted
(wt% <75/~m) (wt%) igniter was connected to the ignition leads, and the
chamber was closed. The chamber was then partially
Gilsonite 82 83 evacuated to 0.4 bar (40 kPa) (absolute) and the disper-
Pittsburgh coal 70 34
Oil shale 62 28 sing air pressure set to 20 bar (2 MPa) (gauge). When the
Pocahontas coal 64 16 solenoid valve between the dust storage chamber and the
test chamber was opened, the air and dust were dispersed
into the chamber, raising its pressure to 1 bar (100 kPa)
(absolute). The ignition source (chemical igniter) was
energized after a time delay of 60 ms. After the test was
finished, the chamber was thoroughly cleaned in
preparation for the next test.
Pressure development during an explosion was mea-
sured by two piezoelectric transducers mounted in the
vessel wall. A personal computer was used to record the
pressure-time data for each run. These data yielded
values of the explosion pressure, Pex, and the rate of
pressure rise, (dP/dt)m. Explosion pressures given in this
paper are gauge pressures that have been corrected for
the contribution due to the igniter; in this form the
explosion pressure is given the symbol Prn. The MEC was
evaluated from the explosion pressure data according to
the criteria in Table 1.
1. Pressure p u k e 5. Exhaust valve
2. Solenoid valve 6. Vacuum valve
RESULTS AND DISCUSSION
3. Dust storqc chamber 7. Pre~ure transducers
4. Vacuum pump 8. Rebound nozzle A typical set of data from the explosion tests is shown in
Figure 3; the example given is gilsonite with an ignition
Figure 2 Schematic diagram of 20 L Siwek chamber energy of 5 kJ.

Fuel 1996 Volume 75 Number 6 655

 Determination of minimum explosible concentration of dusts." N. Chawla et al.

4' is a potential for overdriving at a high ignition energy

3'
1'
o I I I
l0 20 30
Dust Coneentrsflon
Figure 3 E x p l o s i o n o v e r p r e s s u r e a t v a r i o u s gilsonite c o n c e n t r a t i o n s
(5 k J i g n i t i o n e n e r g y )
Minimum explosible concentration
The minimum explosible concentrations of all four
dusts tested using ignition energies of 10, 5 and 2.5 kJ are
given in Table 3. The test results are interpreted
according to both the ASTM and the IEC criteria.
Column a in Table 3 shows the results obtained in the
current work using the IEC criterion, whereas column b
shows the results obtained using the ASTM criterion.
The values from the USBM 20 L chamber and Fike 1 m 3
chamber, using the ASTM criterion, are given in
columns c and d respectively. Data from the Bruceton
Experimental Mine (BEM) and the Lake Lynn Experi-
mental Mine (LLEM) of the USBM are given in columns
e and f respectively.
The values obtained from the ASTM criterion
(column b) show the effect of applied ignition energy
on the MEC: the MEC clearly decreases with an increase
in ignition energy. The decrease is less pronounced for
the high-volatile gilsonite than for the low-volatile oil
shale and Pocahontas coal. This clearly shows that there
such as 10 kJ. (When the ignition energy becomes too
strong relative to the chamber volume, localized burning
may occur, giving the appearance of an explosion. This is
called overdriving the system.) These results support the
observations of Amyotte et al. 7'8 and Cashdollar and
Chatrathi 3, who showed that a 10kJ energy can
overdrive laboratory-scale chambers during MEC mea-
surements, leading to MEC values that are unrealistically
low.
Comparison of the values obtained from the IEC
criterion and the ASTM criterion (columns a and b)
shows that the MEC criterion for explosion overpressure
(see Table 1) is less critical than the recommended
ignition energy; the IEC method recommends 10 kJ and
I I the ASTM method recommends 5 or 2.5 kJ. Whether the
40 50 60 explosion overpressure should be 1 or 0.4bar(g), both
[g/m3] excluding igniter effects, is of secondary importance to
the potential for overdriving at high ignition ener-
gies.This concern has been expressed by Eckhoff 1, who
comments that with 10kJ in the 20L Siwek chamber,
too-low MEC values would be expected for some dusts.
Eckhoff1 further comments that with the same ignition
energy and limit criterion as used in the USBM 20 L
chamber, the Siwek 20 L chamber should be able to yield
comparable MEC results. The data reported here
support this statement, as seen by comparing columns
b and c. The comparison is especially favourable for
Pittsburgh coal but less so for Pocahontas coal. A
possible explanation for this difference is a difference in
the overall yield of volatiles from Pocahontas coal in the
two test chambers. (Here the accepted model of coal dust
explosion propagation by combustion of volatiles is
being used implicitly.) Conti et al. 9 observed that the
volatility of a sample of Pocahontas coal was 18wt%
when measured by the standard ASTM procedure, in
which the sample is heated slowly. The volatility
increased to 33wt% when the coal was subjected to
very rapid heating by a strong CO2 laser flux of 100 or
200Wcm -2. In the Siwek chamber, an ignition delay
time of 60 ms is used, whereas in the USBM chamber it is
400 ms. This difference in ignition delay time means that

Table 3 R e s u l t s o f M E C testing

Ignition M E C (g m - a ) *
energy
Dust (kJ) a b c d e f

Gilsonite 10 10 20 303 362 - 356

5 - 20 322 403 - -
2.5 - 30 362 - - -

Pittsburgh coal 10 30 40 503 802 50-605 606

5 - 50 602 - - -
2.5 - 80 802 - - -

Oil shale 10 130 160 - - 200-3005 -

5 - 300 2504 - - -
2.5 - 400 4004 - - -

Pocahontas coal 10 10 30 - - 1005 -

5 - 50 902 - - -
2.5 - 90 1202 - - -

* a, C u r r e n t w o r k , I E C c r i t e r i o n (Siwek 20 L c h a m b e r ) j b, c u r r e n t w o r k , A S T M c r i t e r i o n (Siwek 20 L c h a m b e r ) ; c, literature, A S T M c r i t e r i o n ( U S B M

20 L c h a m b e r ) ; d, literature, A S T M c r i t e r i o n ( F i k e 1 m s c h a m b e r ) ; e, literature, b a s e d o n l e n g t h o f f l a m e travel ( B r u c e t o n E x p e r i m e n t a l M i n e , U S B M ) ;
f , l i t e r a t u r e , b a s e d o n l e n g t h o f f l a m e travel ( L a k e L y n n E x p e r i m e n t a l M i n e , U S B M ) ; e a n d f , i g n i t i o n e n e r g y >> 10 k J, i g n i t i o n a n d d u s t e n t r a i n m e n t
b y p r i m a r y e x p l o s i o n o f m e t h a n e - a i r m i x t u r e ; s u p e r s c r i p t s d e n o t e reference n u m b e r s ; d a s h e s i n d i c a t e d a t a n o t available.

656 Fuel 1996 Volume 75 Number 6

 Determination of minimum explosible concentration of dusts: N. Chawla et al.
the turbulence intensity at the time of ignition is
significantly higher in the Siwek chamber than in the
USBM chamber. The smaller dispersion reservoir
volume and higher dispersion air pressure used in the
Siwek chamber (see Table 1) also contribute to the higher
turbulence intensity in the Siwek chamber. A well-
established consequence is that the rate of pressure rise
for the same material (which is indicative of the rate of
reaction or heating rate) will be greater in the Siwek
chamber than in the USBM chamber. Hence the overall
yield of volatiles for Pocahontas coal in the USBM
chamber is less than in the Siwek chamber. Therefore the
Siwek MEC values obtained for Pocahontas coal are low
compared with the USBM chamber values. (The fact
that Pocahontas coal is more sensitive to heating rate
than Pittsburgh coal is discussed later in the paper.)
The results from large-scale testing have been included
(columns d, e and f ) to give an indication of what might
be considered the 'true' MECs of the dusts tested. (True
minimum explosible concentrations are independent of
the ignition strength used.) Even though energetic
ignition sources were used in these tests, the problem of
overdriving is eliminated by the concurrent increase in
scale. At the highest energies studied, the MEC values
from the Siwek chamber are lower than those from the
1 m 3 chamber and full-scale mine tests. This is again
evidence of overdriving in the 20 L chamber when a
strong ignition energy is used. Examination of columns
d, e a n d f in the light of columns a, b and c shows that the
comparison between laboratory- and large-scale tests is
generally most favorable when the laboratory-scale
MECs are measured with a 2.5 kJ ignition energy and
the ASTM criterion. An energy of 5kJ is more
appropriate for oil shale because of its high ash content,
leading to reduced ignitability. This behaviour of oil
shale is further discussed in the following section.
Amyotte et al. 7 argued that although the use of a 2.5 kJ
igniter may be appropriate for MEC measurement in
most cases, higher energies may be required to determine
whether a particular dust is actually explosible when it is
non-explosible at lower energies but is known to be
explosible in nature. As an example, they discussed the
case of a pyrite-containing ore that is marginally
explosible with a 5 kJ igniter when tested in a USBM
20L chamber, but definitely explosible with a 10kJ
igniter. In such a case a more energetic ignition source
can be used, such as 5 and 10kJ, for further tests.
However, it should be remembered that these strong
igniters when used in a 20 L chamber may yield overly
conservative MEC data. Therefore it is beneficial to have
confirmatory data from large-scale tests.
Effect of volatile matter
The four dusts tested had different standard volatile
yields (see Table 2), the volatile matter of gilsonite being
the highest. MEC values for gilsonite with all three
ignition energies are thus very low relative to the other
dusts.
The standard volatile yields of the Pittsburgh and
Pocahontas coals were 34 and 16wt% respectively, and
the size distributions of both coals were approximately
the same. However, the MEC values are not greatly
different from each other. Both coals have the same
MEC of 50gm -3 at 5 kJ ignition energy (see Figure 4),
and the Pocahontas coal has a lower MEC at 10kJ
than Pittsburgh coal. The flammability of Pocahontas
100
~ * Pittsburgh Coal (34% VM)
VM)
80
~'m 6o
40
20'
1 2 3 4 5 6 7 8 9 10 I1
Ignition Energy [Ki]
Figure 4 Effect of volatility on m i n i m u m explosible concentration
(Pittsburgh and Pocahontas coals; M E C from A S T M criterion)
coal is therefore greater than that of Pittsburgh coal
at increased ignition energy (leading to high heating
rates).
This behaviour of Pocahontas coal is consistent with
the observations of Hertzberg et alJ , who showed that
at increased oxygen levels, the MEC values of Poca-
hontas coal are lower than those for high-volatile
Pittsburgh coal. In their work, the lean limit of
Pittsburgh coal in air was measured as 130gm -3,
decreasing to 87gm -3 at elevated oxygen concentra-
tions. The lean limit of Pocahontas coal at an oxygen
concentration of 53 vol.% was 68 g m -3, which was lower
than the corresponding value for Pittsburgh coal. The
cause was thought to be higher temperatures and higher
heating rates at higher oxygen levels.
Furthermore, Hertzberg and Zlochower ll showed
from their study of the rate of devolatilization, using
laser beam pyrolysis, that the final volatility of Poca-
hontas coal approached that of Pittsburgh coal with
decreasing particle size (<100 #m). When a laser flux of
200 W c m - 2 w a s used, the volatility of Pittsburgh coal
decreased from 55 to 50 wt% with a decrease in particle
size from 200 to 100 #m. In the case of Pocahontas coal,
the volatility increased from 28 to 34wt% with a
decrease in particle size from 200 to 100#m and
approached that of Pittsburgh coal with further decrease
in particle size. The behaviour of Pocahontas coal
observed in the current work appears to support the
findings of Hertzberg and Zlochower 11.
As mentioned previously, the volatility of Pocahontas
coal is generally found by the standard ASTM proce-
dure, in which the sample is heated slowly. However, in
the 20 L chamber, the heating rates are very high, due to
the sudden release of energy from the strong igniters
used. Therefore the total volatility of Pocahontas coal is
high and appears to approach that of Pittsburgh coal.
Similar behaviour for Prince and Phalen coals can be
seen from the flammability tests performed in a 26L
chamber by Amyotte et al. 8 (these results are given in
Figure 5). The lowering of the lean limit with an increase
in coal volatile matter is more pronounced at lower
ignition energies than at higher ignition energies such as
10 kJ. The overall yield of volatile matter from the coals
is almost the same at higher energies.
Fuel 1996 Volume 75 Number 6 657
Determination of minimum explosible concentration of dusts." N. Chawla et al.

500 inerting requirements were expressed in terms of total

~ Phalm(31% incombustible content, or TIC (ash and moisture in the
Prince (38% VM) coal, plus admixed dolomite), for the Lingan and Phalen
400 coals, the difference in inerting requirements was greatly
reduced (79% TIC for Lingan and 77% TIC for Phalen),
even though the volatile matter of the Phalen coal was
less than that of the Lingan coal. This was due to the high
ash content of the Phalen coal (35 wt% compared with
18 wt% for the Lingan coal). This illustrates that ash is
~ 200
similar in inerting effectiveness to added inert material
such as dolomite. The high MECs of oil shale obtained in
the current work are due to the high ash content of the oil
shale, even though the volatility of the oil shale is greater
100
than that of the Pocahontas coal.

0 I I I I I I I : : ',
0 1 2 3 4 5 6 7 8 9 10 11
CONCLUSIONS
Ignition Energy [kJ] Measurement of the MEC of a combustible dust involves
Figure 5 Effect of coal volatile matter on minimum explosible considerations unique to heterogeneous systems. Igni-
concentration (ref. 8) tion source strength, in particular, plays a key role in
establishing such minimum concentrations. The MEC
500. data presented in this paper for gilsonite, Pittsburgh coal
and Pocahontas coal show that a 2500J igniter is the
Pocahontas toni (16% VM)
most appropriate energy to obtain MEC values in the
400. 20 L Siwek chamber comparable with those from large-
scale testing. As a consequence, the IEC method is overly
conservative for the dusts tested, whereas the ASTM
~ 300. method yields realistic MEC data, primarily due to the
avoidance of overdriving from too energetic an ignition
source.

100.
0 2
1 2 3 4 5 6
Ignition Energy
Figure6 Effect of volatility on minimum explosible concentration
(Pocahontas coal and oil shale; MEC from ASTM criterion)
The MECs of oil shale at various ignition energies are
higher than for Pocahontas coal even though the volatile
matter of oil shale is higher than that of Pocahontas coal
(see Figure 6). This behaviour of the oil shale is due to its
high ash content (the ash content of Pocahontas coal is
5 wt% and that of the oil shale is 70 wt%). Ash, being an
incombustible material, arrests flame propagation by
absorption of thermal energy from the combustion
reaction. This can be understood from the work of
Amyotte et al. 12 on inerting requirements for coal dust.
They demonstrated that the percentage of dolomite
required to inert a coal dust explosion decreased in direct
proportion to decreasing volatile matter, from 80% for a
sample of Prince coal (36wt% volatiles) to 74% for
a sample of Lingan coal (32 wt% volatiles) and 64% for a
sample of Phalen coal (26 wt% volatiles), all three coals
having similar particle size distributions. However, when
REFERENCES
1 Eckhoff, R. K. 'Dust Explosions in the Process Industries', But-
4 terworth-Heinemann, Oxford, 1991
American Society for Testing and Materials. ASTM Standard E
7 8 9 tO 1515-93, 'Standard Test Method for Minimum Explosible Con-
centration of Combustible Dusts', Annual Book of ASTM Stan-
[kJ]
dards, Vol. 14.02, pp. 774-781
3 Cashdollar, K. L. and Chatrathi, K. Combust. Sci. TechnoL
1992, 87, 157
4 Staff, US Bureau of Mines. 'Fire and Explosion Hazards of Oil
Shale', RI 9281, US Bureau of Mines, Pittsburgh, 1989
5 Cashdollar, K. L., Sapko, M. J., Weiss, E. S. and Hertzberg, M.
In 'Industrial Dust Explosions' (Eds K.L. Cashdollar and M.
Hertzberg), STP 958, American Society for Testing and Materi-
als, Philadelphia, 1987, pp. 107-123
6 Greninger, N. B., Cashdollar, K. L., Weiss, E. S. and
Sapko, M. J. In Proceedings of Fourth International Collo-
quium on Dust Explosions, Warsaw, 1991, pp. 208-228
7 Amyotte, P. R., Mintz, K. J. and Pegg, M. J. In Proceedings of
27th Annual Loss Prevention Symposium, AIChE, Houston,
1993, Paper No. llc
8 Amyotte, P. R., Mintz, K. J., Pegg, M. J. and Sun, Y. H. Fuel
1993, 72, 671
9 Conti, R. S., Zlochower, I. A. and Sapko, M. J. Combust. Sci.
TechnoL 1991, 75, 195
l0 Hertzberg, M., Cashdollar, K. L. and Lazzara, C. P. In 'Twenty-
first Symposium (International) on Combustion', The Combus-
tion Institute, Pittsburgh, 1981, pp. 717-729
I1 Hertzberg, M. and Zlochower, A. I. In 'Twenty-third Sympo-
sium (International) on Combustion', The Combustion Insti-
tute, Pittsburgh, 1990, pp. 1247-1255
12 Amyotte, P. R., Mintz, K. J., Pegg, M. J., Sun, Y. H. and
Wilkie, K. I. J. Haz. Mater. 1991, 27, 187

658 Fuel 1996 Volume 75 Number 6


0950-4230(95)00056-9
ELSEVIER
Introduction to the Special Issue on Dust
Explosions. Theme: Hazard Evaluation,
Prevention and Mitigation of Dust
Explosions
In the introduction to his paper in this issue of the jour-
nal, Eckhoff comments that: The total amount ofexisting
knowledge on industrial dust explosions, their origin,
propagation, prevention and mitigation, is vast. And yet,
further information is continually being generated
through on-going research in a large number of coun-
tries. The aim of this special issue is to gather in one
reference some of the results of this on-going research,
consistent with the theme of Huzard Evaluation, Preven-
tion and Mitigation of Dust Explosions. It would, of
course, be impossible to Capture in one instant al1 of the
excellent research on dust explosions being conducted
worldwide. The very definition of research prohibits this
freezing of al1 information. Nevertheless, it is hoped
that the current document wil1 provide the reader with
a sample of recent work on dust explosions and wil1 per-
haps indicate future directions for concerted effort.
The first paper, by Eckhoff, gives a thorough review
of the authors own work and that of others worldwide,
primarily dealing with the period from 1990 onwards.
Al1 elements of the special issues theme are addressed,
with particular emphasis on existing and potential link-
ages between fundamental research and industrial prac-
tice. In many respects, Eckhoff sets the tone for the
remainder of the papers and offers a foundation for
future work in the field.
Siwek continues with the theme of hazard evalu-
ation by drawing on his extensive work at Ciba-Geigy.
He gives practica1 information on laboratory-scale tests
commonly used in Europe and elsewhere to determine
the flammabilitylexplosibility parameters of dust layers
and dust suspensions. An important feature of his paper
is the reference to testing standards developed by organi-
zations such as the International Electrotechnical Com-
mission (IEC) and the International Standardization
Organization (ISO).
Centra1 to the matter of hazard evaluation (and sub-
sequently the mitigation technique of venting) is the
cubic relationship by which the K,, parameter is
obtained. The fact that K,, is not an invariant property
of a material is amply demonstrated in the work of
Dahoe et al. They focus on the role of flame thickness
in pressure development during a dust/air mixture
explosion and present a three-zone model which
accounts for this influence. Their work presents a predic-
1
J. LDSS Prev. Process Ind. Vol. 9. No. 1. pp. 1-2, 1996
Copyright 0 1996 Elsevier Science Ltd
Printed in Great Britain. Al1 rights reserved
095&4230/96 $lWXl + 0.00
tive tool for assessing the pressure development of dust
explosions and suggests a modification of the cubic
relationship based on a function of flame thickness and
vessel radius; there are thus both fundamental and practi-
cal implications of their study.
The work of Lunn et al., as described in the first
of two papers from the UK Health and Safety Labora-
tory, deals with the particular problem of dust explosions
in systems of linked, enclosed vessels. Hazard evaluation
in this case is aimed at gaining a better understanding
of the phenomenon known as pressure-piling; such
knowledge is critical in applying the mitigation tech-
nique of containment, whereby explosion products and
overpressure are contained entirely within the process
unit. The authors begin with a review of recent research
in this area and then describe the results and analysis of
their experimentation with vessels having volumes of 2,
4 and 20 m.
While most of the papers in this issue deal with
dust/air suspensions, the work of Nakajima and Tanaka
offers an interesting look at the potential hazards of dust
layers. In this fundamental study, the critical state for
spontaneous ignition is classified as being of either the
Frank-Kamenetskii type or the oxygen-deficient type.
The authors present guidance for type discrimination
based largely on the bed material and porosity. Bearing
in mind that the primary use of hazard evaluation data
is the design of preventive or protective measures, Naka-
jima and Tanaka illustrate the practica1 significante of
their work through the example of aeration cooling -
an effective technique for Frank-Kamenetskii-type lay-
ers, but one that can actually increase the hazard of oxy-
gen-deficient-type layers.
Cashdollars paper on coal dust explosibility pro-
vides a thorough review of research conducted at the US
Bureau of Mines. Although the work described is spec-
ific to a particular dust, the lessons learned are applicable
to many other materials with respect to how hazard
evaluation is affected by changes in dust composition,
dust particle size and the co-presence of a flammable
gas. Of particular note in this paper is the validation of
laboratory-scale data through comparison with full-scale
test results. The special issue theme of explosion preven-
tion is addressed in Cashdollars work by his examin-
ation of the requirement in coal mines for admixture of
inert rock dust with combustible coal dust.
2 Editorial
Explosion prevention can be achieved by use of an
inert gas (e.g. nitrogen) to ensure that process operation
occurs at oxygen concentrations below the maximum
permissible level. This approach is aimed at removing
the oxidant criterion for explosion. Alternatively,
removal of the heat necessary for combustion can be
achieved by use of an inert dust as described in Cash-
dollars paper. This concept, however, has not been
applied extensively in industries other than coal mining.
The paper by Mintz et al. describes one such application
where the inerting dust is already present in the final
product; product degradation is not, therefore, a prohibi-
tive factor in the use of explosion-inhibiting additives.
The authors provide data which show how the safe prep-
aration of powdered metals (Al/Mg) for refractory
materials can be enhanced by early introduction into the
manufacturing process of inert Mg0 powder.
The theme of explosion mitigation (or protection)
is the subject of the remaining four papers in the special
issue. The first three deal with pressure relief venting
while the final paper covers automatie suppression of
explosions. The common feature of these papers is that
they al1 address the situation where, despite having
evaluated the hazard and taken appropriate preventive
measures, an explosion has occurred and the primary
concern is to mitigate the damage to personnel and plant.
Siweks paper on explosion venting technology
contains a wealth of practica1 information obtained from
industrial-scale testing. Such testing, and the empirical
correlations derived from the data, form the backbone of
the venting guidelines issued by the Verein Deutscher
Ingenieure (VDI) and the Institute of Chemical Engi-
neers (IChemE) in Europe, and the National Fire Protec-
tion Association (NFPA) in North America. It is interes-
ting to note the emergence of regression equations in the
new VDI guideline, with a decreased emphasis on the
use of nomographs.
The absente of adequate venting theories, and the
subsequent reliance on experimentation, is illustrated
further in the paper by Holbrow et al. They describe an
experimental study which extends the work of Lunn et
aE. (this issue) on interconnected enclosed vessel systems
to interconnected vented vessels. As in their previous
paper, the authors first give a thorough review of recent
research in the field and then describe the results and
analysis of their experimentation with vessels having
volumes of 2, 6.3 and 20 m3. Much practica1 guidance
is given on the variables affecting vent area requirements
for interconnected vessels (e.g. diameter and length of
interconnecting pipe); particularly significant is the
authors attention to the relationship between these vari-
ables and vent areas determined by Ks, nomograph
guidelines.
Further advances in venting technology are likely
to arise through a combination of several factors: on-
going programmes of industrial-scale testing, develop-
ment and use of first-principle mathematica1 models,
and continued efforts at computational fluid dynamics
(CFD) modelling. The paper by Tamanini and Valiulis
describes how the first two of these approaches have
been incorporated in a comprehensive and long-standing
programme at Factory Mutual Research Corporation.
The authors draw on their own experimental results and
the test data of other researchers, coupled with the use
of theoretical models, to present a new approach to the
sizing of explosion relief vents. It is significant to note
that in their conclusions they cal1 for contributions to the
advancement of venting technology from organizations
worldwide, with consensus on the conceptual framework
for such work. This fits wel1 with Eckhoffs encourage-
ment (this issue) of current efforts to establish inter-
national cooperation in joint research programmes.
Automatie explosion suppression is the subject of
the final paper, by Moore. The author draws on his
extensive industrial experience to provide a practica1
overview of suppressant choices and the basis for mak-
ing selections. He emphasizes the importante of choos-
ing the best agent for suppression, best often being a
compromise between effectiveness and practica1 accept-
ante. The issue of acceptance is particularly important
with the phase-out of chlorine- and bromine-containing
compounds.
As a final note, 1 wish to express my gratitude to
both the manuscript authors and the reviewers for the
quality of their work and the timeliness of their efforts.
1 would also like to acknowledge with appreciation the
dedication to this special issue of the editorial and pro-
duction staff at Elsevier Science Ltd and the office staff
of the Department of Chemical Engineering at the Tech-
nical University of Nova Scotia.
Paul R. Amyotte
Technical University of Nova Scotia
 J. Lm\ Prev. Prncess Ind. Vol. 9. No. 1. pp. 3-20, 1996
Copyright 0 1996 Elsevier Science Ltd
Printed in Great Britain. All rights reserved
0950-4230(95)00044-5 095&4230/96 $15.00 + 0.00
ELSEVIER

Prevention and mitigation of dust

explosions in the process industries:
A survey of recent research and
development

R. K. Eckhoff
University of Bergen, Department of Physics, 5007 Bergen, Norway

The paper covers three aspects, viz. basic research, applied research and development on pre-
vention and mitigation of dust explosions in industrial practice, and testing of ignitability and
explosibility of dusts. With few exceptions, only works published from 1990 and onwards are
referenced, i.e. publications that are not discussed in the previous survey (Eckhoff: Dust
Explosions in the Process Industries, 1991). Increased knowledge about the numerous facets
of the dust explosion hazard has created a justified request for a more differentiated approach to
the design of preventive and mitigatory measures in industry. In this process, cross-fertilization
between fundamental research and applied research and development is essential. Computer
simulation models, carefully calibrated against experiments, will become useful tools in a not
too distant future. The use of expert systems requires adequate quality assurance.

Keywords: dust explosions; prevention; mitigation

The total amount of existing knowledge on industrial
dust explosions, their origin, propagation, prevention and
mitigation, is vast. And yet, further information is con-
tinually being generated through on-going research in a
large number of countries. A few years ago, 1 had the
opportunity to review some of the knowledge accumu-
lated up to about 1990. About 900 references were
covered, but they almost exclusively originate from
English/American and German literature. There is little
doubt that a lot more interesting and useful information
can be retrieved by screening the large amount of
research reported in languages other than English and
German. Perhaps this could be accomplished by organiz-
ing a joint international translation/edition effort.
The present paper mainly summarizes works pub-
lished from 1990 and onwards, and only more recent
works than those quoted in the above mentioned text-
book are cited explicitly. Earlier versions of the present
survey were presented in Poland in April 1993 (and sub-
sequently published) and in the Peoples Republic of
Chinaj.
There is a growing acceptance of the fact that
adequate dust explosion prevention and mitigation in
practice must be based on a proper understanding of the
basic physical and chemical phenomena involved.
Therefore, when reviewing existing knowledge on the
prevention and mitigation of industrial dust explosions,
it is appropriate to start by addressing the fundamental
aspects. In the second part of the paper, knowledge and
technology related to preventing and mitigating dust
explosions in industrial practice wil1 be addressed
specifically. A separate third part is devoted to the
important area of test methods for assessing the ignit-
ability and explosibility of dusts. One main aim of the
paper is to indicate areas where a more intimate inte-
gration of fundamental research and practica1 develop-
ment in industry is timely.
Status and outstanding problems in
fundamental research related to dwt
explosions
Dust cloud formation processes. What is turbulente
of a dwt cloud?
This is an important and sometimes overlooked aspect
of dust explosions. It is wel1 established experimentally
that the initial state of a cloud of a given dust in a given
gas (dust concentration, degree of dispersion into indi-
vidual particles, dynamic state) has a strong influence on
both the ease with which the dust cloud ignites and the
rate at which it burns.
However, comparatively little fundamental quanti-
tative knowledge exists about the processes by which
dust clouds may be generated. Some work has been done
on entrainment of dust particles by turbulent gas flows
passing over dust layers and ridges in wind tunnels.

3
4 Prevention and mitigation
However, quantitative relationships between essential
parameters of the dust cloud generation processes and
the states of the resulting dust clouds are scarce, and
more work is needed in this area. When designing exper-
iments for fundamental studies, one should select con-
figurations that can be related to industrial practice.
Examples are bulk dust dropped into an air flow, and
entrainment of dust layers by air flows in channels and
galleries with and without flow-obstructing obstacles.
Hauert et ~1.~ measured RMS turbulent velocities,
global velocities and dust concentrations in experimental
dust clouds in the standard 1 m3 ISO vessel, and in a
12 m3 silo. In the latter case, the dust was either blown
into the silo via a pneumatic transport pipe, or dropped
into it from a screw conveyor at the silo top. This kind
of work is essential, e.g. for testing (validating) the
performance of numerical codes for simulating dust
cloud formation and development in real industrial situ-
ations. Such codes, in turn, wil1 be an essential element
in future comprehensive codes for simulating dust
explosion propagation in industrial situations.
Hauert et ~1.~applied laser Doppler anemometry for
measuring the turbulente in their experimental clouds
of maize starch. As tracer particles, they simply used the
starch grains of the experimental cloud itself. It has been
argued against this procedure that such comparatively
large particles are unable to follow the rapid turbulent
movement of the gas phase and, therefore, the measure-
ments performed do not represent the real turbulente of
the cloud.
However, this argument raises a basic question:
What is the turbulente of a dust cloud in the context
of flame propagation and dust explosion? Could it be
that the most important turbulente mechanism is in
some cases the movement of burning particles in relation
to unburnt ones, rather than the movement of the gas
phase? Could it be, therefore, that the movement of the
particles is in fact in some cases a more appropriate mea-
sure of turbulente than the movement of the gas phase?
The answer depends on the nature of the particles. Very
rapid devolatilization, with the combustion occurring
mainly in the bulk of the gas phase, is an extreme case
where the gas movement is indeed important. If, on the
other hand, the combustion mainly occurs close to the
particle surface, the particle movement is of primary
importante.
The state of a dust cloud that influences the ignition
sensitivity of the cloud wil1 clearly be the state at the
moment of ignition (initial state). When it comes to
flame propagation, however, the picture is complicated
further by the strong coupling between the flame propa-
gation process and the state of the unburnt cloud ahead
of the flame. Therefore, the initial state of the dust cloud
at the moment of ignition may only be of secondary sig-
nificance for the flame propagation once the explosion
has developed beyond the initial stages.
As soon as a significant blast wave has been gener-
ated by the primary dust flame, this blast may generate
secondary explosible clouds ahead of the flame from
dust deposits and layers there. Lebecki et ~1.~ investi-
gated such processes in a 100 m long gallery of cross-
section 3 m2. In order to establish an improved under-
standing of these processes, further experimental and
of dust explosions: R.K. Eckhoff
theoretical studies of the influence of blast waves on dust
clouds and dust layers/deposits need to be conducted.
Work on this problem has also been performed by UraP,
Gelfand and Tsyganov7, and others. Gelfand and Tsy-
ganov showed that the presence of dust layers on solid
surfaces exposed to blast waves changed the blast wave
characteristics as compared with the characteristics in
the case of dust-free surfaces. Kauffman et aLx and
Austin et aL9 have summarized their extensive research
on entrainment of dust layers in long tubes by the blast
wave heading a dust explosion propagating along the
tube. Tamanini and Ural have summarized their work
on characterizing the dispersibility of dust layers when
exposed to air blasts.
Increased emphasis should be put on investigating the
connection between the parameters of dust cloud gener-
ation processes and the structures of the resulting dust
clouds. The structures of the clouds produced must be
defined in terms of distribution of dust concentration,
quality of dust dispersion (de-agglomeration), turbulente
leve1 and global velocities. For example, fundamental
theoretical studies are being performed on the generation
of dust clouds behind shock waves sweeping across a
dust layer at 90 to the layer surface. Frolov et al.
developed a mathematica1 diffusion model describing
such a process. A similar model was developed by Lu
et al. *. This kind of work is relevant in the context of
propagation of secondary dust explosions. Nikolova
developed a source code for numerical models for simul-
ation of dust cloud flow. The code can be used for cold
flows as wel1 as flows with combustion. Medvedev et
al. l4 conducted experimental studies of the entrainment
of dust from dust layers by the short-duration flow gen-
erated by the rarefaction wave in a shock tube. Dushin
et al. l5 developed a mathematica1 model of the evolution
of a cloud of entrained dust in the atmosphere, following
a huge explosion on the ground. Nikitin et al. l6 perfor-
med a theoretical numerical study of the evolution of a
dust cloud in a turbulent atmosphere. The cloud could
be generated instantaneously as a result of a sudden
explosion blast, or continuously from multiple moving
sources. Geng et al. l7 studied the influence of clouds of
maize starch in oxygen on the peak pressures of incident
shock waves of Mach numbers 5.4 and 6.0.
In order to guide fundamental research in this area
in the direction of maximum practica1 relevante, infor-
mation about dust cloud structures that are typical in
industrial operation is required. This not only means the
cloud structures in normal plant operation, but perhaps
even more the structures existing during abnormal transi-
ent phases, including plant start-up and close-down, dur-
ing which the occurrence of dust explosions may some-
times seem more likely than under normal steady-state
conditions.
Dust layer and dust cloud ignition processes
Ignition is a broad field of research. The concept of ther-
mul runaway is a common basis for understanding and
describing ignition processes. Theories have been
developed even for complex cases with reactant con-
sumption during the process leading to ignition. How-
ever, it does not seem realistic for the time being to fore-
see the development of one single unified theory, usable
 Prevention and mitigation
in practice, which covers al1 types of ignition sources.
It is instead expected that separate theories, in terms of
dynamic computer models, wil1 be developed for various
categories of ignition sources, such as hot surfaces and
electric sparks. Work conducted on ignition of pre-mixed
gases should be studied to see whether elements of gas
ignition theory can be applied to ignition of dust clouds.
However, development of such theories needs to be
backed by careful experimentation. Many variables must
be considered, which are partly related to the nature of
the potential ignition source itself and partly to the com-
bustible dust layer or cloud that is exposed to the source.
Definition of good parameters and design of correspond-
ing experimental test methods remain a centra1 chal-
lenge.
Hensel and John8,9 provided further insight into
the important relationship between the conditions
required for initiating smouldering combustion of a dust
layer on a hut plate, and the layer thickness. Hensel et
aZ.* presented a mathematica1 model for numerical
simulation of temperature development in powder
deposits enclosing a heat source of constant power. The
model was tested against experiments in which a metal
sphere, heated by a constant power source, was embed-
ded in the powder sample. Further promising work along
this line, focusing on the critical heat flux for ignition
rather than on the critical surface temperature, was
reported by Krause and Hensel*. It was pointed out that
a critical heat flux relates more to real practice than a
critical constant surface temperature.
Kauffman et al. summarized their research on
smouldering combustion of dust heaps and dust layers.
Chernenko et al.* investigated flame propagation along
the surface of layers of metal powders, and mixtures of
metal powders and metal oxides. The influence of the
chemical composition of the powder layer, and to some
extent also of the particle size, on the burning velocity
along the surface of the layer was determined experi-
mentally. Matyukhina and Babushok developed a
mathematica1 model for self-heating in coal deposits,
taking into account the effect of diffusion of air into the
deposit. Sobolev* developed a discrete, two-temperature
mathematica1 model for heat conduction in a dust
deposit. An informative overview of research on self-
heating/self-ignition in dust deposits was given by
Crowhurst25. Zhang and Deng2h studied the combustion
rate of coal dust layers on constant-temperature hot sur-
faces. Influences of hot surface temperature, oxygen con-
tent in the atmosphere and flow rate of air across the
layer on the oxidation rate were investigated. Itagaki and
Matsuda used both differential scanning calorimetry
and accelerating rate calorimetry for measuring the exo-
thermic reactivity of activated carbon dusts. Adsorbed
nitrogen oxides or fluorine on the coal surface increased
the exothermic reactivity markedly, with the onset tem-
perature of exothermic reactions being as low as 30-
40C.
Glinka et al.* conducted an experimental and
theoretical study of ignition of dust layers by thermal
radiation. The ignition process was resolved in detail by
means of high-speed Mach-Zehnder interferometry.
Van der Wel29 conducted a series of experiments
in which laminar dust clouds were ignited by a short
of dust explosions: R.K. Eckhoff 5
light pulse (100 hs or 10 ns) from an Nd:YAG laser
(1064 nm wavelength). This kind of experiment can pro-
vide basic information about dust cloud ignition pro-
cesses, and also about flame propagation processes in
dust clouds. Van der Wel et aLX described how a sim-
ple, modified Semenov theory (no temperature gradients
inside the heated volume) for auto-ignition was used for
transforming the experimental laser-light-pulse ignition
results to predicted minimum ignition temperatures and
energies. The predicted values were in approximate
agreement with measured minimum temperatures and
energies for ignition. Gieras and Klemens studied the
critical conditions for ignition of single coal particles in
air, and in air containing methane. They also investigated
the critical ignition conditions for clouds of coal dust in
air and in methane/air.
In the past, the minimum hot-surface temperature
for ignition qf a dust cloud has often been regarded as
if it were a universal constant for a given cloud. Conse-
quently, results from small-scale laboratory tests were
often applied directly in industrial plant design. How-
ever, it has been known for some time that minimum
ignition temperatures of dust clouds vary significantly
with scale, and this has recently been confirmed by Wol-
anskP2. Further experimental and theoretical work is
needed in this area.
Ignition of dust clouds by smal1 burning metal par-
ticles (impact sparks, metal sparks) generated by mech-
anical impact is a very complex problem. A comprehen-
sive, practically useful theory does not seem to be within
sight. Such a theory must comprise several steps. The
first is the generation and initial heating of the metal
particle by the impact, which is in itself a very complex
problem. The second is the ignition of the flying hot
particle and the subsequent burning process. The third
is the heat transfer to the dust cloud, which ultimately
determines whether ignition occurs or not.
In dust explosion statistics, the frequency data for
occurrence of various categories of ignition sources are
sometimes confusing, because the frequently used term
friction sparks is ambiguous. Sometimes hot surfaces,
generated either by repeated impacts on the same spot
or by sliding friction, are included in the friction spark
or mechanica1 spark category. Furthermore, one does
not always distinguish between burning metal particles
from one single accidental impact and from repeated
impacts in rotationary machinery such as grinding and
cutting equipment. The categorization of ignition by
burning metal particles from grinding and cutting oper-
ations is not always clear either. This is because ignition
may have been caused by the hot surface being produced
at the object being ground, rather than by the burning
metal particles generated in the grinding process.
Ignition of dust clouds by electric/electrostatic dis-
charges is another very complex topic. Theories have
been developed for ignition of dust clouds by electric
sparks between two metal electrodes, which is the sim-
plest case, but even such theories are only rough
approximations. The variables of the electric spark
ignition problem include voltage and current character-
istics across the spark gap, spark gap geometry and elec-
trode material, and dust cloud variables such as particle
material and particle size/shape distributions, dust moist-
6 Prevention and mitigation
ure content, dust concentration, and the dynamic state of
the dust cloud with respect to the spark gap.
Dahn et ~1.~~reviewed some published work on
electric spark ignition of combustible dust clouds. Their
own experiments with clouds of lycopodium confirmed
the dramatic influence of the combination of capacitance
and resistance of the experimental discharge circuit on
the minimum capacitor energy for ignition.
Xu and LirQ4 conducted a semi-quantitative analy-
sis of electric spark ignition of dust clouds. They
proposed a strategy for calculating minimum ignition
energie& in terms of the lowest energy capable of estab-
lishing self-sustained laminar flame propagation in the
dust cloud. This strategy is the same as the one proposed
previously by Klemens and Wojcicki (see Reference 1).
With respect to the ever more complex one-elec-
trode discharge types (corona, brush, propagating brush
etc.), valuable experimental insight has been gained dur-
ing recent years, but so far no attempt at developing dust
cloud ignition theories seems to have been made. Glor5
gave an informative overview of the present status on
theory and experimentation. Some of this work, on poss-
ible incendiary discharges from powders poured into a
heap, was presented by Glor and Maurer36. The question
of whether incendiary lightning type discharges can
occur in dust clouds is stil1 to be answered. Glor37 is
also continuing his work on whether incendiary brush
discharges can occur in dust clouds.
In the context of gas explosion-proof electrical
equipment enclosures, where the maximum experimental
safe gap, MESG, is a centra1 concept, the basic problem
is ignition of an explosible gas cloud by a jet of hot
combustion products. In the case of dust clouds, this may
not be as obvious a problem as for gases, but the concept
of MESG does have some relevante in relation to
explosion isolation (see section on Mitigating and con-
trolling measures below). Pioneering work has been
performed in this area by Schuber (see Reference 1).
Initiation of dust explosions by shock waves has
been studied by several workers, and valuable insight
has been gained. The direct practica1 relevante of this
knowledge is less obvious than for some other aspects
of dust cloud ignition. One possibility is to make use of
the induction time for shock wave ignition in a model
of flame propagation in turbulent dust clouds (see next
section). An informative analysis of shock wave ignition
of dust clouds was given by Wolanski3g and research at
University of Michigan, USA is reported by Kauffman
et uZ.*. Boiko and Papyrin studied dust dispersion
behind a shock wave, using an advanced laser visualiz-
ation method. Ignition delays of different dusts in inci-
dent and reflected shock waves were estimated. Geng et
~1.~~performed a numerical study of the fluid-dynamic
effects of an incident shock wave passing through a dust
cloud on the delay for igniting the dust behind the shock.
Geng et ~1.~used a vertical shock tube for measuring
ignition delays of dust clouds behind an incident shock.
Lu and Fan4* developed a comprehensive analytical
model allowing prediction of ignition delay times of dust
clouds exposed to shock waves. Good agreement
between predictions and experimental data was obtained.
Hu et ~1.~~investigated the fast ignition and combustion
of wheat flour behind a shock wave in a shock tube.
of dust explosions: R.K. Eckhoff
Flume propugution processes in dust clouds
Some centra1 topics are the same as for flame propa-
gation in premixed gases, viz.:
0 laminar flames
?? flame acceleration mechanisms
?? turbulent flames
?? detonation
However, in the case of dust clouds
?? ignition and combustion of single particles in a dust
cloud
is an additional fundamental research topic.
An important differente between dust clouds and
premixed gases is that inertial forces in dust clouds can
produce fuel concentration gradients (displacement of
particles in relation to the gas phase). Furthermore, ther-
mal radiation may contribute significantly to the heat
transfer from the flame to the unburnt cloud, depending
on the type of particles (e.g. light metals).
Understanding luminar Jlame propagation pro-
cesses in dust clouds may be the key to understanding
how dust explosions develop in terms of pressure as a
function of time. Mazurkiewicz and Jarosinski44 studied
the gas composition just upstream of a stationary, stabil-
ized maize starch/air flame front in a burner. It was
found that CO, CO*, H2 and CH, were the main compo-
nents produced during the initial pyrolysis stage. The
CO, content increased with decreasing temperature. In
the gas phase oxidation reaction, burning of CO was
most important. Mazurkiewicz and Jarosinsk?
described an experimental burner for conical dust flames
in air, and measured laminar buming velocities and
flame temperatures for maize starch/air as a function of
dust concentration. The average flow velocities of
unbumt dust cloud through the burner were in the range
0.61465 m s-l, but these values were reduced by the
ratio of the bumer cross-section area and flame surface
area, to comparatively low estimates of laminar burning
velocity in the range 0.15-0.05 m s-. Proust45 described
independent experimental studies of laminar buming
velocities and maximum flame temperatures in clouds of
starch, lycopodium and sulfur in air, whereas Seshadri
et a1.46 studied the structure of laminar dust flames.
Bradley et al.47 investigated the buming of clouds of fine
graphite dust (4 Pm) in premixed methane/air, in a flat
laminar flame. The experiments gave further support to
the hypothesis that active radicals in the gas phase cata-
lyse the char oxidation. This work suggests a basis for
a mathematica1 model of the laminar combustion of
clouds of ultrafine coal dust. Gieras and Klemens48 com-
pared flame propagation in clouds of coal dust in air and
in methane/air, at normal and micro-gravity conditions.
This made it possible to isolate the influence of buoy-
ancy. In the absente of gravity, flame propagation
occurred at velocities very close to the corresponding
fundamental laminar burning velocities.
Van Wingerden49 performed basic work on laminar
flame propagation in Norway, which has disclosed a
need for reconsidering the phenomena involved in
steady-state upwards laminar flame propagation in verti-
cal ducts.
 Prevention and mitigation of dust explosions:
Krishenik and Shkadinskiiso developed a mathe-
matical model for flame propagation in dust clouds of
mixtures of two monosized particle fractions. Both con-.
ductive and radiative heat transfer were incorporated in
the model. Deng et ~1.~proposed that the burning velo-
city of a dust Zayer under specific laminar conditions
be regarded as the fundamental combustion property of
a dust. They did not, however, elaborate the idea to the
extent needed to resolve the implications of the proposal
with respect to dust cloud combustion.
Lee et aLS2 showed that theoretical equilibrium
properties of dust cloud combustion (constant-pressure
adiabatic flame temperatures, and maximum constant-
volume explosion pressures) calculated by standard
computer codes are in good agreement with experimen-
tal data obtained by various workers.
Wlodarczyk et ~1.~conducted experiments in a 5
litre spherical explosion bomb to determine the influence
of the dust concentration in exploding aluminium/air
clouds on the content of aluminium oxide in the reaction
products. Fan et cds4 developed a dynamic numerical
simulation model for the propagation of spherical Al
dust explosions in closed vessels. Good agreement was
found between experiments and theoretical predictions
for the influence of particle size on the pressure develop-
ment in the vessel.
Limiting cloud compositionsfor jame propagation
is an important fundamental research topic for at least
three different practica1 applications, viz. explosible/non-
explosible assessment, assessment of minimum explos-
ible dust concentration, and assessment of maximum
permissible oxygen concentration for inerting. Mintz5j
found evidente for the existente of a maximum explos-
ible dust concentration for dust clouds under certain cir-
cumstances. For a narrow size fraction (106-125 Pm) of
maize starch, a reasonably wel1 defined limit of 800-
1000 g me3 was found. The results were interpreted in
terms of a simple oxygen depletion model. The influ-
ence of the particle size distribution on the minimum
explosible dust concentration was investigated by Pole-
taev and Korolchenko56, using data from experiments
with polysized polyethylene dusts. Promising agreement
between theory and experiments was obtained.
Hertzberg et al.* measured minimum explosible
concentrations, maximum explosion pressures and
maximum rates of pressure rise at constant volume, and
maximum flame temperatures, for clouds in air of dusts
of 14 different metals. They found that for some metals
flame propagation appears to occur in a mixture of metal
vapour and air, similarly to the gas phase flame propa-
gation mechanism in clouds of organic dusts.
Hertzberg et ~1.determined the same parameters
for nine different dusts of solid explosives when dis-
persed as clouds in air in a closed bomb. In the low-
concentration range (6 400 g m-), the dusts behaved as
dusts of normal carbonaceous and plastic materials. At
higher concentrations, they became more hazardous.
It is wel1 known that pulverized coal and coal dust
in mines do not represent a dust explosion hazard unless
the content of volatiles exceeds 7-8%. However, this
does not apply to carbon dusts of specific surface areas
exceeding the order of 100 m* g- (N2 adsorption). Wie-
mann59 showed that dusts of such materials (active
R.K. Eckhoff 7
carbon/active coke) of considerably lower volatile con-
tent than 7-8% could give fully developed dust
explosions in the 1 m3 ISO vessel.
Turbulent combustion in dust clouds has been stud-
ied experimentally and theoretically by several investi-
gators. Eckhoff6 summarized some work on the influ-
ence of initial and explosion-induced turbulente on dust
explosions in closed and vented vessels. Wel et a1.6S62
re-emphasized the important role played by turbulente
in dust explosion propagation in closed vessels. Kauff-
man et uZ.* and Austin et aL9 summarized their research
on turbulent combustion of premixed dust clouds, and
Tamanini and Ural summarized their work on the
effect of initial turbulente on flame propagation in dust
clouds. Scheuermann63 also investigated the influence of
the initial dust cloud turbulente on the development of
dust explosions in vented enclosures. Rzal-Rebire and
Veyssire 64 addressed some centra1 basic aspects of tur-
bulent dust flames. Veyssire65 summarized the funda-
mental studies on flame propagation in dust clouds con-
ducted at LED in Poitiers, comprising laminar flame
propagation in dust clouds, the role of turbulente in
flame acceleration, and the conditions for propagation of
detonation-like, but non-ideal, combustion waves. In
their basic studies Rzal-Rebire and Veyssire% investi-
gated the interaction of a laminar maize starch/air flame
with an obstacle, viz. a sphere, a disk or a vortex ring.
With the ring, flame-quenching phenomena were
observed, which were attributed to centrifugal separation
of dust particles and air in the turbulent eddies. This is
a very important observation, indicating that the burning
rate of a dust cloud may not respond to turbulente in
the same way as the burning rate of a premixed gas.
Further work towards improved understanding of
the relationship between the dynamic state of a dust
cloud and its combustion rate should be encouraged.
Basic parameters need to be identified. Various strategies
have been indicated. For example, an induction time for
ignition may be taken as a global characteristic of the
combustion chemistry (shock tube or stirred reactor).
Geng et aL6measured the ignition delay behind an inci-
dent shock wave of Mach number 4-6 passing through
a cloud of maize starch in oxygen. UraP emphasized
the fact that different induction times are observed with
incident and reflected shock waves, due to different
ignition mechanisms. Boiko et al.69 measured ignition
delays for coal dust clouds exposed to reflected and inci-
dent shock waves.
An alternative approach is to consider the laminar
burning velocity as the fundamental parameter, as sug-
gested by Bradley et al.. Empirical relationships
between turbulent burning velocity and turbulente inten-
sity are then established, using the laminar burning velo-
city as a normalizing parameter. As pointed out by van
Wingerden49, numerical flame libraries can then be for-
mulated and used for closing the positive-feedback loop
combustion-expansion-jow-turbulente-combustion in
numerical dust explosion simulation codes. The on-
going research and debate on numerical modelling of
premixed gas combustion should be watched carefully
to ensure that any elements that may contribute to solv-
ing the dust explosion modelling problem are explored.
8 Prevention and mitigation
Understanding flame acceleration, due to flame dis-
tortion and turbulente produced by the propagating
explosion itself, is centra1 for understanding both dust
and gas explosions. Extensive experimental research
programmes have been conducted to study these
phenomena for gas explosions in obstructed geometries.
If the experimental facilities used in these experiments
are stil1 available, it would seem relatively straightfor-
ward to repeat the experiments in the various vessels,
using dust clouds instead of premixed gas. Systematic
comparison of results with previous data from gas
explosions would yield a valuable overview of simi-
larities and discrepancies, which would help to focus
basic research efforts on important areas where dust
cloud combustion may differ significantly from combus-
tion of premixed gases. Proust7 has given an informative
review of the state-of-the-art on propagation of dust
explosions in pipelines in relation to gas explosion
propagation in pipelines. Pu et aL7* studied experimen-
tally the acceleration of an Al dust/air flame in a one-
end-open horizontal tube of diameter 140 mm and length
5 m. With ignition at the closed end, maximum flame
speeds at the open exit end amounted to 1200 m s-. Hu
and Sun73 studied the mechanisms of fast combustion of
Al powder suspended in atmospheres of different oxygen
contents, using an explosion shock tube technique.
Much work has been, and is currently being done,
on turbulent combustion of sprays and mists74, which is
in part also relevant in the context of dust explosions.
Huang et UZ.75 observed that burning clouds of alu-
minium dust in air are electrically conductive. They attri-
buted this effect to generation of metal vapour by evap-
oration of the particles prior to combustion. Besides
being of fundamental interest, the observed effect also
has implications with respect to industrial safety.
The hope is that adequate mathematica1 codes for
computer simulation of dust JEame propagation pro-
cesses wil1 be available in a not too distant future. Kjld-
man76 has summarized his work on applying compu-
tational fluid dynamics to turbulent dust explosion
propagation. Application of the numerical model to peat
dust explosions in a closed 20 litre vessel showed prom-
ising agreement with experiments. Krause7 presented a
comparatively simple two-dimensional model for
numerical simulation of explosions in vented enclosures.
The turbulente sub-model was empirical, and the
explosible medium was regarded as a homogeneous, pre-
mixed gas. Comparison with more complex models gave
reasonable agreement. Comparison with dust explosion
experimental data was not reported. In a subsequent con-
tribution, Krause7* used this simulation model to predict
the maximum explosion pressure in a vented explosion
as a function of the vent area and the turbulente intensity
in the dust cloud just prior to ignition. He was able to
predict the earlier experimental finding of Tamanini et
al. (see Reference l), that the maximum explosion press-
ure in the vented vessel increases with increasing initial
turbulente intensity in the exploding cloud.
Finally, the singular phenomenon of dust cloud det-
onation should be mentioned. It is now generally
accepted that this can occur, but further work is needed
in order to establish an adequate understanding of the
deflagration-to-detonation-transition process (DDT). An
of dust explosions: R.K. Eckhoff
excellent review of the state-of-the-art and outstanding
problems in dust cloud detonation research has been
given by Kauffman et al. *. The current knowledge on
dust cloud detonations was also summarized by Alex-
ander et al.79. Kauffman et uZ.* and Austin et ~1.~ sum-
marized their extensive work on how detonations can
develop from accelerating turbulent combustion of dust
clouds in long tubes. Sichel and Kauffmannso studied
the transition from deflagration to detonation during dust
explosions in long ducts. The dust was initially deposited
as a layer along the duet floor, and the dust cloud was
generated by the entrainment of the dust layer by the
blast wave propagating ahead of the flame. Khomik et
uE.~~,~* determined experimentally the minimum critical
tube diameter for detonation propagation in suspensions
of a fine aluminium flake dust in air. The critical value
found was in the range 0.040-0.055 m. Korobeinikov83
conducted a theoretical study of the propagation of det-
onation waves in dust clouds. The problem of estab-
lishing adequate staling rules was given particular atten-
tion. Markovg4 presented a new method for numerical
simulation of non-steady detonations in dust clouds.
Two-dimensional computations yield a multi-wave
structure of the detonation process. Ding and Huang8
analysed the mathematica1 theory for the reaction zone
in a detonation wave passing through a dust cloud, and
proposed a new numerical criterion describing the
Chapman-Jouguet condition. Tulis et ~1.~~ conducted
detailed experimental studies of the structure of deton-
ation processes in clouds of aluminium in air. The influ-
ence of particle size and shape was studied, and various
detonation wave structures were identified. Paplinski and
WlodarczykE7 analysed the critical conditions for direct
initiation of detonations in dust clouds of infinite size.
Klemens et al.87 performed experiments in which deton-
ation waves in hybrid mixtures of methane, air and oats
dust were studied.
Blast waves generated by burning dust clouds
One case of practica1 interest is blast waves from
explosions in partly confined geometries, e.g. deliber-
ately vented, or bursting process equipment, and work
rooms. The strength and shape of blast waves from dust
explosions depend on the way in which the dust clouds
burn. For example, Wirkner-Bott et aL8* conducted a
fairly detailed study of the nature of the secondary
explosion, i.e. the explosion of unburnt dust cloud out-
side the vent opening. This phenomenon was discussed
further by Schumann and Wirkner-Botts9. Centra1 vari-
ables influencing blast wave generation, in addition to
type of dust cloud and geometry of system, include the
dynamic state of the dust cloud at the moment of
ignition, ignition point in relation to vent, the vent size,
and the vent cover opening pressure. Van Wingerden
gave an informative overview of pressure and flame
effects in the direct surroundings of installations pro-
tected by dust explosion venting. Some basic studies of
shock wave emission from burning dust clouds were per-
formed by Gelfand et ~1.~.
Medvedev et a1.92 studied, experimentally as wel1
as theoretically, the blast wave generated by sudden
expansion of a dust-filled enclosure, such as a hopper or
a pipe. The same authors 93 also studied the interaction
 Prevention and mitigation
between blast waves and dust deposits, using a specially
developed shock tube technique. The experiments
revealed a strong dependence of the pressure amplitude
transmitted through the dust on the duration of the com-
pression phase of the primary air shock wave. Smirnov
et aLy4 presented a new mathematica1 model of shock
wave propagation in dust clouds comprising poly-dis-
perse systems, i.e. a range of different particle sizes
within the same cloud. Inter-particle collision was not
considered. An experimental and numerical study of the
supersonic flow behind a shock wave passing through a
dust cloud was performed by Boiko et al.. Gelfand et
aLy6 investigated experimentally the attenuation of shock
waves propagating through dust clouds in a 50 mm
diameter shock tube. Reasonable agreement between
experimental data and analytical and numerical predic-
tions was found for incident shock waves of Mach num-
ber < 3.
A useful condensed introduction to the complex
field of properties and effects of blast waves from
explosions has been given by Harmanny.
The effect of a given blast wave on humans, build-
ings and process equipment is an important area where
more research is needed. Valuable reviews are given by
MercxcJx and LAbb. Britan er al. 0studied the inter-
action of shock waves with layers of water-based foam
used in fire fighting. Induction times and time constants
for foam layer destruction were determined. They also
analysed the features of the transmitted shock wave and
the waves reflected from the air/foam boundary and from
the walls of the experimental channel.
Status and outstanding problems in
preventing and mitigatingkontrolling dwt
explosions in industrial practice
The role of fundamental knorvledge in developing
practica1 solutions
The status on the fundamental side, outlined above, is
necessarily also the status as regards the genuine under-
standing of the various corresponding phenomena in
industrial practice. indicated in Table 1. As already
emphasized, the strengthening of the links between fun-
damental research and industrial research and develop-
ment should be encouraged. This wil1 promote the use
of available basic knowledge in the design of practica1
measures, as wel1 as direct the basic research into areas
useful for practice.
Table 1 Means for preventing and mitigatingicontrolling dwt
and development
Prevention
Preventing ignition sources Preventing explosible
Self-heating in dust deposits Inerting by inert gas
Open flames Inerting by inert dust
Hot surfaces Keeping dust concentration
Burning metal particles explosible range
Electric sparks and arts
Electrostatic discharges
of dust explosions: R.K. Eckhoff 9
Deng and He pointed out the need to use thermo-
dynamics, chemical reaction kinetics, and fluid dynamics
for proper description of ignition and flame propagation
phenomena in dust clouds and layers. These topics con-
stitute centra1 elements of classica1 chemical reaction
engineering, and Deng and He proposed a corresponding
concept Dust explosion reaction engineering (DERE)
for the dust explosion domain.
Siwekio3 presented a concentrated overview of
current methods for dust explosion prevention and miti-
gation in the process industries, based mainly on experi-
mental research and development performed within the
Swiss/German domain. Siweks paper reflects the
important fact that industry needs practicable solutions
for today, and cannot wait for more ideal solutions that
may become available in the future. However, industrial
pragmatism must not, on the other hand, black the con-
stant strive for better solutions based on improved basic
understanding of the phenomena involved. It seems as
if the mutual understanding and respect between the two
parties, industry and the researchers, is growing.
EckhoffOJ reviewed the state-of-the-art on pre-
venting and mitigating dust explosions in the ferro-alloys
industry. In general, both the electric spark ignition sen-
sitivity and the explosion violente (K,,) of metal dusts
increase with decreasing particle size, right down to the
1 Pm region. In the past, this has not always been taken
into account. Often particle size was merely specified in
terms of less than 74 Pm or less than 63 Pm, which
is by no means satisfactory. More research is needed, in
particular on alloys, where the most hazardous compo-
nents may sometimes accumulate in the fine tail of the
particle size distribution.
Generation and states of industriul dust clouds
Comparatively little quantitative knowledge of practica1
use has been generated. It is known, however, that the
dynamic state of a dust cloud dramatically influences
both its ignition sensitivity and its combustion rate.
Therefore, it is clear that systematic quantitative assess-
ment of this state is of great practica1 significante.
Experimental investigation of typical processes of gener-
ation of dust clouds and the resulting states of the clouds,
in various types of process equipment and modes of
operation, should be encouraged. This is a typical area
where close interaction of fundamental and applied
research can be highly beneficial. The work of Hauert
explosions in the process industries. Topics for applied research
dust cloud Mitigation/Control
Explosion-pressure-resistant process
equipment
outside Isolation
Partial inerting by inert gas
Explosion venting
Explosion suppression
Preventing secondary explosions (good
house-keeping for preventing dust layer
formation)
10 Prevention and mitigation
et ak4, discussed above, constitutes an important step in
the right direction.
Preventing ignition sources
A considerable amount of fundamental knowledge is
available, and the time is ripe for accepting that the con-
cept of ignition comprises a range of very complex pro-
cesses. Simple parameters such as a minimum ignition
energy or temperature are not true constants for a given
dust, but vary significantly with the geometry and other
properties of the ignition source, as wel1 as with the state
of the dust cloud. Further development of correlations
between results from standardized tests and actual criti-
cal ignition conditions in practice should be encouraged.
Gibson gave a valuable summary of methods for
preventing ignition of powders and dusts in drying oper-
ations. ZockollOh described a new system for early
detection of se(f-heating/self-ignition in deposits of
organic powders in spray driers, based on detection of
low concentrations of CO.
Xu et al. found that very weak ekctric spark dis-
charges, in the range 0.1-1.0 mJ, are probably able to
initiate smouldering combustion in linen flax. Electro-
static hazards in connection with industrial use of flex-
ible big bags were discussed by Rogers and by Dahn
et al. 09. Pratt presented three case histories in which
electrostatic spark discharges were generated during
pneumatic transport of powders. Sun Keping con-
structed a new type of electric charge eliminator for
charged powders being pneumatically conveyed in pipes.
The method was tested out successfully in industrial
practice during a period of one year. Wang and Lou~
discussed the electrostatic hazards in powder handling
and storage in general, and proposed methods for
preventing/reducing this hazard.
An example of commercially available equipment
for preventing ignition in industrial plants was given by
Kleinschmidt3, who reported that the GreCon system
for detection and extinction of sparks in terms of Jlying
burning particles is being developed further with respect
to optimizing system performance. The same applies to
the Firefly system, which, according to Janssen, offers
an adjustable lower particle temperature limit of detec-
tion, down to 150C. A multi-zone checkpoint system
prevents false alarms, and indicates the size of the hot
object (single particle, several particles or extensive
flame). Depending on the detection temperature and the
nature of the industrial process, detection of a hot object
may give rise to either activation of an extinction system,
close-down of plant, or simply adjustment of plant run-
ning conditions to prevent further hot object generation.
A similar spark detection and extinguishing system has
also been developed by T&B Electronic .
Lloyd and Laar presented the state-of-the-art
on international standardization of classication ofareas
containing combustible dust, and guidelines for selection
of electrical equipment for such areas. Centra1 organiza-
tions involved in this work are the International Electro-
technical Commission (IEC), and CEN and CENELEC
of the European Community.
of dust explosions: R.K. Eckhoff
Preventing explosible dust clouds
Inerting of the dust cloud can be accomplished by
adding gases such as nitrogen or carbon dioxide to lower
the volume percentage of oxygen in the atmosphere to
a leve1 at which the dust cloud can no longer propagate
a self-sustained flame. A fair amount of data exists for
maximum permissible oxygen content in the atmosphere
for inerting. However, there is room for improving the
test methods by which such data are obtained. Further-
more, most data are for atmospheric pressure and normal
temperature. Data for other conditions, in particular for
elevated temperatures and pressures, are sometimes
required, and adequate test methods should be
developed. Glor determined maximum permissible O2
contents for inerting clouds of coal dusts at elevated tem-
peratures and pressures. Wolinski and Hayashi18
determined standard explosibility parameters of dusts
and various rare metal alloys in air containing halon
130 1 and added nitrogen. Theories for flame propagation
limits (see discussion above with reference to the third
column of Table 1) would be useful even in this context,
and their development should be encouraged. Krause et
al. 19 proposed a simplified, approximate method for
first-order estimation of the minimum oxygen concen-
tration for flame propagation in dust clouds. Maddison*O
summarized some important aspects of inerting powder
handling plants using nitrogen, carbon dioxide and other
inert gases.
Gao er al. 12described a new system for injection
of pulverized bituminous coal into a blast furnace for
steel production, for which nitrogen inerting constitutes
a centra1 safety measure. Full or partial inerting by inert
gas also constitutes a primary element in the strategy for
reducing the coal dust explosion hazard in melted slag
reduction furnaces, as outlined by Wang Junyi et al.22.
Whilst reducing the oxygen content in the atmosphere
prevents dust explosions, it can introduce a suffocation
hazard. However, it has been shown that adding a few
vol% CO2 to the gas mixture considerably reduces the
critical oxygen threshold for suffocation. An inert gas
mixture (INERGEN) utilizing this effect is now being
marketed by Dansk Fire Eater A/S23. Further work to
identify gas mixtures that keep the dust cloud inert with-
out presenting a suffocation hazard should be welcomed.
Inerting by adding non-combustible dust is not gen-
erally applicable, because the inert dust wil1 in most
cases cause unacceptable contamination. However, there
are cases where the dust/powder processed is already a
mixture of combustible and non-combustible dusts, and
where control of the composition ensures that dust
clouds are non-flammable. It is then essential to avoid
segregation of combustible and non-combustible compo-
nents throughout the process. Zhang Zhancheng et al. 24
studied the possibility of preventing grain dust
explosions in grain storage facilities by adding talc pow-
der to the grain stream. The method works, as far as
explosion prevention is concerned, but the industrial
hygiene aspects require careful consideration.
Keeping the dust concentration below the minimum
explosible concentration is a third means of maintaining
dust clouds non-explosible. More work is needed to
establish procedures for the determination of dust con-
centrations in industrial situations. Zockoll25 described
 Prevention and mitigation
an apparatus, based on infra-red light attenuation, for
determining rhe dusr concenrrarion of dust clouds gener-
ated in various situations in industry. Typical measure-
ment results were reported, demonstrating the feasibility
of the method. Xu Bowen et al. 26proposed a systematic
method for estimating the dust explosion hazards in
industrial plants. The method implies an empirical
relationship between local dust concentration and the
local rate of dust deposition from the cloud [mass (time
area)-]. Properties of the dust emission source must be
known. Also, research is needed to establish the mini-
mum mass of dust deposit per unit of surface area that
is required for maintaining a self-sustained explosive
combustion along the surface. The assumption that
entrained dust wil1 become distributed evenly throughout
the available space may not hold. For example, a thin
dust layer on a floor may be dispersed into just a shallow,
dense dust cloud close to the floor, through which the
flame can sweep. The conditions required for producing
this kind of self-sustained shallow sweeping Aames need
to be investigated further in the context of the frst col-
umn of Table 1.
Wolanski27 suggested that the concept of mini-
mum hazardous mass of dust be introduced in the evalu-
ation of dust explosion hazards in practice. This para-
meter is defined as the minimum mass of the actual dust
that can generate a dust explosion of destructive
strength. The parameter is not a constant for a given
dust, but depends also on the characteristics of the actual
enclosure in which the explosion takes place. The basic
idea is as follows: a quantity of dust, unable to generate
a cloud of concentration above the minimum explosible
value when being dispersed evenly throughout the entire
enclosure volume, can present a significant explosion
hazard when being dispersed in the enclosure as a
smaller and correspondingly denser cloud.
Xu Bowen et al. 2Xfound that the minimum explos-
ible concentration of linen flax dust in air was inde-
pendent of particle size up to about 100 Pm. For larger
particles, a systematic increase of MEC with increasing
particle size was found. Mittal12 discussed various
mathematica1 models for calculating minimum explos-
ible concentrations of dust clouds. Shao and Wang130
reviewed some non-intrusive methods for measuring the
dust concentration in dust clouds, primarily in pipes
and ducts.
Mitigating and controlling measures
The use of explosion-pressure-resistant process equip-
ment for full explosion conjinement is limited because
of high equipment costs. Current experimental methods
allow sufficiently accurate prediction of maximum
explosion pressures in simple vessels with point source
ignition. This also applies to explosions at elevated
initial pressures. Torrent and Menndez found that the
proportionality between initial and final pressure found
previously in smal1 laboratory-scale vessels also holds
in vessels of 1 m3 volume, up to 12 bar initial pressure.
However, if complex dynamic pressure development,
e.g. with pressure piling, is to be expected, closed-bomb
test data are of limited value. There is also room for
further improvement in the design of the process equip-
ment itself, with respect to minimizing its heaviness. The
of dust explosions: R.K. Eckhoff 11
German concept of pressure-shock-resistant design
Should be further developed. Crowhurst* gave a USefUl
overview of the state-of-the-art on desig? of equipment
to withstand a given overpressure caused by a fully con-
fined or vented explosion. Bartenev et al. 33 developed
a mathematica1 model for the failure of a process vessel
subjected to an internal exothermic, comparatively slow
process creating an overpressure, allowing the maximum
initial velocities of fragments or vent covers to be esti-
mated. Bartenev et al. 34extended this work to the case
where the pressurized vessel is filled with dust. Exper-
iments revealed that the presence of dust can have a sig-
nificant effect on the pressure development inside the
bursting vessel, and on the kinematic parameters of the
ejected fragments.
The objective of explosion isolation is to prevent
dust explosions from spreading from the primary
explosion location to other process units, workrooms,
etc. Van Wingerden et al. reported on dust explosion
experiments in integrated systems of various process
equipment connected with ducts. Due to pressure piling,
jet-initiated high initial turbulente and turbulent jet
ignition, very high pressure peaks can be generated even
in generously vented vessels. This investigation is a
further demonstration of the need for effective means of
explosion isolation in coupled systems. Various passive
and active techniques for explosion isolation have been
developed and are being used, but there is room for
further improvement. If adequate performance can be
achieved, passive techniques are clearly more attractive
than active ones.
A basic understanding of flame propagation and
pressure build-up in coupled geometries (interconnected
vessels) is important for the prediction of the perform-
ance of various active and passive isolation equipment.
Valuable large-scale experimental work in this area was
reported by Lunn and by van Wingerden and Alfert.
The understanding of flame propagation in long ducts
and pipes is also very important. Vog138 presented
results from a comprehensive experimental investigation
of the propagation of dust explosions in pipes for pneu-
matic powder transportation. Pipe lengths up to 48 m and
pipe diameters up to 200 mm were used. The influences
of a range of experimental parameters on the flame speed
and explosion pressure were studied. The parameters
were: initial air velocity in the pipe, pipe diameter,
location of the ignition source, dust concentration in the
pipe, and the Ks, value of the dust. Zellweger
described further work to improve passive and active
isolation valves of the VENTEX type. Closing times
(from sensing of the explosion to valve fully closed)
down to 12 ms were obtained for active valves. A sim-
plified VENTEX valve, operating in one direction only,
has also been developed. Passive explosion interrupters
based on venting at a bend have been in use for some
time. However, there is room for further exploration of
the potential of this attractive, simple principle of
explosion isolation. A new low-pressure-drop design
was described by Alfert and Fuhre140 and by van Wing-
erden and Alfert13. Glor reported work on the per-
formance of explosion barriers in ducting connected to
vessels with venting or automatie explosion suppression.
Klincewicz and Kordylewski described a new, prom-
12 Prevention and mtgaton of dust explosons:
ising design of an explosion diverter for interrupting
explosions in pipe lines. The new design avoids the
pressure drop created in normal operation by passive
diverters, but requires active triggering. Cybulski et
al. 42, addressing the problem of coal dust explosions in
coal mines, used solar panels for automatie detection of
the coal dust flame and simultaneous actuation of water
barriers. The water was contained in plastic bags, which
were opened by a detonating cord triggered by the flame-
generated power from the solar panel.
Partial inerting by inert gas is a promising means
for mitigating dust explosions, which deserves further
attention. The idea is that as the oxygen content in the
atmosphere is decreased, there is a gradual decrease of
both ignition sensitivity and combustion rate of the dust
cloud. In some cases, the explosion hazard may be
reduced substantially by only a moderate reduction of
the oxygen content in the gas. However, more research
seems necessary in this area to establish a correlation
between the oxygen content in the gas and various ignit-
ability and explosibility parameters. Interesting new
information is currently being produced in a joint Euro-
pean research programme (CREDIT), focusing on the
influence of oxygen content of the atmosphere on the
minimum energy and temperature for ignition of dust
clouds.
Dust explosion venting remains a complex and con-
troversial subject. Adequate understanding of flame
propagation processes in dust clouds is essential for the
design of optimal venting arrangements for industry.
Useful reviews of various aspects of dust explosion vent-
ing in practice were given by Scholl4 and Lunn44*45.
The basic understanding of flame propagation processes
inside and outside vented enclosures is stil1 unsatisfac-
tory. This implies that neither the processes by which
dust clouds of given structures are generated, nor the
way in which clouds of given initial structures burn, are
wel1 understood. Consequently, adequate venting the-
ories do not exist, and one must rely on experiments.
During the last few years, the need for differentiating
vent area requirements in view of the different turbu-
lente levels, degrees of dust dispersion and concen-
tration distributions of dust clouds which occur in prac-
tice has become widely accepted. Eckhoff46 discussed
this problem with particular reference to venting of large
silo cells. Deng et al. 14conducted vented maize starch
explosions in a 95 m3 vertical experimental silo of
UD = 3. The 3.4 m2 vent was located in the silo roof.
The maximum vented explosion pressures were com-
paratively low. Hchst et al. 4X conducted systematic
maize starch/air explosion experiments in a top-vented
silo of volume 50 m3 and UD = 4. Both dust concen-
tration and initial turbulente (prior to ignition) were
monitored. The dust injection process as wel1 as the
ignition point location were varied. Flame propagation
and pressure build-up during the vented explosion were
measured. Comparative experiments with methane/air in
the silo were also performed.
Molkov et al.14 studied experimentally the influ-
ence of the inertia of the vent cover on the gas dynamics
of the venting process. Good agreement was obtained
between pressure-versus-time traces from experiments
and corresponding traces obtained using the numerical
R.K. Eckhoff
simulation code DYNAMICS. Fan Xisheng et al. Is0
studied, theoretically as wel1 as experimentally, the
influences of details of the mounting flange arrangement
on the static bursting pressure of bursting
panels/membranes. They observed two distinctly differ-
ent bursting patterns, viz. bursting of the membrane in
its centra1 region, and bursting along the edge.
The paper by Crowhurst2 on the design ofenclos-
ures to withstand a given maximum explosion pressure
also applies to vented enclosures. Harmanny pre-
sented a new formula for predicting the duration of
vented dust explosions in enclosures of volumes from
10 to 60 m3. This is useful for evaluating whether static
pressure considerations or impulse considerations apply
when predicting the response of the enclosure structure
to the explosion load.
A further dimension of complexity is added to the
venting problem if the initial pressure (and/or
temperature) deviates from atmospheric. Results from
venting of dust explosions in air of elevated initial press-
ure were reported by Siwek et al. 15*. The effect of press-
ure piling and turbulent flame jet ignition on vent area
requirements in systems of interconnected vessels were
studied by Lunn6.
In dust explosion venting, maintaining the integrity
of the enclosure is not the only concern. Venting implies
that both pressure waves and flames are emitted into the
surroundings, and this may present a hazard, depending
on the size of the emitted flame and the magnitude of
the blast wave. Crowhurst et al. reported a series of
large-scale venting experiments (40 m3 and 20 m)
where external blast waves and flame lengths were
determined and compared with existing empirical corre-
lations (see also the section above on Preventing explos-
ible dust clouds). Peng et al. Is4 developed a labyrinth
type flame arrestor for mitigating flame and pressure
effects from dust explosion venting. Experiments using
a 2.7 m3 enclosure and a 370 mm diameter vent opening
suggested that this arrestor concept works satisfactorily
for Stl dusts. The maximum explosion pressures in the
vented enclosure, with the arrestor mounted in the vent,
were only a few per cent higher than the pressures with-
out the arrestor, whereas both flame and pressure effects
were substantially reduced. Li Gang et al.55 presented
a new version of the quenching tube first developed in
Norway (see Reference l), in which the conventional
bursting diaphragm had been replaced by a re-usable
vent cover. This solution is of interest in situations where
the expected frequency of explosions is comparatively
high.
Venting of industrial buildings requires special con-
siderations. A useful overview was given by
Crowhurst.
The injluence of vent ducts on the maximum
explosion pressure in the vented vessel has been studied
experimentally by several workers. Recently Ural pre-
sented a theoretical model for vented gas explosions by
which he has been able to calculate pressure-versus-time
characteristics in the vented vessel that agree wel1 with
corresponding experimental data. It remains to be inves-
tigated, however, whether this theory can reproduce
existing experimental data.
 Prevention and mitigation
Dust explosion venting remains an area in which
considerably more work is required. The new German
draft VDI-3673 venting guideline, issued in 1992, has
met some objections in other countries. This in particular
applies to the concept of heterogeneous dust clouds.
Nevertheless, computer-based software for easier use
of this draft guideline was presented by Alfertls8.
Hattwig and Hensel 59 discussed deficiencies in the new
VDI draft guideline on the basis of dust explosion exper-
iments in 4.5 m3 (UD = 2) and 85 m3 (UD = 4) vented
silos of square cross-sections.
Automatie explosion suppression is an active, com-
paratively sophisticated method of dust explosion
mitigation/control, which is used when simpler and less
expensive methods cannot be applied. Although this
method has been in use for many years, there is stil1 a
need for research and development. Moore60 reported
that the number of suppressant bottles of a given size
required for suppressing explosions of a given dust in a
given vessel was reduced by a factor of 0.2-0.3 when
the dust clouds were generated by industrial pneumatic
injection rather than by the VDI method used in previous
experiments. G10r~ and Moore6 reported current work
on the possibility of applying this method even in the
case of highly explosible organic dusts of K,, > 300
bar m s-. In the case of aluminium powders, satisfactory
suppression has not yet been achieved for powders of
K,, > 200 bar m ss, which means that only dust
explosions in clouds of relatively coarse aluminium
powders can be suppressed. The influence of the
dynamic state of the dust cloud at the moment of sup-
pressant injection, the influence of the suppressant injec-
tion on this state, and development of improved sup-
pressants are some of the areas where further work
would seem useful. Recent experiments have indicated
that water can be an effective suppressant, if injected at
a temperature > 180C. According to Tyldesleyh2, opti-
mum suppression is found to require about 0.5 litres of
water per m of vessel volume: 16-18% of the super-
heated water is flashed to steam, the remainder forms
very smal1 droplets. Reduced explosion pressures of 0.3-
0.4 bar(g) were obtained in experiments in a 28 m3
experimental vessel, using a dust of K,, = 150 bar m ss.
Gieras et af. conducted a series of experiments in order
to optimize the shape and mass of the explosive charge
used for automatie release of suppressant (powder) for
suppressing dust explosions. The overall aim of the
research was to minimize the powder ejection time.
Siwek described experiments where a combi-
nation of explosion venting and automatie suppression
was adopted for mitigating/controlling dust explosions
in various enclosures. Sliz et a1.1h4investigated the per-
formance of such a combined system for mitigating grain
dust explosions in an 8 m experimental vessel. Different
types of vent covers and suppressors were tested.
Prevention of secondary explosions outside process
equipment remains an important issue. Proper house-
keeping is an esstntial means of achieving this aim.
However, further work is required for assessment of the
maximum acceptable mass of deposited dust per unit
area of surface for preventing secondary dust flame
propagation under various conditions. Cybulski et al. M
showed that comparatively weak secondary dust
of dust explosions: R.K. Eckhoff 13
explosions in short, narrow tunnels in grain elevators can
be extinguished by properly designed, actively triggered
water barriers. Cybulski et al. M presented results of
experiments with propagation of weak coal dust
explosions in a network of full-scale mine galleries. A
main conclusion was that, under the conditions prevail-
ing, the possibility of flame penetration into blind gallery
branches was small. This kind of work may also be of
relevante to the analysis of flame propagation in large
industrial systems, e.g. in grain storage and handling
plants.
Gieras et al. 67 studied the development of com-
bustible gases (H2, CO and CH,) during combustion of
fuel-rich clouds (up to 5 kg mmR)of grain dust in air.
This is an important aspect in the context of industrial
safety, because mixtures of combustible gases and air
can give rise to severe secondary gas explosions, follow-
ing comparatively slow primary fires in fuel-rich dust
clouds.
Quantitative risk analysis (QRA) is receiving
increasing attention as a potential means of controlling
explosion risks in the process industries. This also
applies to the dust explosion hazard. The subject is com-
plex, and in part controversial, not least because of a
lack of relevant quantitative failure rate data. However,
attempts are nevertheless being made to apply QRA to
the dust explosion problem, as demonstrated by
Wagner6x. Mittal69 discussed the risk of dust
explosions in pneumatic powder transportation systems.
Potential sources of ignition were identified, with the
main focus on electrostatic discharges. A systematic risk
evaluation and risk reduction approach was outlined. Li
and Wangi7 proposed a quantitative, computer-based
decision support system for assessing the explosion
hazard and taking appropriate actions to prevent and
mitigate coal dust explosions in metal production plants.
Moore and Frehill7 outlined a philosophy for system-
atic analysis of the risks of dust explosions in industrial
plants. Specific technical solutions for preventing and
mitigating explosions for some typical cases were pro-
posed.
Status and outstanding problems in testing
of dust ignitability and explosibility
Historica1 background and standardization of methods
When some of the older test methods were designed, the
ambition was in fact quite modest. The original intention
was just to establish some relative measures of properties
of practica1 relevante to preventing and control-
lingimitigating dust explosions. Later some of these
methods were adopted as official standards, and test data
were sometimes treated as basic physical constants for
a given dust to an extent far beyond the original purpose
of the test. As more knowledge from systematic research
became available, the lack of justification for this use of
these test data was pointed out, and the arbitrary, relative
nature of the various test methods was brought to light
again.
The situation today is complex. It is realized that
only a few of the dust parameters that are currently being
used for characterizing ignition sensitivity and explos-
ibility of dusts can be regarded even as approximate
physical constants for a given dust. In most cases, a
 14 Prevention and mitigation
great number of variables are involved and a differen-
tiated view is required. Typical examples are the mini-
mum ignition energy and the explosion violente of
dust clouds.
A number of standard test methods have been
developed through the years, e.g. by IEC (International
Electrotechnical Commission) and ISO (International
Standardization Organization). In the USA, ASTM
(American Society for Testing and Materials) has issued
a number of standards in this area. Recently, the stan-
dardization organization of the European Union, CEN,
has launched a multi-year programme to produce a series
of standard methods for the testing of explosion propa-
gation limits, ignition sensitivity and explosion violente
of combustible dust clouds. Reference 1 gives a compre-
hensive presentation of the wide range of test methods
in use in Europe and the USA.
Two alternative approaches for achieving
differentiation
A first approach for handling this situation is to have
several test methods for any given parameter, allowing
test conditions to be selected according to the practica1
use of the test result. In the case of minimum ignition
energy, such an approach has in fact been incorporated
in the new IEC standard issued in 1994. Whereas an
appreciable inductance is to be included in the capacitive
discharge circuit in the case of standard testing, to obtain
the most incendiary sparks, it is argued that this is not
relevant if the test result is to be used for assessing the
electrostatic spark discharge hazard.
A similar situation should arise when testing for
explosion violente. It is now appreciated that the stan-
dard ISO Ks, value of a given dust reflects a rather
extreme combustion rate in the conservative direction,
because the turbulente leve1 and the degree of dust dis-
persion in the test are rather extreme. If, for example,
K,, values are stil1 to be used for sizing of explosion
vents, differentiation by varying the intensity of the dust
cloud formation process used in the test to fit the practi-
cal situation of interest could be worthwhile considering.
The work by Liu et al. 17is relevant in this context. They
described introductory studies of the turbulente structure
in experimental dust clouds in the 1.2 litre Hartmann
bomb under various conditions of dust injection. This
apparatus was traditionally used for characterizing the
explosion violente to be expected from various dusts.
Because of its smal1 size, the Hartmann bomb has in
recent years been replaced by larger test vessels.
However, there is a second, alternative approach for
handling the need for differentiation. In this case, the
measured dust ignitability or explosibility parameter
should be of a more basic nature, and differentiation to
meet various practica1 situations in industry should be
via adequate theory, using the basic parameter as input.
In the case of explosion violente assessment, the exist-
ing K,, concept may be replaced by the laminar burning
velocity of the dust cloud, for example. It is clear, how-
ever, that theories for transforming such basic para-
meters to flame propagation rates in practice must
involve complex fluid dynamic computer simulation
codes.
of dust explosions: R.K. Eckhoff
In the case of minimum ignition energy, the trans-
formation would imply that the basic test value be con-
verted to minimum discharge energies for ignition of the
same dust cloud by various kinds of discharges, such
as direct electrostatic two-electrode sparks, break flashes
(e.g. live wire rupture) and one-electrode discharges.
However, theories are scarce and the time may not yet
be ripe for this approach. Siwek and Cesana173 presented
a new test apparatus for the determination of minimum
electric spark ignition energies of dust clouds down to
below 1 mJ spark energy. The measurements in the low-
energy range are demanding and require carefully
designed equipment and skilled personnel. Zhou
Benmou et aZ.74 discussed the features of an apparatus
for determining the minimum electric spark energy for
igniting very sensitive explosives such as lead azide.
Spark energies as low as 1 PJ were attainable, using a
travelling-electrode system.
Limits of JEame propagation - a special problem of
scale
Determining limits of flame propagation constitutes an
important test objective (explosible/non-explosible
assessment, and assessment of minimum explosible dust
concentration and maximum permissible oxygen con-
centration for inerting). However, special care must be
exercised in designing flame-propagation-limit tests.
One basic problem is that near the limits self-sustained
flame propagation cannot be established unless a con-
siderable amount of energy is supplied for initiating
flame propagation. Hence, if the volume of the experi-
mental dust cloud is too small, it is difficult to assess
whether observed flame propagation is truly independent
of the ignition source. Some recent results by Cashdollar
et aZ.175 and Cashdollar and Chatrathi76 are of funda-
mental significante in this context. They found that
clouds in air, at normal ambient conditions, of an anthra-
cite coal dust of 8% volatile matter did not show self-
sustained flame propagation in a 1 m3 test chamber, even
when being exposed to a 30 kJ chemical igniter. How-
ever, in a 20 litre chamber, fully developed explosions
were generated even with a 5 kJ chemical igniter. The
reason for this could be that in the smal1 chamber, due
to the initial combustion and expansion of the dust cloud
directly affected by the ignition source, the pressure and
temperature in the unbumt cloud ahead of the flame
increase significantly before flame propagation no longer
receives support from the ignition source. Consequently
the self-sustained flame propagation, if any, occurs in an
adiabatically pre-compressed dust cloud, rather than in
a cloud of normal ambient temperature and pressure.
These results suggest that great care must be exercised
whenever comparatively smal1 chambers, in particular
closed ones, are used for any explosion limit determi-
nation.
Zhou Congzhang et a1.77 proposed a new, interest-
ing procedure for determining minimum explosible dust
concentration values in closed-bomb explosion exper-
iments. They presented experimental evidente indicating
that at the minimum explosible concentration, the time
interval from ignition to the pressure peak has its highest
value. They proposed that this criterion be used instead
of some arbitrary pressure rise criterion of explosion.
 Prevention and mitigation
Matsuda and Itagaki, comparing dust explosions in a
30 litre explosion bomb, and in a 1 m3 vessel, found that
the ranges of explosible concentrations in the 30 litre
vessel were considerably wider than those in the 1 m3
vessel for the same dust. A marked increase of the
explosible range was found in the 30 litre bomb when
increasing the ignition energy from 1 to 10 kJ. This
effect was practically absent in the 1 m vessel in the
ignition energy range 4-20 kJ. Tian Renqu er al. ,
using a 20 litre explosion bomb, found that the minimum
explosible concentrations of coal dusts decreased by a
factor of two or more when the ignition energy was
increased from 2.5 tol0 kJ. When using a 2.5 kJ igniter,
and adding 2 vol% methane to the air, the minimum
explosible dust concentration dropped by at least a factor
of two, compared with the values for dust in air. This
hybrid effect has been studied previously by several
other workers. Xu Tianrui et al. also arrived at the
conclusion that the apparent minimum explosible dust
concentration determined in a 20 litre bomb depends
markedly on the ignition energy. A value for ignition
energy of 10 kJ was found to be too high to yield
realistic results.
Pu et al. concluded that the turbulente structure
of experimental dust clouds in a standard 20 litre spheri-
cal dust explosion test bomb had little resemblance to
turbulente structures in dust clouds in accidental dust
explosions in industry.
Miscellaneous
Wang and ZhanglX determined the minimum ignition
energy, the minimum explosible concentration, and the
maximum explosion pressure for clouds of TNT dusts
in air. The values are similar to those of natura1 organic
materials. The results confirm that dilute clouds of dusts
of explosives do not exhibit explosive properties, but
behave as clouds of ordinary combustible dusts.
Expert systems - friends or enemies?
During recent years, there has been an increasing interest
in developing sophisticated computer-based expert sys-
tems for evaluation of dust explosion hazards and assess-
ment of optimal safety design features. Haefen and
Scheckerix presented such a system for assessment of
dust explosion hazards in industry and selection of
appropriate means of prevention and mitigation, which
is in al1 essentials based on the German protection philo-
sophy. Wachix4 presented another expert system
designed for the same purpose, but the technical and
philosophical basis was not explicitly stated. A compre-
hensive expert system developed in the UK was pre-
sented by Tyldesleys, and the need for quality assur-
ante of such systems was emphasized.
The development of this kind of expert system is a
natura1 consequente of two main factors. The first is the
almost explosive development of the performance of
personal computers. The second is the steadily increas-
ing knowledge about ignition and explosion phenomena,
which demands a steadily more differentiated and com-
plex approach for solving the practica1 design problems.
As long as this development is conducted by people
who are not only experts on computers, but also on the
of dust explosions: R.K. Eckhoff 15
physics and chemistry of the phenomena treated, expert
systems should indeed be welcomed. However, there
may be a possibility of the future market place being
offered software that is not up to acceptable standards
with respect to the physics and chemistry. As long as the
interior of the system is not fully exposed, deficiencies in
the basics may not be obvious to the user.
The time may very soon be ripe for making
adequate quality assurance of this kind of system com-
pulsory. A need may emerge for establishing some inter-
nationally recognized body of experts that can ensure
that expert systems offered in the area of explosion pre-
vention and mitigation are up to acceptable standards.
The human hazard factors
The present survey deals with the chemistry, physics and
technology of dust explosion prevention and mitigation.
However, a brief mention should also be made of the
importante of the human factors in this effort. This
aspect has been discussed by Fernandoxh.
Joint research efforts in Europe, and
research and development in the Peoples
Republic of China
Over recent years, a steadily growing potential for
organizing joint European research efforts has emerged
within the EU/EFTA/EUREKA system. This also
applies to dust explosion research. The British Materials
Handling Board (BMHB) in the UK has played a centra1
role in this processx7. A number of parallel research pro-
grammes have been started within the European Unions
Credit Project. Gibsonxx summarized the areas requir-
ing further work under the headlines:
?? combustion processes in dust clouds (experiments,
theoretical models)
?? identification and control of ignition sources
?? design of methods to prevent/protect against dust
explosions
These headlines in fact cover most of the research needs
identified in the present review. In addition, mechanisms
of generation of dust clouds, which are important
because they set the stage for subsequent ignition and
combustion, require further exploration.
Wang DongyanXy, characterizing the Peoples
Republic of China as a developing country, emphasized
the need for increasing the efforts to prevent dust
explosion accidents in Chinas rapidly growing industry.
Of the number of dust explosions recorded in this coun-
try during the decade 1980-1989, 65% were in the grain
industry, 17% in the textile industry, 12% in the coal
industry and 6% in the metallurgical industry. With the
rapid development of the chemical and metallurgical
industries, the annual number of explosions may easily
rise if adequate precautions are not taken. There is a
strong need for education and training on al1 levels, and
for adequate safety technology. The 6th International
Colloquium on Dust Explosions in Shenyang, Peoples
Republic of China, in August/September 1994, demon-
strated that research and development on dust explosion
prevention and protection in this enormous country is
growing at great pace.
 Prevention and mitigation
Conclusion
Initiation and propagation of industrial dust explosions
are, from a fundamental scientific point of view,
extremely complex phenomena. Comprehensive math-
ematical theories for predicting ignition and combustion
of dust clouds in industrial environments from funda-
mental physical and chemical principles are at present
beyond reach.
It is not surprising, therefore, that existing know-
ledge of dust explosion-related phenomena is to a large
extent fragmented. It is believed, however, that more and
more fragments will, step by step, become tied together,
and steadily increasing domains of coherente emerge.
Powerful computers are invaluable tools in this process.
However, experiments wil1 remain indispensable for
calibration of the mathematica1 models, because such
models wil1 remain approximate and require careful tun-
ing in the foreseeable future. It is necessary to continue
the execution of realistic industrial-scale experiments. At
the same time, the more basic research and mathematica1
modelling should continue at full pace.
The current efforts to establish international co-
operation in joint research programmes should be
encouraged.
Acknowledgements
The author wishes to express his sincere gratitude to al1
those who kindly provided essential information without
which this paper could not have been written. Sincere
thanks are due also to Aaslaug Mikalsen for typing
the manuscript.
References
Eckhoff, R. K. Dust explosions in the process industries, But-
terworth-Heinemann, Oxford, 1994
Eckhoff, R. K. J. Hazardous Mater. 1993, 35, 103-117 (Also
reprinted in: Archivum Combustionis 1993, 13, 135-147)
Eckhoff, R. K. Prevention and mitigation of dust explosions in
the process industries. A survey of recent research and develop-
ment, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. China,
29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 5-34
Hauert, F., Vogl, A. and Radandt, S. Measurement of turbulente
and dust concentration in silos and vessels, Proc. 6th Int. Coll.
Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept., 1994,
eds Deng Xufan and Wolanski, P., pp 71-80
Lebecki, K., Sliz, J., Dyduch, Z. and Wolanski, P. Critical Dust
layer thickness for combustion of grain dust, American Institute
of Aeronautics and Astronautics, 1990, pp 51-58
Ural, E. A. Plan&Qerations Progr. 1992
Gelfand, B. and Tsyganov, S. Private communication, Semenov
Institute of Chemical Physics, Academy of Sciences of Russia,
Kosygin Street 4, Moscow 117977, Russia
Kauffman, C. W., Sichel, M. and Wolanski, P. Powder Technol.
1992, 71, 188
Austin, P. J., Girodroux, F., Li, Y.C., Alexander, C. G., Kauff-
man, C. W. and Sichel, M. Recent progress in the study of dust
combustion phenomena at The University of Michigan, Proc. 5th
Int. Coll. Dust Exolosions. Pultusk near Warsaw, Poland, 19-22
April 1993, pp 21-214
10 Tamanini, F. and Ural, E. A. Powder Technol. 1992,71, 135-151
11 Frolov, S. M., Mack, A. and Roth, P. Diffusion model of dust
lifting behind a shock wave, Proc. 5th Int. Coll. Dust Explosions,
Pultusk near Warsaw, Poland, 19-22 April 1993, pp 301-310
12 Lu, S. X., Fan, B. C., Pu, Y. K. and Gong, C. C. Numerical
investigation of boundary layer behind a shock passing over a
dust deposit, Proc, (suppl.) 5th Int. Coll. Dust Explosions, Pul-
tusk near Warsaw, Poland, 19-22 April 1993, pp 47-56
13 Nikolova, 1. P. Two-phase (source) code application for hetero-
geneous combustion modeling, Proc. 5th Int. Coll. Dust
Explosions, Pultusk near Warsaw, Poland, 19-22 April 1993,
pp 185-190
of dust explosions: R.K. Eckhoff
14 Medvedev, S. P., Geng, J. H. and Grnig, H. Shock tube study
of dust layer dispersion by rarefaction wave, Proc. 5th Int. Coll.
Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
1993, pp 311-320
15 Dushin, V. R., Nikitin, V. F., Smirnov, N. N., Sverev, N. I., Mach-
viladze, G. M. and Yakush, S. E. Mathematica1 modelling of
particle cloud evolution in the atmosphere after a huge explosion,
Proc. 5th Int. Coll. Dust Explosions, Pultusk near Warsaw,
Poland, 19-22 April 1993, pp 287-292
16 Nikitin, V. F., Smirnov, N. N., Dushin, V. R. and Zverev, N. 1.
Numerical simulation of particles evolution in turbulent strati-
fied flow, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R.
China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P.,
pp 61-70
17 Geng, J. H., Tang, M. J. and Grnig, H. Pressure front of an
incident shock propagating into a combustible particlesoxidative
gas mixture, Proc. 5th Int. Coll. Dust Explosions, Pultusk near
Warsaw, Poland, 19-22 April 1993, pp 335-344
18 Hensel, W. and John, W. Zusammenhang zwischen Glimmver-
halten und Staubschichtdicke, Paper presented at VDI Collo-
quium Sichere Handhabung brennbarer Stube, 4-6 November
1992, Nrnberg, Germany
19 Hensel, W. and John, W. The dependence of minimum ignition
temperatures of dust layers upon layer thicknesses, Proc. 5th Int.
Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
1993, pp 35-57
20 Hensel, W., Krause, U., John, W. and Machnov, K. Critical para-
meters for the ignition of dust layers at constant heat flux bound-
ary conditions, Proc. 28th AIChE Ann. Loss Prev. Symp., Ses-
sion No. 13 on Dust Explosions, Atlanta, USA, 17-21 April 1994
21 Krause. IJ. and Hensel. W. Hazards arisine from electrical
devices surrounded by deposits of flammable dcst, Proc. 6th Int.
Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept.
1994, eds Deng Xufan and Wolanski, P., pp 146-157
22 Chemenko, E. V., Afanasyeva, L. F. and Lebedeva, V. A. Flame
propagation along the surface of metal powders and powdered
mixtures of metal oxides, Proc. Joint Meeting of the Russian and
Japanese Sections of The Combustion Institte, Chemogolovka,
Moscow Reeion. 2-5 October 1993. DD 124-125
23 Matyukhina: 0. and Babushok, V. Seltheating of coal layer,
Proc. 5th Int. Coll. Dust Explosions, Pultusk near Warsaw,
Poland, 19-22 April 1993, pp 111-122
24 Sobolev, S. L. Two-temperature discrete model for heat conduc-
tion in dust systems, Proc. 5th Int. Coll. Dust Explosions, Pultusk
near Warsaw, Poland, 19-22 April 1993, pp 123-129
25 Crowhurst, D. Self-heating and spontaneous combustion charac-
teristics of bulk powders and powder layers, Paper given at the
European Summer School on Dust Explosion Hazards: Their
Assessment and Control, Cambridge, UK, organized by IBC
Technical Services Ltd in association with BMHB and IELG
26 Zhang, T. and Deng, X. Combustion kinetics of coal dust layers
on heated surface, Proc, (suppl.) 5th Int. Coll. Dust Explosions,
Pultusk near Warsaw, Poland, 19-22 April 1993, pp 9-19
27 Itagaki, H. and Matsuda, T. Thermal ignition of activated carbon
dusts, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R.
China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P.,
pp 141-145
28 Glinka, W., Klemens, R. and Wolanski, P. Experimental and
theoretical studies on radiative ignition of dust layer, Proc. 5th
Int. Coll. Dust Explosions, Puhusk near Warsaw, Poland, 19-22
April 1993, pp 69-86
29 van der Wel, P. G. J. Ignition and propagation of dust
explosions, Doctorate Dissertation, Delft University Press, 1993
30 van der Wel, P. G. J., Lemkowitz, S. M., Leschonski, S. et al.
Ignition of dust clouds using pulsed laser beams, Proc. 6th Int.
Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept.
1994, eds Deng Xufan and Wolanski, P., pp 125-140
31 Gieras, M. and Klemens, R. Experimental study of the ignition
and mechanism of flame propagation in dust and hybrid mixtures,
Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29
Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 158-
178
32 Wolanski, P. Powder Technol. 1992, 71, 197-206
33 Dahn, C. J., Reyes, B. N. and Kashani, A. Electrostatic discharge
(ESD) enerev initiation of dust cloud. Proc. 5th Int. Coll. Dust
xplosions,~Pultusk near Warsaw, Poland, 19-22 April 1993,
pp 87-97
34 Xu Tianrui and Lin Daze A calculating model of the minimum
ignition energy of dust clouds, Proc. 5th Int. Coll. Dust
Explosions, Pultusk near Warsaw, Poland, 19-22 April 1993,
pp 99-108
35 Glor, M. Static electricity - theory. Electrostatic hazards in pow-
der handling operations, Paper given at the European Summer
School on Dust Explosion Hazards: Their Assessment and Con-
 Prevention and mitigation
trol, Cambridge, UK, organized by IBC Technical Services Ltd
in association with BMHB and IELG
36 Glor, M. and Maurer, B. Zndversuche mit Schttkegelentladun-
gen, Paper presented at VDI Colloquium Sichere Handhabung
brennbarer Stube, 4-6 November 1992, Nrnberg, Germany
37 Glor, M. Private communication, Ciba-Geigy AG, Basle, Switzer-
land, 1993
38 Wolanski, P. Deflagration and detonation combustion of dust
mixtures, American Institute of Aeronautics and Astronautics,
1990, pp 3-3 1
39 Boiko, V. M. and Papyrin, A. N. Laser diagnostics processes of
mixture formation and combustion of dusts in shock waves, Proc.
6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-
2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 302-305
40 Geng, J. H., Liao, S. P. and Tang, M. J. Dynamics effects on
ignition of a dust suspension, Proc. 6th Int. Coll. Dust
xplosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
Dene Xufan and Wolanski. P.. DD 366-370
41 Ge& J. H., van de Ven, A., Zhang, F. and Gronig, H. A new
setup to measure ignition delay of a dust suspension behind an
incident shock wave, Proc. 6th Int. Coll. Dust Explosions, Shen-
yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and
Wolanski, P., pp 309-314
42 Lu, S. X. and Fan, B. C. Dust combustion in shock-induced
boundary layer, Proc. 6th Int. Coll. Dust Explosions, Shenyang,
P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski,
P., pp 329-335 - Puhusk near Warsaw, Poland, 19-22 April 1993, pp 321-328
43 Hu Dona. Wana Tianfu. Zhana Guanren and Sun Zhumei Studies
on highSpeed riaction behaviiur of wheaten flour, Proc. 6th Int.
Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept.
1994, eds Deng Xufan and Wolanski, P., pp 355-359
44 Mazurkiewicz, J. and Jarosinski, J. Gas composition near a com-
starch dust-air flame front, Proc. 5th Int. Coll. Dust Explosions,
Pultusk near Warsaw, Poland, 19-22 April 1993, pp 153-160
45 Proust, Ch. Experimental determination of the maximum flame
temperatures and of the laminar buming velocities for some com-
bustible dust-air mixtures, Proc. 5th Int. Coll. Dust Explosions,
Pultusk near Warsaw, Poland, 19-22 April 1993, pp 161-184
46 Seshardi, K., Berlad, A. L. and Tangirala, V. ComDust. Flame
1992,89, 333-342
47 Bradley, D., Chen, Z., El-Sherif, S., El-Din Habik, S. and John,
G. Combust. Flame, 1994, 96, 80-96
48 Gieras, M., Klemens, R., Wolanski, P. er al. Suppression of dust
explosion triggered by explosive charge, Proc. 6th Int. Coll. Dust
Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
Deng Xufan and Wolanski, P., pp 484490
49 van Wingerden, K. Private communication, Christian Michelsen
Research AS, Fantoftvegen 38, N-5036 Fantoft, Bergen, Nor-
way, 1993
50 Krishenik, P. M. and Shkadinskii, K. G. Modeling of combustion
modes in Bi-size and Bi-material dust-air mixture, Proc. 5th Int.
Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22
April, 1993, pp 191-198
51 Deng Xufan, Xu Renxian, Xie Lin, Dang Junxian, Zang Tingan,
Gao Jun and Tan Feng Gui Explosibility and ignitability of 16
types of dust and some opinions on fundamental research for dust
explosions, Proc. 5th Int. Coll. Dust Explosions, Pultusk near
Warsaw, Poland, 19-22 April 1993, pp 217-224
52 Lee, J. H. S., Zhang, F. and Knystautas. R. Powder Technol. 1992,
71, 153-162
53 Wlodarczyk, E., Maranda, A., Nowaczewski, J., Trebinski, R. and
Sliz, J. Influence of the composition of gas-dust mixture on the
content of aluminium oxide in explosion products, Proc. 5th Int.
Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
1993, pp 279-285
54 Fan, B. C., Ding, D. M. and Tang, M. J. An aluminium dust
explosion in a spherical closed vessel, Proc. (Suppl.) 5th Int.
Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
1993, ~~21-31
55 Mintz, K. J. Combusr. Flame 1993, 34, 125-130
56 Poletaev, N. L. and Korolchenko, A. Ya. A note on the relation-
ship between the lower explosibihty limit of dust, and particle
size, Proceedings of Joint Meeting of the Russian and Japanese
Sections of the Combustion Institute, Chemogolovka, Moscow
Region, 2-5 October 1993, pp 116117
57 Hertzberg, M., Zlochower, 1. A. and Cashdollar, K. L. Metal dust
combustion: Explosion limits, pressures, and temperatures, Proc.
24th Symp. (Int.) on Combustion, The Combustion Institute, Pitts-
burgh, USA, 1992, pp 1827-1835
58 Hertzberg, M., Cashdollar, K. L., Zlochower, 1. A. and Green, G.
M. Explosives dust cloud combustion, Proc. 24th Symp. (Int.)
on Combustion, The Combustion Institute, Pittsburgh, USA,
1992,pp1837-1843
of dust explosions: R.K. Eckhoff 17
59 Wiemann, W. Rauchgasreinigung mit brennbaren Stuben, VDI
Betichte No. 975, 1992, pp 607-625
60 Eckhoff, R. K. Powder Technol. 1992, 71, 181-187
61 van der Wel, P. G. J., van Veen, J. P. W., Lemkowitz, S. M.,
Scarlett, B. and van Wingerden, K. Powder Technol. 1992, 71,
207-215
62 van der Wel, P. G. J., Lemkowitz, S. M., Timmers, P. and Scar-
lett, B. The role of turbulente on the propagation mechanism and
behaviour of dust exolosions. Proc. 5th Int. Coll. Dust
Exnlosions, Pultusk near Warsaw, Poland, 19-22 April 1993,
pp 199-209
63 Scheuermann, K. Studies about the influence of turbulente on the
course of exolosions. Proc. 28th AIChE Ann. Loss Prev. Svmp.,
Section No. 13 on Dust Explosions, Atlanta, USA, 17-21 April
1994
64 Rzal-Rebire, F. and Veyssire, B. Interaction of a vortex ring
with gaseous and dust flames, Paper submitted for presentation
at the 24th Symposium (International) on Combustion, Sydney,
Australia, 5-10 July 1992 (International Combustion Institute)
65 Veyssire, B. Powder Technol. 1992, 71, 171-180
66 Rzal-Rebire, F. and Veyssire, B. Propagation mechanisms of
starch particles-air flames, Proc. 6th Int. Coll. Dust Explosions,
Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan
and Wolanski, P., pp 186200
67 Geng, J. H., Tang, M. J. and Grnig, H. Shock-induced ignition
delay of cornstarch dusts, Proc. 5th Int. Coll. Dust Explosions,
68 Ural, E. A. Private communication, Factory Mutual Research,
PO Box 9102, Norwood, MA 02062, USA, 1992
69 Boiko, V. M., Papyrin, A. N. and Poplavski, S. V. On peculiari-
ties of coal dust ignition in incident shock waves, Proc. 5th Int.
Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
1993, pp 329-334
70 Bradley, D., Chen, Z. and Swithenbank, J. R. Buming rates in
turbulent fine dust/air explosions, Proc. 22nd Symp. (Int.) on
Combustion, The Combustion Institute, Pittsburgh, USA,
pp 1767-1775
71 Proust, Ch. Dust explosions in pipelines, Paper given at the
European Summer School on Dust Explosion Hazards: Their
Assessment and Control, Cambridge, UK, organized by IBC
Technical Services Ltd in association with BMHB and IELG,
1993
72 Pu, Y. K., Hu, S., Yang, Q. Z., Bao, G. G. and Wang, Z. M.
Combustion tube studies of dust flame acceleration, Proc.
(suppl.) 5th Int. Coll. Dust Explosions, Pultusk near Warsaw,
Poland, 19-22 April 1993, pp 33-46
73 Hu Dong and Sun Zhumei Studies of the behaviour of aluminium
powder reaction in the gas phase reaction environment, Proc. 6th
Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-
2 Sent. 1994. eds Dena Xufan and Wolanski. P.. DD 349-354
74 Eckhoff, R. K. Generition. ignition, combustion and explosion
of sprays and mists of flammable liquids in air. A literature sur-
vey, Report No. CMI-91-A25014 from Christian Michelsen Insti-
tute. N-5036 Fantoft. Norwav. 1991
75 Huang Wanli, Pu Yikang and Ding Hua The electrical conduc-
tion phenomenon in the process of combustion of aluminium
powder air mixtures, Proc. 6th Int. Coll. Dust Explosions, Shen-
vang, P.R. China. 29 Atm-2 Seot. 1994, eds Dena Xufan and
Woanski, P., pp 201-205- 1
76 Kjldman, L. Powder Technol. 1992, 71, 163-169
77 Krause, U. A two-dimensional model for the numerical simul-
ation of explosions in vented vessels, Proc. 5th Int. Coll. Dust
Explosions, Pultusk near Warsaw, Poland, 19-22 April 1993,
pp42430
78 Krause, U. Numerical investigation of the influence of velocity
fluctuation on ventina of vessels. Proc. 6th Int. Coll. Dust
Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
Deng Xufan and Wolanski, P., pp 441-452
79 Alexander, C. G., Harbaugh, A. S., Kauffman, C. W., Li, Y. C.,
Cybulski, K., Dyduch, Z., Lebecki, K., Sliz, J., Klemens, R., Wol-
anski, P. and Zalesinski, M. The establishment of dust deton-
ations. Proc. 5th Int. Coll. Dust Explosions, Pultusk near War-
saw. Poland. 19-22 April 1993. DD365-381
80 Sichel, M. and Kauffman, C. W. Transition from deflagration to
detonation in layered dust explosions, Proc. 28th AIChE Ann.
Loss Prev. Symp., Section No. 13 on Dust Explosions, Atlanta,
USA, 17-21 April 1994
81 Khomik, S. V., Gelfand, B. E. and Knyazev, M. V. On the criti-
cal diameter of detonation propagation in dust suspensions, Proc.
Joint Meeting of the Russian and Japanese Sections of the Com-
bustion Institute. Chernogolovka, Moscow Region. 2-5 October
1993, p 188
82 Khomik, S. V., Gelfand, B. E. and Knyazev, M. V. Experimental
determination of a critical diameter of detonation propagation in
18 Prevention and mitigation
dust suspensions, Proc. 6th Int. Coll. Dust Explosions, Shenyang,
P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski,
P., pp 315-319
83 Korobeinikov, V. P. The analysis of basic parameters for deton-
ation of dusty gases, Proc. 5th Int. Coll. Dust Explosions, Pultusk
near Warsaw, Poland, 19-22 April 1993, pp 351-364
84 Markov, V. V. A new numerical method for two-phase flows,
Proc. 5th Int. Coll. Dust Explosions, Pultusk near Warsaw,
Poland, 19-22 April 1993, pp 383-390
85 Ding Hua and Huang Wanli On the Chapman-Jouguet condition
of dust detonation, Proc. 6th Int. Coll. Dust Explosions, Shen-
yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and
Wolanski, P., pp 341-348
86 Tulis, A. J., Sumida, W. K., Heberlein, D. C., Patel, D. L. and
Egghart, H. Detonation tube studies of particle size and RDX
sensitization of aluminium powder-air with regard to spinning
and/or multiple-front detonations, Proc. 5th Int. Coll. Dust
Explosions, Pultusk near Warsaw, Poland, 19-22 April 1993,
pp 391400
87 Paplinski, A. and Wlodarczyk, E. About estimation of the effec-
tive energy for detonation initiation in combustible gaseous sus-
pensions, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R.
China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P.,
pp 320-328
87 Klemens, R., Teodorczyk, A., Wolanski, P. and Wolinski, M.
Detonation parameters of hybrid mixtures containing grain
dusts, Proc. <suppl.) 5th Int. Coll. Dust Explosions, Pultusk near
Warsaw. Poland. 19-22 A~til 1993, DD5769
88 WirknerlBott, 1.; Schumann, S. anditock, M. Dust explosion
venting: investigation of the secondary explosion, Paper pre-
sented at 7th Int. Symp. Loss Prev. & Safety Prom. Proc. Ind.,
Italy, 4-8 May 1992. Copy received from Battelle Europe, Frank-
furt - Main, Germany
89 Schumann, S. Wirkner-Bott, 1. Dust explosion venting: Second-
ary explosion for vessel volumes up to 250 m3, EuropEx News-
letter, Edition 22, September 1993, pp 2-5
90 van Wingerden, K. J. L.oss Prev. Proc. Ind. 1993, 6, 241-249
91 Gelfand, B., Medvedev, S., Polenov, A. and Bartenev, A. Com-
busrion Explosion Shock Waves 1990, 3, 85-91
92 Medvedv, S. P., Polenov, A. N. and Gelfand, B. E. Blast waves
induced by sudden expansion of pressurized dusty systems, Proc.
6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-
2 Sept. 1994, eds De& Xufan and Wolanski, P., pp 289-296
93 Medvedev. S. P.. Polenov. A. N. and Gelfand. B. E. On the
amplification of blast wave transmitted through a dust deposit,
Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29
Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 297-
301
94 Smimov, N. N.. Kuksenko, B. V. and Chen Dongqing Mathemat-
ical modelling of shock induced explosion in piydispersed dust
suspended in the oxidizer, Proc 6th Int. Coll. Dust, Explosions,
Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan
and Wolanski, P., pp 306-308
95 Boiko, V. M., Kiselyov, V. P., Kiselyov, S. P. et al. On gas
parameter profiles at-non-stationary shck wave interaction with
dust cloud. Proc. 6th Int. Coll. Dust Exolosions, Shenvang, P.R.
China, 29 ug.-2nd Sept. 1994, eds Deng Xufan andwoianski,
P., pp 336-340
96 Gelfand, B. E., Khomik, S. V. and Polenov, A. Quenching of
shock waves in dusty medium, Proc. 6th Int. Coll. Dust
Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
Deng Xufan and Wolanski, P., pp 478483
97 Harmanny, A. Explosion effects, Proc. 1st World Seminar on
the Explosion Phenomenon and on the Application of Explosion
Protection Techniques in Practice. Arranged by EuropEx, Brus-
sels, 17-21 February 1992
98 Mercx, W. P. M. Critical appraisal of damage criteria for build-
ings and process equipment, Proc. 1st World Seminar on the
Explosion Phenomenon and on the Application of Explosion Pro-
tection Techniques in Practice. Arranged by EuropEx, Brussels,
17-21 February 1992
99 LAbb, R. J. Explosion effects on people, Proc. 1st World Sem-
inar on the Explosion Phenomenon and on the Application of
Explosion Protection Techniques in Practice. Arranged by Euro-
pEx, Brussels, 17-21 February 1992
100 Britan, A. B., Levin, V. A., Mitichkin, S. Y. et al. Investigation
of relaxation zone under interaction of shock wave with protective
aqueous foam layer, Proc. 6th Int. Coll. Dust Explosions, Sheny-
ang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wol-
anski, P., pp 36&365
101 Britan, A. B., Levin, V. A., Mitichkin, S. Y. et al. Influence of
aqueous foam screen upon propagation of shock waves, Proc.
6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-
2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 472477
of d lti exprosions: R.K. Eckhoff
102 Deng Xufan and He Jicheng A brief review of dust explosion
reaction engineering, Proc. 6th Int. Coll. Dust Explosions, Shen-
yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and
Wolanski, P., pp 96-115
103 Siwek, R. Latest development in explosion protection tech-
nology, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R.
China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P.,
pp 35-60
104 Eckhoff, R. K. Dust explosion hazards in the ferro-alloys indus-
try, Proc. 52nd Electric Fumace Conference, 13-16 November,
bon and Steel Society, Inc., Warrendale, PA, USA, 1994
105 Gibson, N. Precautions against fires and explosions in drying
operations, Paper given at the European Summer School on Dust
Explosion Hazards: Their Assessment and Control, Cambridge,
UK, organized by IBC Technical Services Ltd in association with
BMHB and IELG, 1993
106 Zockoll, C. Preventive protection against fire and explosion of
organic processes by example of spray driers in the milk industry,
Preprints for seminar on Explosion Satety and Related Risk Con-
trol, Gent, Belgium, 23-24 March 1994, organized by Techno-
logica1 Institute-KVIV, in cooperation with EuropEx (Kontich,
Belgium), pp 181-193
107 Xu Bo-Wen, Cui Yun-Shen and Xu Wan-Qing Research on fee-
ble electrical spark igniting deposited fibre dust, Proc. 5th Int.
Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
1993, pp 59968
108 Rogers, R. L. Fire, explosion and electrostatic hazards during the
use of big bags for powder handling, Preprints for seminar on
Explosion Safety and Related Risk Control, Gent, Belgium, 23-
24 March 1994, organized by Technological Institute-KVIV, in
cooperation with EuropEx (Kontich, Belgium), pp 169-179
109 Dahn, C. J., Reyes, B. N. and Kashani, A. Static electricity haz-
ards of flexible intermediate bulk containers, Proc. 28th AIChE
Ann. Loss Prev. Symp., Session No. 12 on Electrostatic Hazards,
Atlanta, USA, 17-21 April 1994
110 Pratt, T. H. Static electricity in pneumatic transport systems.
Three case histories, Proc. 28th AIChE Ann. Loss Prev. Symp.,
Session No. 12 on Electrostatic Hazards, Atlanta, USA, 17-21
Aoril 1994
111 Stut Keping. Investigation on statics explosion-proof of aldehyde
resin powder in pneumatic pipe, Proc. 6th Int. Coll. Dust
Explosions, Shenyang, P.R. China, 29. Aug.-2 Sept. 1994, eds
Deng Xufan and Wolanski, P., pp 42@425
112 Wang Mingzhong and Lou Renjie The main electrostatic hazards
and its elimination in powder pneumatic conveying system, Proc.
6th Int. Coll. Dust Explosions, Shenyang, 29 Aug.-2 Sept. 1994,
eds Denp Xufan and Wolanski, P., pp 426-433 __
113 Kleinschmidt, H.-P. Private communication, Fagus-GreCon
Greten GmbH & Co., Postfach 1243, D-W-3220 Alfeld-Hann-
over, Germany, 1992
114 Janssen, L. Private communication, Firefly AB, Huddinge,
Sweden, 1993
115 T&B Electronic Spark detection and extinguishing systems.
Detailed technical description of system features and appli-
cations, T&B Electronic GmbH, 31061 Alfeld, Germany, 1994
116 Lloyd, F. C. Electrical equipment in plant handling combustible
powder, Paper given at the European Summer School on Dust
Explosion Hazards: Their Assessment and Control, Cambridge,
UK, organized by IBC Technical Services Ltd in association with
BMHB and IELG, 1993
117 van Laar, G. F. M.Area classification for dust explosion hazard-
ous environments, Preprints for seminar on Explosion Safety
and Related Risk Control, Gent, Belgium, 23-24 March 1994.
organized by Technological Institute-KVIV, in cooperation with
EurooEx (Kontich. Beleium), po 127-141
118 Wol&ski,M. and Hayashi, T.Explosibility of powders of rare
earth metal alloys, Proc. 5th Int. Coll. Dust Exolosions, Pultusk
near Warsaw, Poland, 19-22 April 1993, pp 225-236
119 Krause, U., Weinert, D. and Wohm, P. Diagrams for the determi-
nation of the limiting oxygen concentration of dust/air mixtures,
Proc. 5th Int. Coll. Dust Explosions, Pultusk near Warsaw,
Poland, 19-22 April 1993, pp 257-266
120 Maddison, N. Avoidance of flammable atmosoheres: use of inert
gas in powder handling plant, Paper given at ihe European Sum-
mer School on Dust Explosion Hazards: Their Assessment and
Control, Cambridge, UK, organized by IBC Technical Services
Ltd in association with BMHB and IELG, 1993 .__
121 Gao Guangchun, Chen Zhandong, Tang Qinghua and Luo Ruquan
New technique and equipment for fire and explosion-protection
of blast fumace bituminous injection, Proc. 6th Int. Coll. Dust
Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
Deng Xufan and Wolanski, P., pp 407-411
122 Wang Junyi, Zhang Jingjie and Liu Jingxian Prevention of coal
dust explosion in the fuming fumace, Proc. 6th Int. Coll. Dust
 Prevention and mitigation of dust explosions: R.K. Eckhoff

Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds 145 Lunn, G. A. Dust explosion prevention and protection, Part 1,
Deng Xufan and Wolanski, P., pp 412419 Venting, 1 Chem E, Rugby, K, 2nd edition,- 1992
123 Dansk Fire Eater A/S INERGEN. Anlcegsbeskrivelse & Design, 146 Eckhoff. R. K. Safetv Sci. 1993. 16. 511-525
Dansk Fire Eater A/S, Skovlytoften 14. DK-2840 Holte, 147 Deng Xufan, Zang Tingan, Dang Junxian and Xie Lin Maize
Denmark, 1992 dust explosion in the 94.4 m3 experimental silo for venting of
124 Zhang Zhancheng. Zhao Ping and Shi Jianye A laboratory study deflagrations in low strength silos, Proc. 5th Int. Coll. Dust
on grain dust explosion suppression by adding inorganic powder, Explosions, Pultusk near Warsaw, Poland, 19-22 April 1993,
Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 pp 4034 11
Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P.. pp 491- 148 Htichst, S., Leuckel, W. and Eibl, J. (1993) Chem-Ing.-Techn.,
499 65, 1488-1490
125 Zockoll, C. Concentration and ignitability of dust clouds during 149 Molkov, V. V., Nikitenko, V. M., Filippov, A. V. and Korolch-
the discharge of bulk material, Proc. 6th Int. Coll. Dust enko, A. Ya. Dynamics of gas explosion in a vented vessel with
Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994. eds inertial vent covers, Proc. Joint Meeting of the Russian and
Deng Xufan and Wolanski, P., pp 263-277 Japanese Sections of The Combustion Institute, Chemogolovka,
126 Xu Bowen, Li Zhaojiang, Cui Yushen and Xu Wanqing Esti- Mscow Region, 2-5 October 1993, pp 183-185
mation of dust explosion hazard in plants. Estimation method with 150 Fan Xishene. Wu Jianxine and Li Li The edge effect in the static
factors of static and dynamic dust depositing. Proc. 6th Int. Coll. bursting of;ent closure: Proc. 6th Int. Col. Dust Explosions,
Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan
eds Deng Xufan and Wolanski, P., pp 403406 and Wolanski, P., pp 553-559
127 Wolanski, P. Minimum explosive concentration of dust-air mix- 151 Harmanny, A. Dration of vented dust explosions, EuropEx
tures, Proc 6th Int. Coll. Dust Explosions, Shenyang, P.R. China, Newslerter, Edition 23. December 1993. DD5-9
29 Aug.LZ Sept. 1994, eds Deng Xufan and Wolanski, P., 152 Siwek, R., Glor, M. nd orreggiani Duit explosion venting at
pp 206219 elevated initial pressure: 5 reprints 7th Int. Symp. Loss Prev. &
128 Xu Bowen, Zheng Haijun and Xu Wanqing The effect of particle Safety Prom. Proc. Ind.. Italy,!4-8 May 1992 _SRP-Partners,
. .. -_
size of libre dust on LEL, Proc. 5th Int. Coll. Dust Explosions, Roma ; r.i:
Pultusk near Warsaw, Poland, 19-22 April 1993, pp 237-246 153 Crowhurst, D., Colwell, S. A. and Hoare, D. P. The extemal
129 Mittal, M. Mathematica1 models for minimum explosible concen- effects of vented dust explosions, Proc. 6th Int. Coll. Dust
tration of dusts, Proc. 5th Int. Coll. Dust Explosions. Pultusk near Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
Warsaw, Poland, 19-22 April 1993, pp 247-256 Deng Xufan and Wolanski, P., pp 510-525
130 Shao Fuqun and Wang Shi Concentration measurement of pow- 154 Peng Dianhua, Wu Fianying, Xu Tianrui and Fin Xiang. A devel-
dery materials in gas/solid two phase flow, Proc. 6th Int. Coll. opment of flame arrestor on dust vessel vents, Proc. 6th Int. Coll.
Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994,
eds Deng Xufan and Wolanski, P., pp 237-244 eds Deng Xufan and Wolanski. P.. DD 526528
13 1 Torrent, J. Cl. and Menndez, E. Explosion tests at elevated initial 155 Li Gang: Deng Xufan, Liu Wenxin et al. Development of a
pressures. Eurol>Er Newslerter, Edition 22, September 1993, quenching venting door (QVD), Proc. 6th Int. Coll. Dust
PP 6-8 Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
132 Crowhurst, D. The strength f equipment to be used for con- Deng Xufan and Wolanski. P., pp 530-534
tained and vented explosions P Paper given at the European Sum- 156 Crowhurst, D. Explosion protection of industrial buildings,
mer School on Dust ExpIpsion Hazards: Their Assessment and Paper given at the European Summer School on Dust Explosion
Control, Cambridge, UK, organized by IBC Technical Services Hazards: Their Assessment and Control, Cambridge, UK,
Ltd in association with BMLl%3- and IELG organized by IBC Technical Services Ltd in association with
133 Bartenev, A. M., Gelfand, B. E. and Frolov, S. M. The rupture BMHB and IELG
of vessels containing explosive mixtures, Proc. 5th Int. Coll. Dust 157 Ural, E. A. J. Lus Prev. Proc. Ind., 1993, 6, 3-10
Explosions, Pultusk near Warsaw, Poland, 19-22 April 1993, -l+++Alfert, F. PC-Vent. A software solution to control explosion haz-
pp 465472
134 Bartenev, A. M., Medvedev, S. P., Polenov, A. N. et al. The
effect of dust filler on the rupture of high pressure vessels, Proc.
6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-
2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp I 16-124 159 Hattwig, M. and Hensel, W. Applicability of the new VDI-guide-
35 van Wingerden, K., Pedersen, G. H., Teigland, R. and Eckhoff, line 3673 to silos of rectangular geometry, Proc. (suppl.) 5th Int.
R. K. Violente of dust explosions in integrated systems, Proc. Coll. Dust Explosions, Pultusk near Warsaw, Poland 19-22 April
28th AIChE Ann. Loss Prev. Symp., bession No. 13 on Dust 1993, pp 71-81
Explosions, Atlanta, USA, 17-21 April 1994 160 Moore, P. E. Explosion suppression for protecting systems
136 Lunn, G. A. The explosion venting of interconnected vessels, against the explosion hazard presented by pneumatic filling tech-
Proc. 1st World Seminar on the Explosion Phenomenon and on niques, Proc. 1st World Seminar on the Explosion Phenomenon
the Application of Explosion Protection Techniques in Practice. and on the Application of Explosion Protection Techniques in
--*Arranged by EuropEx, Brussels, 17-21 February 1992 Practice. Arranged by EuropEx, Brussels, 17-21 February 1992
137 van Wingerden, K. and Alfert, F. Dust explosion propagation in 161 Moore, P. E. Develooments in exolosion suuoression. Paper
connected vessels, VDI Berichte No. 975. 1992. DIJ LI 507-528 given at the European Nummer School on DuStExplosion Haz-
138 Vogl. A. The course of dust explosions in pipes of pneumatic ards: Their Assessment and Control, Cambridge, UK, organized
systems, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. by IBC Technical Services Ltd in association with BMHB and
China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P., IELG. 1992
pp 535-552 162 Tyldesley, A. Private communication, 16 November 1993
139 Zellweger, J. Private communication, Rico-Sicherheitstechnik 163 Siwek, R. The combination of explosion venting and explosion
AG, Lindenstrasse 77, CH-9006 St. Gallen, Switzerland, 1992 suppression. Explosion suppression in very small;olumes, Proc.
140 Alfert, F. and Fuhre, K. Venting of dust explosions in filters 1st World Seminar on the Explosion Phenomenon and on the
and integrated systems, Report 92-A25021, Christian Michelsen Application of Explosion Protection Techniques in Practice.
Institute, N-5036 Fantoft, Norway, 1992 Arranged by Europx, Brussels, 17-21 Februaj 1992
141 Klincewicz, M. and Kordylewski, W. A new explosion diverter 164 Sliz, J., Lebecki, K. and Dyduch, Z. Venting and suppression of
for pipelines protection, Proc. 5th Int. Coll. Dust Exolosions. Pul- grain dust explosions - experiments in 8 ml chamber, Proc. 5th
tusk near Warsaw, Poland, 19-22 April 1993, pp 45 1436 Int. Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22
142 Cybulaki, K., Dyduch, Z., Lebecki, K. and Sliz, J. The tests on April 1993, pp 413420
triggered barriers in cross-roads of mining galleries, Proc, 6th 165 Cybulski, K., Dyduch, Z., Lebecki, K. and Sliz, J. Suppression
Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.- of grain dust explosions with triggered barriers, Proc. 5th Int.
2 Sept. 1994, eds. Deng Xufan and Wolanski, P., pp 500-509 Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
143 Scholl, E. W. The technique of explosion venting. Much more 1993, pp 437447
than just a set of nomographs, Proc. 1st World Seminar on the 166 Cybulski, K., Dyduch, Z., Lebecki, K. and Sliz, J. Weak coal
Explosion Phenomenon and on the Application of Explosion Pro- dust explosion propagation in a mine workings network, Proc.
tection Techniques in Practice. Arranged by Europx, Brussels, 6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-
17-21 February 1992 2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 371-380
144 Lunn, G. A. Paper given at the European Summer School on 167 Gieras, M., Klemens, R. and Wolanski, P. Pyrolysis processes
Dust Explosion Harards: Their Assessment and Control, Cam- during grain dust-air mixture explosions, Proc. 5th Int. Coll.
bridge, UK, organized by IBC Technical Services Ltd in associ- Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
ation with BMHB and IELG. 1993 1993, pp 137-152
20 Prevention and mitigation
168 Wagner, K. A case study of applied risk analvsis in dust
explosion hazardous envirnments; Preprints for - seminar on
Explosion Safetv and Related Risk Control. Gent. Belaium. 23-
24 March 1994,-organized by Technological Insbtute-iVV:m
cooperation with EuropEx (Kontich, Belgium), pp 155-167
169 Mittal, M. Dust explosion risk analysis of pneumatic transport
systems, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R.
China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P.,
pp 381-390
170 Lin Jingxian and Wang Junyi The DSS to prevent and control
coal dust explosion, Proc. 6th Int. Coll. Dust Explosions, Shen-
yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and
Wolanski, P., pp 434440
171 Moore, P. E. and Frehill, R. M. Dust explosion protection - the
choices, Proc. 6th Int. Coll. Dust Exposions, Shenyang, P.R.
China, 29 Aua.-2 Sent. 1994. eds Dena Xufan and Wolanski. P..
pp453471 - -
172 Liu Wenxin, Sheny Zhongquan, Deng Xufan et al. Measurement
of turbulente in the Hartmann bomb, Proc. 6th Int. Coll. Dust
Exnlosions, Shenvane, P.R. China. 29 Aua.-2 Sent. 1994. eds
Ding Xufan and Fvoanski, P., pp 81-95 - * 19-22 April 1993, pp 487496
173 Siwek, R. and Cesana, C. Ignition behaviour of combustible
dusts, Proc. 28th AIChE AnnLoss Prev. Symp., Session No. 12
on Electrostatic Hazards. Atlanta. USA. 17-21 Aoril 1994
174 Zhou Benmou, Tan Fenggui, Yu Xuesheng et al. A new type of
movable electrode electrostatic ignition energy apparatus, Proc.
6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-
2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 257-262
175 Cashdollar, K. L., Weiss, E. S., Greninger, N. B. and Chatrathi,
K. Plant/Operations Progr. 1992, 11, 247-255
176 Cashdollar, K. L. and Chatrathi, K. Combus. Sci. Technol. 1992,
87, 157-171
177 Zhou Congzhang, Zhang Ruiping and Yu Yongfang On the cri-
terion of lower explosion limit of dust cloud, Proc. 6th Int. Coll.
Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994,
eds Dene Xufan and Wolanski. P.. OD220-226
178 Matsuda: T. and Itagaki, H. ffectsof ignition energy on dust
explosion parameters, Proc. 6th Int. Coll. Dust Explosions, Shen-
yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and
Wolanski, P., pp 245-256
179 Tian Renqu, He Chaoyuan and Zhang Yinghe Effect of different
ignition energies and the existente of methane gas on the mini-
of dust explosions: R.K. Eckhoff
mum explosive concentration of coal dusts, Proc. 6th Int. Coll.
Dust Explosions, Shenvana. P.R. China. 29 Aua-2 Seot. 1994.
eds Deng Xufan and Wolanski, P., pp 227-236 1
180 Xu Tianrui, Ding Banqing, Gong Youcheng and Wang Xingqun
Thoughts on some dust explosibility test methods in 20 1appar-
atus, Proc. 6th Int. Coll. Dust Exnlosions. Shenvang. 29 Auz.-
2 Sept. 1994, eds Deng Xufan and Wolanski, P.,*pp278-283
181 Pu, Y. K., Jarosinski, J., Johnson, V. G. and Kauffman, C. W.
Turbulente effects on dust explosions in the 20-litre spherical
vessel, Proc. 23rd Symp. (Int.) on Combustion, The Combustion
Institute, Pittsburgh, PA, USA, pp 843-849
182 Wang Baoming and Zhang Jinhn Investigation on the process of
TNT dust explosion and explosion vent systems of TNT pro-
duction process, Proc. 6th Int. Coll. Dust Explosions, Shenyang,
P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski,
P., pp 391-402
183 von Haefen, E. and Schecker, H.-G. DUSTEXPERT - an expert
system for the assessment of explosion hazards and the selection
of explosion protection methods for dust handling plants, Proc.
5th Int. Coll. Dust Explosions, Puhusk near Warsaw, Poland,
184 Wach, J. Expert system for designing protection measures against
accidental dust explosions, Proc. 5th Int. Coll. Dust Explosions,
Pultusk near Warsaw, Poland, 19-22 April 1993, pp 497-502
185 Tyldesley, A. Dust-Expert - Much more than just a programme
for running the vent size calculations, Paper given at the Euro-
pean Summer School on Dust Explosion Hazards: Their Assess-
ment and Control, Cambridge, UK, organized by IBC Technical
Services Ltd in association with BMHB and IELG. 1993
186 Femando, D. Dust explosion hazards: the human element, Paper
given at the European Summer School on Dust Explosion Haz-
ards: Their Assessment and Control, Cambridge, IJ-K, organized
bv IBC Technical Services Ltd in association with BMHB and
ILG, 1993
187 Middleton, P. G. Private communication, British Materials Hand-
ling Board, Index House, Ascot, Berkshire SL5 7EU, UK, 1992
188 Gibson, N. A review of dust explosion research projects for
BMHB research project panel. An interim report, November
1991
189 Wang Dongyan Hazards and control countermeasures in China,
Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29
Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 1-4
 J. l.oss /re\: Procms Ind. Vol. 9. No. 1. pp. 21-31, 1996
Copyright 0 1996 Elsevier Science Ltd
Printed in Great Britain. Al1 rights reserved
0950-4230(95)00057-7 0950-1230/96 $15.00 + 0.00
ELSEVIER

Determination of technical safety indices

and factors influencing hazard evaluation
of dusts

Richard Siwek
Explosion Technology, Ciba-Geigy Ltd, K-32.3.04, 4002-Basle, Switzerland

A knowledge of material safety properties is an essential requirement for the safe handling of
unit operations. Incidents in plants can often be traced to an insufficient knowledge of the haz-
ardous properties of combustible substances. Unforeseen violent reactions in which destructive
energies or large amounts of hazardous materials were released have resulted. Safety properties
provide information on the reaction behaviour of the substances and possible fire and explosion
hazards, presupposing that they are determined carefully and applied properly. The paper pro-
vides definitions of the material safety properties and brief descriptions of their methods of
determination, which should be standardized internationally if at all possible, and describes
various evaluation criteria. The analysis distinguishes between dust layers and dust suspen-
sions.
Keywords: material safety propetties; hazard evaluation; combustible substances

For the safe handling of combustible substances, it is
imperative to know the dangerous properties of the pro-
duct. The reliable way to characterize the combustible
and explosive properties of a product is to subject a sam-
ple of the product to various tests and classify the results
in accordance with the technical safety characteristics.
The determination of these characteristics of combust-
ible dusts is often complicated, even more complicated
than with gases or vapours, because, in contrast to gases
and vapours dusts are particulate solids and therefore
have additional characteristics: particle size distribution,
ageing degree, humidity and density. These character-
istics can influence the reaction behaviour and therefore
complicate the classification.
In order to get a sufficient genera1 view of the
danger of a product, it is necessary to carry out a large
number of tests. The mosaic-like nature of these tests
results in a more or less extensive picture of a given
problem.
The extent of the investigations is determined by
the problem. This depends on the goal of the protection
system, i.e. the chosen protection measures, whether
against the beginning of combustion or explosion or
against the explosion effects. This decision is made
cooperatively between the tester and the user. In the fol-
lowing work, some chosen test methods and their
resulting safety technical characteristics are described,
with their effective magnitudes (if they are sufficiently
known and relevant for the particular practice). Th
analysis wil1 distinguish between dust layers and
dust suspensions.
The substances are tested using an unambiguously
defined sample following a preparation tailored to the
requirements of the selected test method. This usually
includes careful drying of the product (e.g. 50C in vac-
uum or 75C at standard pressure) and preparation of a
specified particle size fraction of < 250 Pm for the dust
layer or < 63 Pm for suspended dus?. Tests are perfor-
med under initial atmospheric conditions.
Dust layers
Ignitability
Dust layers are recognized as combustible dusts if they
can be ignited with a foreign ignition source and the
local fire wil1 propagate sufficiently after the outside
source is taken away.
For determination of the genera1 ignition and com-
bustion behaviour of dust layers, dust deposits are
exposed in pre-tests to the effect of different ignition
sources. These sources play an important part in indus-
trial incidents.
The ignition sources include: sparks, mechanica1
sparks (cerium iron sparks), combustible matches, glow-
ing cigarettes and gas flames (Figure 1).
Burning behaviour (burning test)
This test involves investigation of whether and to what
extent a fire initiated by an external ignition source can
propagate in a dust deposit of width ca. 2 cm and length
4 cm (Figure 2, left). The test is carried out at room
temperature (20C) and at elevated product temperature

21
22 Hazard evaluation
Figure 1 Test on ignitability of dwt layers
(e.g. 100C). The flammability of the product is rated in
accordance with the reaction progress by a burning class
numbe?. Essentially this involves distinguishing
between no propagation of a fire (class 1-3) or spread
of a fire (class 4-6) in the dust layer. Products in class
6 can no longer be controlled after ignition (reference
substance = black powder). If products in classes 4 and
5 are processed, it is advisable to incorporate a water
deluge installation in the apparatus.
If the 4 cm long dust heap is lengthened to 25 cm,
the burning time can be measured (Figure 2, right). If
the buming time is less than 45 s, according to the
dangerous substance ordinance and EC guideline4, the
substance is considered as an easily ignitable substance
and must be assigned to class 4.1 according to the test
criteria of the United Nations (UN recommendation)5.
One version of the test is used to determine the
packaging group at the same time: a material is classed
in packaging group 11 if the burning time is less than
45 s and the flame crosses a moistened zone at the end
of the train. Packaging group 111contains materials with
a burning time of less than 45 s whose flame is stopped
by the moistened zone.
Relative self-ignition temperature, RSIT (test for
exothennic reaction in an air steam)
This is understood to mean the lowest temperature at
which the temperature of the test substance starts to
exceed that of an inert reference substance when both
are heated in the same air stream. If the temperature of
test on ccmmic plate
platinum win2 [-1000eC)
V
test in glass tube
Figure 2 Test apparatus to determine the burning behaviour
of dusts: R. Siwek
the sample increases to 400C or more, this is known as
self-ignition, if not self-heating3.
The determination is performed in a Grewer oven
in which 8 ml of the test substance and the same amount
of graphite powder as reference are heated up to 350C
in fine wire mesh baskets at a rate of 1 K min- in an
air stream of 2 1 min-. The temperature determined in
this manner is specific to the measurement procedure.
The Grewer test corresponds to one of the tests
described in the EC guideline4 and is used when a new
substance is registered under the chemicals act.
For products which are briefly exposed to a heated
air stream (e.g. in spray or pneumatic-conveyor dryers),
under certain conditions it is often sufficient to select
a safety margin of at least 50C between the operating
temperature and the relative self-ignition temperature.
The measured RSIT is not only dependent on the pro-
duct, but is also influenced by the volume of the dust
deposit. With low relative self-ignition temperatures
(e.g. RSIT < 22OC), experiments should thus be perfor-
med with larger test volumes (heat ageing tests) at
lower temperatures.
Self-ignition temperature, SIT (heat ageing tests in a
wire basket)
The self-ignition temperature determined through heat
ageing tests is the lowest ambient temperature at which
a substance just fails to exhibit exothermic behaviour
(AT < 5C) 72 h after reaching the temperature of its
surroundings.
It is used for an accurate assessment of the hazards
associated with the handling of relatively large amounts
of product. The samples under investigation are added
to cylindrical wire mesh containers and stored in a dry-
ing oven through which a stream of air flows at a con-
stant temperature. Usually, wire baskets of content 400-
3000 ml are used3s6. When combustible dusts are stored,
the dependence of the self-ignition temperature on the
quantity must be taken into consideration and this can
be estimated by extrapolation from experiments with dif-
ferent amounts of dust.
According to UN recommendations, substances
which reach or exceed a temperature of 400C (200C
in Japan) in a 1000 ml wire basket at surrounding tem-
peratures less than 140C must be assigned to transport
class 4.2.
(left) and the burning velocity (right)
 Hazard evaluation
Relative decomposition temperature, RDT (test for
exothermic reaction in an open cup)
This is defined as the lowest temperature at which the
temperature of the test substance starts to exceed the
temperature of a thermally stable reference substance
when both are heated simultaneously with exclusion of
air.
In addition to differential thermal analysis (DTA),
the method described by Ltolf3 has wel1 proved its
worth as it allows not only determination of the relative
decomposition temperature but also assessment of the
amount of gases released in the exothermic reaction and
whether they are flammable (Figure 3). It is used for
products or reaction masses which are exposed to an
elevated temperature in part in the absente of fresh air
(e.g. in vacuum ovens, fluidized-bed dryers).
An approximately 2 g sample and a reference subst-
ante (graphite) are heated continuously up to 350C at
a rate of 2.5 K min-. If an exothermic reaction occurs,
a fresh sample is tested at constant temperature. The test
is repeated, each time with fresh sample, at decreasing
temperature intervals of 10 K until the substance no
longer exhibits exothermic behaviour over a period of at
least 8 h.
The relative decomposition temperature by the
Ltolf method is the temperature at which, in the exper-
iment at a constant oven temperature, the substance just
fails to show an increase in temperature compared with
the reference substance over an 8 h period.
The results of these investigations can also be use-
ful in an assessment of the consequences of decompo-
sition reactions in installations as they allow estimation
of the possible formation of readily ignitable and very
violently reacting hybrid mixtures.
For reasons of safety, at least 50C are subtracted
from the experimentally determined relative decompo-
sition temperature and the temperature so obtained is
specified as the maximum admissible medium heating
temperature. For a more accurate determination of the
admissible temperature exposure, heat accumulation
tests must be performed in a Dewar vessel at lower tem-
peratures if the measured relative decomposition tem-
perature is less than 200C.
to recorder
to qas meter+
3 ,Nz/ventilation
heating black, temperafure controlled
Figure3 Test apparatus described by Ltolf
the relative decomposition temperature
of dusts: R. Siwek 23
Decomposition temperature, DT (heat accumulation
experiments in a Dewar jlask)
The substance under test is stored in a Dewar flask (heat
accumulation tests) together with a reference substance
in a thermostated box or drying oven at a constant tem-
perature of the surroundings. Usually, Dewar flasks of
200-2000 ml content are used. Self-heating of the pro-
duct is shown by the temperature of the test substance
exceeding that of the reference substance. The test is
repeated with fresh samples and a temperature decrease
of 10 K each time until exothermic behaviour can no
longer be observed (AT < 5C) over a period of at least
72 h. As the determined induction time of the decompo-
sition (time between attainment of the oven temperature
by the sample and the maximum sample temperature)
is generally linearly dependent on the reciprocal of the
absolute temperature, a reliable storage temperature for
a specified storage time can be determined.
Spontaneous decomposability, SD
A substance is regarded as spontaneously decomposable
if it undergoes complete decomposition with a progress-
ive reaction zone after local initiation by a sufficiently
powerful ignition source, also without the presence of
(atmospheric) oxygen3.
The test substance (5-10 ml) is placed in a crys-
tallizing dish in a Witt jar (V= ca. 1 litre) and ignited
with the glowing, integral platinum wire (Figure 4). If
the sample burns, the jar is evacuated. If the product
continues to react in a vacuum, it is said to undergo
spontaneous decomposition. The test is repeated, but this
time ignition is not initiated until the vessel has been
evacuated. If the product reacts in vacuum and
undergoes a propagating decomposition reaction, spon-
taneous decomposition is said to have occurred. The test
is conducted at room temperature (20C) and at an elev-
ated product temperature (e.g. 100C).
A spontaneous decomposition can be initiated by,
for instance, a heated foreign body. However, it can also
result from a fire initiated locally in which the fire
changes into a spontaneous decomposition owing to an
insufficient supply of air. A spontaneous decomposition
can be neither prevented by inerting nor stopped by
smothering.
As large amounts of decomposition gases can be
evolved within a short space of time in a spontaneous
decomposition, a pressure build-up hazard can exist
evacuoted
glans pot
I-e
--r
vacuum
IWittscher pot)
t-
platinum wire
id=
(-1000C 1
test sample
c.
for determining
Figure 4 Test for spontaneous decomposition
24 Hazard evaluation
which can lead to the apparatus rupturing. For this rea-
son, products which decompose spontaneously can only
be processed batchwise and in smal1 amounts, keeping
the residence time to a minimum and the temperature as
low as possible.
A laboratory apparatus is at present being
developed by Ciba and Khner to allow accurate investi-
gations of spontaneous decomposition.
Impact sensitivity (falling hammer test)
This test determines whether a decomposition reaction
or an explosion can be initiated by impacting the sample.
It does not correspond to the test required by the codes
for explosives.
The sample under investigation is encased in a die
comprising two coaxial steel cylinders one above the
other and a hollow cylinder made of steel as a guide
ring. This die, which must contain a particular amount
of sample, is placed on an anvil and subjected to the
action of different drop weights. Depending on the test
method, six or ten trials are conducted. If a reaction
(smoke, fire, sparks) is observed, the sample must be
assessed as impact-sensitive. However, if a detonation
(bang) is heard, the product has explosive character and
thus must not be milled when dry. In such a case, more
exhaustive investigations are necessary3.
Minimum ignition temperature of dust layer (glowing
temperature)
The minimum ignition temperature of a dust layer is a
characteristic which describes the ignition hazard of a
horizontal dust layer on hot surfaces. It is defined as the
lowest temperature of a heated, open, exposed surface
on which a 5 mm thick dust layer wil1 be ignited. These
investigations are done with a temperature-controlled,
electrically heated plate. If the tests are carried out with
a dust layer more than 5 mm thickness, then this charac-
teristic is called glowing temperature. Thicker dust lay-
ers ignite at lower temperatures.
For surfaces which cannot be kept effectively clear
of dangerous dust accumulations that are susceptible to
glow (smouldering), the temperature must be at least
75C lower than the minimum ignition temperature of
the specific dust layer.
Dwt suspensions
The international standards-3 describe the 1 m3 vessel
as the test apparatus. In recent years, increasing use has
been made of the more convenient and less expensive
20 litre apparatus as the standard equipment. This is
admissible because the results obtained in the 20 litre
apparatus using a wide variety of dusts are the same as
those found in the 1 m3 vessel. At present, over 70 are
in use worldwide. Investigation of the explosion behav-
iour of combustible materials (combustible dusts, flam-
mable gases, solvent vapours) must be performed in
accordance with internationally recognized test pro-
cedureslm6. For the determination of combustible gases
and combustible vapours, the test is generally done in a
quiescent state (n, = 0 s), whereas, with combustible
dusts or a hybrid mixture, a standardized, moderate mix-
ture turbulente is specified (Figure 5). The ignition
of dus&: R. Siwek
ExdoBioll
Figure 5 Dust investigation procedure in internationally
accepted test equipment (schematic)
delay time (turbulente level) in the 1 m3 vessel is in gen-
era1 tv = 0.6 s and in the 20 litre apparatus it is
tv = 0.06 s.
In both procedures, the dust is stored under air
pressure in a dust container. After activation of the dust
container outlet valve, the dust is dispersed within the
explosion vessel through a dispersion system. When the
defined ignition delay time tv has elapsed, the explosion
is initiated by activation of a suitable ignition source.
This dust testing procedure results in fairly homogenous
dust-air mixtures that enable one to work at constant
dust turbulente. Ignition sources arranged in the middle
of the test apparatus depend on the substance being
tested, and are as follows:
0 combustible dusts: two chemical ignitors with a total
energyE=lOkJ(2x5kJ)
0 flammable gases and solvent vapours: inductive con-
tinuous spark gap or capacitor discharge spark with
an energy E = ca. 10 J.
During the last 20 years, intensive efforts have been
made to make the 20 litre apparatus a useful too17-*, not
only for determination of explosion data but also for
other test purposes. The goal of these numerous investi-
gations was to ensure good correlation of the results with
those of the standard 1 m3 vessel13-15.
In the 20 litre apparatus (Figure 6), important
explosion characteristics of combustible dusts such as
lower explosion limit, maximum explosion overpressure,
maximum explosion constant, and limiting oxygen con-
centration can be determined with adherente to stan-
dardized test procedures.
Lower explosion limit, LEL
The lower explosion limit, LEL, of a dust is the highest
concentration at which a dust explosion is not detected
in three successive tests.
In industrial practice, the LEL of dusts can be of
importante if, for example, the explosion hazards of
combustible products have to be assessed in dust extrac-
tion (filtration) systems.
In the 20 litre apparatus, the concentration is sys-
tematically reduced (10 g me3 steps) until a concen-
tration is reached at which no ignition of the dust-air
mixture is observed in three successive tests2,16. For
assessment of whether or not a reaction occurred, the
pressure criterion was used, taking into account the
 Hazard evaluation of dusts: R. Siwek 25

Figure 6 20 litre apparatus for the determination of explosion characteristics of fuels

pressure effect of the standard ignition source (2 x 5 kJ). over a wide range of concentrations (Figure 8) using
There is no ignition if: two chemical ignitors (2 x 5 kJ). The two peak values
normally occur at different dust concentrations.
?? less than 2 bar overpressure in the 20 litre apparatus The explosion indices P,,, and (dP/dt),,, are
is observed, and defined as the mean values of the maximum values of al1
?? less than 0.2 bar overpressure in the 1 m3 vessel is three seriesom2. Subsequently, the maximum explosion
observed.
constant K,,, is calculated from (dP/dt),,, by use of the
As can be seen from Figure 7, there is a good agreement cubic law:
between the LEL values in the two types of test appar-
V (dP/dt),,, = K,,, (1)
atus. They are either equal or generally lie within a con-
centration step of each other. Despite the very powerful where K,,, is the maximum explosion constant specific
ignition source, no influence of the volume of the test to the dust or gas and dependent on the test procedure
apparatus on the LEL of combustible dust is observed. and is numerically equal to the value for the maximum
rate of pressure rise in the 1 m3 vessel, and the maximum
Maximum explosion overpressure P,,, maximum rate explosion overpressure P,,, of al1 combustible subst-
of explosion pressure rise (dP/dt),, and maximum ances is independent of the volume in closed near-
explosion constant K,,, spherical vessels of adequate size.
The peak values which occur with the explosion of a As there are so many dusts produced and processed
combustible dust are the maximum explosion overpres- in industrial practice, it is appropriate to assign this
sure P,,, and the maximum rate of explosion pressure maximum explosion constant to one of several dust
rise (dP/dt),,,. These peak values are obtained from tests explosion classes and to use these as a basis for the
dimensioning of constructional protective measures
(Table 1).
120 _./
IE=Zx5kJ ,: Limiting oxygen concentration, LOC
i
a, The limiting oxygen concentration, LOC, is that oxygen
Y 90.
concentration below which an ignition of a mixture can
3 no longer take place in three successive tests. In general,

b 60= nitrogen is used as an inert gas, therefore the following
a test conditions are based on nitrogen only.
9 According to Figure 9, the LOC of a product
$j 3Qm
decreases linearly with the logarithm of the ignition
4
_: energy, ZE. This decrease is independent of the volume
o ,/ L 1 of the test apparatus, i.e. - within the range of accuracy -
0 30 60 90 120 a product-independent decrease. The oxygen concen-
LEL: 1-mSvessel [ Q/II+] tration giving IE = 10 kJ is called the LOC, where no
ignition of a dust-air mixture is possible7-20.
Figure 7 Comparison between the lower explosion limit (LEL)
of combustible dusts in the 1 m3 vessel and in the 20 litre appar- In the case of the 20 litre apparatus, the ignition
atus energy value for determining the LOC must be limited
26 Hazard evaluation of dusts: R. Siwek

Figure 8 Determination of the maximum (peak) explosion characteristics of combustible dusts

Table 1

-KV
Relation between K,,,,, and dust explosion class St

K max (m bar s-1

1-200
Dust explosion class

st 1
._..
._/
..

201-300 st 2 . /
..
>300 st 3 t _/
./

20
ii
8
d
l //..
e
43L
8
d 10 12 14 16 18
LOC(l-m,IE- 1Ok.J) [Vd.-%]
16
Figure 11 Influence of the vessel volume and ignition energy
;ii (IE) on the limiting oxygen concentration (LOC) of combust-
ible dusts
? 10

i have to be repeated with a second ignition energy. Then,

s in a semi-logarithmic plot (see Figure 9), the LOC value
0
0.01 0.1 1 1000 loooo
IgnM:
Energ$
[J ]
Figure 9 Limiting oxygen concentration (LOC) as a function of
the ignition energy (IE) and the volume of the apparatus
(medium turbulente of the mixture)
to ZE = 250 J because higher energies have a strong
influence on the explosion behaviour, resulting in lower
LOC values compared to the ones measured in the 1 m3
vessel. Therefore, in the 20 litre apparatus, the LOC can
be properly determined only by extrapolation3*8*9,21.
Determination of the LOC of combustible dusts in
the 20 litre apparatus is carried out at medium turbulente
of the mixture, which is equivalent to the turbulente
used for the determination of explosion characteristics.
The test procedure is shown in Figure 10. These tests
can be extrapolated to an ignition energy leve1 of
ZE = 10 kJ. This procedure ensures comparison with the
loooal values measured in the 1 m3 vessel according to the stan-
dard test procedure.
The LOC measured in the 1 m3 vessel and in the
20 litre apparatus under different ignition conditions are
compared in Figure lZ*. It becomes apparent that, even
within the limits of experimental error, it is not possible
to reach the LOC of the 1 m3 vessel through selection
of a single ignition energy in the 20 litre apparatus, and
shows again that the LOC can be determined only prop-
erly by extrapolation3*8*19,21.
Figure 1Z leads to the conclusion that when an
ignition energy of ZE = 2.5 kJ is used in the 20 litre
apparatus, within the limits of the experimental error,
agreement can be reached with the LOC values in the
1 m3 vessel when this is 210 vol% in nitrogen, and if
it is accepted that for products with a relatively high

8
1

a6
-4

E
2

0 -
0 2% soo 7iitl IOW 1FBO

Figure 10 Influence of oxygen content upon explosion characteristics

 Hazard evaluation of dusts: R. Siwek 27

1ixloo

16

z 100
L 12
0

10
;

1
0 0 250 500 750 1ocaJ 1250
s -4 4
log MIE :rl
+(MITi& )} [ J ]2 Dust concentration [ g/n-P]

Figure 12 Correlation of limiting oxygen concentration (LOC) Figure 14 Influence of the dust concentration on the minimum
with minimum ignition energy (MIE) and minimum ignition ignition energy (MIE)
temperature (MIT)

limit value in the 20 litre apparatus LOC values lower

generation, has been specially developed by Khner
by ca. 1-2 vol% are measured. The ignition energy AG8,%17,2S?
ZE = 2.5 kJ in the 20 litre apparatus provides values that
The MIE is usually quoted as a range: the lower
are too high when the LOC measured according to the
value represents the highest energy at which no ignition
standard in the 1 m3 vessel is cl0 vol% and must there-
is found in at least ten experiments. The higher value, on
fore be increased to at least IE = 5 kJ (2 x 2.5 kJ).
the other hand, is the lowest energy at which the dust-air
Figure 12 shows the LOC as a function of the product
mixture is just ignited:
of the minimum ignition energy (MIE) and the minimum
ignition temperature (MIT). As can be seen, a very clear no ignition < MIE < ignition
correlation is given between the limiting values. This
The dust concentration and turbulente leve1
finding is also valid for combustible gases and
(ignition delay time) is varied until the most readily
vapours,. If the MIE and the MIT (measured in the
ignitable concentration and the optimum ignition delay
BAM oven (BAM: Bundesanstalt fr Materialprfung;
time for the MIE is found (Figures 14 and 15).
German Federal Institute for Materials Testing)) of a
dust are known, then the LOC can be calculated accord-
ing to the following empirical numerical equation9.?:
lwo
LOC = 1.62 log MIE (1 + (MIT/273)) + 12.9 (2)
An experimental determination of the LOC is thus not
- 100
necessary.
E
Minimum ignition energy, MIE g 10
The minimum ignition energy, MIE, of a combustible .__ ..-
r
substance is the lowest value of the electrical energy
stored in a capacitor, which, on discharge, just suffices 1
0
I 260 600 760 looo 1260 lsoo
to ignite the most readily ignitable dust-air mixture at
atmospheric pressure and room temperature22. Dust concentration [ g/ma]
To ensure a standardized test procedure16, a test Figure 15 Influence of the ignition delay time tv on the mini-
apparatus known as MIKE 3 (Figure 13), of the third mum Ignition energy (MIE)

Figure 13 MIKE 3 apparatus for determination of the minimum ignition energy (MIE) of dlxts
28 Hazard evaluation
MIE with L [ mJ ]
Figure16 Influence of the inductance L on the minimum
ignition energy (MIE)
Inductance L causes the capacitor to discharge in a
protracted manner. As Figure 16 shows, dust-air mix-
tures are more readily ignited by time-extended dis-
charges than by purely capacitive discharges8~o-2.
If the MIE is used for assessment of ignition haz-
ards resulting from electrostatic charges, it must be
determined without an inductance in the discharge cir-
cuit. Only in this manner is a realistic assessment of the
incendivity of electrostatic discharges towards dust-air
mixtures possibles*o-2.
Minimum ignition temperature, MIT (ignition test for
airborne dwt)
The minimum ignition temperature, MIT, is defined as
the lowest temperature of a heated surface at which the
most readily ignitable mixture of a dust with air just
ignites. It provides information on the ignition behaviour
of a dust suspension when quickly passing over a hot
surface such as in spray and fluid-bed dryers,.
In Europe, the MIT is usually determined in the
BAM furnace2,. This apparatus is enjoying increasing
popularity (Figure 17).
The BAM furnace is heated to a maximum tem-
perature of ca. 600C. The furnace temperature is
allowed to fa11 and a dust sample (1-2 ml) is blown
against the test plate axially at intervals of 50 K.
This is repeated until ignition is no longer found.
The furnace is then reheated to the next higher tempera-
ture stage. The test is repeated, again with decreasing
furnace temperature, but this time at temperature inter-
vals of 10 K until no ignition (flame appearance, audible
reaction) is found in at least three experiments. The
appearance of glowing dust particles alone or ignition
later than 5 s after dispersion of the dust is ignored.
Heating
Test plak
1 v
Figure 17 BAM furnace
of dusts: R. Siwek
In order to compare MIT values measured in the
BAM with the ones measured in the Godbert Greenwald
(GG) furnaces24, the MIT values of over 100 different
dusts from various laboratories have been investi-
gated7~9~2The
. following equation (3) for the mutual
dependence of the two MIT values allows a comparative
estimation and has been calculated by regression over
al1 measured values:
MIT(GG) = 1.1 MIT(BAM) - 10C (3)
Hybrid mixtures
Products which contain flammable solvents can form
hybrid mixtures. These are easier to ignite and explode
with greater severity than solvent-free dust-air mixtures.
This increased hazard can become important when the
solvent vapour concentration exceeds 20% of the lower
explosion limit. With products which contain no more
than 0.5 wt% flammable solvents, this is not anticipated.
Dusts with a higher solvent content require detailed
investigation3. The special features of hybrid mixtures
are that
Non-explosible dust-air mixtures and non-explosible
flammable gas (solvent vapour)-air mixtures can
form explosible hybrid mixtures. The LEL of the
hybrid mixture lies between the LEL of the combust-
ible dust and the LEL of the flammable gas, e.g. pro-
pane (Figure 18).
While addition of flammable gas to the combustible
atmosphere raises P,,, to some extent (Figure 19,
left), the explosion constant K,,,,, of the combustible
dust is increased considerably, even though its con-
centration is below the LEL. When the optimum gas
concentration is present, al1 dusts in question wil1 be
classified into the dust explosion class St 3 (Figure
19, right). The K,,, of the hybrid mixtures lies
between the K,,,,, of the combustible dust and the
K max of the flammable gas, e.g. propane (turbulent
condition).
3. The MIE and the LOC of the hybrid mixture are
determined by the combustible substance with the
lowest limit value (Figures 20 and 21).
Influencing factors
Dust accumulation
The self-ignition temperature of a specific dust not only
depends on the type of the dust (chemical nature, particle
size, moisture content, etc.) but also on the size of the
pile (Figure 22).
40
0
0 0.3 0.6 0.8 1.2 1.5
C3H6 in air [Vol. -% ]
Figure 16 Lower explosion limit (LEL) of hybrid mixtures
 Hazard evaluation of dusts: R. Siwek 29

Figure 19 Explosion characteristics of hybrid mixtures

Temperature
The temperature has sometimes a marked influence on
ulwo
the buming class3. Table 2 shows some examples of pro-
lW0
ducts whose buming class is dependent on the tempera-
7 100
ture.
t; 10
A temperature increase from 20 to 100C essen-
1
f
tially changes the reaction progress from IZOpropagation
ai of a jre (BC: 1-3) to spread of a jire (BC: 4-6).
OOI An increase of the temperature wil1 linearly reduce
0.001 the value of the LEL of the combustible dusts. This
t@w
.-I

0 al5 1 1.5 2 26 3 3.6 4 4.5

influence is more pronounced the higher the value is at
Roprw- W%l
room (ambient) temperature (T0)3~7-9.The influence of
the temperature (7) on the LEL can be estimated accord-
Figure20 Minimum ignition energy (MIE) of hybrid mixtures
(dust and propane)
ing to the following equation:
LEL(T) = LEL(T,)( 1 - O.O027(T- T,,)) (4)
The maximum explosion overpressure (P,,,)
decreases linearly with the reciprocal value of the
- 14
increasing temperature at ambient (normal) pressure.
r
According to present knowledge, the K,,,,, wil1 only
3 change insignificantly6,8m0.
- 12
The temperature dependence of P,,, can be
8 Lot: Roparle
described according to equation (5):
10 _______________________________
P,,,(r) = P,,,(T,,)(0.88(To + 273)/(T + 273)

8 + 0.12) (5)
0 2 4 6 8
Propane concentration [ Vol.-% ] The LOC is also influenced by temperature. An
increase of temperature results in a linearly decreasing
Figure 21 Limiting oxygen concentration (LOC) of hybrid mix-
tures (dust and propane) of the LOC6,8~o.
The temperature influence on the LOC for different
combustible dusts and the inert gases nitrogen and car-
bon dioxide can be estimated according to equation (6):
LOC(Z-) = LOC(T,) - a(T- T,) (vol%) (6)

Table 2 Influence of temperature on the burning class, BC

Product BC @ 20C BC @ 100C

Toluene-4-sulfonic acid 2
chloride
Tetrahydrophthalic acid 2
anhydride
1-Diazo-2-naphthol-4-sulfonic 2
acid
Hydroxyquinaldine-4-carbonic 3
1 10 100 1000 1WW acid
8moulltQfdust [cnis] Dextrin 3
Butter powder 3
Figure 22 Correlation of self-ignition temperature (SIT) of Milk powder 3
cylindrical dust piles with dust quantities
30 Hazard evaluation of dusts: R. Siwek

The disadvantage of equation (6) is that both the dust

and the inert gas are dependent of the temperature coef-
ficient (Y(vol% KV).
By analysing al1 available test results, it was found
that by increasing the temperature by lOOC, the LOC
wil1 decrease by an average of 1.3 vol% if nitrogen as an
inert gas is used6. Based on these findings, the following
simplified equation can be used for estimation of the
temperature dependency on the LOC if nitrogen is used 8
10 1
as inert gas: 0 1 2 3 4 6 8 7 B

Figure24 Influence of the initial pressure (Pi) on the limiting

LOC(T) = LOC(7,) + 0.26 - 0.013 T (7) oxygen concentration (LOC)
The temperature has an appreciable influence on the
MIE of a dust4~6*8-10. 1OOW
In a log-log plot, the MIEs of dusts are systemati-
1000
cally lowered with increasing temperature and meet at
a point (lOOOC, 0.088 ml). This relation can also be
g 100
described by the following equation:
MIE(T) = 10 exp(- 4.056 + (1.873 - 0.624 log T)
x (log MIE(T,) + 4.056)) (8)
Equations (5)-(8) can be used to estimate the safety 0.1 I
characteristics in a temperature range from 20 to 300C. 0 0.5 1.5 2

The corresponding experiments at elevated temperatures Pi [ lm:. abs. ]

are thus not needed.
Initial pressure
An increase in initial pressure, Pi (this is the pressure
existing at the moment of ignition), causes a proportional
rise in the P,,, and K,,, 4-6. Reducing the initial pressure
in a piece of equipment wil1 cause a decrease in the
explosion violente (Figure 23).
The correlation between the initial pressure, Pi, and
the LOC has not yet been investigated for many differ-
ent substance@.
While, with increasing initial pressure, coal dust
and peat show an increase of the LOC, brown coal shows
the opposite phenomenon, i.e. the determined LOC
decreases slightly (Figure 24). The reason for this dis-
crepancy could be that application of the standard test
procedure at elevated initial pressures creates some dif-
ficulties.
Figure 25 shows the correlation between initial
pressure (Pi) and the MIE for two different dusts4s6. In
the range of 1-1.6 bar (abs) for the initial pressure, there
is no noticeable influence. For negative pressure [Pi < 1
bar (abs)], however, the readiness for dusts to ignite
decreases. This is understandable because the most read-
ily ignitable dust concentration decreases with decreased
Figure 25 Influence of the initial pressure (pi) on the minimum.
ignition energy (MIE)
initial pressure which, in turn, may cal1 for an increase
in the MIE value.
The MIT of dusts is particularly important in the
case of operating areas which are subject to an explosion
hazard due to dusts. Here, only equipment whose surface
temperatures reach a maximum two-thirds of the MIT
in continuous operation may be used.
Obligatory testing
Every product which has not been tested must be investi-
gated by a suitable test laboratory in accordance with
the relevant regulations using a sample taken from the
current or proposed production process. With products
being processed for the first time, this test must be per-
formed before any unit operation is begun, using a
reliable average sample of the lot in question.
If the processing of a product involves the addition
of other substances, an appropriate mixture must be
tested as the handling safety of mixtures can not be
easily derived from the properties of the individual
components.

10

0
0 0.2 0.4 08 a0 1 1.2 1.4 1.6 3.8 2
inMalpmumPI [bar,abe.l

Figure 23 Influence of the initial pressure pi on the maximum explosion characteristics

 Hazard evaluation
To assess the safety of a unit operation, the test
laboratory must know not only the content of possible
solvents, but also the usual product properties such as
empirical formula, melting point and toxicity.
A test must be repeated if a change in the hazard
situation is anticipated, for instance as a result of
a change in the production process, including work-
UP,
a change in the specifications or origin of starting
materials and additives,
irregularities in the production, or
production relocation.
References
1 VDI Guideline 2263: Dust Fires and Dust Explosions, Hazards-
Assessment-Protective Measures, Beuth, Berlin und Kln, May
1992
2 VDI Guideline 2263, Part 1: Dust Fires and Dust Explosions, Haz-
ards-Assessment-Protective Measures, Test Methods for the
Determination of the Safety Characteristics of Dusts, Beuth, Berlin,
May 1990
3 Bartknecht, W. Explosion Protection, Basics and Application,
Springer, Berlin, 1993 (in German)
4 Guideline 84/449/EC. OfJicial J. Eur. Communiry, L251, 27 A.10
1984
5 UN Recommendations on the Transport of Dangerous Goods,
1986, Chapter 14, p 3
6 Bowes, P. C. Self-heating: Evaluating and Controlling the Haz-
ards, Her Majestys Statinery Office, London, 1984 -
7 Bartknecht. W. Dust Exulosions. Course Prevention. Protection.
Springer, erlin, 1989 L
8 Siwek, R. Reliable Determination of the Safety Characteristics in
the 20-1 Apparatus, Proceedings of the Conference Flammable
Dust Exolosion. St. Louis. Missouri. USA. 2-4 November 1988
9 Cesana, Ch. and Siwek, R. KSEP 332 - Measurement and Control
System for the 20-1 Apparatus, Khner, Birsfelden, 1991
10 Siwek, R. Explosion Indices, Khner, Birsfelden, 1991
11 Cesana, Ch. and Siwek, R. Screening Tests: Test Procedures,
Khner, Birsfelden, 1991
of dusts: R. Siwek 31
12 Siwek, R. and Cesana, Ch. Assessment of the Fire & Explosion
Hazard of Combustible Products for Unit Ouerations. Proceedinns
of the International Conference and Exhibibon on Safety, Healyh
and Loss Prevention in the Oil, Chemical and Process Industries,
Singapore, 1993, Butterworth-Heinemann, London
13 ISO/DIS 61840: Explosion protection systems - Part 1: Determi-
nation of explosion indices of combustible dusts in air, Inter-
national Organization for Standardization, 1985
14 ISO/DIS 6184/2: Explosion protection systems - Part 2: Determi-
nation of explosion indices of combustible gases in air, Inter-
national Organization for Standardization, 1985
15 ISO/DIS 6184/3: Explosion protection systems - Part 3: Determi-
nation of explosion indices of fuel/air mixtures other than dust/air
and gas/air mixtures, International Organization for Standardiz-
ation, 1985
16 IEC. Electrical apparatus for use in the presence of ignitable dust.
Part 2 Test Methods. Sheet 2-5. Method for determining the mini-
mum explosible concentration of dust/air mixtures, 1990
17 Siwek, R. et ul. Ignition behavior of dusts, Proceedings of the 28th
Annual Loss Prevention Symposium, Atlanta, Georgia, USA, 17-
21 April 1994
18 Siwek, R. Latest development in explosion protection technology,
Proceedings of the 6th International Colloquium of Dust
Explosions, Shenyang, PR China, 29 Aug.-2 Sept. 1994
19 Siwek, R. and Cesana, Ch. Ignition behavior of dusts: meaning and
interpretation. Process Safety Progress, April 1995
20 Bartknecht, W. and Siwek, R. Meaning of Ignition Energy and
Minimum Ignition Temperature for the Assessment of the Ignit-
ability of Ignition Sources in Dust-Air Mixtures, Part 1: Minimum
Ignition Energy and Minimum Ignition Temperature, Staub-Rein-
haltung der Luft 54 (1994) Springer, Berlin, 1994 (in German),
pp 325-330
21 Bartknecht, W. and Siwek, R. Meaning of Ignition Energy and
Minimum Ignition Temperature for the Assessment of the Ignit-
ability of Ignition Sources in Dust/Air-Mixtures, Part 2: Assess-
ment of the ignition efficacy, Staub-Reinhaltung der Luft 54
(1994) Springer, Berlin, 1994 (in German), pp 325-330
22 IEC. Electrical apparatus for use in the presence of combustible
dust, Part 2, Test methods, Section 3: Method for determining mini-
mum ignition energy of dustfair mixtures (draft), 1992
23 Cesana, Ch. and Siwek, R. MIKE 3 Minimum Ignition Apparatus,
Khner, Birsfelden, 1992
24 IEC. Electrical apparatus for use in the presence of ignitable dust,
Part 2, Test methods, Sheet 2-2: Method for determining the mini-
mum ignition temperature of dusts - Dust cloud in a furnace at
constant temperature, 1984
 J. Loss Prrv. Process Ind. Vol. 9. No. 1.pp.91-103, 1996
Crown copyright 0 1996 Published by Elsevier Science Ltd
Printed in Great Britain
0950-4230(95)00055-0 09504230/96 $15.00 + 0.00
ELSEVIER

Dust explosions in interconnected vented

vessels

P. Holbrow, S. Andrews and G, A. Lunn

Health and Safety Laboratory, Health and Safety Executive, Harpur Hill, Buxton,
Derbyshire SKI7 9JN, UK

Explosion venting is used often to protect industrial dust-handling plant from pressures gener-
ated by accidental dust explosions. Many calculation methods are available to estimate a safe
vent area, but these are without exception applicable to single vessels. In industry, vessels are
usually linked and flames can travel through connecting pipework from one vessel to another.
The behaviour of flames in such configurations is not well understood and the pressures that
can be developed are not well known. As a result, the venting requirements are difficult to
estimate. This report describes a project in which the behaviour of dust explosions in linked
vented vessel systems was studied. Vessels ranging in size from 2 to 20 m3 were used, linked
via pipes up to 15 m long with diameters of 0.15, 0.25 and 0.5 m. Tests were carried out using
coal dust, toner dust and anthraquinone dust. The work has shown that the passage of flame
from the primary ignition vessel can result in a secondary explosion which produces a much
higher pressure than expected from a vented single vessel. Secondary explosion pressures up
to 6.2 bar were recorded. The pressure in the first vessel was also enhanced but to a lesser
degree. These effects were dependent on the pipe diameter, the pipe length, the vessel volumes
and volume ratio, the vessel in which the explosion was initiated, the explosibility of the dust,
and the vent area. Effectively, the K,,nomograph approach for sizing the vent area was found
to be adequate for vessels linked via the 0.15 m diameter pipe using coal or toner dust. However,
explosions in vessels linked by pipes having larger diameters (0.5 and 0.25 m) and a shot? length
generated enhanced reduced explosion pressures. The pressure enhancement was greatest
when the primary ignition occurred in the larger of the two vessels. The vessel volume ratio
tended to increase the pressure enhancement. Also, the larger the vessel volume for a volume
ratio of l:l, the greater was the pressure enhancement. The more reactive dusts, toner and
anthraquinone, generally produced significantly higher reduced explosion pressures in equival-
ent vessel/pipe configurations when compared with coal. Larger vent areas generated lower
absolute pressure than the smaller vent areas.

Keywords: dust explosions; vented vessels; secondary explosions

More than 70% of dusts or powders processed in indus-
try are combustible, and a dust explosion hazard may
arise in any plant handling these dusts. It has been esti-
mated that there is one dust explosion on average every-
day in Europe, in the wood, food, chemicals, fuel and
pharmaceuticals industries. Recent data from the USA
for grain dust explosions show 13 in 1993, and 154 from
1984 onwards. The relative frequency of incidents in
each industry depends very much upon the source of
statistical data3.
Dust explosions cause a rapid increase in pressure
when confined in vessels or other process equipment.
Such pressures cannot be contained by most plant and
measures have to be taken either to prevent the explosion
or to protect the plant against the destructive effects.
A number of systematic studies of dust explosions
have been carried out and the results used to produce
guidance on the prevention and mitigation of dust
explosion hazards. This guidance is published as VDI
36734, NFPA 6gs and the IChemE guide? Dust
Explosion Prevention and Protection Parts 1 to 3.
However, this guidance usually applies to isolated
vessels, whereas, in industry, plant items are connected
by pipelines or conveyors and an explosion may propa-
gate from the first vessel to the second causing a second-
ary explosion. The conditions under which this second-
ary explosion takes place are generally unknown, but
there is enough evidente from the literature on gas
explosions to show that in some circumstances there can
be a significant increase in the rate of combustion and
violente of the explosion, and a significant increase in
the explosion pressures7-9. Similar effects have also been
noted with dust explosionstOst .
The reasons for this increase in the rate of combus-
tion are the generation of turbulente as the explosion
passes through the interconnecting pipe and the entry
into the second vessel of a relatively large jet flame
ignition source. If protection techniques such as venting
or suppression are applied to these secondary explosions,
the degree of protection calculated on the basis of a
single, isolated vessel may be insufficient to cape with
a more violent secondary explosion.

91
92 Dust explosions in interconnected
If dust-handling plant does consist of several con-
nected items of equipment, consideration must be given
to isolating the various units to prevent transmission of
flame or other ignition sources, to strengthening the plant
to cape with high pressures and to increasing the leve1
of explosion protection. However, there is little work on
the pressure changes that can occur, and guidance for
the protection of such systems - apart from the instal-
lation of isolation devices - is sparse.
The British Materials Handling Board (BMBH) has
sponsored a project with the Health and Safety Executive
to investigate the behaviour of dust explosions in sys-
tems of connected vessels. The main purpose of the pro-
ject has been to measure the pressure changes that occur
when a dust explosion is ignited in one vessel and the
flame transmits into the other. The aim is to provide
industry with guidance as to when, and to what degree,
additional venting is required on linked vessel systems
and whether alternative precautions are required.
Literature review
There are a number of accepted methods for calculating
the area of vent required for effective and safe explosion
venting4-6.
Generally, the wel1 known calculation methods
such as the K,, nomograph approach produce vent areas
on the conservative side, i.e. the vent area is larger than
is strictly necessary to protect the plant from the real
explosions likely to occur inside it. Nevertheless, there
are conditions and circumstances in which intense
explosions could occur and with which the normal
degree of venting could not cape.
The intensity of a dust cloud explosion is affected
by the turbulente. The more turbulente, the more rapid
the combustion rate, and the more violent the explosion.
This .becomes particularly apparent when explosions
propagate along pipes and ducts where the confinement
greatly enhances turbulente in the dust cloud ahead of
the flame and within the flame front itself. This wil1
increase the rate of pressure rise, and therefore more
venting is required to protect the plant. For dust
explosions in pipes connected to vessels, whilst the
flame is stil1 within the vessel the pressure gradually
increases in the pipe as dust-laden air is pushed into it.
Any dust previously deposited in the pipe wil1 be raised
into suspension. Once the flame enters the pipe, the rate
of pressure rise wil1 be faster than that in the vessel13.
This could be especially important where two ves-
sels are interconnected. An explosion in the first vessel,
by driving material through the interconnecting pipe,
wil1 not only enhance the turbulent conditions in the
second, but produce a large energetic jet flame ignition
source there.
However, turbulent flow in dust-air mixtures is
likely to generate strong concentration gradients, giving
rise to rich and lean zones. These may have considerable
influences on the propagation mechanisms of turbulent
dust explosions. The dust particles cannot always follow
the streamlines of a turbulent flow, especially in eddies,
even at low eddy velocities. The higher the Reynolds
number and turbulent flow velocity, the more pro-
nounced this effect becomes14. This effect has been seen
in pulverized coal dust clouds emerging from 20 m long
vented vessels: P. Holbrow et al.
150 and 250 mm diameter pipes connected to a 1 m3
vessel15. Wel1 demarcated, symmetrical zones of lean
and rich coal dust were observed. Upon arrival of the
flame front, the flame intensity in the rich zones was
greater. The effect seemed to be closely related to the
flow conditions in the pipe, and hence the explosion
severity in the vessel. Pineau and Ronchail15 have
assumed that the structure of the dust cloud and the
flame just at the end of the pipe is representative of that
inside the pipe, so that the extent to which the explosion
propagates down the pipe is dependent upon the local
dust concentrations within it.
When an explosion in a vessel enters a pipe, the
change in geometry causes a rapid acceleration of the
flame front16. Depending on the length and diameter of
the pipe, this flame front may lose its momentum and
peter out, or may continue to accelerate, possibly to the
point of transition to detonation. Where heat losses are
great, e.g. when the flame propagates along a narrow
pipe, the flame may eventually be quenched, and for a
given dust and experimental conditions a critical quench-
ing diameter exists .* . Where the pipe is above this
critical diameter, the flame wil1 accelerate due to wal1
friction turbulence19 and may reach a steady velocity of
the order of the speed of sound in the combustion pro-
ducts. If the pipe is long enough, deflagration to deton-
ation transition (DDT) may occur. Changes in pipe diam-
eter along its length promote flame acceleration as do
bends which allow pressure piling.
Oscillatory flame behaviour is reported under cer-
tain conditions for gas mixtures in pipes. These oscil-
lations of the flame front are triggered by the first contact
between the flame and the pipe walls, when rapid cool-
ing takes place, setting up hydrodynamic instabilities in
the flame. Changes in the burning rate of the flame front,
flame-front instability and longitudinal acoustic vibra-
tions al1 influence oscillatory combustion. Oscillatory
combustion has also been reported for vented gas
explosions from single vessels, particularly when the
vent area is large *J* . Harrison and Eyre* found that the
oscillatory flames always began when the pressure of the
external explosion peaked, often with a consequent rise
in pressure within the vessel (the external explosion
invariably peaked before that inside the vessel).
Studies of gas explosions have been carried out in
interconnected totally enclosed vessels7-9,23. Where the
vessels are of the same volume, not only can the
maximum explosion pressure in the second vessel be
higher than that in the first, but the rate of pressure rise
in both vessels can be significantly increased. This is
particularly apparent in the second vessel. The diameter
and to a lesser degree the length of the interconnecting
pipe were the important variables16.
If vessels of different sizes were used, the
maximum pressures and rates of pressure rise depended
on the vessel in which the primary ignition took place.
When this vessel was smaller than the second, the values
of the maximum rate of pressure rise were similar in
each vessel and increased as pipe diameter increased.
The maximum explosion pressure also increased in both
vessels as pipe diameters increased, but for smal1 pipe
diameters this pressure was lower in the second vessel
if it was larger than the first. When ignition was in the
 Dust explosions in interconnected vented vessels: P. Holbrow et al. 93

larger vessel, the maximum pressures and rates of press- Table 1 Explosion characteristics of dusts
ure rise in the second vessel were considerably higher
than those of single vessels. Optimum
Singhg noticed that the maximum explosion press- K,, value P concentration
Dust (bar m s-1 maXg)
(bar (kg m-?
ure reached a maximum value as the volume ratio
between the vessels increased up to a point where back- Toner dust 222 7.1 0.25-0.50
venting from the second vessel caused the pressures to Coal dust 168 7.7 0.25
leve1 out or even decrease, depending on the gas mixtare Anthraquinone 308 8.4 0.25-0.375

and pipe diameter. Larger pipe diameters allowed more

back-venting. In addition, the pressures generated in the
two vessels became more similar as pipe diameter
The vessels
increased, because the pipe became less effective at par- Five vessels have been used in this programme. Table
titioning the vessels. Thus pressure piling was reduced 2 lists the vessel volume configurations employed in the
as pipe diameter increased.
tests. The dimensionless vent area is defined as AW3,
Phylaktou and Andrews found also that the pos- where A is the vent area and V is the volume of the
ition of ignition source was important. The further back
vessel. Figure 1 shows an example of an interconnected
in the first vessel it was, the more violent the explosion
vented system.
in the second vessel. Singh suggests that beyond a given
distance from the pipe the position of the ignition source
wil1 cease to have an effect on the pressure in the second
Table 2 Vessel volume configurations used in the tests
vessel. This is because the amount burnt in the first
vessel before the flame extends into the pipe wil1 reach Volume (m3) Dimensionless vent area
a constant value.
Vessel 1 Vessel 2 Minimum Maximum

Experimental
2 2 0.08 0.42
The dusts 2 6.3 0.08 0.28
6.3 2 0.08 0.28
Three dusts have been used in experiments: coal dust, 2 20 0.08 0.42
toner dust and anthraquinone. The explosion character- 20 2 0.08 0.42
6.3 20 0.08 0.26
istics of the dusts are given in Tuble 1. The measure-
20 6.3 0.08 0.26
ments were conducted in the standard 20 litre sphere 20 20 0.08 0.41
dust-explosibility test apparatus.

Figure 1 2m3-6.3m3 vessels, 0.25 m diameter x 10 m pipe

94 Dust explosions in interconnected vented vessels: P. Holbrow et al.

Pressure transducer ports are positioned in each 20m3 - 2m3 vessels

vessel, as wel1 as at intervals along the length of the
interconnecting pipe. As an example, the positions of the
transducers in the 2-20 m3 system are shown diagram-
matically in Figure 2.
The dust is injected from extemal pressurized reser-
voirs, the number of reservoirs depending on the volume
of the explosion vessel. An amount of dust necessary to
produce a concentration equal to the concentration of Vent area: 0.812m2/ 0.19m2
A/V:O.ll 10.12
dust in the primary vessel is laid along the connecting
pipe.
Each vessel is fitted with a flanged opening of 0.5 m
intemal diameter. This is the connection for the pipeline,
and the vessels can be connected by up to 15 m of three Vent area: l.25m2 / 0.26m2
AlV:0.17/0.164
diameters of pipe: 0.5, 0.25 and 0.15 m.
The front of each vessel consists of a framework
onto which either blanking plates or membranes can be
fastened, according to the vent area required for a parti-
cular test. Figures 3 and 4 show diagrammatically vent
area configurations for the test series in two of the linked
vessel systems. Vent panel membranes were either pro-
prietary PTFE membranes with a bursting pressure (P,,,,)
of 1.12 bar a or a 200 Pm thick polythene membrane
which also had a P,,,, of 1.12 bar a. A 1V== : 0.259 / 0.247

Vented interconnected vessels

Initial tests
The potential explosion hazard that can arise when two
vented vessels are connected is shown in Figure 5. These
particular tests were carried out with two 2 m3 vessels,
connected by a 0.5 m diameter X 10 m pipe. Figure 5
shows the reduced explosion pressure (Pred) in both ves-
sels and the changes that occurred as the vent area in
Vesse12 was altered. The vent area in Vessel 1 was held
constant at 0.126 m2. The results show that the reduced
explosion pressure in Vessel 2 can substantially exceed
that in Vessel 1. Only when the vent area in Vessel 2
was increased to 0.6 m2, approximately five times the
vent area in Vessel 1, were the reduced explosion press-
ures equal.
N.B. A / W3 values do not include pipe volume
Figure 3 Vented vessel vent area configurations
As the vent area in Vessel 2 decreased, not only
did Pred in Vessel 2 increase, but so did P,, in Vessel
1. This effect was the result of the flow of gas back
into Vessel 1 due to the high pressures generated by the
explosion in Vesse12. Figure 6 shows the pressure-time
trace measured in one of the tests. A double peak struc-
ture in the trace in Vessel 1 is evident and Figure 6

r - Pressure Pressure
Heasurement Measurement
1
1

2nd dispersers

measurement

Ignition
position

J I
1 Ven

Figure 2 2 m3 vessel connected to 20 m3 vessel

 Dust explosions in interconnected vented vessels: P. Holbrow et al. 95

20m3 - 6.3m3 vessels

2m3 - 2m3 VESSELS

TONER DUST : 30 BAR 600 MS

Pipe Length : lom

Rpe Diameter : 0.5m
Vent area: 0.596m2 / 0.271 m2
A / V2 : 0.081 / 0.079

Vent area: 0.81 2m2 / 0.41 mz

A/V:O.ll /0.12

Vent area: 1.25m2 / 0.547m2

AIV2R:0.17/0.16

Vent area: 1.47m2 / 0.658m2

AIVZ:0.2/0.193

Vent area: 1.91 m2 / 0.89m2

A / V : 0.259 / 0.261
A

I I I
N.B. A / V2 values do not include pipe volume 0 I l I I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Figure 4 Vented vessel vent area configurations
Vent Area : 2nd Vessel (m3
Figure 5 Plot of reduced explosion pressure against vent area

shows also how the pressure peaks in each vessel are

related in time. These initial experiments have been
reported by Lunn.
Main test programme
The pressures shown in Figure .5 are too high for a lot
of dust-handling equipment to contain, but as Figure 5
also shows, altering the division of the vent area between
the vessels can affect the reduced explosion pressure.
The example shown in Figure 5 demonstrates that when
the total vent area is divided equally, the reduced
explosion pressures are much less than when it is divided
in the ratio 1:3. In the main test programme, the vent
area in each connected system was divided so that each
vessel had the same effective vent area, i.e.
A,,IVI = A,,lfiJ3
where A, is the vent area, V is the vessel volume and
the subscripts refer to the vessels.
Throughout the test programme, in successfully
transmitted explosions, the double peak structure in the
pressure trace from Vessel 1 was repeated to a greater
or lesser extent. In some cases the two peaks were hardly
separable and in others they were quite distinct. Systems
having short, relatively large-diameter interconnecting
pipes fitted to vessels with relatively smal1 explosion
vent areas tended to produce closely spaced pressure
peaks in Vessel 1, and a short time lapse between the
primary and secondary explosions. At the other
extreme - narrow pipes and large vent areas - relatively
long delays occurred between the initial explosion in
Vessel 1 and the secondary explosion in Vessel 2.
In the latter case, the reduced driving force behind
the flame produced a slower flame in the connecting
pipe. Figure 7 is a typical pressure-time trace showing
this behaviour.
Delays between the primary and secondary
explosions tended to reduce the violente of the second-
ary explosion.
EfSect of vent area on reduced explosion pressure. The
effect of vent area on the reduced explosion pressure is
demonstrated in Figures 8-20, which are a sample of
the total set of results and demonstrate the variation in
pressure enhancement as changes in such factors as ves-
sel volume, vessel volume ratio, pipe length and pipe
diameter and explosibility of the dust occur. The
maximum reduced explosion pressure in the system is
plotted against A,W3.
An envelope of the Preddata has been superimposed
on the plots. In Table 3, the experimental maximum
reduced explosion pressures are compared to predictions
from available guidance of reduced explosion pressures
in vented single vessels with the same dimensional vent
area, AJv2J3.
In general, the Pr_, results in the interconnected
system are in excess of the single-vessel predictions.
Although some individual results can be identified which
show experimental Pred to be lower than predicted, these
96 Dust explosions in interconnected vented vessels: P. Holbrow et al.

, , I I I

200 400 600 600 1000 200 400 800 800 1000
Time (msec) lme (msec)
Pressure - Time Trace Pressure - Time Trace
Location - Vessel 1 Localion - Vessel 2

Vessel Volumes : 2m3 - 2m3

Pipe Diameter : 0.5m
Pipe Length : lom
Vent Area : Vessel 1 : 0.126m2
Vent Area : Vessel2 :0.126m2

I , I I I
200 400 600 800 1000
Time (msec)
Pressure - Time Traces

Figure 6 Pressure-time traces

could be attributed to experimental scatter. The impli-
cation is that the explosion venting requirements for
interconnected systems generally exceed those required
for single vessels.
Tuble 4 shows in more detail the extent to which the
explosion pressure enhancement for the interconnected
vessel system increases the venting requirements for
reduced explosion pressures of 0.2, 0.5, 1.0 and 1.5 bar.
Experimental data from the tests were used and the ratio
of the experimental to predicted areas, A,,,#AKS,, was
calculated. This ratio indicates the extent to which the
interconnected system required increased venting com-
pared with predictions by the single-vessel K,, nomo-
graph method. As Pred increases, there is a progressive
increase in the ratio A,,,~AvKs,. The predicted values of
AVIV for single vessels are shown in Table 5.
Table 6 shows in summary the enhancement factor
for various configurations of vessel volume and pipe
diameter. For each configuration, the worst-case result
has been used for the tests with different pipe length.
Because in practice this factor would not be allowed to
be below unity, any value below 1 has been included in
Table 6 as unity.
Effect of pipe length. Pipe length had an influence on
the degree of pressure enhancement and the maximum
reduced pressure. The 5 m long pipe produced the high-
est enhancement in most cases. However, this was not
always the case; 10 and 15 m pipes produced high
enhanced pressures in some tests. It was not possible to
determine a single optimum pipe length for enhancing
the explosion violente.
The number of transmissions were few when ves-
sels were connected by the 0.15 m diameter pipe and no
transmissions were achieved with the 0.15 m diameter
X 15 m pipe with any of the vessels or explosible dusts.
Pipe length had an effect on the pressure profile and
the peak pressures recorded. The longer the pipe length,
the greater the separation of the pressure peaks.
The effect of backflow into Vessel 1 decreased with
increasing pipe length. There are two likely reasons for
this: the longer the pipe, the more the pressure pulse
 Dwt explosions in interconnected vented vessek: P. Holbrow et al. 97

1
2.28

2.16

1.80 ! I I I I 1
0.399 0.799 1.198 1.597 1.996 0 0.399 0.799 1.198 1.597 1.996
sec sec
VESSEL 1 PIPE

2.28 -I

2m3 - 2m3 vessels

0.5m diameter x 1 Om pipe
Av/Vw = 0.25
Toner dust
Primafy ignition in vessel 1

1.80-y I I I I I
0 0.399 0.799 1.198 1.597 1.996
sec
VESSEL 2

Figure 7 Pressure-time traces

2m3 -
2171~VESSELS 2m3. 6.3m3 VESSELS
TONER DUST TONER DUST
0
0 0.5177 dia x 5177 PIPE 0 OSm dia x 5m PIPE

A 0.5m diax lom PIPE A OSm dia x 10~ PIPE

D 0.5117 dia x 15m PIPE IJ OSm dia x 15m PIPE

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

AVIV=

Figure 8 Maximum reduced explosion pressure against Figure 9 Maximum reduced explosion pressure against
A Y /P A,lP

decreases as it travels back along it, and because the
time between primary ignition and arrival of the
returning pressure pulse at Vessel 1 increases, the less
likely it is that this backflow wil1 act to increase the rate
of combustion in the primary explosion. When pipes are
short, the back flow can occur before the primary
explosion is completed and influence the final stages of
combustion in Vessel 1.
Effect of pipe diameter. The tests involving the 0.5 m
diameter pipe generally produced the highest pressure
enhancement, and, as the pipe diameter was decreased,
mis enhancement decreased also. This is demonstrated
in Table 4 when similar vessel configurations but with
different pipe diameters are compared. Figures 13 and
14 demonstrate the lower explosion pressure that arises
for the same AVIV3 as the pipe diameter is decreased
98 Dust explosions in interconnected vented vessels: P. Holbrow et al.

6
t
6.3m3-2m'VESSELS 20m - 6.3m3 VESSELS
TONERDUST
;5 0 OSm dia x 5m PIPE
TONER DUST

0 0.5m dia x 5m PIPE

f A OSm dia x 1Om PIPE A 0.5m dia x lom PIPE

a ? ?0.5m dia x 15m PIPE m 0.5m dia x 15m PIPE

5 4
'8 \
0
f

8
a" 3
0

i
A\

2 2

1 ??

0 0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 050

AvA@ AvNu3

Figure 10 Maximum reduced explosion pressure against Figure 12 Maximum reduced explosion pressure against
AJVZ A,IV=

f- , 8 I I I I 1 I l

6.3mP. 20m3 VESSELS 2m - 2m3 VESSELS

TONER DUST TONER DUST

0 O.Sm dia x 5m PIPE Q 0.25m dia x 5m PIPE

A 0.25m dia x lom PIPE

,&,
0

, ,
*
i
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.w 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

AVNZ

Figure 11 Maximum reduced explosion pressure against Figure 13 Maximum reduced explosion pressure against
AJVZR 4wP

from 0.5 m diameter to 0.15 m diameter. This effect is explosion pressures generally in accordance with Pred
generally mirrored in other sets of data regardless of the predictions from the K,, nomographs.
vessel in which ignition took place.
Only a smal1 proportion of the tests carried out with Ejfct of vessel conjguration. Figures 8-12 and 15-19
the 0.15 m pipe resulted in transmission of an explosion. demonstrate the variation in pressure enhancements that
When successful transmissions did occur they produced occur as the vessel volume changes, and more especially
 Dust explosions in interconnected vented vessels: P. Holbrow et al. 99

7 I I I 1 I I I I I 7 I l 1 I I I I I I

6 6

2mJ 2m VESSELS 20m3 - 20m3 VESSELS

TONER DUST COAL DUST 1

0 0.15m dia x 5m PIPE 0 0.5m dia x 5m PIPE

A 0.15m dia x lom PIPE A 0.5m dia x 10m PIPE

?? 0 5m dia x 15m PIPE

1

-,
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.00
I

0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

AvN~ AvN~

Figure 14 Maximum reduced explosion pressure against Figure 16 Maximum reduced explosion pressure against
AJVZ

7 I I I I I I I , I

2m3 2m3 VESSELS 20mJ - 6.3m3 VESSELS

; COAL DUST GOAL DUST

0m 5 5
0 0.5m dia x 5m PIPE o 0.5m dia x 5m PIPE
e
a A 0.5m dm x lom PIPE A 0.5m dia x 10m PIPE
B
0 0.5m dia x l5m PIPE m 0.5m dia Y i5m PIPE
a
c
.Q, 4 4 - d
s?

3
a
1
$ 3 3 -
.

0
0.00 0.05 0.10 0 15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

AvN AVlv2R

Figure 15 Maximum reduced explosion pressure against Figure 17 Maximum reduced explosion pressure against
AVIV A,lP

as the ratio of vessel volume changes. The data in Tuble
4 show that the A,/AvKst ratio is generally greater when
the explosion is ignited in the larger vessel of a particular
configuration. As an example, the 20-2 m3 configuration
with a 0.5 m diameter connecting pipe generated the
highest absolute pressures and pressure enhancement of
virtually al1 the configurations for the coal dust
explosion tests.
The scale effect was tested at a volume ratio of 1: 1
with 2 m3 vessels and 20 m3 vessels using coal dust. The
results are shown in Figures 15 and 16. The pressures
in the larger-scale system are higher than in the smaller
 Dust explosions in interconnected vented vessels: P. Holbrow et al.

20m -
2m3 VESSELS 2m3 . 20m3 VESSELS
COAL DUST COAL DUST

0 0.5m dia x 5m PIPE o 0.5m dia x 5m PIPE

A 0.5m dia x lom PIPE A 0.5m dia x lom PIPE

? ? 0.5m dia x 15m PIPE D 0.5m dia x 15m PIPE

o-
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

AvN AvN*

Figure 18 Maximum reduced explosion pressure against Figure 19 Maximum reduced explosion pressure against
A,lL@ AvW3

Table 3 Comparison of experimental fred data envelope with predicted Pred

System configuration Radant curve K,, curve

Coal dust
2-2 vessels, 0.5 m diameter x 5,10,15 m within on upper boundary
2-6.3 vessels. 0.5 m diameter X 5,10,15 m within within
2-6.3 vessels, 0.25 m diameter X 5,10,15 m within within
6.3-2 vessels, 0.5 m diameter X 5,10,15 m below within
6.3-2 vessels, 0.25 m diameter X 5,10,15 m on lower boundary on upper boundary
2-20 vessels, 0.5 m diameter X 5,10,15 m on lower boundary within
20-2 vessels, 0.5 m diameter x 5,10,15 below below
2-20 vessels, 0.25 m diameter X 5,10,15 within within
20-2 vessels, 0.25 m diameter X 5,10,15 on lower boundary within
2-20 vessels, 0.15 m diameter x 5 m on lower boundary on upper boundary
20-2 vessels, 0.15 m diameter x 5 m below within
6.3-20 vessels, 0.5 m diameter X 5,10,15 m below on lower boundary
20-6.3 vessels, 0.5 m diameter X 5,10,15 m below within
20-6.3 vessels, 0.25 m diameter X 5,10,15 m below within
6.3-20 vessels, 0.25 m diameter X 5.10 m on lower boundary within
20-20 vessels, 0.5 m diameter X 5,10,15 m below within
20-20 vessels, 0.25 m diameter X 5,10,15 m below within

Toner dust
2-2 vessels, 0.5 m diameter x 5,10,15 m below within
2-2 vessels, 0.25 m diameter X 5,lO m on lower boundary within
2-2 vessels, 0.15 m diameter X 5.10 m on lower boundary within
2-6.3 vessels, 0.5 m diameter X 5.10.15 m below within
2-6.3 vessels, 0.25 m diameter X 5,10,15 m below within
6.3-2 vessels, 0.5 m diameter X 5,10,15 m below below
6.3-2 vessels, 0.25 m diameter X 5,10,15 m on lower boundary within
2-20 vessels, 0.5 m diameter X 10 m below below
20-2 vessels, 0.5 m diameter X 5.10 m below below
6.3-20 vessels, 0.5 m diameter X 5,10,15 m below within
20-6.3 vessels, 0.5 m diameter X 5.10.15 m below below

Anthraquinone
2-2 vessels, 0.25 m diameter X 5,lO m n.a.

On envelope boundary = predicted Predequal to the experimental fred along the boundary. Within envelope = some of the experimen-
tal Prsd exceed the predicted Pred. Below envelope = all experimental Pred exceed predicted Pred.
n.a. = not available.
 Dust explosions in interconnected vented vessels: P. Holbrow et al. 101

Table4 Ratio of experimental vent area to K.,-predicted vent area (single vessel) for Pred of 0.2, 0.5, 1 and 1.5 bar

Pred = 0.2 bar Pred = 0.5 bar Pred = 1.0 bar Pred = 1.5 bar

System configuration

Coal dust
2-2 vessels, 0.5 m diameter x 5 m 0.33 1.1 0.15 0.08 0.8 n.m. -
X IOm 0.18 0.64 0.13 A.87 n.m. - n.m. -
X 15m 0.16 0.57 0.13 0.87 n.m. n.m.
2-6.3 vessels, 0.5 m diameter x 5 m 0.17 0.61 0.15 1 0.07 1
-
X lom 0.17 0.61 0.12 :8 :::9 0.9 n.m.
X 15m 0.17 0.61 0.11 0173 n.m. - n.m. -
2-6.3 vessels, 0.25 m diameter X 5 m 0.18 0.64 0.15 n.m. n.m. -
X IOm 0.2 0.71 0.15 : 0.12 .2 n.m. -
X 15m n.m. 0.09 0.6 n.m. - n.m.
6.3-2 vessels. 0.5 m diameter X 5 m 0.17 ;61 0.15 1 0.11 1.1 0.1 ;42
X 10m 0.21 0.75 0.17 1.13 0.13 1.3 0.11 1.57
X 15m 0.16 0.57 0.12 0.8 0.1 1 0.09 1.28
6.3-2 vessels, 0.25 m diameter x 5 m 0.17 0.61 0.12 n.m. n.m. -
X IOm 0.2 0.71 0.12 0:: 0.09 ;9 n.m. -
x 15m 0.16 0.57 n.m. n.m. -
2-20 vessels, 0.5 m diameter x 5 m 0.25 0.89 0.19 1.27 En 1 n.m.
X IOm 0.25 0.89 0.14 0.93 0.1 1 0.08 1.14
X 15m 0.25 0.89 0.16 1.07 0.11 1.1 0.08 1.14
20-2 vessels, 0.5 m diameter x 5 m 0.21 0.75 1.33 0.12 1.2 0.1 1.43
x IOm 0.2 0.71 o.; 1.33 0.13 1.3 0.11 1.57
X 15m 0.2 0.71 0:2 1.33 0.14 1.4 0.11 1.57
2-20 vessels, 0.25 m diameter x 5 m 0.21 0.75 0.17 1.13 0.12 1.2 n.m. -
x IOm 0.29 0.21 1.4 n.m. n.m. -
X 15m 0.24 i86 0.14 0.93 1 0.09 1.29
20-2 vessels, 0.25 m diameter X 5 m 0.23 0:82 0.19 1.26 0:: 1 n.m.
X 10m 0.24 0.86 0.21 1.4 0.1 0.09 1.29
X 15m 0.21 0.75 0.2 1.33 0.08 0.8 n.m. -
2-20 vessels. 0.15 m diameter x 5 m n.m. - 0.15 1 n.m. n.m. -
20-2 vessels; 0.15 m diameter X 5 m n.m. 0.18 1.2 0.08 ;8 n.m.
6.3-20 vessels, 0.5 m diameter x 5 m 0.24 .86 0.2 1.33 0.14 1.4 0.1 1.43
X lom 0.21 0.75 0.15 n.m. n.m. -
X 15m 0.21 0.75 0.13 :87 0.08 ;8 n.m. -
20-6.3 vessels, 0.5 m diameter X 5 m 0.26 0.93 0.2 1:33 0.13 1.3 0.11 1.57
X lflm
._ 0.24 0.86 0.16 1.07 0.11 0.09 1.29
X 15m 0.2 0.71 0.12 0.8 0.09 Er 0.09 1.29
20-6.3 vessels, 0.25 m diameter X 5 m 0.25 0.89 0.18 1.2 0.12 1.2 0.1 1.43
x IOm 0.21 0.75 0.15 1 n.m. - n.m. -
X 15m n.m. n.m. n.m. n.m. -
6.3-20 vessels, 0.25 m diameter X 5 m 0.24 F86 0.18 1.2 0.1 7 n.m. -
X IOm 0.19 0.67 n.m. n.m. n.m.
20-20 vessels, 0.5 m diameter x 5 m 0.26 0.93 0.16 1.07 0.11 Tl 0.1 1.43
X IOm 0.71 0.12 0.8 0.09 0.9 0.08 1.14
x 15m 0:26 0.93 0.15 0.11 1.1 0.1 1.43
20-20 vessels, 0.25 m diameter X 5 m 0.2 0.71 0.14 0.93 0.09 0.9 0.08 1.14
X IOm 0.2 0.71 0.16 1.07 n.m. n.m.
x 15m 0.3 1.1 0.14 0.93 0.1 t 0.08 1.14
Toner dust
2-2 vessels, 0.5 m diameter x 5 m 0.4 0.37 1.85 0.31 2.3 0.22 2.44
X lom 0.25 :62 0.18 0.9 0.12 0.92 0.1 1.11
X 15m 0.3 0:75 0.16 0.8 0.12 0.92 0.12 1.33
2-2 vessels, 0.25 m diameter X 5 m 0.41 0.26 1.3 0.2 1.53 0.17 1.89
X IOm 0.3 i.75 0.26 0.12 0.92 n.m.
2-2 vessels, 0.15 m diameter x 5 m 0.29 0.72 0.17 0::5 0.12 0.92 0.08 ;89
X IOm 0.4 1 0.19 0.95 0.13 n.m. -
2-6.3 vessels, 0.5 m diameter x 5 m n.m. - 0.28 0.22 : 69 0.17 1.89
X IOm n.m. 0.25 ::;5 0.17 1:31 0.15 1.67
X 15m 0.2 .5 0.21 1.05 0.15 1.15 0.11 1.22
2-6.3 vessels, 0.25 m diameter X 5 m n.m. 0.21 1.05 0.18 1.38 0.16 1.78
X IOm 0.3 o;r5 0.21 1.05 0.15 1.15 0.14 1.55
X 15m n.m. - 0.21 1.05 0.11 0.85 n.m.
6.3-2 vessels. 0.5 m diameter x 5 m n.m. - 0.29 1.45 0.26 2 0.24 ;67
X lom 0.28 0.25 1.25 0.22 1.69 2.22
X 15m 0.25 0:;2 0.21 1.05 0.15 1.15 0.:4 1.55
6.3-2 vessels, 0.25 m diameter x 5 m n.m. 0.25 1.25 0.16 1.23 0:13 1.44
X 10m 0.29 ?72 0.26 1.3 0.22 1.69 0.16 1.78
X 15m n.m. - 0.21 1.05 n.m. n.m. -
2-20 vessels, 0.5 m diameter x 10 m n.m. - 0.25 1.25 0.15 715 n.m.
20-2 vessels, 0.5 m diameter x 5 m n.m. - 0.25 1.25 0.21 1.61 0.19 Gl
2-20 vessels, 0.15 m diameter x 5 m n.m. - n.m. n.m. - 0.15 1.67
20-2 vessels, 0.15 m diameter x 5 m n.m. - 0.13 K65 n.m. n.m.
6.3-20 vessels, 0.5 m diameter x 5 m n.m. - 0.3 0.24 ;85 0.17 ;89
X IOm n.m. - 0.25 E5 0.22 1.69 0.19 2.11
X 15m n.m. - 0.2 1 0.12 0.92 n.m.
20-6.3 vessels, 0.5 m diameter x 5 m n.m. - n.m. 0.26 2 0.22 F44
x IOm n.m. - 0.24 1.2 0.2 1.54 0.17 1.89
X 15m n.m. - 0.24 1.2 0.21 1.61 0.17 1.89
Anthraquinone
2-2 vessels, 0.25 m diameter x 5 m n.m. - 0.43 1.53 0.24 1.26 0.23 1.76
X lom n.m. - 0.45 1.61 0.25 1.32 n.m. -

n.m. = not measured.

102 Dust explosions in interconnected vented vessels: P. Holbrow et al.

Table 5 Table 5 Predicted values of AJl@ from single-vessel K., nomographs

Dust Pred = 0.2 bar P re3 = 0.5 bar Pred = 1 bar Pred = 1.5 bar

Coal 0.28 0.15 0.1 0.07

Toner 0.4 0.2 0.13 0.09
Anthraquinone 0.54 0.28 0.19 0.13

Table 6 Vent areas required for interconnected systems (constant X K., value)

x Kat vent area

Vessels Volume ratio Pred 0.2 bar Pred 0.5 bar Pred 1.0 bar Pred 1.5 bar

Coal 0.5 m diameter pipe

2-2 1:l 1.1 1 1 -
20-20 1:l 1 1.07 1.1 1.43
6.3-2 3.15:1 1 1.13 1.3 1.57
20-6.3 3.17:1 1 1.33 1.4 1.57
20-2 1O:l 1 1.33 1.4 1.57
Coal 0.25 m diameter pipe
20-20 1:l 1.1 1.07 1 1.14
6.3-2 3.15:1 1 1 1.2 -
20-6.3 3.17:1 1 1.2 1.2 1.43
20-2 1O:l 1 1.33 1.2 1.29
Coal 0.15 m diameter pipe
20-2 1O:l - 1.2 1 -
Toner 0.5 m diameter pipe
2-2 1:l 1 1.85 2.3 2.44
6.3-2 3.15:1 1 1.45 1.89 2.67
20-6.3 3.17:l - 1.5 1.85 2.44
20-2 1O:l 1 1.25 1.61 2.11
Toner 0.25 m diameter pipe
2-2 1:l 1 1.3 1.53 1.89
6.3-2 3.15:1 1 1.3 1.69 1.78
Toner 0.15 m diameter pipe
2-2 1:l 1 1 1 1
20-2 1O:l - 1 - -
Anthraquinone
2-2 1:l - 1.61 1.32 1.76

at low values of vent area. However, the maximum

pressure results for coal dust shown in Figures 15-19
do not show overall a great deal of variation as the ves-
sel-to-vessel configuration changes. This is distinct from
the toner dust results, where the vessel-to-vessel con-
2m3. 2m3 VESSELS
figuration has a significant effect on the maximum ANTHRAQUINONE DUST
explosion pressure. 0 0.25m dia x 5m PIPE

h 0.25m dia x lom PIPE

Eflect of dust type. The more explosible a dust tends to
be, the greater is the ease of transmission of the
explosion to the second vessel and the higher the levels
of pressure enhancement. A comparison of the results in
Figures 8, 15 and 20 demonstrates this effect.

Conclusions
The following conclusions are relevant to two linked
vessels connected by straight lengths of unobstructed
pipe.
1. The results have demonstrated that enhanced
explosion pressures can occur in interconnected
vented vessels, and that the important variables
affecting the venting requirements of both vessels
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 050
were:
(i) the diameter and to a lesser extent the length of AvNzR

the interconnecting pipe, Figure 20 Maximum reduced explosion pressure against

(ii) the volume ratio of the two vessels, A,W
 Dust explosions in interconnected vented
(iii) the dust explosibility, 9 Singh, J. Gas explosions in interconnected vessels: pressure piling.
(iv) the vent areas themselves.
10
Under some circumstances, when the connecting pipe
is relatively long and has a smal1 diameter, and the 11
degree of interconnectedness is low, the K,, nomo-
12
graph approach for vent sizing of single vessels can
be applied to the vessel in which the initial ignition
occurs. When the degree of interconnectedness is
13
high, both vessels wil1 require the vent areas to be
increased compared to the nomograph approach. 14
Pressure enhancements in linked vented vessel sys-
tem are generally low when the reduced explosion
,l5
pressure does not exceed 0.2 bar g.
16
References
1 Gibson, N. Dust explosion hazards: the problem and the options
for safety, European Summer School on Dust Explosion Hazards:
Their Assessment and Control, 21-25 September 1992, Cam- 17
bridge, UK
2 Schoeff, R. W. News release and summary for 1993. Dust
exulosion statistics - USA. Eurooex Newslett. March 1994. 24 18
3 Proust, Ch. and Pineau, J. P. Dusiexplosions: risk assessment. Dust
Explosion Protection, Europex Int. Symp, Antwerp, 1989 19
4 VDI Guideline 3673, Pressure lease of dust explosions. VDI-
Commission Reinhaltung der Luft. Beuth-Verlag, W. Germany, 20
1984
5 National Fire Protection Association Guide for venting of defla-
grations. NFPA 68-1988. Quincy, MA, USA, 1988 21
6 Lunn, G. Dust Explosion Prevention and Protection. Part 1 - Vent-
ing. 2nd edn. IChemE, Rugby, UK, 1992
7 Abdullin, R. K., Babkin, V. S. and Senachin, P. K. Combustion of 22
gas in connected vessels. Comb. Expl Shock Waves 1988, 24, 123
8 Phylaktou, H. and Andrews, G. E. Gas explosions in linked vessels. 23
Presented at 13th ICDERS Meeting, Japan, 1991
vessels: P. Holbrow et al. 103
Presented at.iChem E Hazards X11 Svmoosium.
, I 19-21 A~til
. 1994.
UMIST, M&chester, UK
Lunn, G. A. The explosion venting of interconnected vessels. First
World Seminar on the Explosion Phenomenon and on the Appli-
cation of Explosion Protection Techniques in Practice, EUROPEX,
Brussels, 1992
Van Wingerden, K. and Alfert, F. Dust Explosion Propagation in
Cannected Vessels. VDI Betjchte, Nr. 975, -5@,1992
Gardner, B. R., WintexR. J. and Moore, M. J. Explosion develop-
ment and deflagration-to-detonation transition in coal dust/air sus-
pensions. 2lst Symp. Int. on Combustion, The Combustion Insti-
tute, 1986, pp 355-343
Pineau, J. P. Dust Explnsions in Vessels Connected to Ducts. CpI-
Berichte Nr 494; 1984
Bond, F., Fresko, M., Knystautus, R. and Lee, J. H. S. Influence
of turbulente on dust explosions. 1st Int. Colloq. on Explosibility
of Industrial Dusts, 1984, pp 37-50
Pineau, J. P. and Ronchail, G. Propagation of Coal Dust
Explosions in Pipes. Industrial Dust Explosions, @TM STP 958d
eds K. L. Cashdollar and M. Hertzberg. American Society for Test-
ing and Materials, Philadelphia, 1987, pp 74-89
Bouhard, F., Veyssiere, B., Leyer, J.-C. and Chaineaux, J,
Explosion in a vented vessel connected to a duct.Prog. Astronaut
Aekonaut 134. AIAA A. L. Kuhl et al. Eds Dynamics of Deton-
ations and Exolosions: Exolosion Phenomena. 12th ICDERS. Ann
Arbor. MI, U:S.A, July 1989
Proust, C. Dubt explosions in pipelines. EUROPEX European Sum-
mer School on Dust Explosions Hazards - Their Prevention and
Control, Cambridge, UK, September 1992
Zhang, F. and Gronig, H. Transition to detonation in torn starch
dust oxygen and air mixtures. Combust. Flame 1991, 86, 21-32
Eckhoff, R. K. Dust Exptosions in the Process Industries. But-
terworth Heinemann, Oxford, 199 1
Phylaktou, H. N., Andrews, G. E. and Herath, P. Fast flame speeds
and rates of pressure rise in the initial period of gas explosions in
large L/D cylindrical vessels. J. Loss Prev. Process Ind. 1990, 3
McCann, D. P. J., Thomas, G. 0. and Edwards, D. H. Gas dynam-
ics of vented explosions. Part 1: experimental studies. Combusf.
Flame 1985. 59, 233-250
Harrison, A. J. and Eyre, J. A. External explosions as a result of
explosion venting. Combust. Sci. Technol. 1987. 52, 91-106
Bartknecht, W. Explosions. Course, Prevention, Protection.
Springer. Berlin, 198 1
Problems in experimental measurements of dust explosions
K.J. Mintz*
Canadian Explosives Research Laboratory, CANMET, 5.55 Booth St., Ottawa, ON KIA OGI, Canada

Received 1 October 1994: accepted 15 January 1995

Abstract

The experimental method employed for measuring the parameters of dust explosions was
critically examined using methane/air as a model system. Because of the rapid rate that air
enters the test vessel (in order to disperse the dust), the system is not at thermodynamic
equilibrium which causes the actual concentration of air to be about 15% less than indicated by
the measured pressure at the time of ignition, thus creating a significant error in the determina-
tion of explosion pressures. For measuring the lower explosible limit, the ignitor must be of
optimum strength, otherwise too high or low limit values are obtained. Pyrotechnic ignitors,
which are usually used for igniting dusts, produce a significant pressure by themselves;
a method for accurately taking this effect into account is given. Some unusual problems
encountered with particular dusts are discussed.

1. Introduction

The parameters of dust explosions are commonly carried out in vessels of the order
of 0.020 m3. Although the exact method of creating the dust cloud varies from vessel
to vessel, the fundamental principle is that a strong pulse of air disperses the dust
(which is usually placed in another smaller vessel outside the test vessel). Thus, the
dust explosion testing is carried out under turbulent, dynamic conditions. In order to
ensure the meaningfulness of the results, a detailed study was carried out, using
methane-air explosions to compare dynamic with static conditions.

2. Experimental

The 0.020m3 vessel was essentially the same type as used by Cashdollar and
Hertzberg [ 11. The dust was placed in a chamber underneath the vessel and the vessel

* Fax: ( + 1-613) 995-1230. Tel.: ( + 1-613) 995-1413.

1995 Elsevier Science B.V.

SSDI 0304-3894(95)00011-9
178 K.J. MintzJJournal of Hazardous Materials 42 (1995) 177-186

1.20

1.15
2
..
.j
pI 1.10
.s
2
1.os

40 60 80 100 120 140 160 180 200

Fig. 1. The effect of different tank pressures (I,& on the Psir/Pri.n, ratio, where Pai, is the peak pressure
and Pfinal is the pressure at equilibrium: (0) P,,,, = 1.1 MPa; (0) Ptank = 1.3 MPa.

evacuated. A 0.016 m3 pressure vessel containing air, usually at 1.1-1.3 MPa, was
connected to the dust chamber through a solenoid valve. A control panel allowed the
selection of time that the solenoid valve remained open as well as the selection of delay
time before firing (and, for ignition by an electric arc, the length of time that the arc
discharges). Generally, the aim was to have the explosion test carried out at as close to
normal atmospheric pressure as possible, so as to be applicable to most industrial
applications. A CEClOOO pressure transducer, connected to a Nicolet 4094 digital
oscilloscope, was used to measure the pressure throughout the experimental sequence.
Methane-air mixtures were prepared in a binary mixer from technical grade methane
and dry air, and an infrared analyzer was used to verify the methane concentration.
Also, a paramagnetic oxygen analyzer was used to measure the concentration of
oxygen remaining in the vessel after the trial. Two types of ignition were used: an
electric arc and Sobbe pyrotechnic ignitors of nominal energy from 250 to 5000 J.

3. Results and discussion

3.1. Calibration of air concentration

These tests were carried out without dust and an ignition source, i.e. only air was
involved. After the solenoid valve was opened, the pressure inside the vessel increased
rapidly (from 0 kPa absolute) until the valve was closed, then slowly decreased to
reach a constant value approximately 15% smaller after about 15-70 s. The initial
pressure rise was close to linear because only a small fraction of the air in the pressure
tank was required to pressurize the test vessel to 1 bar. Fig. 1 shows that the ratio of
the peak pressure, Pair to the final pressure Prinar is nearly independent of normal
 K.J. MintzjJournal of Hazardous Materials 42 (1995) 177-186 119

variations of the pressure in the pressure vessel (which provide different flow rates)
and solenoid valve open times (to provide different final pressures). In addition,
different initial vessel pressures (which are used by other workers) do not substantially
affect this ratio.
Because of the rapid rate at which the air enters the vessel, it is not at thermody-
namic equilibrium (i.e. does not follow the Boltzmann distribution law) and therefore
does not have a definable temperature. Although the pressure in the explosion vessel
decreases after the solenoid valve closes, the concentration, which is a more funda-
mental quantity, must remain constant. The final pressure, therefore, must be used as
the actual pressure at which the explosion test is carried out, even though ignition of
the dust cloud is usually done about 0.1-l s after the solenoid valve is closed, i.e. when
the measured pressure is still near Par,. In an explosion test, it is impossible, of course,
to measure a Prinalfree from the effects of the explosion; however, the fact that the ratio
P,i,/Pri,,i is quite constant means that it is valid to divide the measured value of Pair
by 1.17 f 0.01 (95% confidence limit) to yield an equivalent pressure, the pressure
that would be measured for that concentration of air if the air were at equilibrium.
A large body of evidence exists that for gases and dusts the explosion pressure is
directly proportional to the initial pressure (at constant initial temperature). In
addition, the explosion pressure of dusts has been found to be directly proportional to
the reciprocal absolute temperature at constant initial pressure [2]. These two
relationships can be combined to state that the explosion pressure is directly propor-
tional to the initial concentration of gas. This conclusion is quite reasonable because it
implies a reaction which is first order in oxygen concentration.
The implication of the above is that explosion tests carried out (in 0.02 m3 vessels)
at an apparent (measured) pressure of 1 bar are actually carried out at the equivalent
of 0.85 bar, and therefore a correction factor of 1.17 must be applied to the measured
explosion pressures to yield the correct values for 1 bar.

3.2. CH,-air explosion tests: electric arc ignition

First, tests were carried out under quiescent conditions by preparing a mixture in
the 0.020 m3 vessel by partial pressures and measuring the methane concentration by
the infrared analyzer, to an accuracy of f 0.1% absolute. (The total pressure in all
tests was 1 bar.) Fig. 2 shows that the explosion pressure of this system reaches
a broad maximum of 700 + 10 kPa when an electric arc is used, close to the value of
720 kPa obtained by Hertzberg et al. [3]. Fig. 3 shows that the peak rate of pressure
rise also rises smoothly to a maximum of 24.8 f 0.6 MPas-, close to the value of
25 MPa SC of Hertzberg et al. [3].
Second, tests were carried out under turbulent conditions similar to those when
dusts are tested. A quantity of methane to give an approximately stoichiometric
mixture was metered into the vessel; the solenoid was opened so that air could flow
quickly from the pressure tank into the vessel to raise the total equivalent pressure to
1 bar, and the arc fired at various delay times. Fig. 4 shows that the explosion pressure
decreases slightly with delay time; the extrapolated value to infinite delay time, which
should be a quiescent state, is 710 f. 20 kPa, close to the value obtained under true
180 K.J. MintzJJournal of Hazardous Materials 42 (1995) 177-186

I/,,,,,,,,,/,,,,,/,,,,,,,

2 3 4 5 6 7 8 9 10 11 12
% CH,
Fig. 2. Explosion pressure as a function of methane concentration under quiescent conditions, for different
ignition sources: (0) arc ignitor; (0) 250-J Sobbe ignitor; (A) 5000-J Sobbe ignitor.

2 3 4 5 6 7 8 9 10 11 12
% CH,

Fig. 3. Rate of pressure rise as a function of methane concentration under quiescent conditions, for different
ignition sources: (0) arc ignitor; (0) 250-J Sobbe ignitor; (A) 5000-J Sobbe ignitor.

quiescent state. If the apparent (measured) pressure were used instead of the equiva-
lent pressure, then the explosion pressure would be only about 600 kPa, which would
clearly be in error. With the rate of pressure rise, the scatter is much greater, as can be
seen in Fig. 5. More work would have to be carried out to establish the functional
dependence of rate on delay time. Since the physical processes governing the effects of
turbulence on dust-air flames are mostly the same as those involved in gas flames, an
estimate of the zero-turbulence parameters of dust-air explosions can be made by
extrapolation from the values obtained at the usual delay time of 100 ms. For the
 K.J. MintzJJournal of Hazardous Materials 42 (1995) 177-186 181

60060 5 10 15 20

Reciprocal of ignition delay time Ii

Fig. 4. Peak explosion pressure for stoichiometric methane-air mixtures as a function of turbulence, with
the 95% confidence limits shown as dotted lines.

0 5 10 15 20
Reciprocal of ignition delay /s-l

Fig. 5. Peak rate of pressure rise for stoichiometric methane-air mixtures as a function of turbulence, with
the 95% confidence limits shown as dotted lines.

explosion pressure, that value would be about 2-4% lower than that measured; for
the rate of pressure rise, the value would be of the order of half.

3.3. CH,-air explosion tests: Sobbe pyrotechnic ignitors

The data in Figs. 2 and 3 show that the electric arc tests yield an apparent lower
flammability limit (LFL) of about 5.9%, well above the established value of 5.0%.
182 K.J. Mintz/Joumal of Hazardous Materials 42 (1995) I77-186

05
0 50 100 150 200
Initial Pressure /kPa

Fig 6. Peak explosion pressure from the 5000-J Sobbe pyrotechnic ignitor in 20 1vessel at different initial
pressures: (0) quiescent mixtures; (0) turbulent mixtures.

Therefore, a stronger ignitor is required to measure the LFL. Tests were carried out
with two strengths of pyrotechnic ignitors: 250-J and 5000-J, as shown in Figs. 2 and
3. The problem in using these types of ignitors is allowing for their effects on the
system. When ignited in air at 1 bar, the 250-J ignitor produces 9 kPa maximum
overpressure. Since this value is of the same order as the precision of the results, it does
not affect significantly the accuracy of the explosion pressure measurements. (Never-
theless, 9 kPa was subtracted from the apparent explosion pressures to yield the best
estimate of the true explosion pressure.) On the other hand, as shown in Fig. 6, the
5000-J ignitor produces a significant pressure, which is a function of total pressure,
particularly at low pressures. Turbulence increases the peak pressure, but also in-
creases the rate at which the pressure decreases after the maximum, probably because
of faster heat transfer to the walls of the vessel. The usual way of accounting for the
5000-J ignitor is simply to subtract its maximum pressure from the total overpressure.
There is a logical problem with that method: the maximum pressure of the ignitor
occurs very soon after ignition, well before the maximum pressure due to the dust or
gas, thus, that method overcorrects for the ignitor. Furthermore, the time to max-
imum explosion pressure varies with the concentration of ignitor and also varies from
dust to dust, so that even relative values of explosion pressure will be in error due to
this effect. In order to resolve that problem, the pressure trace for the ignitor by itself
was subtracted from the experimental pressure trace to yield the net pressure trace in
each trial. The latter was used to obtain the maximum overpressure and the maximum
rate of pressure rise. However, it should be realized that the ignitor also causes an
increase in temperature of the gas, but likely not uniformly. Therefore, use of the
Sobbe ignitors increases the uncertainty of explosion tests; to reduce this uncertainty,
it is best to use as small an ignitor as possible.
 K.J. Mintz/Joumal of Hazardous Materials 42 (1995) 177-186 183

0
2 3 4 5 6 7 8
% CH,

Fig. 7. Percentage oxygen consumed as a function of percent methane in methane-air mixtures: (0) tests
with a 250-J Sobbe ignitor; (0) tests with a 5000-J Sobbe ignitor.

The results from tests under quiescent conditions with the 250-J and 5000-J ignitors
(Fig. 2) yield completely different results for the LFL, with the former yielding
a definitive value of 5.1 f O.l%, close to the accepted value of 5.0%, whereas the latter
does not yield a sharp cut-off, but appears to be at less than 3%. The shape of the
curve for the 5000-J ignitor tests is indicative of overdriving, i.e. the system does not
reach a steady state in the limited vessel size, but still seesthe effect of the ignitor.
Hertzberg et al. [33 have attributed the cause of the lower peak explosion pressure
obtained with a pyrotechnic ignitor (compared to an electric discharge) to it being
a nonpoint source and thus causing a less adiabatic reaction. Alternatively, it can be
considered that a fraction of the methane reacts during the time that the ignitor is
active (due to the overdriving) and thus does not contribute to the normal explosion
pressure. From the difference in peak explosion pressures, this hypothesis leads to an
estimate of the percentage thus reacted: 3% for the 250-J ignitor, 10% for the 1000-J
ignitor and 16% for the 5000-J ignitor. (These values, of course, are specific to
a particular set of ingredients.) The peak maximum rate of pressure rise using 250-J
ignitors is distinctly less than that using either electric arc ignition or 5000-J ignitors.
This parameter is measured earlier in the process than the maximum pressure, thus is
more susceptible to effects from the ignitor. Furthermore, the time to ignition is much
shorter for tests with the 5000-J ignitor; therefore, the pressure peaks from the ignitor
and the dust explosion will overlap more causing a greater uncertainty in the results.
Further work would have to be carried out to determine if the apparent decrease in
rate of pressure rise is in fact real.
Fig. 7 shows the percentage of oxygen consumed for different methane-air mixtures
using 250-J and 5000-J ignitors. In both cases, the result is a straight line which
extrapolates to between 0% and 1% CH4 at one end and close to the stoichiometric
184 K.J. Mintz/Journal of Hazardous Materials 42 (1995) 177-186

mixture at the other end. The break-point, i.e. the point at which the points deviate
from the straight line, may be indicative of a nonself-sustaining explosion. Finding the
break-point on the oxygen consumption curve may be a more decisive method than
using the peak pressures (cf. Fig. 2). The methane concentration in the vessel measured
after the test was close to zero for all tests that fall on the line. The data in Fig. 2 show
that the quantity of oxygen consumed was twice that of methane (within 5%), which is
consistent with the simple reaction: CH4 + 202 + COZ + 2Hz0.
The LFL was also measured under low-turbulent and high-turbulent conditions
(570 and 20 ms delay times, respectively), using a 250-J ignitor. The results were
4.8 f 0.1% and 4.9 + O.l%, respectively, which are slightly lower than those obtained
in the quiescent mixture (5.1%) and close to the accepted value (5.0%).

3.4. Dust tests: blank runs

In order to separate the physical effects of dust from the chemical, tests were carried
out using finely powdered CaC03 and a 5000-J ignitor in exactly the same way as
trials with explosible dusts were carried out. The peak pressure decreased significantly
with increasing dust concentration (49 kPa with no dust to 41 kPa with 100 g/m and
35 kPa with 500 g/m3). The decrease is probably caused by the dust absorbing energy
that would otherwise go to the air; thus, the gas does not heat up as much and
produces less pressure. A second factor is the backpressure caused by the dust which
decreases the rate of air entering the vessel; the resultant lower level of turbulence
affects the shape of the pressure trace after the peak pressure, because the transfer of
heat to the walls decreases. (At higher loadings of dust, the solenoid valve must be left
open longer to achieve the same 1 bar pressure in the explosion vessel.) Therefore, for
accurate work, the pressure trace of a trial using the same concentration of an inert
dust as that used for the explosible dust and with the same ignitor must be used for
subtraction from the experimental pressure trace before further analysis is carried out.
In principle, the physical size of the inert dust should be matched to the combustible
dust; however, that is a second-order effect and is much smaller than normal trial-to-
trial variability.

3.5. Efsects of dust dispersion

Considerable work has shown that dust particles above 0.5 mm are nonexplosible
[4]. Tests were carried out in the standard way (using 5000-J Sobbe ignitors) on
a sample of petroleum coke that had particles of between 1.6 and 9.5 mm diameter.
The minimum explosible concentration was found to be 410 g/m3 and the maximum
explosion pressure to be 380 kPa, which would seem to contradict previous know-
ledge. However, dispersing the sample under the same conditions as the explosion
tests, but without any ignition source, produced a fine powder. Therefore, the necess-
ity of applying a high-pressure air blast to disperse a dust sample can significantly
affect the physical characteristics of the dust and thus its explosibility characteristics.
This material had actually been involved in an explosion in a factory. Later tests
proved that normal handling of the material (e.g. dropping a bag) could generate
 K.J. MintzjJoumal of Hazardous Materials 42 (1995) 177-186 185

ot,..,,~..,,,,,.,..,.,,.~.,.,,.;o
0 50 100 150 200 250 300
Particle Diameter /pm

Fig. 8. Minimum ignition energy of pyrite dust (0) and minimum explosible concentration using a 5000-J
Sobbe ignitor (0) as a function of particle diameter.

powder similar to that observed in the test without an ignition source. Therefore,
knowing the friability of a dust sample is important both for accurate measurement
and interpretation of its explosibility characteristics as well as for its hazard in use.

3.6. Sulphide dusts

These dusts are of interest because explosions of sulphide dust clouds occur in
mines only when initiated by a blast, i.e. an extremely strong ignition source. In
laboratory-scale apparatuses, electric arc ignition does not ignite an iron disulphide
(pyrite) dust cloud and does not even produce any reaction, unless artificially stim-
ulated by operating in pure oxygen. However, use of the Sobbe ignitors does produce
explosions when ordinary air is used. Sulphide dusts are much denser than other
explosible dusts, hence, a higher dispersion pressure (13 bar rather than 11 bar) and
a shorter ignition delay time (30 ms rather than 100 ms) must be used to produce
a reasonably homogeneous cloud. Use of different ignitor strengths allowed measure-
ment of the minimum ignition energy as a function of particle size for pyrite (Fig. 8),
showing the increase in energy required to initiate pyrite of larger sizes. Fig. 8 also
shows that the minimum explosible concentration increases with particle size. The
explosions are rather weak compared to other explosible dusts: the explosion pressure
varies from 200 to 300 kPa, as the particle size decreases from 280 to 60 urn.
The mechanism of the reaction was explored by examining the solid products
produced and by analysing the gas after the trials by infrared spectrophotometry.
When the explosion is weaker (lower concentration of dust or less oxygen in the
atmosphere), FezO, (haematite) is the predominant product. When the explosion is
stronger, then FeJ04 (magnetite) is produced. The reason is the thermodynamics of
186 K.J. Mintz/Journal of Hazardous Materials 42 (1995) I77-I86

the system: the net energy release is slightly greater for the haematite reaction, which
means that the equilibrium shifts towards the magnetite reaction at higher tempera-
tures, which are produced during stronger explosions. The quantity of SOz produced
(measured by using the IR band at 9 urn) was approximately twice the quantity of
oxygen consumed, which is consistent with simple oxidation. With lead sulphide, lead
sulphate was formed predominantly; therefore, the amount of SOz produced was
much less. When tested in pure oxygen, no SOz was produced, presumably because
the excess oxygen favoured the sulphate production.
A particular sulphide ore dust from a mine produced a small overpressure at
concentrations as low as 500 g/m3, but its lower explosible concentration was
3500 g/m3 (using 5000-J ignitors). Since this is much greater than for other dusts, it
can be easily missed using standard techniques, yet is important because such
a concentration and possible explosion can occur in the mine. An indication of
a samples possible explosibility is the observance of some overpressure at lower
concentrations.

4. Conclusions

This work has shown that the characteristics of the dust dispersal system and the
ignition system must be studied in detail for any particular apparatus before it is
possible to generate accurate data. The methane-air model system was effective for
this purpose. In particular, explosion pressures generated in 0.020 m3 vessels may be
underestimated by 17%. (For hazard quantification studies in deep mines, the in-
creased air pressure must also be taken into account.) The effects due to the pressure
generated by pyrotechnic ignitors and the heat absorbance of the dust must also be
properly taken into account.

References

[l] K.L. Cashdollar, and M. Hertzberg, Rev. Sci. Instrum., 56 (1985) 596-602.
[Z] W. Wiemann, in: K.L. Cashdollar and M. Hertzberg (Eds.), Industrial Dust Explosions, American
Society for Testing and Materials, Philadelphia, 1987, pp. 33-44.
[3] M. Hertzberg, K.L. Cashdollar and I.A. Zlochower, Proc. 21st Symp. (Int.) Comb., 1986, pp. 303-313.
[4] P. Field, Dust Explosions, Elsevier, Amsterdam, 1982, p. 15.
Journal of Hazardous Material, 35 (1993) 103-117 103
Elsevier Science Publishers B.V., Amsterdam

Dust explosion research. State-of-the-art and

outstanding problems

Rolf K. Eckhoff
University of Bergen and Christian Michelsen Research AS, Fantoftveien 38,
IV-5036 Fantoft-Bergen (Norway)

(Received October 22, 1992; accepted in revised form April 2, 1993)

Abstract

Increased knowledge creates a justified request for a steadily more differentiated approach
to design of preventive and mitigative measures against dust explosions in industry. In this
process, cross-fertilisation between fundamental research and applied research and develop-
ment is beneficial. Central research topics include formation of, ignition of and flame propaga-
tion in dust clouds, blast waves from exploding dust clouds, and test methods. Due to the
complexity of the various problem areas, comprehensive theories are scarce. However, com-
puter simulation models, carefully calibrated against experiments, may become useful tools in
the future. Expert systems should be welcomed, but adequate Quality Assurance is essential,

1. Introduction

The present paper relates to the dust explosion hazard in the process
industries. Dust explosions in mines is outside the scope of the paper.
The total amount of existing knowledge on industrial dust explosions, their
origin, propagation, prevention and mitigation, is vast. And yet, further in-
formation is continually being generated through on-going research in a large
number of countries. Not long ago the present author had the opportunity to
review some of the knowledge accumulated up to about 1990 [I]. About 900
references were covered, but they almost exclusively originate from
English/American and German literature. There is little doubt that a lot
more interesting and useful information can be retrieved by screening the
large amount of research reported in other languages than English and
German. Perhaps this could be accomplished by organizing a joint interna-
tional translation/edition effort.
Predicting the future has never been easy. This also applies to dust explosion
research. The present paper is an attempt at systematizing some recently
published work, as well as some information about on-going work that was
kindly made available to the author, and to indicate some possible future
trends. It seems useful to take a dual view. The f&t focuses on the fundamental

0304~3894/93/$06.00 @ 1993 Elsevier Science Publishers B.V. All rights reserved.

104 RX EckhoffjJ. Hazardous Mater. 35 (1993) 103-117

aspects, whereas the second considers knowledge related directly to practical

prevention and mitigation of explosions in the process industry.
Whenever reference to existing knowledge is made in the present paper
without mentioning any specific source, the appropriate references can in most
cases be found in the book mentioned above [XJ.

2. Fundamental and applied dust explosion research. A thematic

framework

2.1 Fundamental research

The fundamental aspects of dust explosion research may be grouped under
the four main headlines shown in Table 1.

2.2 Applied research and development

The practical aspects of dust explosion research and development may be
systematized as shown in Table 2. The fundamental and the applied aspects of
dust explosion research, i.e. Tables 1 and 2, are intimately related. This is
because fundamental knowledge is essential for proper understanding of the
practical aspects. Experience has shown that sometimes development of good
practica1 solutions can be hampered by not drawing upon available funda-
mental knowledge. In recent years the appreciation of the benefits that can be
harvested from cross-fertilization between fundamental and applied research
has been increasing.

3. Status and outstanding problems in fundamental research related

to dust explosions

3.1 Dust cloud formation processes

The status in this field up to 1989/90 is covered in Chapter 3 of ref. [l]. This is
an important and sometimes overlooked aspect of dust explosions It is well
established experimentally that the initial state of a cloud of a given dust in
a given gas (dust concentration, degree of dispersion into individual particles,
dynamic state) has a strong influence on both the ease with which the dust
cloud ignites and the rate at which it burns.
As soon as a significant blast wave has been generated by the primary dust
flame, this blast may generate secondary explosible clouds ahead of the flame
from dust deposits and layers there. Lebecki et al. 121investigated such pro-
cesses in a 100 m long gallery of cross-section 3m2. In order to establish an
improved understanding of these processes, further experimental and theoret-
ical studies of the influence of blast waves or dust clouds and dust
layers/deposits need to be conducted. Work on this problem has also been
performed by Ural [3], Gelfand and Tsyganov [4], and others. Gelfand and
TABLE 1
Fundamental aspects of dust explosion research
Dust cloud formation processes
l Inter-particle forces
in dust deposits
(cohesion).
l Entrainment of particles from
dust deposits by turbulent
gas flows.
l Transport of dust particles
turbulent gas flows.
l Measurement and
characterization of state of
turbulence of dust clouds.
Dust cloud ignition processes
l General ignition theory.
Ignition of single
particles and clouds.
l Ignition by:
- Smouldering combustion in
dust layers/deposits.
- Hot surfaces.
- Flying burning metal particles.
- Electric discharges.
- Hot gas jets.
- Shock waves.
- Focused light-beam
hot spots.
in l Influences of dust cloud
properties on ignition
sensitivity (composition,
size and shape of particles,
turbulence, dust concentration,
gas phase composition).
Flame propagation processes in Blast waves generated by
dust clouds burning dust clouds
0 Single-particle ignition l Blast wave properties as
and combustion in a a function of properties
flame front. of burning dust clouds.
0 Laminar and turbulent l Effects of blast waves
flames in dust clouds. on humans and
Mechanisms of heat transfer mechanical structures.
(conduction, convection,
radiation).
a Limit conditions for flame l Ability of blast waves
propagation through dust to transform dust layers
clouds (particle properties, dust into dust clouds.
concentration, oxygen
concentration, geometry).
0 Flame acceleration processes
in dust clouds. Influence of
concentration gradients, and
initial and explosion-
generated turbulence.
0 Detonation of dust clouds.
106 R.K. Eckhoff/J. Hazardous Mater. 35 (1993) 103-117

TABLE 2

Means for preventing and mitigating/controlling dust explosions in the process industries

Prevention Mitigation/Control

Preventing Preventing
ignition sources explosible dust cloud

l Self-heating in dust a Inerting by inert gas. 0 Explosion-pressure-

deposits.
l Open flames.
l Hot surfaces.
a Burning metal particles.
a Electrical discharges.
l Inerting by inert dust.
l Keeping dust concentration
outside explosible range.
resistant process
equipment.
l Isolation.
l Partial inerting by inert
gas.
l Explosion venting.
0 Explosion suppression.
l Preventing secondary
explosions (good house-
keeping for preventing
dust layer formation).

Tsyganov [4] also showed that the presence of dust layers on solid surfaces
exposed to blast waves, changed the blast wave characteristics as compared
with the characteristics in the case of dust-free surfaces.
Increased emphasis should be put on investigating the connection between
the parameters of dust cloud generation processes and the structures of the
resulting dust clouds. The structures of the clouds produced must be defined in
terms of distribution of dust concentration, quality of dust dispersion
(de-agglomeration), turbulence level and global velocities. In order to guide
fundamental research in this area in the direction of maximum practical
relevance, information about dust cloud structures that are typical in
industrial operation is required,

3.2 Dust cloud ignition processes

In the context of accidental dust cloud ignition, the ignition processes
indicated in the first column of Table 2 are central. Although much insight has
already been gained, many significant problems remain unsolved.
Ignition is a broad field of research. The concept of thermal run-away is
a common basis for understanding and describing ignition processes. However,
it does not seem realistic for the time being to foresee the development of one
single unified theory, usable in practice, which covers all types of ignition
sources. It is rather expected that separate theories, in terms of dynamic
computer models, will be developed for various categories of ignition sources,
such as hot surfaces and electric sparks.
RX. Eckhoff/J. Hazardous Mater. 35 (1993) 103-117 107

Klemens [5] reported ongoing research on initiation of combustion/smoul-

dering in dust layers by thermal radiation, whereas Hensel and John [6]
provided further insight into the important relationship between the condi-
tions required for initiating smouldering combustion of a dust layer on a hot
plate, and the layer thickness. An informative overview of the state-of-the-art
in this field was given by Crowhurst [7].
In the past, the minimum hot-surface temperature for ignition of a dust cloud
has often been regarded as if it were a universal constant for a given cloud.
However, it has been known for some time that minimum ignition temper-
atures of dust clouds vary significantly with scale, and this has recently been
confirmed by Wolanski 181. Further experimental and theoretical work is
needed in this area.
Ignition of dust clouds by small burning metal particles (impact sparks,
metal sparks) generated by mechanical impact, is a very complex problem.
A comprehensive, practically useful theory does not seem to be within
sight.
Ignition of dust clouds by electric/electrostatic discharges is another very
complex topic. Theories have been developed for ignition of dust clouds by
electric sparks between two metal electrodes, which is the simplest case, but
even such theories are only rough approximations.
With respect to the ever more complex one-electrode discharge types (co-
rona, brush, propagating brush), valuable experimental insight has been
gained during the last years, but so far no attempt at developing dust cloud
ignition theories seems to have been made. Glor [9] gave an informative
overview of the present status on theory and experimentation. Some of this
work, on possible incendiary discharges from powders poured into a heap, was
presented by Glor and Maurer [lo]. The question of whether incendiary light-
ning type discharges can occur in dust clouds is still to be answered. Glor [11] is
also continuing his work on whether occurrence of incendiary brush
discharges can occur in dust clouds.
Initiation of dust explosions by shock waves has been studied by several
workers, and valuable insight has been gained. An informative analysis of
shock wave ignition of dust clouds was given by Wolanski [12] and research at
University of Michigan, USA is reported by Kauffman et al. [13].

3.3 Flame propagation processes in dust clouds

Some central topics are:
l Ignition and combustion of single particles in a dust cloud

l Laminar flames

l Flame acceleration mechanisms

l Turbulent flames

a Detonation
Understanding the flame propagation processes in dust clouds is the key to
understanding how dust explosions develop in terms of pressure as a function
of time. A recent review was given by Lee et al. [14].
108 R.K. EckhofflJ. Hazardou Mater. 35 (1993)103-117

Both laminar and turbulent combustion in dust clouds have been studied
experimentally and theoretically by many workers. Examples of recent re-
search are given in the papers by Seshadri et al. [15] on the structure of laminar
dust flames, and by Rzal and Veyssiere [lS] on basic aspects of turbulent
dust flames. Proust 1171reports on interesting continued work on both laminar
and turbulent dust flames in France, and Van Wingerden [lSj reports on
forthcoming similar basic work in Norway.
Further work is needed on the relationship between the dynamic state of
a dust cloud and its combustion rate. The induction time for ignition may be
a useful parameter, describing the fundamental, global chemistry. However,
Ural [19] emphasized the fact that different induction times are observed with
incident and reflected shock waves, due to different ignition mechanisms. An
alternative approach is to consider the laminar burning velocity as the funda-
mental parameter, as suggested by Bradley et al. [ZO].Empirical relationships
between turbulent burning velocity and turbulence intensity are then estab-
lished, using the laminar burning velocity as a normalizing parameter. As
pointed out by Van Wingerden [18],numer-ical flame libraries can then be
formulated and used for closing the loop expansion-flow-turbulence-combus-
tion in numerical dust explosion simulation codes. Understanding flame accel-
eration due to flame distortion and turbulence produced by the propagating
explosion itself is central for understanding both dust and gas explosions.
Proust [Zl] gave an informative review of the state-of-the-art on propagation of
dust explosions in pipelines in relation to gas explosion propagation in pipe-
lines. The influence of turbulence on dust explosion propagation was also
discussed by Tamanini and Ural [22].
Much work has been, and is currently being done, on turbulent combustion
of sprays and mists [23], which may be worth while considering in the present
context.
Finally the singular phenomenon of dust cloud detonation should be men-
tioned. It is now generally accepted that this can occur, but further work is
needed in order to establish adequate understanding of the deflagration-to-
detonation-transition process (DDT). An excellent review of the state-of-the art
and outstanding problems in dust cloud detonation research was given by
Kauffman et al. 1131.

3.4 Blast waves generated by burning dust clouds

The case of interest in practice is blast waves from explosions in partly
confined geometries (e.g. deliberately vented, or bursting process equipment
and work rooms). The strength and shape of blast waves from dust explosions
depend on the way in which the dust clouds burn. For example, Wirkner-Bott et
al. [24] conducted a fairly detailed study of the nature of the secondary
explosion, i.e. the explosion of unburnt dust cloud outside the vent opening.
Central variables influencing blast wave generation, in addition to type of dust
cloud and geometry of system, comprise the dynamic state of the dust cloud at
the moment of ignition, ignition point in relation to vent, the vent size, and the
RX. Eckhff/J. Hazardous Mater. 35 (1993) 103-117 109

vent cover opening pressure. Wingerden [25] gave an informative overview of

pressure and flame effects in the direct surroundings of installations protected
by dust explosion venting.
Some basic studies of shock wave emission from burning dust clouds were
performed by Gelfand et al. [26]. More work, experimental as well as theoret-
ical, is needed in this area. A useful condensed introduction to the complex
field of properties and effects of blast waves from explosions was given by
Harmanny [27]. The effect of a given blast wave on humans, buildings and
process equipment is an other important area where more research is needed.
Valuable reviews are given by Mercx [28] and LAbbi? [29].

4. Status and outstanding problems in preventing and

mitigating/controlling dust explosions in practice

4.1 Generation and states of industrial dust clouds

Little quantitative knowledge of practical value has been generated. It is
known, however, that the dynamic state of a dust cloud dramatically influen-
ces both its ignition sensitivity and its combustion rate. Experimental invest-
igation of typical industrial processes of dust cloud generation, and the result-
ing states of the clouds, in various types of process equipment and modes of
operation, should be encouraged.

4.2 Preventing ignition sources

See the first column of Table 2. A considerable amount of fundamental
knowledge is available, as discussed in Section 3.2. Time is ripe for accepting
that the concept of ignition comprises a range of very complex processes.
Simple parameters such as a minimum ignition energy or temperature are not
true constants for a given dust, but vary significantly with the geometry and
other properties of the ignition source, as well as with the state of the dust
cloud. In the future this must be accounted for in practice.
Kleinschmidt [30] reported that the GreCon system for extinction of
sparks in terms of flying burning metal particles or organic material, is being
developed further with respect to optimizing system performance.

4.3 Preventing explosible dust clouds

Reference is made to the second column of Table 2. Inerting of dust clouds by
adding inert gas such as nitrogen or carbon dioxide, implies that the volume
percentage of oxygen in the atmosphere is reduced to a level at which the dust
cloud no longer can propagate a selfsustained flame. A fair amount of data for
maximum permissible oxygen content in the atmosphere for inerting exists.
However, there is room for improving the test methods by which such data are
obtained. Furthermore, most data are for atmospheric pressure and normal
temperature. Data for other conditions, in particular for elevated temperatures
110 R.K. EckhofflJ. HazardousMater. 35 (1993) 103-117

and pressures, are sometimes required, and adequate test methods should be
developed. Glor [ll] reported work on the determination of maximum permis-
sible &-contents for inerting clouds of coal dusts at elevated temperatures and
pressures.
Whilst reducing the oxygen content in the atmosphere prevents dust ex-
plosions, it can introduce a suffocation hazard. However, recent research has
shown that adding a few vol.% COZ to the gas mixture reduces the critical
oxygen threshold for suffocation considerably. An inert gas mixture
(INERGEN) utilizing this effect is now being marketed by Dansk Fire Eater
A/S [31], Further work to identify gas mixtures that keep the dust cloud inert
without presenting a suffocation hazard, should be welcomed.
Inerting by adding non-combustible dust is not generally applicable, because
the inert dust will in most cases cause unacceptable contamination. However,
there are cases where the dust/powder processed is already a mixture of
combustible and non-combustible dusts, and where control of the composition
ensures that dust clouds are non-flammable. It is then essential to avoid
segregation of combustible and non-combustible components, throughout the
process.
Keeping the dust concentration below the minimum explosible concentra-
tion is a third means of maintaining dust clouds non-explosible, There are,
however, at least two problem areas requiring further research and develop-
ment. First, more work is needed to establish test procedures for the determina-
tion of the minimum explosible concentration that is relevant in industrial
situations. Secondly, more research is needed to establish the minimum mass of
dust deposit per unit of surface area that is required for maintaining a self-
sustained explosive combustion along the surface. A thin dust layer on a floor
may be dispersed into a shallow, dense dust cloud close to the floor, through
which the flame can sweep, The conditions required for producing this kind of
self-sustained shallow sweeping flames need to be investigated further.

4.4 Mitigating and controlling measures

Attention is drawn to the third column of Table 2. The use of explosion-
pressure-resistantprocess equipment is limited because of high equipment costs.
Current experimental methods allow sufficiently accurate prediction of max-
imum explosion pressures in simple vessels with point source ignition. How-
ever, if complex dynamic pressure development, e.g. with pressure piling, is to
be expected, such test data are of limited value. There is also room for further
improvement in the design of the process equipment itself, with respect to
minimizing its heaviness. The German concept of pressure-shock-resistant
design should be further developed. Crowhurst [32] gave a useful overview of
the state-of-the-art on design of process equipment to withstand a given over-
pressure caused by a fully confined or vented explosion.
The objective of ezpbsion isolation is to prevent dust explosions from
spreading from the primary explosion location to other process units, work-
rooms etc. Various passive and active techniques have been developed and are
R.K. EckhoffjJ. Hazardous Mater. 35 (1993) 103-117 111

being used, but there is room for further improvement. If adequate performance
can be achieved, passive techniques are clearly more attractive than active
ones. Basic understanding of flame propagation and pressure build-up in
coupled geometries (interconnected vessels) is important for the prediction
of the performance of various active and passive isolation equipment. Valuable
large-scale experimental work in this area was reported by Lunn [33].
Zellweger [34] reports on further improvement of passive and active isolation
valves of the VENTEX type. Closing times (from sensing of the explosion to
valve is closed) down to 12 ms are now being obtained for active valves.
A simplified VENTEX valve, operating in one direction only, has also been
developed. Passive explosion interrupters based on venting at a bend have
been in use for some time. However, there is room for further exploration of the
potential of this attractive, simple principle of explosion isolation. A new
development was described by Alfert and Fuhre [35]. Glor [ll] reports on
ongoing work on performance of explosion barriers in ducting connecting to
vessels with venting or automatic explosion suppression.
PartiaE inerting by inert gas is a means for mitigating dust explosions, which
deserves further attention. The idea is that as the oxygen content in the
atmosphere is decreased, there is a gradual decrease of both ignition sensitiv-
ity and combustion rate of the dust cloud. In some cases the explosion hazard
may be reduced substantially by only a moderate reduction of the oxygen
content in the gas. However, more research seems necessary in this area to
establish correlations between the oxygen content in the gas and various
ignitability and combustion parameters.
Explosion uenting remains a complex and in part controversial subject.
Adequate understanding of flame propagation processes in dust clouds is
essential for the design of optimal venting arrangements in practice. Useful
reviews of various aspects of dust explosion venting in practice were given by
Scholl [36] and Lunn [37]. The basic understanding of flame propagation
processes inside and outside vented enclosures is still unsatisfactory. This
implies that neither the processes by which dust clouds of given structures are
generated, nor the way in which clouds of given initial structures burn, are
well understood. Consequently, adequate venting theories do not exist, and
one must rely on experiments. During the last few years the need for differenti-
ating vent, area requirements in view of the different turbulence levels, degrees
of dust dispersion and concentration distributions of dust clouds, which occur
in practice, has become widely accepted.
A further dimension of complexity is added to the venting problem if the
initial pressure (and/or temperature) deviates from atmospheric. Results from
venting of dust explosions at elevated initial dust cloud pressure were reported
by Siwek et al. [38]. The effect of pressure piling and turbulent flame jet
ignition on vent area requirements in systems of interconnected vessels, were
studied by Lunn [33].
The influence of vent ducts on the maximum explosion pressure in the vented
vessel has been studied experimentally by several workers. Recently, however,
112 R.K. &&off/J. HazardousMater.35 (1993) 103-117

Ural [39] presented a theoretical model for vented gas explosions by which he
has been able to predict pressure-versus-time characteristics in the vented
vessel that agrees well with corresponding experimental data. It would be
interesting to see whether a similar theory could reproduce existing data for
dust explosion venting via ducts.
The prediction of pressure and flame effects in the direct surroundings
of installations protected by dust explosion venting was discussed by Van
Wingerden [%I.
Dust explosion venting remains an area in which considerably more work is
required. An extensive review of existing knowledge is given by Lunn [40],
whereas the German Verein deutscher Ingenieure [41] issued a new draft
version of their venting guideline VDI 3673.
Automatic explosion suppression, being an active, comparatively sophisti-
cated method of dust explosion mitigation/control, is used when simpler and
less expensive methods cannot be applied. Recently Moore [42] reported that
the number of suppressant bottles of a given size required for suppressing
explosions of a given dust in a given vessel, was reduced by a factor of O-2-0.3
when the dust clouds were generated by industrial pneumatic injection, rather
than by the traditional VDI method used in previous experiments.
Siwek [43] described experiments where a combination of explosion venting
and automatic suppression was adopted for mitigating/controlling dust ex-
plosions in various enclosures. Although automatic suppression has been in
use for many years, there is still a need for research and development. Glor [ll]
and Moore [44] reported on current work on the possibility of applying this
method even in the case of highly explosible organic dusts of KS, > 300 bar-m/s.
In the case of aluminium powders, satisfactory suppression has not yet been
achieved for powders of Ks, ~200 bar m/s, which means that only dust ex-
plosions in clouds of relatively coarse aluminium powders can be suppressed.
The influence of the dynamic state of the dust cloud at the moment of sup-
pressant injection, the influence of the suppressant injection on this state, and
development of improved suppressants, are some of the areas where further
work seems useful. Recent experiments have indicated that water can be
effective as a suppressant, if injected at a temperature > 180 C.

5. Status and outstanding problems in testing of dust ignitability

and explosibility

5.1 Historical background

When some of the older test methods were designed, the ambition was in fact
quite modest. The original intention was just to establish some relative
measures of properties of practical relevance to preventing and control-
ling/mitigating dust. Later some of these methods were adopted as official
standards, and test data were sometimes treated as basic physical constants for
a given dust to an extent far beyond the original purpose of the test. As more
R.K. Eckhoff /J. HazardousMater.35 (1993)103-l 17 113

knowledge from systematic research became available, the lack of justification

for this use of these test data was pointed out, and the arbitrary, relative
nature of the various test methods was brought to light again.
The situation today is complex. It is realized that only a few of the dust
parameters that are currently being used for characterizing ignition sensitiv-
ity and explosibility of dusts, can be regarded even as approximate physical
constants for a given dust. In most cases a great number of variables are
involved and a differentiated view is required. Typical examples are the
minimum ignition energy and the explosion violence of dust clouds.

5.2 Limits of flame propagation - A special problem of scale

Determining limits of flame propagation constitutes an important test
objective. However, special care must be exercised in designing flame-propaga-
tion-limit tests. The basic problem is that near the limits self-sustained flame
propagation cannot be established unless a considerable amount of energy
is supplied for initiating flame propagation. Hence, if the volume of the
experimental dust cloud is too small, it is difficult to assess whether observed
flame propagation is truly independent of the ignition source. Some recent
results by Cashdollar and Chatrathi [45] are of fundamental significance in this
context. They found that clouds in air at normal ambient conditions of an
anthracite coal dust of 8% volatile matter, did not show self-sustained flame
propagation in a 1m3 test chamber, even when being exposed to a 30 kJ
chemical igniter. However, in a 20 litre chamber, fully developed explosions
were generated even with a 5 kJ chemical igniter, The reason for this could be
that in the small chamber, due to the initial combustion and expansion of the
dust cloud directly affected by the ignition source, the pressure and temper-
ature in the unburnt cloud ahead of the flame increase significantly before
flame propagation does no longer receive support from the ignition source.
Consequently the self-sustained flame propagation, if any, occurs in adiabati-
cally pre-compressed dust cloud, rather than in a cloud of normal ambient
temperature and pressure.
The results of Cashdollar and Chatrathi suggest that great care must be
exercised whenever comparatively small chambers, in particular closed ones,
are used for any explosion limit determination (explosible/non-explosible as-
sessment, minimum explosible dust concentration, maximum permissible
oxygen concentration for inerting).

6. Computer models and expert systems

During the last years there has been an increasing interest in developing
sophisticated mathematical models and expert systems for evaluation of dust
explosion hazards and assessment of optimal safety design features.
114 RX. Eckhoff/J. Hazardous Mater. 35 (1993) 103-117

This development is a natural consequence of two main factors. The first is

the almost explosive development of the performance of personal computers.
The second is the steadily increasing knowledge about ignition and explosion
phenomena, which demands a steadily more differentiated and complex
approach for solving practical design problems.
As long as this development is conducted by people who are not only experts
on computers, but also on the physics and chemistry of the phenomena treated,
models and expert systems should indeed be welcomed. However, there may be
a possibility of the future market place being offered software that is not up to
acceptable standards with respect to the physics and chemistry. As long as the
interior of the code is unknown, deficiencies in the basics may not be obvious
to the user.
A time may come where it could be useful and necessary to introduce the
concept of Quality Assurance even in this context. A need may emerge for
establishment of some internationally recognized body of experts that can
ensure that the software offered is up to acceptable standards. Software having
passed the investigation of this body could receive a certificate of approval.

7. Joint research efforts in Europe

Over the last years a steadily growing potential for organizing joint Euro-
pean research efforts has emerged within the EEC/EFTA/EUREKA system.
This also applies to dust explosion research.
British Materials Handling Board (BMHB) in UK has played a central role
in this process [46]. A number of research programmes have been started.
Gibson [47] reviewed the BMHB dust explosion research projects existing by
the date of his report, and some of the work on this topic being conducted or
planned in Europe and USA on the whole. He summarized the areas requiring
further work under the headlines:
l Combustion processes in dust clouds (experiments, theoretical models)

l Identification and control of ignition sources

l Design of methods to prevent/protect against dust explosions

These three headlines covers most of the research needs identified in the
present review, apart from mechanisms of generation of the dust cloud, which
are important because they set the stage for subsequent ignition and combus-
tion.

8. Conclusion

Initiation and propagation of industrial dust explosions are, from a funda-

mental scientific point of view, extremely complex phenomena. Comprehensive
theories for predicting ignition and combustion of dust clouds in industrial
RX. Eckhoff/J. Hazardous Maim-. 35 (1993) 103-117 115

environments from fundamental physical and chemical knowledge, are so far

beyond reach.
It is not surprising, therefore, that existing knowledge is to a large extent
fragmented. Nevertheless the hope is that more and more fragments will, step
by step, become tied together, and steadily increasing domains of coherence
emerge. Powerful computers are invaluable tools in this process. However,
experiments will remain indispensable for calibration of the mathematical
models, because such models will remain approximate and require tuning in
the foreseeable titure.
It is necessary to continue the execusion of realistic industrial-scale experi-
ments, at the same time as the more basic research and mathematical model-
ling should continue at full pace.
The current efforts to establish international co-operation in joint research
programmes should be encouraged.

References

1 R.K. Eckhoff, Dust Explosions in the Process Industries. Butterworth-Heinemann,

Oxford, UK, 1991.
2 K. Lebecki, J. Sliz, 2. Dyduch and P. Wolanski, Critical Dust Layer Thickness for
Combustion of Grain Dust. American Institute of Aeronautics and Astronautics, New
York, 1990, pp. 51-58.
E.A. Ural, Dust entrainability and its effect on explosion propagation in elongated
structures. Plant/Operations Progress, 1992.
B. Gelfand and S. Tsyganov, Private communication, Semenov Institute of Chemical
Physics, Academy of Sciences of Russia, Kosygin Street 4, Moscow 117977, Russia, 1992.
R. Klemens, Private communication, Politechnika Warszawska Instytut Techniki Ciep-
lnej, ul. Nowowiejska 25, Warszawa, Poland, 1992.
W. Hensel and W. John, Zusammenhang zwischen Glimmverhalten und Staubschich-
tdicke. Paper presented at VDI Colloquium Sichere Handhabung brennbarer St&be,
4-6 November 1992, Niirnberg, Germany, available from VDI, Diisseldorf (in German).
7 D. Crowhurst, Self-heating and spontaneous combustion characteristics of bulk powders
and powder layers. Paper given at the European Summer School on Dust Explosion
Hazards: Their Assessment and Control, Cambridge, UK, organized by IBC Technical
Services Ltd. in association with BMHB and IELG, 1992.
8 P. Wolanski, Dust explosion research in Poland. Powder Technol., 71(2) (1992) 197-208.
9 M. Glor, Paper given at the European Summer School on Dust Explosion Hazards:
Their Assessment and Control, Cambridge, UK, organized by IBC Technical Services
Ltd. in association with BMHB and IELG, 1992.
10 M. Glor and B. Maurer, Ziindversuche mit Schiittkegelentladungen. Paper presented
at VDI Colloquium Sichere Handhabung brennbarer Stiiube, 4-6 November 1992,
Niimberg, Germany, VDI-Berichte 975, available from VDI, Diisseldorf (in German).
11 M. Glor, Private communication, Ciba-Geigy AG, Basle, Switzerland, 1992.
12 P. Wolanski, Deflagration and Detonation Combustion of Dust Mixtures, by the Ameri-
can Institute of Aeronautics and Astronautics, New York, 1990, pp. 3-31.
13 C.W. Kauffman, M. Sichel and P. Wolanski, Dust related detonations, In: Dynamic
Structure of Detonation in Gaseous and Dispersed Media. Kluwer, Boston, MA, 1991; see
also C.W. Kauffman, M. Sichel and P. Wolanski, Research on dust explosions at the
University of Michigan, Powder Technol., 71(2) (1992) 119-134.
116 RX. EckhoffjJ. Hazardous Mater. 35 (1993) 103-117

14 J.H.S. Lee, F. Zhang and R. Knystautas, Propagation mechanisms of combustion waves

in dust-air mixtures, Powder Technol., 71 (1992) 153-162.
15 K. Seshadri, A.L. Berlad and V. Tar&ala, The structure of premixed particle-cloud
flames, Comb. Flame, 89 (1992) 333-342.
16 F. Rzal and B. Veyssiere, Interaction of a vortex ring with gaseous and dust flames, In:
Proc. 24th Int. Symp. on Combustion, Sydney, Australia, July 5-10, 1992. The Interna-
tional Combustion Institute, Pittsburgh, PA, 1992.
17 C. Proust, Private communication, INERIS, Part Technologique ALATA B&P. No. 2,
F-66550 Verneuil-en-Halatte, France, 1992.
18 K. van Wingerden, Private communication, Christian Michelsen Research AS, Fantof-
tvegen 38, N-5036 Fantoft-Bergen, Norway, 1992.
19 E.A. Ural, Private communication, Factory Mutual Research, P.O. Box 9102, Norwood,
MA 02062, USA, 1992.
29 D. Bradley, Z. Chen and J.R. Swithenbank, Burning rates in turbulent fine dust/air
explosions, In: Proc. 22nd Symp. (Int.) on Combustion. The Combustion Institute, Pit-
tsburgh, PA, 1988, pp. 1767-1775.
21 C, Proust, Paper given at the European Summer School on Dust Explosion Hazards:
Their Assessment and Control, Cambridge, UK, organized by IBC Technical Services
Ltd. in association with BMHB and IELG, 1992.
22 F. Tamanini and E.A. Ural, FMRC studies of parameters affecting the propagation of
dust explosions, Powder Technol., 71 (1992)135r151.
23 R.K. Eckhoff, Generation, ignition, combustion and explosion of sprays and mists of
flammable liquids in air. A literature survey. Report No. CMI-91.A25014 from Chr.
Michelsen Institute, N-5036 Fantoft, Norway, 1991.
24 I. Wirkner-Bott, S. Schumann and M. Stock, Dust explosion venting: Investigation of the
secondary explosion. Paper presented at 7th. Int. Symp. Loss. Prev. Saf, Prom. Proc. Ind.,
Taormina, Sicily, Italy 4-8 May 1992. Copy received from Battelle Europe, Frankfurt/
Main, Germany.
25 K. van Wingerden, Prediction of pressure and flame effects in the direct surroundings
of installations protected by explosion venting. Paper given at the European Summer
School on Dust Explosion Hazards: Their Assessment and Control, Cambridge,
UK, organized by IBC Technical Services Ltd. in association with BMHB and IELG,
1992.
26 B. Gelfand, S. Medvedev, A. Polenov and A. Bartenev, Shock waves from expansion of
burning dust clouds, Comb. Expl. Shock Waves, 3 (1990) 8591.
27 A. Harmanny, Explosion Effects. Proc. 1. World Seminar on the Explosion Phenomenon
and on the Application of Explosion Protection Techniques in Practice. Arranged by
EuropEx, in Brussels 17-21 February, 1992.
28 W.P.M. Mercx, Critical appraisal of damage criteria for buildings and process equip-
ment, In: Proc. 1st World Seminar on the Explosion Phenomenon and on the Application
of Explosion Protection Techniques in Practice. Arranged by EuropEx. in Brussels
17-21 February 1992.
29 R.J. LAbbb, Explosion effects on people, In: Proc. 1st World Seminar on the Explosion
Phenomenon and on the Application of Explosion Protection Techniques in Practice.
Arranged by EuropEx, in Brussels 17-21 February 1992. Available from EuropEx,
Kontich, Belgium.
30 H.-P. Kleinschmidt, Private communication, Fagus-GreCon Greten GmbH & Co.,
Postfach 1243, D-W-3230 Alfeld-Hannover, Germany, 1992.
31 Dansk Fire Eater A/S, INERGEN, Anlaegsbeskrivelse & Design. Skovlytoften 14,DK-
2840 Holte, Denmark, 1992 (in Danish).
32 D. Crowhurst, Paper given at the European Summer School on Dust Explosion Haz-
ards: Their Assessment and Control, Cambridge, UK, organized by IBC Technical
Services Ltd. in association with BMHB and IELG, 1992.
R.K. Eckhoffl J. Hazardous Mater. 35 (1993) 103-117 117

33 G.A. Lunn, The explosion venting of interconnected vessels, In: Proc. 1st World Seminar
on the Explosion Phenomenon and on the Application of Explosion Protection Tech-
niques in Practice. Arranged by EuropEx, in Brussels 17-21 February 1992.
34 J. Zellweger, Private communication; Rico-Sicherheitstechnik AG, Lindenstrasse 77,
CH-9906 St. Gallen, Switzerland, 1992.
35 F. Alfert and K. Fuhre, Venting of dust explosions in filters and integrated systems.
Report 92-A26021, Chr. Michelsen Institute, Dept. of Science and Technology, N-5036
Fantoft, Norway, 1992.
36 E.W. Scholl, The technique of explosion venting. Much more than just a set of nomo-
graphs, In: Proc. 1st World Seminar on the Explosion Phenomenon and on the Applica-
tion of Explosion Protection Techniques in Practice. Arranged by EuropEx, in Brussels
17-21 February 1992.
37 G.A. Lunn, Paper given at the European Summer School on Dust Explosion Hazards:
Their Assessment and Control, Cambridge, UK, organized by IBC Technical Services
Ltd. in association with BMHB and IELG, 1992.
38 R. Siwek, M. Glor and Torreggiani, Dust explosion venting at elevated initial pressure.
Prepr, 7 Int. Symp. Loss. Prev. Safety Prom. Proc. Ind., Taormina, Sicily, Italy, 4-8 May
1992, SRP Partners, 66161 Rome, Viale di Villa Massimo 37, Italy.
39 E.A. Ural, A simplified method for predicting the effect of ducts connected to explosion
vents, J. Loss Prev. Process Ind., submitted.
40 G.A. Lunn, Dust Explosion Prevention and Protection. Part I - Venting. Institution of
Chemical Engineers, Rugby, UK, 1992.
41 Verein dcutscher Ingenieure, Pressure Release of Dust Explosions, Draft revised Guide-
line, VDI, Dusseldorf, Germany, 1992 (in German).
42 P.E. Moore, Explosion suppression for protecting systems against the explosion hazard
presented by pneumatic filling techniques, In: Proc. 1st World Seminar on the Explosion
Phenomenon and on the Application of Explosion Protection Techniques in Practice.
Arranged by EuropEx, in Brussels 17-21 February 1992.
43 R. Siwek, The combination of explosion venting and explosion suppression. Explosion
suppression in very small volumes, In: Proc. 1st World Seminar on the Explosion
Phenomenon and on the Application of Explosion Protection Techniques in Practice.
Arranged by EuropEx, in Brussels 17-21 February 1992. Available from EuropEx,
Kontich, Belgium.
44 P.E. Moore, Paper given at the European Summer School on Dust Explosion Hazards:
Their Assessment and Control, Cambridge, UK, organized by IBC Technical Services
Ltd. in association with BMHB and IELG, 1992.
45 K.L. Cashdollar and K. Chatrathi, Minimum explosible dust concentrations measured in
20 litre and 1 m3 chambers. Comb. Sci. Technol., 87 (1992) 157-171.
46 P.G. Middleton, Private communication, British Materials Handling Board, Index
House, Ascot, Berkshire SL5 7EU, UK, 1992.
47 N. Gibson, A review of dust explosion research projects for BMHB research project
panel. An interim report. BMHB, Ascot, Berkshire, UK, November 1991.
Safety Science, 16 (1993) 511-525 511
Elsevier

Influence of initial and explosion-induced

turbulence on dust explosions in large vented silo
cells

R.K. Eckhoff
University of Bergen/Christian Michelsen Research, Bergen, Norway

Abstract

The violence, or the rate, of a dust explosion is a deciding factor as to whether venting of the
explosion will be effective. The turbulence of the dust cloud, initial as well as explosion-induced,
plays a central role in determining the explosion violence. This was demonstrated in dust explo-
sion experiments in large vented silo cells of 500 and 236 m3. In vented silo cells of L/D > 6, explo-
sion-induced turbulence can increase the explosion violence substantially. Further development
of mathematical models for simulation of dust explosions in silo cells and other process equipment
should be encouraged.

R&urn6

La violence, ou la force, dune explosion de poussikre est un facteur dkcisif pour savoir si oui non
le percage de lexplosion peut Btre efficace. La turbulence du nuage de poussikre, initiale et nourrie
par lexplosion, joue un r81e elk dans la determination de la violence de lexplosion. On a pu faire
la demonstration de cela dans des explosions de poussikre dans de grands silos a&es de 500 et de
236 m3. Dans les silos a&es de L/D> 6, la turbulence nourrie par lexplosion peut augmenter la
violence de lexplosion de man&e significative. 11faudrait encourager la mise en place de nouveaux
modkles mathematiques destines a la simulation des explosions de poussikre dans des silos et
dautres equipements analogues.

Zusammenfassung

Die Gewalt oder Geschwindigkeit einer Staubexplosion ist von entscheidender Bedeutung fur
die Frage, ob die Enthiftung im Explosionsfall effektiv sein wird. Die Turbulenz der Staubwolke,
von Ursprung an oder infolge der Explosion, spielt eine zentrale Rolle bei der Bestimmung der
Explosionsgewalt. Dies wurde gezeigt bei Staubexplosionsversuchen in umfassend geltifteten Sil-
ozellen von 500 und 236 ma Inhalt. In geltifteten Silozellen, wo L/D > 6, kann die durch Explosion
hervorgerufene Turbulenz die Explosionsgewalt betrachtlich erhijhen. Weitere Entwicklung von
mathematischen Modellen zur Simulation von Staubexplosionen in Silozellen und anderen
Prozecanlagen sollte stimuliert werden.
512

1. Introduction

Industrial enclosures, such as conventional process equipment, are normally

too weak to withstand the pressures exerted even by only partly developed,
confined dust explosions. Consequently a primary objective of fighting an ex-
plosion, after it has been initiated, is to prevent the building-up of destructive
overpressures. Several techniques for preventing destructive overpressures are
in current use in industry. Probably, the most widely used is uerttirzg.
Large storage silos of up to several thousand m3 volume are common in the
grain, feed and flour industries. In order to protect the wall and roof of such
silo cells against excessive internal overpressure in the case of a dust explosion
inside the silo, the silos are often provided with an explosion vent in the top
part, normally in the roof. However, the sizing of such vents remains a complex
and partly controversial issue.
The rate of combustion of the dust cloud, i.e. the rate of heat generation
inside the enclosure where the explosion is initiated, is a deciding factor as to
whether a given vent design will perform adequately. In view of the fact that
the combustion rates of dust clouds in industrial practice, even of the same
dust, vary substantially, it is important to base the design of the vent arrange-
ment on the best possible estimate of the combustion rate, or explosion viol-
ence, that will occur in practice.

2. Factors influencing the explosion of dust clouds

There are several important factors, and the comprehensive list may look as
follows:
(a) Chemical composition of the dust, including its moisture content.
(b) Chemical composition, and initial pressure and temperature of the gas
phase.
(c) Distributions of particle sizes and shapes in the dust, determining the spe-
cific surface area of the dust in the fully dispersed state.
(d) Degree of dispersion, or de-agglomeration, of dust particles, determining
the effective specific surface area available to the combustion process in
the dust cloud in the actual industrial situation.
(e) Distribution of dust concentration in the actual cloud.
(f) Distribution of initial turbulence in the actual cloud.
(g) Possibility of generation of explosion-induced turbulence in the still un-
burnt part of the cloud. (Location of ignition source important parameter.
Flame jet ignition will create turbulence.)
(h) Possibility of flame front distortion by other mechanisms than turbulence.
(i) Possibility of significant radiative heat transfer (highly dependent on flame
temperature, which in turn depends on particle chemistry).
 513

Factors (a), (b), (c) and (i) can be regarded as basic parameters of the
explosible dust cloud. Factors (d) to (h) are, however, influenced by the actual
industrial dust cloud generation process and explosion development. These in
turn depend on the nature of the industrial process (flow rates, etc.) and the
geometry of the system in which the dust cloud burns.
In view of the wide spectrum of dust cloud concentrations, degrees of dust
dispersion and turbulence, and locations of potential ignition sources in in-
dustry, a correspondingly wide spectrum of possible dust cloud combustion
rates must be expected for any given dust.
This complex reality of the process industry is also shared by laboratory
experimentation and represents a constant challenge in the design of adequate
experiments and the interpretation of experimental results.

3. Role of turbulence in increasing the combustion rate of dust clouds

3.1. Initial and explosion-induced turbulence

In practical terms, turbulence may in the present context, be described as a

state of rapid, more or less random, movement of small elements of the dust
cloud relative to each other in three dimensions. If the cloud is burning, tur-
bulence will give rise to mixing of the hot burnt and burning parts of the cloud
with the cold unburnt parts, and the cloud will become a three-dimensional
laminate of alternating hot/burning and cold unburnt zones. Therefore a tur-
bulent cloud will burn much faster than a quiescent cloud through which a
smooth, comparatively thin flame sheet is propagating.
In relation to dust explosions in industry two kinds of turbulence, differing
by their origin, have to be considered. The first is turbulence generated by the
industrial process in which the dust cloud is formed, whether an air jet mill, a
mixer, a cyclone, a bag filter, a pneumatic transport pipe, or a bucket elevator.
This kind of turbulence is often called initial turbulence. The second kind is
generated by the explosion itself by expansion-induced flow of unburnt dust
cloud ahead of the propagating flame. The level of turbulence generated in this
way depends on the speed of the flow and the geometry of the system. Vent
openings and obstacles, like the buckets in a bucket elevator leg, enhance the
turbulence generation under such conditions.
In long ducts or galleries a positive feed-back loop can be established by
which the flame can accelerate to very high speeds and even transit to a
detonation.
Further details and references to published work are given by Eckhoff (1991) .
514

3.2. Influence of initial turbulence on the violence of dust explosions in closed

vessels.

The marked influence of initial turbulence on the combustion rate of a dust

cloud is exposed clearly in dust explosion experiments in closed vessels, using
transient dust clouds.
A series of such experiments were described by Eckhoff (1977 ). Lycopodium
dust was dispersed into a cloud in the Hartmann bomb by a short blast of air.
Ignition of the cloud was accomplished by an electric spark of a few J net
energy, or by a 100 J chemical igniter, fired at a pre-set delay after onset of
dust dispersion.
Some results are given in Fig. 1. The ignition delay is a relative measure of
the initial turbulence. In the early stages of dust dispersion, the dust cloud was
quite turbulent, but the turbulence faded away with time as the dispersion air
flow diminished. Therefore, when explosion experiments with the same dust
were repeated in the vessel, using different delays between initiation of dust
dispersion, and ignition, the dust clouds had different levels of initial turbu-
lence at the moment of ignition.

20 SO 100 200 500 1000

DELAY BETWEEN DUST DISPERSION AND IGNITION Id

Fig. 1. Influence of initial turbulence on the explosion rate of a dust cloud. Experiments with 420
g mP3 of lycopdoium in air in the 1.2 1 Hartmann bomb. 5 experiments per delay. Bars indicate
& 1 standard deviation. Note that bar (g)denotes gauge pressure, i.e. pressure above atmospheric
pressure. After Eckhoff (1977).
 515

p 0.6
.c
->
04

0
0 LO 60 80
TIME AFTER OPENING AIR BLAST VALVE [ml

Fig. 2. Variation of rms turbulence velocities within 5 ms windows in a 1.2 1 Hartmann bomb
with time after opening of air blast valve, and with initial pressure in dispersion reservoir. Air
only, no dust. After Amyotte and Pegg (1989).

As Fig. 1 shows, the explosion violence, in terms of the maximum gradient

of pressure vs. time, decreased markedly, by at least an order of magnitude, as
the initial turbulence faded away. However, the maximum explosion pressure
remained fairly constant up to delays of about 200 ms. This is due to the fact
that the maximum pressure is essentially a thermodynamic property, reflect-
ing the total heat released in the combustion of the dust, whereas the rate of
pressure rise contains a strong kinetic component. However, even the maxi-
mum pressure will start to decay when ignition delays become very long, due
to settling-out of the dust, as is also indicated in Fig. 1.
The direct measurements of the rms (root mean square) turbulence inten-
sity as a function of time after opening the dispersion air valve in a Hartmann
bomb, by Amyotte and Pegg (1989)) and their comparison of the data with the
data from Hartmann bomb explosion experiments by themselves and Eckhoff
(1977), give further evidence. The results are shown in Fig. 2. It is seen that a
decay by a factor of almost ten of the turbulence intensity occurs within the
same time frame from about 40 to 200 ms as a corresponding decay of (dP/
dt),,, in Eckhoffs experiments in Fig. 1. It is also seen that the turbulence
intensity increases systematically with the initial pressure in the dispersing air
reservoir, i.e. with increasing strength of the air blast, in accordance with
expectations.
Further discussion of these and similar experiments conducted by other
workers is given by Eckhoff (1991).

4. The role of dust cloud turbulence in vented dust explosions

4.1. General outline

In the venting of dust explosions the pressure as a function of time P(t) in

the vented enclosure is the net result of two competing processes:
516

- burning of the dust, causing heat production and hence pressure rise;
- discharge of unburnt cloud and combustion products through the vent open-
ing, causing pressure decrease.
Therefore, the maximum explosion pressure in a given enclosure equipped
with a given vent increases with the rate of burning of the dust cloud, or rate
of heat release by combustion, in the enclosure. This not only implies that
different dusts burn with different rates, but indeed also that the combustion
rate of a given dust in a given enclosure can vary considerably, depending on
the physical state of the cloud, in terms of dust concentration distribution,
degree of dispersion or de-agglomeration of the particles, and last but not least,
the turbulence level. In the present context the influence of turbulence is of
primary interest.
It is to be expected in general that the maximum pressure in an explosion of
a given dust in a given vented enclosure increases systematically with increas-
ing turbulence level (e.g. expressed as an rms velocity) in the burning dust
cloud. (At very high levels of turbulence, flame quenching phenomena may
arise, but this aspect will not be considered here.)

4.2. Experiments by Tamanini and co-workers

The influence of dust cloud turbulence on explosion venting processes was

studied specifically by Tamanini (1989). He and his co-workers conducted
vented dust explosion experiments in a 64 m3 rectangular enclosure, the vent
being a 5.6 m2 square opening in one of the four 14 m2 walls of the enclosure.
Details of the experiments were given by Tamanini and Chaffee (1989).
The dust injection system consisted of four pressurized-air containers con-
nected to 16 perforated dust dispersion nozzles. The dust was placed in four
canisters, one for each of the pressurized air containers, located in the lines
between the pressurized containers and the dispersion nozzles. On activation
of high-speedvalves, the pressurized air released from the containers entrained
the dust and dispersed it into a cloud in the 64 m3 chamber via the 16 nozzles.
As discussed above for closed-vessel experiments, this dust dispersion method
generates transient dust clouds characterized by a comparatively high turbu-
lence intensity during the early stages of dust dispersion, and a subsequent
marked fall-off of the turbulence intensity with increasing time from the start
of the dispersion. This means that the turbulence level of such a dust cloud at
the moment of ignition can be controlled by controlling the delay between start
of dust dispersion and activation of the ignition source.
By varying the delay between onset of dust dispersion and ignition, Taman-
ini (1989) and Tamanini and Chaffee (1989) used this effect to study the
influence of the turbulence intensity at the moment of ignition on the maxi-
mum pressure generated by explosion of a given dust at a given concentration
in their 64 m3 vented chamber. The actual turbulence intensity in the large-
 517

0.6 -

05 -

04 -

03 r

02 -

01 _

00 I J
0 2 i 6 8 10 12 1L

AVERAGE RMS OFINSTANTANEOUS

VELOCITY [m/s1

Fig. 3. Influence of initial turbulence intensity of burning dust cloud on maximum pressure in
vented maize starch explosion in a 64 m3 rectangular chamber. Starch concentration 250 g/m3.
Vent size 5.6 m. Ignition at the chamber centre. From Tamanini (1989).

scale dust cloud at any given time was measured by a bi-directional fast-re-
sponse gas velocity probe, in terms of the rms (root-mean-square) of the in-
stantaneous velocity.
Figure 3 gives a set of data showing a clear correlation between the maximum
pressure in the vented explosion and the average rms of the instantaneous
fluctuating turbulence velocity at the moment of ignition, as measured by the
pressure probes.
The contribution of Tamanini and co-workers is particularly valuable be-
cause it suggests that a quantitative link between systematic venting experi-
ments, in which the initial turbulence is quantified, and real industrial explo-
sion situations may be obtained via measurement of characteristic turbulence
levels in dust clouds in industrial process equipment in which vented explo-
sions may occur.

5. Influenceof initial turbulence on violence of dust explosions in a

500 m3 vented silo of L/D= 4

The experiments to be reported were described in detail by Eckhoff and

Fuhre (1984). The silo used is illustrated in Fig. 4. The dust was injected up-
wards into the silo through a 200 mm diameter pipe coinciding with the silo
axis. In the final, destructive experiment, high initial turbulence was ensured
in the upper part of the silo by maintaining turbulent injection of explosible
dust cloud right through the entire explosion process.
The resulting pressure vs. time history is shown in Fig. 5. It consists of three
phases. Following an initial small pressure rise of about 10 mbar after about
0.8 s, the onset of a normal rise of pressure ( z 0.17 bar s-l) was observed at
1.3 s. The pressure continued to rise in this normal manner up to about 55
 518

VENT
-_.._..i -

22 m

,) ,.
.,I , .. 1
,I .,.
.._..
i . :
:

Fig. 4. Cross section of 500 m3 steel plate silo used in the turbulent dust/air jet explosion experi-
ment described by Eckhoff and Fuhre (1984)

ONSET OF

TIME AFTER IGNITION bl

Fig. 5. Pressure VS. time during explosion of clouds of maize starch in 500 m3 steel plate silo in the
presence of a turbulent dust/air jet. 300 kg maize starch injected into the silo. 8.8 m2 open vent in
the silo roof. Ignition at the silo bottom. After Eckhoff and Fuhre (1984).
 519

mbar at about 1.5 s. The overpressure developed at this point was right in the
middle of the range of peak pressures obtained in the six preceding normal
experiments (with initially quiescent maize starch clouds). However, at this
moment the abnormal, very fast pressure rise began, resulting in a further
pressure increase of more than 500 mbar in less than 0.2 s. The steepest part
of this pressure rise phase had a slope of 14 bar s-l, i.e. 80 times that in the
normal phase. The maximum overpressure recorded before the silo ruptured
was about 580 mbar. From the very high rate of pressure rise at the moment of
rupture, it seems reasonable to deduce that had the silo been sufficiently strong
to withstand the explosion, the pressure in the silo would probably have risen
considerably higher than about 0.6 bar(g) before reaching its peak, in spite of
the generous venting.
Figure 6 summarizes the results for the initially quiescent dust clouds (dust
injection terminated a few seconds before ignition) and the single experiment
with ignition while dust injection was maintained. The maximum explosion-
pressure vs. vent-area relationship predicted by VDI 3673 (1979 edn.) is also
included for comparison.
Figure 6 suggests that the combustion rate of a relatively quiescent dust
cloud in a large empty vented silo of L/D=4 is considerably lower than that
implied in the VDI 3673 (1979 edn. ) recommendations. However, the result
from the experiment with high initial turbulence shows that not even these
conservative recommendations constitute a universally valid worst-case. On
the other hand, it can be argued that, as far as large silos are concerned, the

EXPERIMENTAL
MAX EXPLOSION

OF TURBULENT

K.jk = 115 bar m/s

Pstat: 0.1 bar (gl

EXPERIMENTAL
MAX. EXPLOSION

ABSENCE OF
TURBULENT JET
1
0 02
t
001 I
0 5 10 15

VENT AREA Lm21

Fig. 6. Comparison of experimental maximum explosion pressures from experiments in vented

500 m3 silo with and without initial turbulence, and predictions by VDI 3673 (1979 edn.). After
Eckhoff and Fuhre (1984).
520

dust cloud generation process illustrated in Fig. 4 is not likely to be represent-

ative of industrial practice.

6. Influence of explosion-induced turbulence on violence of dust ex-

plosions in a 236 m3 vented silo of L/D= 6

These experiments were reported in detail by Eckhoff et al. ( 1987,1988). A

cross section of silo, indicating alternative dust injection and ignition points,
and various diagnostics, is shown in Fig. 7.
In the case of Eckhoff et al. (1987) the vent was located in the silo roof,
whereas in the experiments by Eckhoff et al. (1988) the vent opening was in
the silo wall just below the top, the roof itself being fully closed. Dust was
injected pneumatically into the silo either from the bottom or the top as indi-
cated in Fig. 7. Air of the desired flow rate was supplied by a Roots blower, and

STRONG STEEL GRID SUPPORTING

STEEL COVER PLATES-, II , ,

HIGHEST IGNITION POINT

OUST INLET
PIPE, TOP
PRESSURE PROBE INJECTION

OUST CONCENTRATION
t
PROBE
,,,,CONCENTRATlONj_ i

INTERMEDIATE IGNITION
POINT

OUST CONCENTRATIO
PROBE
OUST CONCENTRATIO
PROBE

PRESSURE PROBE

INTERMEDIATE IG
POINT
OUST CONCENTRATI
PROBE

OUST CONCENTRA
PROBE

PRESSURE PROBE

LOWEST IGNITION POINT 0123

METRES

OUST INLET, I.
BOTTOM INJECTION

Fig. 7.236 m3 silo in Norway for dust explosion experiments, with provisions for roof venting. L/
D=6.Aftereckhoffetal. (1987).
 521

dust was fed into the air flow at the desired mass flow rate. Before ignition the
air flow was stopped and the dust cloud allowed to calm down for a few seconds.
Therefore, the initial turbulence of the dust cloud at the moment of ignition
was negligible.
Figure 8 shows the influence of the location of the ignition point along the
silo axis, on the maximum explosion pressure in the vented silo for three dif-
ferent vent configurations. Some influence of the vent configuration was ob-
served. However, the main message of Fig. 8 is that there is a very marked
increase of the maximum explosion pressure in the silo, from the order of 0.01
to the order of 1.0 bar(g), when the ignition point is shifted downwards in the
silo from the top to the bottom. Some of this effect is due to different vented
gas densities. In the case of top ignition the vented gas was essentially hot, low-
density combustion products, whereas cold, high density dust cloud was vented
during the first decisive phase with bottom ignition. This effect was studied
theoretically by Nagy and Verakis ( 1983 ) .
However, with bottom ignition the upwards flame front speed close to the
vent exceeded 100 m SC and there is little doubt that explosion-induced tur-
bulence also played a central role. Thus quite early in this century, gas explo-
sion experiments in one-end-open pipes had shown that as the explosion prop-
agated along the pipe, it could undergo a dramatic change from harmless laminar
flames, via a wide spectrum of turbulent combustion, to detonation. The secret
of the dramatic flame acceleration was also resolved on a qualitative basis.
After ignition, the first combustion phase occurs rather slowly as an approxi-
mately laminar flame spread. However, as a result of the heat produced by the
combustion, the combustion products expand, and this leads to the still un-

1.4

0 2 6 6 8 10 12 l'+ 16 18 20

LOCATION OF IGNITIONPOINT ABOVE SILO FLOOR [ml

Fig. 8. Influence of location of ignition point in a 236 m3 vented silo of L/D = 6,on the maximum
explosion pressure in the silo. Maize starch of 11% moisture and concentration 5 400-500 g m-.
After Eckhoff et al. (1987,1988).
burnt gas ahead of the flame being pushed towards the open end of the pipe.
Depending on pipe diameter, flow rate, and the roughness of the pipe walls, the
flowing unburnt gas will be rendered more or less turbulent. When the com-
bustion zone approaches these turbulent regions further along the pipe, the
combustion rate will increase. This leads in turn to faster expansion of the
burnt gas, which in its turn increases yet further the flow rate of the unburnt
gas ahead of the flame, and so on. In other words, there is a self-accelerating
explosion process, or a positive feed-back loop, with the generation of flow-
induced turbulence in the gas ahead of the flame as a key element.
Extensive experimentation in several countries with coal dust explosions in
long pipes, ducts and large-scale galleries has confirmed that this rather mul-
tifarious picture indeed also applies to dust explosions, and explains why the
location of the ignition point with respect to the vent was of such crucial im-
portance for the pressure build-up during dust explosions in the 236 m3 silo. If
ignition occurs at the silo top, close to the vent, there is no possibility of gen-
erating high-velocity flow and high turbulence levels in the unburnt cloud fur-
ther down. These aspects were discussed in greater detail by Eckhoff (1987).
Figure 8 shows that a 4.4 m2 wall vent in fact gave somewhat lower explosion
pressures than a larger 5.7 m2 vent in the silo roof. A possible explanation of
this result was given by Eckhoff et al. (1988).

7. Initiation of dust explosions in large silo cells by strong flame jets

Alfert and Fuhre (1992) showed that flame jet ignition of dust clouds in a
generously vented 2 m3 vessel gave rise to much higher maximum pressures in
the vessel than those expected from even quite conservative venting guidelines.
In principle a similar effect should be expected if a dust cloud in a silo cell is
ignited by a flame jet.
As pointed out by Eckhoff (1987), analyses of accidental dust explosions
involving silos indicate that most often the silo explosion is a secondary event,
following a primary explosion elsewhere in the plant. For this reason, flame
jets entering the silo from the outside through available openings seem to be a
likely type of ignition source for silo explosions. In view of Fig. 8, however, it
is crucial whether flame jet ignition takes place near the bottom of the silo cell
or near the vented top.

7.1. Is flume jet ignition in the bottom part of the silo cell a credible event?

It has been suggested that some silo explosions in the past may have been
initiated by a flame jet from a violent explosion below the silo that entered the
silo through the discharge spout at the bottom. This suggestion implies first
that the discharge spout is open, or can be blown open by an explosion below
 523

the silo. Secondly, it implies that the explosible dust cloud in the silo is large
enough to generate a significant explosion pressure. An initially open dis-
charge spout would normally imply that the silo is empty, apart from dust
layers on the wall and roof. If the dust cloud is to be generated from such layers,
it must be dispersed either by mechanical vibrations or by an air blast preced-
ing the flame jet.
The scenario that the silo spout at the bottom is open at the same time as
the silo is being filled from the top seems rather unlikely for normal operational
reasons. One could in theory envisage a situation where a dust explosion prop-
agated in the tunnel below the silo at the same time as filling of the silo with
dust from the top had just started with the bottom spout left open by mistake.
In such a case, strong worst-case explosions could occur that no normal silo
would be able to withstand, even if generous venting was provided. However,
one must consider whether an accidental coincidence of the three events nec-
essary to create this situation would be credible.
Whether or not it is likely that a closed bottom spout can be blown open by
an overpressure from below also needs to be considered. In addition to blowing
the spout open, the overpressure must also be able to lift any bulk material
stored in the silo.
Another mechanism for generating and igniting dust clouds in silos from
below could be that a dust explosion in the tunnel generates and compresses a
dust cloud ahead of the flame, so that an unburnt explosible dust cloud at a
higher than atmospheric pressure is pushed into the silo through any open
bottom spout, and subsequently ignited inside the silo by the flame jet that
follows it. If the size of the spout opening and the pressure drop across it are
known, the amount of dust cloud entering the silo per unit time can be esti-
mated. However, it does not seem obvious that large volumes of unburnt cloud
can be generated in the silo in this way before the flame reaches the spout.
A more likely situation would be that the dust cloud was generated in the
empty silo by dispersion of dust layers on the internal surface of the wall and
roof. Dispersion of a 1 mm layer of grain dust on the internal wall of a 7 m
diameter silo would give an average dust concentration of about 300 g/m3 in
the silo. Dispersion of such dust layers could be due to mechanical shaking
caused by preceding explosions elsewhere in the plant. If an explosion had first
occurred in the tunnel below the silo, ignition at the silo bottom could be ac-
complished by a flame jet entering the silo through the open bottom spout.

7.2. Flame jet ignition in the top part of the silo cell seems more credible

It is nevertheless felt that the most likely scenario for flame jet ignition in a
large silo is a jet entering the silo in the top region, either through a dust ex-
traction opening or through the product feeding system. This is likely both
because these openings would be open during dust filling, and because the or-
524

igin of dust explosions in grain elevators and feed and flour mills is often either
the milling and grinding system or mechanical transport units such as bucket
elevators.
As shown in Fig. 8, top ignition in a large slim silo, provided with a reason-
able size vent at the top, only generates very low explosion pressures, of the
order of a few tens of millibars, even with a large, effective ignition source, and
with the entire silo filled with a worst-case concentration dust cloud. Flame jet
ignition near the silo top would be expected to give somewhat higher, but not
excessive pressures.
Further discussions of possible modes of ignition of dust clouds in large silos
are given by Eckhoff (1987 ) and Alfert et al. (1989). The need for considering
this problem exposes the fact that sizing of dust explosion vents is also indeed
a problem of risk analysis, as outlined by Eckhoff (1986).

8. Computer modelling of turbulent dust explosions

This aspect has been reviewed by Eckhoff (1991). Considerable progress has
been made during the last decade in predicting the violence of turbulent gas
explosions in various industrial geometries by advanced computer simulation
techniques. The pioneering work by Hjertager (1982,1984,1986), carried out
at CMI, should be mentioned specifically.
The parameters of the explosible gas mixture used as input to the compu-
tational codes are the standard physical constants, and the laminar and tur-
bulent burning velocities, or a global chemical induction time for ignition, as
functions of gas composition, pressure, and temperature. In addition, an ade-
quate description of the geometry of the system in which the explosion takes
place must be provided to the computer. The same applies to a suitable repre-
sentation of the composition distribution of the gas cloud throughout the ge-
ometry in question at the moment of ignition, and the location of the ignition
source. On the basis of this information the computer code is in principle able
to estimate explosion pressure and flame structure as functions of time and
space.
The question in the present context is whether a similar approach is appli-
cable to dust explosions. The two-phase nature of dust clouds complicates the
modelling problem, but it seems that a comprehensive theoretical method for
prediction of dust explosion violence in industrial plants may in fact be within
reach. One promising attempt at developing such a model is that of Kjaldman
(1992). As shown by Eckhoff (1989), the scaling of vented dust explosions in
silo cells is very complex and a comprehensive computer model seems required
even for this purpose.
 525

References

Alfert, F. and Fuhre, K., 1992. Venting of Dust Explosions in Filters and Integrated Systems,
CMI-report No. 92-A25021, April 1992.
Alfert, F., Eckhoff, R.K. and Fuhre, K., 1989. VDI-Berichte No. 701, Vol. 1, pp. 303-319.
Amyotte, P.R. and Pegg, M.J., 1989. J. Loss Prev. Process Ind., 2: 87.
Eckhoff, R.K., 1977. Staub-Reinhalt. Luft, 37: 110.
Eckhoff, R.K., 1986. Bulk Solids Handling, 6: No. 5.
Eckhoff, R.K., 1987. in K.L. Cashdollar and M. Hertzberg (eds.), Industrial Dust Explosions,
ASTM Special Techn. Publ. 958, ASTM, Philadelphia, USA, pp. 265-280.
Eckhoff, R.K., 1989. Archivum Combustionis, 9: 327.
Eckhoff, R.K., 1991. Dust Explosions in the Process Industries, Butterworth-Heinemann, Oxford.
Eckhoff, R.K. and Fuhre, K., 1984. J. Occup. Accid., 6: 229.
Eckhoff, R.K., Alfert, F., Fuhre, K. and Pedersen, G.H., 1988. J. Loss Prev. Process Ind., 1: 16.
Eckhoff, R.K., Fuhre, K. and Pedersen, G.H., 1987. J. Occup. Accid., 9: 161.
Hjertager, B.H., 1982. Combust. Sci. Technol., 27: 159.
Hjertager, B.H., 1984. J. Hazard. Mater., 9: 315.
Hjertager, B.H., 1986. Handbook of Heat and Mass Transfer, Gulf Publishing, Houston, pp. 1303-
1350.
KjXklman, L., 1992. Powder Technol.. 71: 163.
Nagy, J. and Verakis, H.C., 1983. Development and Control of Dust Explosions. Marcel Dekker,
New York.
Tamanini, F., 1989. Turbulence effects on dust explosion venting, AlChE Loss Prevention Sym-
posium, Paper 12a Session 8, (April) Houston.
Tamanini, F. and Chaffee, J.L., 1989. Large-scale vented dust explosions - effect of turbulence on
explosion severity, Technical Report FMRC J.I. OQ2E2.RK (April) Factory Mutual Research.
266 Book Reviews/J. Hazardous Mater. 34 (1993) 261-270

Gas, Dust and Hybrid Explosions (Fundamental Studies in Engineering Vol. 51,
by W-E. Baker and M.J. Tang, Elsevier, Amsterdam, 1991, ISBN 0-444-
88150-6, xx+256 pp., DA. 185.00.

The first author of the text, W.E. Baker, was a distinguished scholar in
the area of the science of explosions and a member of the editoral advisory
board. Unfortunately, he passed away as this book was being published.
His demise was a great loss to this Journal, as well as to the field of
explosion science.
The authors goal in writing the book was an attempt to gather and present
the available material for explosions in air of the non-ideal explosion sour-
ces most often encountered in industrial accidents. With this purpose they
discussed both theoretical and experimental approaches to these classes of
explosions with rather complete descriptions of the physical processes known
or believed to occur. Types of explosion sources discussed include: Dust ex-
plosions, (including grain elevator explosions), reactor gas explosions (includ-
ing unconfined vapor cloud explosions), hybrid explosions (such as may occur
when combustible dusts and gases are mixed with air in strong, vented or
unvented enclosures and then are ignited), non-reactive gas and flash evapo-
rating fluid explosions (often referred to as a single pressure burst), and finally
liquid propellant explosion.
Each chapter follows the general form when the authors discuss funda-
mentals (theory) scaling (or scale-up), testing methods, and control methods.
This reviewer is clearly not an expert in the field of explosion science. But
Baker was and this most excellent book is a testimony to his expertise.

GARY F. BENNETT

Surface Impoundments: Design, Construction and Operation, by R.P. Hartley,

Noyes Data Corp., Park Ridge, NJ, 1992, ISBN 0-8155-1302-X, 183 pp., $45.00.

This book was produced by the author, a member of the consulting firm of
P.E.I. Associates, for the U.S. Environmental Protection Agency. The goal of
the book is to provide up-to-date information on the design, construction,
and operation of surface impoundments used for the treatment, storage, or
disposal of hazardous and non-hazardous wastes.
The topic is an important one as in a 1983 study; the U.S. EPA identifies over
180,000 surface impoundments. The concern of course is their potential impact
on groundwater quality if they leak. Leak prevention is the key to their safe
operation and that is accomplished by preservation of the integrity of their
hydraulic barriers; Polymeric membranes and low permeability soil. To this
end, the Resource Conservation and Recovery Acts minimum technology
requirements specify double-lined structures with a leak collection and
removal layer.
memid Engineming Scirncr, Vol. 48, No. 3, pp. 629431.1992.
p~rgamon Rcsa Ltd. Printed In Great Britain.
Ibe Science of Crystallization. By W. A. TILLER. Cambridge
University Press, Cambridge, 1991, 391 pp., $69.50/27.95
On its back cover this book claims to be a complete
course . . . to understand . . , crystallization. And indeed, a
well accessible and elegant treatment of many complicated
subjects in crystal growth is given. But, of course, it cannot
be really complete.
The treatment of the equilibriwn and growth form is good
and several effects arc treated which are often neglected, like
surface stress, dislocations and reconstruction. What is
lackin& on the other hand, is some modern theory and
experiments on surface roughening (only mean field theory,
which is now known to give several incorrect predictions, is
treated in detail).
A nice discussion is devoted to the description of interface
shapes, e.g. the shape of the triple junction solid/liquid/vapor
in crystal pulling systems. Attention is paid to the effect of
anisotropy of the surface free energy as well as of the surface
kinetics. EspeciaIly, the discussion of an atomically smooth
surface orientation, like Si( 11 l), near interface junctions and
near the tip of dendrites is quite revealing.
Adsorption and surface diffusion are introduced in rela-
tion to thin-film growth and to impurity effects, but I have
missed a discussion of the Burton, Cabrera and Frank
surface diffusion model for step growth.
The stabilization at T = 0 K of different crystal and sur-
face structures by different interaction potentials is given, but
Dnst Exploaict~ in the Procerrs Indmt&% By RQLF K.
FXKHOFF. Butterworth Heinemann. 1991,590 pp., $75
This is a good book which will rightfully claim its place as a
classic treatise on dust explosions alongside the earlier works
of Palmer (1973) and Field (1982). It is comprehensive in
coverage, and well set out in its treatment of the principle
parameters of dust cloud formation, ignition and flame
propagation. Treatment of the subject does tend to ret&t
the approach of the researcher and academic, but I was
pleased to find that Prof. Eckhoff also laced his text with
much nseful advice. for plant designers and process engineers.
The book starts by giving an overview of the subject,
although in these first 150 pages the author is able to provide
a valuable insight into the main factors which inllued~e
ignitability and explosion violence, followed, importantly, by
a useful guide to prevention and mitigation and how these
might be achieved in practice. This is most helpful, and will
steer students, researchers and designers alike to the later
more detailed chapters on these issues.
As if to punch home the basic messages from the overview,
Eckhoff then follows with an interesting and well-chosen set
of case histories from several countries. These translate the
theory into real life situations, and so form a valuable
platform from which to launch the more detailed treatment
of the science in the chapters which follow.
629
simulations at T > 0 K with such potentials are left out. In
particular, the phenomenon of surface melting is not even
mentioned in the book.
Concerning impurity effects, there is a good and rather
complete discussion of adsorption and desorption and the
influence on the step velocity. The kinematic wave theory is
shown to give insight into the phenomenon of step bunching.
By now well-established results in the literature that this
theory cannot explain the firmntion of macrosteps, but
merely their shape, as well as several alternative models for
step bunching are, however, hardly mentioned.
S-a&in& this is a nice and useful book as an in-
troduction to the understanding of several phenomena re-
lated to crystal growth. The fact that it does not cover all
modem developments is not unexpected and probably un-
avoidable. Taking this point of view, I can advise this book
to scientists who intend to be alerted to general trends in
crystal growth, without being anxious to be well-informed
about all recent developments. A further positive point is
that the author has brought together experimental and
theoretical evidences in a book whose size is not excessive.
J. P. VAN DER EERDEN
Lab VCWT
Grensvlakken en Thermydynamica
P~~~~OUPI a
3584 CH Utrecht
l&e Netherlands
First is a welcome examination of the formation mech-
anisms of dust clouds, which takes the reader through the
relevant aspects of particle dynamics in air flows, and how
these are able to provide the concentration gradients so
necessary for flammable and explosive behaviour. This is
followed by a look at flame propagation, again fundamental
to those reaction mechanisms which control the generation
of disruptive energies in ignited dust clouds. In these
chapters Prof. Eckhoff both describes and explains the ex-
perimental work and the mathematical modelling needed in
its application, and I was pleased to see his critical remarks
about some of the current methodology in this area.
The nw steps from laboratory to plant are then
explored in more depth, particularly with respect to how
dusts may be ignited, and the parameters which influence
safe venting an& hence, its sizing. Most of this material is
well up-to-date, and indeed the author speculates where it
might lead in the future. This is important because it demon-
strates that our understanding of the science is still develop-
ing, via both laboratory-scale and pilot trials (an area where
my own organisation under Lunn is making an active
contribution using plant size equipment at Buxton).
Finally, there is a comprehensive examination of laborat-
ory methods for assessing dust ignitability, explosibility and
related properties, and again the authors critical views are
informative. This is especially helpful in an area where the
630 Book Reviews
variability of results can often say as much about the test
method as the parameter under evaluation.
All in all, a worthwhile and valuable book, which through-
out contains, what appear to be, exhaustive topic biblio-
graphies for those who want to delve deeper into any
particular aspect of the subject. Inevitably in an area as
involved as this, some material gets more comprehensive
treatment than others, but I was not left with the feeling that
anything significant to the science had been left out. How-
ever, one line could use.fully have been more deeply explored,
and this perhaps illustrates the difference between the
worldly academic and the practical engineer doing real
Liqoid-Vapor Pbaae-ehange Phenomena. By VAN P. CAREY.
Hemisphere Publishing Corporation, Washington,
Philadelphia, London, 1992, 645 pp., El07 (E41 paperback)
Professor Carey has written a substantial and wide-ranging
book on a topic that is so broad, and of which our under-
standing is still so incomplete, that compromises in coverage
are inevitable. The book is subtitled An Introduction to the
Thermophysics of Vaporization and Condensation Pro-
eesses in Heat Transfer Equipment. The first 200 pages
introduce the basic physics (thermodynamic equilibrium,
surface tension and wetting, interfacial boundary conditions
and thermal resistances, waves, homogeneous nucleation) to
be used in the following 400 pages on the modelling and
correlation of nucleate and film pool boiling, condensation
on plates and cylinders, two-phase flow in round tubes,
condensation and boiling inside tubes. This treatment is
mainly confined to pure, single-component fluids. A final
section on special topics provides brief discussions of flow
instabilities in tubes, boiling and condensation of binary
mixtures and in finned passages. Little is done to relate the
calculations of heat transfer and pressure drop to the design
of heat transfer equipment. The reader will not learn from
this book what industrial boilers and condensers look like
and may wonder why anyone cares about film boiling on
cylinders and spheres; convective two-phase heat transfer on
the shell side of tube bundles and practical problems of
corrosion and fouling are not discussed. I would have wel-
comed an introductory chapter to put the subject in its
engineering context.
The general plan to proceed from basic phenomena to the
modelling of heat transfer is to be applauded, although the
present incomplete understanding of multiphase heat trans-
fer means that it cannot always be carried through. Thus,
ultra-thin films (a welcome addition to the list of topics in
two-phase textbooks), interfacial resistances and Marangoni
flows all appear in the first part of the book but do not
contribute much to subsequent applications. In part, this is a
consequence of the emphasis on single-component fluids
with local equilibrium and consequently uniform state at the
liquid-vapour interface. The influence of an additional com-
ponent on the interfacial temperature slips is unheralded in a
section on the evaporation of a liquid tilm into an air stream.
Given the practical importance of mixtures of volatile fluids,
designs for industrial process plant. I refer to the dimensions
of cost and practicability of some of the solutions, since these
are as likely to inform the actual selection of methods and
plant as the sophisticated experimental data. A small criti-
cism of a major new text book perhaps, but one that is
probably realistic nevertheless.
I. G. JONES
Health & Safety Executive
St Annes House
Stanley Precinct
Bootle, Merseyside L20 3MF, U.K.
it would have been helpful to have in the first part of the
book a simple example, say of bubble growth in a uniformly
superheated mixture, illustrating the effect of mass dif-
fusional resistances.
The standard models for heat transfer across continuous
films of liquid or vapour are given in full. Idealized models
for more complicated situations like nucleate boiling and
critical heat flux are sketched out at intermediate levels
broadly representing the state of understanding 5-10 years
ago. For elaboration of these models readers are given
references to follow up and make their own judgments.
Fifteen per cent of the references are to publications in the
decade immediately preceding the publication of this book.
Some substantial bodies of work, e;g. in Germany on nu-
cleate boiling and in Japan on film boiling, are not re-
ferenced. Models often originate from photographic obser-
vations and students should be aware of the often significant
differences between primary data and model. The photo-
genic nature of two-phase flow is one of its attractions and it
is disappointing that the book contains only three photo-
graphs, two of which (showing static sessile drops with high
and low contact angles) were apparently chosen to avoid
over-exciting readers. Professor Carey draws attention to the
weaknesses and contradictions sometimes present in models.
He provides, by worked examples, useful guides to the
application of empirical correlations that must then supply
the answers required by designers.
The book is intended for graduate-level instruction and
could be supplemented by texts such as Smith, R. A., Popor-
isers (1986). Longman 8cimtifrc and Technical, to provide
the engineering context. Problems are provided at the end of
each chapter, in addition to the worked examples in the text.
Research students would find the book a reasonable starting
point but would have to do a lot of additional reading,
starting perhaps from the critical and topical reviews-in the
proceedings of the quadrennial International Heat Transfer
Conferences. For undergraduate instruction there are less
detailed but more readable alternatives.
D. B. R. KENNING
Department of Engineering Science
Oxford University
Parks Road
Oxford OX1 3PJ, U.K.
Safety Science, 15 (1992) 207-208 207
Elsevier

Book Reviews

Computer Systems for Occupational Safety and Health Management, by C.W.

Ross, 2nd Edition, Marcel Dekker, New York, 476 pages, $99.75 (USA and
Canada), $114.50 (all other countries).

This book consists of 203 pages of the authors text and 202 pages of ap-
pendices. The text is in three parts.
I Preparation and analysis: Where are you now?
II Developing a system
III System introduction and applications.
The parts, and the chapters within them, are poorly structured with little
logical flow. Computers are hardly mentioned and there are major omissions.
For example on the computing side there is no reference to microcomputers,
the growing range of software specifically designed for safety professionals or
the safety uses of computer communications. On the safety side, there is no
mention of, for example, safety auditing, chemical databases or inventory
software.
The appendices consist mainly of computer safety systems from a number
of companies. Page after page of coding schemes.
This second edition claims to be a revised and expanded version of that
published in 1984. However, there are no references later than 1976, apart from
some rewrites of the authors own articles first published in 1981.
There have been major developments in computing and safety thinking over
the last ten years, none of which is referred to in this book.
In 1984 this book would not have lived up to its title, in 1992, it would be
positively misleading for its target audience. In my view, the publishers should
withdraw it before it has the opportunity to mislead people into thinking that
this is what computing in health and safety consists of in 1992.

A.J. BOYLE
Birmingham, UK

Gas, Dust and Hybrid Explosions, by the late W.E. Baker and M.J. Tang, Fun-
damental Studies in Engineering, 13, Elsevier, 1991.256 pages, ISBN 0-444-
88150-6,25 tables, 151 figures, 256 literature references, Dfl. 185.00.

The book complements an earlier text by the late W.E. Baker, Explosions
in Air. It collects, for the first time in text book form, the journal articles de-
scribing the propagation of blast from accidental explosions with which Baker

0925-7535/92/$05.00 0 1992 Elsevier Science Publishers B.V. All rights reserved.

208

and the late R.E. Strehlow had association, together with related work by other
authors.
The start to each section, except the final one, is a study of scaling laws based
on dimensional analysis. For dust explosions 45 dimensionless groups are iden-
tified. Some conclusion is drawn relating to suitable scale factors for laboratory
experiments but the most general conclusion is that it would be a formidable
task to satisfy all the parameters in a single experiment. The authors compare
the comprehensive set of scaling parameters with the scaling rules that have
been found to be useful in analysing experiments and adopted for predicting
accident events.
One of the important problems with dust explosion has been to classify the
hazards of a flammable dust. The authors describe the test apparatus for meas-
uring the explosion and ignition characteristics of dust and the limitations of
each apparatus. The section concludes with a survey of explosions in grain silos
and means of mitigating the consequences.
The section on reactive gas explosions includes a review of the basic physics
of flammability, flames and explosions, including vented explosions and det-
onation. The authors proceed to survey the methods in use to predict the blast
effects of vapour cloud explosion, both deflagration and detonation, of approx-
imately hemispherical vapour clouds. The influence of obstacles on flame ac-
celeration is briefly considered. Data for predicting blast effects from detona-
tion of non-spherical (pancake) clouds is presented, but the reference to papers
dealing with deflagration of such flat clouds, or of cigar shaped clouds, is omitted.
The authors remark on the paucity of data on hybrid explosions which ex-
plains why they could manage only five pages of text plus tables and figures.
The inclusion of hybrid explosion in the title of the book is hardly warranted.
The section on non-reactive gas and flash-evaporating fluid explosions deals
almost exclusively with pressure vessel failure and the blast wave generated by
the expanding gases and rapidly evaporation liquids. Lacking is any descrip-
tion of the mechanisms for generating, by accident, the pressure needed to
burst the enclosure, such as run-away reaction, or the rapid mixing of a hot
liquid with a cold volatile liquid. Missing also is a description of rapid phase
transition explosions such as occur when molten metal is dropped into water,
which have caused destructive blast in foundries.
The section on liquid propellant explosions is well presented. The material
is all available in report form but this is probably the first time for it to appear
as part of a text book. There is much common ground for both these explosions
and vapour cloud explosions so that the latter gains important lessons from
the former.
In spite of the omissions noted above the book forms a useful introduction
to accidental industrial dust and gas explosions, especially to the experimental
basis for the prediction of blast effects. However, anyone looking for a com-
prehensive treatment would still need to search the literature in journals.

H. PHILLIPS
Buxton. UK
Powder Technology, 71 (1992) 119-134 119

Research on dust explosions at the University of Michigan

C. W. Kauffman, M. Sichel and P. Wolanski*

The University of Michigan, Ann Arbor, MI 48109-2140 (USA)

Abstract

Dust explosion research carried out at the University of Michigan during the last two decades has been summarized.
Significant results are presented on the smoldering combustion of dust heaps, turbulent combustion of premixed
dust clouds, entrainment and combustion of layered dust, and on shock wave ignition of particles and shock
wave initiated detonative combustion. Also, information on the detonation of hybrid mixtures and gaseous mixtures
containing nonreactive particles is given.

Introduction that detonative combustion of dust may occur relatively

The rapid development of powdered material and
products as applied to modem technology has signif-
icantly increased the explosion hazards in industry,
since many fine metallic and organic materials are very
explosive substances. The explosion of dust mixtures
is, however, not a new problem. It first emerged many
years ago when very violent dust explosions caused
significant human and proper losses in mines and in
the grain industry [l-4].
Combustible dust can react with air or with other
gaseous oxidizers in three different modes: during smol-
dering it can be slowly oxidized and propagate with
velocities of mm or cm per hour; it can bum in the
deflagration mode with velocity ranging from cm s-l
to hundreds of m s-l; or it can burn in the detonating
mode with a velocity of km s-l.
Each of these modes can be very dangerous and can
result in great human and property losses. A great
human and property loss due to smoldering combustion
of grain dust occurred in the huge grain elevator in
Kuybyshev, USSR, where smoldering combustion of
grain started in early 1988 and continued until 1990.
It caused several explosions, killing at least four people,
and created a very large environmental problem.
Deflagrative combustion of dust mixtures is the most
frequent cause of explosive damage in industry and
mining, so the greatest interest is always focused on
this mode of dust mixture combustion. However the
most violent combustion can occur during the detonative
mode of combustion [5-121. There is growing evidence
*Visiting research scientist. Permanent address: Warsaw Uni-
versity of Technology, 00-65 Warsaw, Poland.
easy, and simultaneously there is a growing number of
incidents of the reported detonative combustion of
dust-air mixtures [13-191.
In this paper we will present the contributions of
the University of Michigan to the investigations of these
problems.
Smoldering combustion of dust mixtures
The smoldering combustion of dust is a slow, non-
flaming, oxidation process limited by the rate of oxygen
diffusion through the dust. It can be initiated by a hot
surface beneath the dust layer, by a hot object entrained
within a dust layer, or by self ignition. Once initiated,
smoldering combustion can propagate through a dust
bed, and under favorable conditions it can change to
glowing, flaming, and even explosive combustion of the
dust.
Experiments on the smoldering combustion of a dust
layer carried out at the University of Michigan were
directed at evaluation of the critical conditions necessary
for ignition, studying the influence of convective flow
above the layer surface on the reaction front propa-
gation, and at the conditions for transition to glowing
and flaming.
Experiments were carried out in a specially arranged
wind tunnel, Fig. 1. The tunnel has a square cross-
section with a dimension of 0.61 m. The smoldering
box is attached to the tunnel bottom. Inside this box,
which has insulated walls, the dust is placed, and a
set of thermocouples and an ignition source are installed.
Layer density can be adjusted when the dust is placed
in the box. Flow velocity can be changed by adjusting

Q 1992 - Elsevier Sequoia. All rights reserved

120

Fan Plexiglas
Square Cross Section

Exhaust

I-
Combustion 20.3 cm
Box L
k 66.0 cm1
\l
P
0
II Inlet
It-----_ 243.8 cm ,-d

Air Velocity O-18 m/s

Fig. 1. The University of Michigan wind tunnel facility for the study of the smoldering combustion of dust.

the fan speed and by the setting of a damper. All 100 -

measurements are controlled and recorded by the data 90 -

acquisition system connected to a PC computer. More 60 -

details concerning this experimental set-up can be found p 70-

in [20]. % 60-
It was found that the ignition energy for dust in the ;: 50-
z 5.0
layer depends very much on the ignition source power w 40-
0
and location. Also the convective flow above the surface B 30-
layer influences the ignition. The typical dependence a 20-
of the minimum ignition energy on the specific surface
10 -
power of the ignition source is shown in Fig. 2. Ignition I I I I
0
of dust in the layer has a statistical nature. The prob- 0 10 20 30 40 50 60 70 60 90 100 110 120

ability of ignition thus depends on the specific surface TOTAL ENERGY INPUT (calories)

energy of the igniter and also on the total energy. This Fig. 3. Probability of the ignition of grain dust in a layer as a
is shown in Fig. 3. function of total energy input, for different specific surface power
Following ignition, the smoldering front propagates of the igniter [20].
through the dust bed. The smoldering velocity depends

very strongly on the convective flow direction and

velocity, and on the bed arrangement (dust packing
density, or layer porosity). In general, the air flow over
the layer surface increases the smoldering propagation
velocity in grain dust by 2 to 2.5 times. The typical
Ignition
propagation velocity for organic particles is in the order
P of 10-5-10-4 m s-l. Convective flow can increase the
propagation velocity by more than an order of mag-
nitude.
In some circumstances smoldering combustion of dust
can change to glowing or flaming. Glowing or flaming
occurs on the surface of the dust bed. Generally the
transition time to glowing and flaming increases with
increasing bed density, but it also depends on the igniter
Maximum Power/Area
for No Ignition location and the bed porosity (structure of bed material).
I
I Typical data for the transition time in bees wings (a
I
I material derived from corn) and in wood shavings is
I I I I I c
5.0 10.0 15.0 20.0 25.0 presented in Fig. 4. Closer location of the igniter to
POWER I AREA (Watts I cm2) the surface usually decreases the time necessary to
Fig. 2. The minimum ignition energy for a grain dust layer as obtain flaming combustion. The biggest influence, how-
a function of ignition source power [20]. ever, is from the convective flow. Relatively small velocity

6
I i ; 4 5 initial turbulence level can be very high. For these
v tm
Fig. 4. Transition from combustion to glowing for different
WI*
increases can lower the time necessary for the initiation
of flaming by more than an order of magnitude.
Smoldering combustion is controlled by heteroge-
neous oxidation on the particle surface, and by the
processes of heat and mass transfer through the bed.
The bed density (porosity) influences the heat and mass
transfer processes, while the specific reactive surface
is only one of the factors controlling the heat release.
The reaction wave spreads when the conductive and
convective heat losses are balanced by the heterogeneous
heat release. The conductive heat loss depends basically
on the bed properties, while convective losses depend
on bed porosity (density) and on the induced flow
velocity. In a large, well insulated bed (as for instance
in grain elevators or in a large coal pile), smoldering
combustion can be initiated due to a self ignition
processes, and it can propagate undetected for a long
time. It should be mentioned that smoldering com-
bustion can proceed even for a local oxygen concen-
tration in the bed as small as 1%.
It should be noted also that, under actual conditions,
the smoldering, glowing or flaming dust bed can rel-
atively easily result in an explosion. Weak blast waves
in an unconfined space can transform such a layer to
a large fire-ball, or in a confined space to a very
dangerous explosion.
Combustion of dust mixtures
Understanding of dust combustion is the key element
needed to understand the nature of dust explosions,
since the combustion rate is one of most important
controlling factors in all explosions. During the last
two decades, much fundamental work has been focused
on this problem.
It is most important to understand the influence of
turbulence intensity and, probably, scale on the flame
propagation velocity. For gaseous mixtures, the velocity
of turbulent combustion can be related to the laminar
burning velocity and the turbulence parameters. The
laminar burning velocity for gaseous mixtures can be
121
easily measured, so that the turbulent burning velocity
may be predicted. In contrast to gaseous flames, laminar
dust flames are practically nonexistent, since the pres-
ence of dust in the gaseous oxidizer (at normal gravity
conditions) will always result in the generation of some
turbulence (due to sedimentation of the dust). Tur-
bulence is thus an inherent property of dust mixtures.
In practice dust mixtures are created by the interaction
of a turbulent flow with dust, so in such cases the
reasons only so called quasi-laminar dust flames can
dusts be obtained. Even these flames can generate additional
turbulence due to the burning of nonuniformly dispersed
dust clouds [23].
It was recognized very early that high turbulence
intensity increases dramatically the rate of dust com-
bustion (explosion), but the early studies were carried
out only concerning the influence of unquantified tur-
bulence intensity on the rate of pressure rise [4], [24-271.
Such studies and others lead to the introduction of the
Cubic Law factor, Kst, which is by definition equal to
the measured maximum rate of pressure rise in the
confining vessel times the volume of the vessel raised
to the one third power, which relates the rate of pressure
increase to the arbitrarily chosen and unknown tur-
bulence intensity. Thus, Kst is an arbitrarily chosen
factor, and it depends very strongly on the initial and
boundary conditions. It is especially sensitive to the
shape and size of the explosion chamber, initial tur-
bulence (scale and intensity), uniformity or nonuni-
formity of the created mixture, and the type of ignition
source. For these reasons the Kst factor evaluated for
the same dust can vary very significantly if it is evaluated
in different test stands, in different laboratories, or
even by different researchers. A much better and more
objective comparison of the explosive severity of dust
mixtures can be made by the comparison of the flame
propagation velocity measured under controlled tur-
bulence intensity. For this reason a special jet-stirred
reactor was built at the University of Michigan to study
constant volume turbulent flame propagation [28].
A schematic diagram of the jet-stirred reactor is
shown in Fig. 5. The combustion vessel is a 1.22 m
diameter sphere, designed to withstand 10 atm pressure.
It is also possible to study in this vessel the combustion
of homogeneous gaseous mixtures, as for instance, a
methane-air mixture. The combustible mixture is sup-
plied to the chamber through six 1 inch diameter hoses
of equal length connected from the inlet flow divider.
These inlets are positioned symmetrically at the north
and south poles and at 90 along the equator in order
to create uniform isotropic turbulence within the sphere.
The outlet hoses are located in each of eight quadrants
of the sphere and are then manifolded into the exhaust
pipe. Two ball valves are located at the inlet and at
 I
Inlet Ports (6)

Dnmeter=1.22m ; Volumc=O.95 d

Fig. 5. Jet-stirred factor.

the exhaust manifolds. The dust mixture of required 9r

concentration and turbulence intensity is created by

8-
mixing the dust supplied from a dust feeder with the
inlet air. The selection of the feeding rate and air flow
7-
rate make it possible to control the dust concentration
and the turbulence intensity independently. The dust 6-
air mixture is passed through the sphere under steady 5
state conditions for a few seconds. Then both the inlet =5
I
and exhaust ball valves are closed and ignition is d
initiated. In most cases an exploding wire of 100 J 4-

energy with 7.8 g of gun powder located at the center x Cornstarch 4% u = 3.3
A Cornstarch 4% u = 1.5
is used as the ignition source. 3-
+ Cornstarch 1% u = 3.3

The dust concentration is measured by a light at-

2-
tenuation probe and turbulence intensity by both laser
and hot wire anemometers. Ionization probes allow II I I I I I I I
measurement of the flame propagation velocity, and a 200 300 400 500 600 700 800 900
photodiode is used to measure the flame thickness. A CONCENTRATION (g/ms)

pressure transducer is used to monitor the pressure
variation during the explosion, required to find the
combustion velocity and ultimately to normalize the
burning velocity to atmospheric conditions. A more
detailed description of the apparatus and experimental
techniques can be found in [28].
This experimental apparatus allows measurement of
the influence of turbulence intensity up to u -4.2 m
s-l on the explosion parameters. Much data for different
dust mixtures was obtained. Typical data obtained are
shown in Figs. 6 to 10. Increased turbulence has a
relatively small effect on the maximum explosion pres-
sure (Fig. 6), but a rather significant effect on the rate
of pressure rise (Fig. 7). Since the rate of pressure
rise depends directly on the combustion velocity there
is a similar influence here. In the sphere, the flame
spreading velocity is measured by the ionization probes.
This velocity is the result of the burning velocity and
of the convective flow velocity generated ahead of the
flame front by the expanding combustion products. So
to get the burning velocity the convective component
should be subtracted from the measured flame velocity.
Fig. 6. Effect of turbulence on P,,,, for different dusts, and
different moisture content of cornstarch dust.
To calculate the convective flow velocity, the flame
front position, the pressure, and the rate of pressure
rise should be known. Then the burning velocity of the
dust flame can be calculated [21, 29, 301. Such cal-
culations allow one to obtain the burning velocity at
the given pressure. Since the pressure in the sphere
is continuously changing, measurement of the burning
velocity at different pressures is necessary for the ex-
trapolation of results to atmospheric conditions. Fig.
8 shows the burning velocity corrected to atmospheric
pressure for different dusts and different turbulence
intensities.
From the measured burning velocities obtained at
different turbulence intensities, the laminar burning
velocity for dust mixtures can be obtained by extrap-
olation. Corrected laminar burning velocities for a few
different dust mixtures are presented in Fig. 9. Iaminar
or quasi-laminar burning velocities cannot be measured
directly as the dust would not remain in suspension
 123

r
96 36

x Cornstarch 4%

A Cornstarch 4%

0 I I I I I I I
2c)O 300 400 500 600 700 800 900
CONCENTRATION (g/m3)

Fig. 9. Corrected burning velocity for different dusts (cornstarch

0 I I I I I dust of different moisture content).
200 350 500 650 800 950
CONCENTRATION (glms)
250 -
Fig. 7. Effect of turbulence on (dpldt),, for different dusts, and
for different moisture content of cornstarch dust.
200 - (= oh
Concentration: 400 g/m3
14??
440
r z
3
4 150-
3
, +y4)
385 - 3 ~p~3t~ylene 3
3 Polyethylene &
u= 1.5
330 - 9 loo-
n Cornstarch 1%
u = 3.3 B 0 50 ps
b Cornstarch 1% 031 ps
: 275 - u=1.5 A 12.1 ps
50 -
+ Cornstarch 4% 0 7.4 ps
u = 3.3
Ll
3 Cornstarch 4%
u= 1.5
01 I I I I I
0 5 IO 15 20 25
165
POWER (watts)
Fig. 10. Minimum ignition energy for cornstarch as a function
of ignition source power for different turbulence intensities.

01 I I I I I
200 350 500 650 800 950
CONCENTRATION ( g/m3) Much data concerning the rate of pressure rise was
Fig. 8. The effect of turbulence on corrected burning velocity obtained during the last decade in the 20 1 sphere, so
for different dusts, and for cornstarch dust of different moisture it was of great interest to compare the turbulence
content. intensity and structure generated in the 950 1 sphere
to the turbulence generated in the small sphere [31].
under such conditions. For the cornstarch dust the It was found that the turbulence characteristics are
influence of moisture content on burning velocity can more constant in the larger sphere. In the 950 1 sphere,
be seen. It should be noted that the corrected laminar turbulence intensity in stationary flow is about twice
burning velocity obtained for different organic dusts as large as in nonstationary flow. After interruption of
are of the same order as the burning velocity of a the flow, the turbulence intensity decays relatively slowly
methane-air mixture. The thickness of the flame is during the first 0.5 s and then more rapidly while the
actually much bigger for a dust than for a gas. turbulence scale increases smoothly even up to 1 s
The turbulence also has a major influence on the after dust ignition. By contrast, dissipation processes
minimum ignition energy of dust mixtures. As expected, in the small sphere are much faster. Turbulence intensity
increasing turbulence intensity increases the minimum is highest immediately after dust dispersion and de-
energy level necessary for ignition. Figure 10 presents creases rapidly to less than 10% of its maximum value
data obtained for cornstarch at different ignition source in 15040 ms. It is obvious that the turbulence can
powers (exploding wire) at different turbulence inten- be more easily controlled with less transient effects in
sities. the big sphere.
124

,
-I
i

7 29
c
2.46 m -J

l-
x

I\\\\\\\\\\\\

Fig. 11. Flame Acceleration Tube (FAT): (1) main tube; (2) exhaust system; (3) initiator (auxiliary detonation tube); (4) dust
dispersion system at the beginning; (5) glow plug; (6) mylar diaphragm.

I: I
I
,,,,,,,,,,,,,,,,~,1,,,,,,,,,,~,,,,,,,,,,,,,,,,~,.,~,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,.,,,,~,~~~,~,~~~~.~~,~~~.~~~~~~~~~~

Time (1 msldtv)

Fig. 12. The effect of the passage of a M=1.4 shock wave on the suspension of a 4.0 mm layer of corn starch with shock wave
velocity and intensity indicated as well as reflected light intensity.

In the above experiments the generated turbulence The work carried out in this sphere is now focused
was of relatively small scale (the order of cm) and low not only on measurement of burning velocities, but
intensity, but in real industrial conditions much larger also on more detailed measurements of the turbulent
scale turbulence probably controls the flame propa- flame structure, since a better understanding of the
gation. For that reason explosion research must also mechanism of turbulent flame propagation in the dust
be carried out under conditions closer to those which mixture is essential to understand the nature of all
are encountered in industry. dust explosions.

Fig. 13. mica1 variation of dust concentration in the tube center
as a function of time.
=t! a00 400 600 .300
,
1000
Nominal Secondary Dust Concentration (g/m3)
Fig. 14. The effect of the layer thickness and grids on the
maximum flame velocity. Moist baghouse dust under 75 pm,
primary cont. = 1000 g mS3, initiator pressure = 1.013 X 16 N m-*;
0, thin layer for 24.39 m - with grids; X, thin layer for 24.39
m - without grids; +, medium layer for 36.58 m - with grids;
A, medium layer for 36.58 m - without grids; 0, thick layer
for 36.58 m - with grids; Cl, thick layer for 36.58 m - Without
grids.
Accelerating dust flames
Higher turbulence intensity can be obtained when
a flame accelerates in a tube. In many gaseous mixtures,
the flame accelerating in a tube can eventually change
to a detonation. It is thus of great interest to find how
the flame will accelerate, particularly to layered dust
flames in relatively long ducts.
Experiments on flame acceleration in long ducts have
been carried out for a long time, especially for different
coal dusts, but also for many other organic or inorganic
125
4 -I
-b 6 12 18 a4 30 36
Tube Location (ml
Fig. 15. Maximum explosion pressures for different dusts and
different layer thicknesses. Primary cont. =500 g mo3, secondary
cont. = 366 g mW3, initiator pressure = 1.013 x l@ N m-*, bead
length = 24.39 m; 0, dry Mira gel - thin layer; 0, wet Mira gel
- thin layer; A, dry Mira gel - thick layer; +, dry PFP -
thin layer; X, wet PFP - thin layer; 0, dry PFP - thick layer.
dusts. Such experiments range in scale from small
laboratory tubes to the big (real scale) surface or
underground mine galleries. These latter ones are re-
lated directly to explosion safety and are aimed at
finding effective countermeasures to prevent or suppress
uncontrolled explosions in mining and in industry. The
results of these studies can be found elsewhere [3-4].
For studies of flame acceleration resulting from lay-
ered combustible dusts, a special Flame Acceleration
Tube (FAT) was constructed at the University of Mich-
igan. The tube is of 0.305 m internal diameter and was
initially 36.6 m long, but was later extended to 70 m.
A schematic diagram of this tube is shown in Fig. 11
1321.
Dust can be deposited inside the tube in layers of
three different widths: 13 mm layer (called thick layer)
90 mm layer (called medium layer) and a thin layer
for which the dust is uniformly distributed inside the
tube using a centrifugal blower. The layers are of varying
thickness to give different dust concentrations. However,
the entrainment process is sensitive to both thickness
and width so the same concentration of layered dust
per unit length will not give the same concentration
of suspended dust. Ignition is provided by a small
auxiliary detonation tube filled with a hydrogen+xygen
mixture. Just in front of the initiator, a 3.66 m long
V-shaped dispersion system is placed, to provide dust
dispersion in the fist part of the tube prior to the
126
x lnitiai shock
//
0
I
0.5 1.0 1.5 2.0
J
2.5
tit,
Fig. 16. Typical pressure histories at the different tube locations,
and the flame front position for the case of flame acceleration
to a quasi-detonation in a medium corn dust layer with 1%
addition of Cabosil. Corn concentration in the layer: 500 g mV3,
layer length: 51.5 m.
initiation of the H,+ l/20, detonation in the auxiliary
detonation tube. The explosion of a hydrogen-oxygen
mixture provides the precursor shock which is respon-
sible for the slight initial dispersion of the dust layer
and the initiation of the primary dust explosion (com-
bustion of the dust predispersed in the first 3.66 m of
the tube). As a result of this primary explosion the
flame accelerates in the tube.
Dust dispersion is monitored by a concentration
probe, and a high speed camera can be used for the
qualitative observation of the variation of dust dispersion
with time. Direct pictures give general information about
the dust dispersion process while the optical probe
measures the dust concentration variation in the tube
center. Quantitatively information such as rate of en-
trainment may be obtained from backscattered and
transmitted laser light as layered dust behaves optically
differently from suspended dust [33]. The interaction
of a 4.0 mm dust layer with an M- 1.4 shock wave is
shown in Fig. 12. The relaxation time for the dust layer
after the passage of the pressure increase can be clearly
noted (tl) as well as the steady erosion of the dust
0
I
/
03 1.0 1.5 2.0
1
2.5
111,
Fig. 17. Mach number of flow at different tube locations and
the flame front trajectory for corn a dust layer explosion. The
same condition as in the previous figure.
from the layer as indicated by the decrease in the
intensity of the reflected (diRused) light (t2). From
Fig. 13 one can see that the precursor shock causes
initial suspension of the dust into the tube center. At
this stage the dust concentration in the tube center is
relatively low, but it increases dramatically just before
flame arrival, and then the dust concentration decreases
due to bum out of the dust particles.
Static and dynamic pressures and temperature are
measured at eight different locations along the tube.
Additionally the temperature of burning dust particles
is measured by a four wavelength pyrometer. At the
end of the tube a special exhaust section provides for
the ventilation of the explosion. The whole system is
controlled by an IBM 9000 microcomputer. The suitably
amplified outputs from all instruments are fed into a
32-channel analog-digital converter. After each run all
data is processed by the microcomputer. A detailed
description of the whole system can be found in [34-351.
In the early stage of research, the tube had a length
of 36.6 m. Some results obtained from these studies
are presented in Figs. 14 and 15. In Fig. 14 the maximum
flame velocity is presented as a function of the nominal
 127

convective flow ahead of the flame; dust can be carried

even a dozen meters ahead of the original end of the
dust layer. It can be seen that in some cases the highest
flame velocities are observed beyond the original end
of the dust layer. In the short tube the maximum
explosion pressures did not exceed 8 bars, but for the
70 m long tube the maximum explosion pressure as-
sociated with layered dust explosion exceeded 26 bars.
The maximum pressure in the FAT is always related
to the flame propagation velocity. This ratio is nearly
linear for deflagrative combustion and becomes highly
nonlinear when the flame accelerates to a quasi-det-
onative combustion process. For the latter case the
Secondary Concentration = 500 gm/m3 maximum pressure rise is observed in the region where
Initiator Pressure = 1.01 x 105 N/m*
the transition to quasi-detonation occurs, in the region
of the so-called explosion within the explosion.
Large 100 80 60 40 20 0
Small 0 20 40 60 80 100 Typical pressure profiles recorded at different lo-
Mass Percentage of Small and Large Corncobs cations and the flame front trajectory in the tube are
in Secondary Mixtures
shown in Fig. 16. It can be seen that the flame accelerates
(4 continuo&ly until it reaches quasi-detonation velocity.
Transition to quasi-detonation occurs between 45 m
Primary Dust: Corn Starch and 57 m, and then the flame propagates with a velocity
1.2 Primary Concentration = 500 gmlm3
Secondary Concentration = 500 gmlm3
of about 1200 m s-l. Gas velocity at different tube
Initiator Pressure = 1.Ol x 105 N/m2 locations is shown in Fig. 17. Just behind the shock
2 3.0 t A
wave generated by the initiator, the gas Mach number
is about 0.5, and it then decreases for some period.
Later, the influence of the flame generated flow can
be observed, and the gas velocity ahead of the flame
increases. Before transition to quasi-detonation the gas
velocity increases continuously, and some distance be-
hind the flame front a steep decrease can be found.
After transition to quasi-detonation a sharp velocity
increase is detected followed by a rapid flow reversal.
The reversed flow in the tube is not monitored due
to the omni directional measurement of the dynamic
pressure. The reverse flow is created by the shock wave
I I I I I I
Large 100 80 80 40 20 0 reflected from the tube elbow.
Small 0 20 40 60 80 100
e of Small and Large Corncobs
Quasi-detonation in the FAT up to now has been
Mass Percenta
in I econdary Mixtures observed for corn dust with the addition of Cabosil
@I j and for Mirage1 dust. Propagation velocities are in the
Fig. 18. Maximum velocity (a), and pressure (b), obtained for range of 1000-1200 m s-l, and the associated pressure
different bimodai particle percentage compositions. Corn starch in the shock front ranges from 9 bar to 12 bar. Much
dust concentration in the layer: 0.5 kg m-3.
higher pressures are observed in the transition region.
Continuous propagation of a quasi-detonation sup-
dust concentration in the tube for different arrange- ported by layered dust is not possible. For continuous
ments. A thin layer without grids results in the maximum propagation of a quasi-detonation or detonative com-
propagation velocity while for the medium and thick bustion the dust-air mixture should be created ahead
layer the grid almost always accelerates the flame of the leading detonation (quasi-detonation) wave. Ob-
propagation. Fig. 15 shows the maximum pressure as viously dust must be dispersed by a precursor shock.
a function of downstream location. For slowly accel- The precursor shock propagates, however, with a much
erating flames the maximum pressures remain nearly smaller velocity than the detonative (quasi-detonative)
constant along the whole tube length, but for a rapidly combustion, so when the detonative front reaches the
accelerating flame the maximum pressure is observed region of undisturbed mixture the detonation (quasi-
at the end of the dust layer or even beyond it. This detonation) will fail, and the shock wave will decouple
is due to the process of dispersion of the dust in the from the associated combustion wave. The flame will
128
* Baghouse
+ Bees wings
+ Mira Gel
1C
0 20 40 60
VOID LENGTH (tube diameters)
Fig. 19. Effect of voids on the tube exit pressure as a function
then slow down and the shock will again disperse dust
ahead of the flame. The flame will accelerate again
and the quasi-detonation (or detonation) from will be
reestablished. So in very long tubes one can expect
conditions similar to a galloping detonation.
Another interesting study carried out in the FAT
tube concerns the influence of a bimodal particle size
distribution and also the effect of a dust void on the
explosion deveIopment [36]. For the bimodal particle
experiments, small corn cob material with a diameter
of less than 75 pm and large particle corn cob material
with diameters between 425-800 pm were used. For
the total dust concentration of 0.5 kg me3 a different
ratio between the small and large particles was tested.
Also separate experiments for the small and for the
large particles were carried out. It was found that the
highest propagation velocity and maximum pressure
rise occur when the small particles comprise 60% of
the mixture (Fig. 18). However, the addition of Cabosil
to the mixture shifts the optimum value to 80% of the
small particles and significantly increases the maximum
explosion pressure. This can be explained by the in-
creased dispersive properties of the dust with the ad-
dition of the larger particles or the Cabosil.
60
of dimensionless void length (normalized to tube diameter).
The effect of a void on layered dust explosions has
been studied for three different kinds of dust: baghouse,
bees wings, and Miragel. Baghouse dust is the dust
literally removed from the baghouse of a dust collection
system at a grain elevator which handled only corn,
bees wings are a red colored flake type material which
comes from corn cobs, and Mirage1 is a commercial
starch product. Details are given in [38]. Despite the
differences in the explosibility of these dusts, the general
influence of the void on the pressure measured at the
end of the tube (exit pressure) was the same. It can
be seen in Fig. 19 that the exit pressure decreases
continuously as the void length increases. When the
void reaches approximately 65 tube diameters the ex-
plosion is no longer transmitted through the void area.
From this data it is difficult to conclude whether the
linear distance or number of tube diameters is important
in isolating the primary explosion from the secondary
explosion of the layered dust. It will probably be a
stronger function of the linear distance. The safe void
distance will depend on the time of burning of the
dust particles and the velocity of the induced flow. So,
if burning particles (or still sufficiently hot burned gases)
will reach the fresh dust dispersed from the next layer,
the newly dispersed cloud can be ignited and the flame
 129

6-

- 125
1 i

1+ : : : : : : : 4
.18 .19 .20 -21 .P .P .24 .a .a6

2 295/T

- 25 Fig. 21. Ignition delay time of dust particles as a function of

(reciprocal) temperature behind the incident shock wave. Lines
represent the mean square fit of the experimental ignition delay
data. Equivalent activation energy in kcai (g mol)-, 1, coal 15.3;
2, graphite 18.2; 3, Oats 18.4; 4, RDX-A* 10.5; 5, RDX-E* 16.4;
0 1 2 3 4 5
6, RDX-E 18.0; 7, diamond* 6.2; 8, coal 18.2; 9, graphite 19.8.
Normalized Time (t/t,) *implies addition of 10% Cabosil by weight. l-7 were obtained
(a) by inertial injector, 8, 9 were obtained by air injection.

Dry Mixed Dust Under 75 microns

Primary Concentration 500 g/m3
Secondary Concentration 366 g/m3
I\ Initiator Pressure 1 atm

u c
6.8 4
0.9 1.0 1.1 1.2 1.3
100WTEMPERATURE

Fig. 22. The variation of characteristic thermal and chemical

time of coal particles as a function of inverse particle surface
temperature. Coal for 53 pm particle, -, math No. 4.2; ---,
math No. 4.6.

portant and difhcult phenomena to model are the

I I I I I
0 1 2 3 4 5
Normalized Time (t/t,)
@)
Fig. 20. Variations of pressure and flame position at different
tube locations as a function of dimensionless time (normalized
to the time of the shock wave passage through the entire tube);
(a) numerical calculations; (b) experimental results.
will start to accelerate again. However, for a more
conclusive answer more studies are necessary.
Modeling layered dust explosions is a complex and
dithcult task, since it involves many nonlinear and
nonstationary phenomena. In this case the most im-
entrainment. of dust from a surface by the boundary
layer flow with the successive mixture formation, and
the coupling of the burning velocity with the intensity
and scale of the turbulence ahead of the flame. In the
tirst attempt to model such an explosion, the Chi and
Perlee model [37] was modified by Srinath [38]. By the
incorporation into this model of the submodel of dust
dispersion developed by Mirels [39] and the turbulence
combustion submodel of Clark and Smoot [40], nu-
merical calculations relating to layered grain dust ex-
plosions were possible. Calculations were done for a
grain dust size of 49 pm and a dust layer loading of
0.5 kg m-. The calculated pressure histories at different
130

down stream locations are given in Fig. 20(a), while

Glow plus ,
experimental results obtained for similar conditions are
Feeder
presented in Fig. 20(b). It can be seen that the initial Driver ,-
pressure rise and flame trajectory are comparable, but
Ball valve
the subsequent calculated pressure rise did not follow
/
the experimental results closely. This indicates that this
model should be improved, and additional work on the b
problem is in progress.
Rressure switches

Shock wave initiated dust combustion

Shock tube

A shock wave can be the source of ignition for a

dust particle or it can initiate detonative combustion
Shock tube
of dust mixtures. Studies of such processes has been cross section
carried out at the University of Michigan for a long
time, and there are many publications on this subject
[13, 14, 18, 41-471. Also theoretical investigations and
numerical calculations have been pursued for many
related problems, [43, 48-501. Recently a survey paper
on this subject matter was written [51], so only a brief
general description will be given here.
Ignition of dust particles has been studied experi-
mentally and theoretically at the University of Michigan
[13, 41-45, 49, 501. Experiments were conducted in
horizontal shock tubes, and the ignition process for
particles was analyzed analytically and numerically.
Ignition delay data for many different dusts are reported.
The most important findings from the experiments are
as follows. Many organic and metallic dusts can be
easily ignited by a shock wave, some even easier than
gaseous mixtures. The size and structure of the dust
particle affects the ignition strongly. There is an optimum
dust diameter for ignition. A theoretical analysis shows Fig. 23. Schematic diagram of the University of Michigan vertical
that physical considerations (particle heating) rather detonation tube.
than chemical considerations are the most important
elements controlling the process of particle ignition. behind the shock wave, a bow shock associated with
Typical experimental ignition delay data obtained in the particle is formed. The high temperature gases
the horizontal shock tube are shown in Fig. 21. It can between the bow shock and particle cause rapid heating
be seen that despite very big differences in dust particles, of the particle surface. Simultaneously the aerodynamic
the data obtained ditfer little and that the activation drag causes the particle to accelerate, thus reducing
energies are of the same order of magnitude. Theo- the gas temperature in the region between the bow
retically calculated thermal and chemical characteristic shock and the particle and subsequently the rate of
times for a coal particle subjected to a shock wave of particle convective heating. Ignition occurs when the
different Mach numbers are presented in Fig. 22. These surface temperature increases to the critical value nec-
and other calculations show that, for the conditions of essary for a fast chemical reaction on the surface which
shock wave ignition of a dust particle, the physical leads to ignition. The rate of particle heating depends
process of thermal heating of the particle is the con- on the particle size and properties. Very small particles
trolling factor in dust particle ignition, so that the global can be heated more rapidly than larger ones, but the
activation energy for ignition is basically very similar smaller particles rapidly accelerate to the convective
for most dusts. flow velocity, so that the high temperature stagnation
The relatively short ignition delay of many dust region will exist only for a short time. Larger particles
particles is due to the peculiar behavior of the dust accelerate more slowly but also absorb more heat prior
particles behind the incident shock wave. When a dust to ignition. So one is led to the conclusion that there
particle is subjected to the supersonic flow existing is an optimum particle diameter providing the most

(a) Oats/air V= 1550 m s-l, c-245+20 g mT3
(b) Soybean/enriched air V= 1650 m s-r, c=315 f25 g mm3
(c) Oats/oxygen V= 1680 m s-l, c=305 &-25 g me3
Fig. 24. Streak photographs and pressure traces of detonation
wave in organic dust mixtures: (a) oats dust/air mixture, oats
dust concentration: 0.245 kg rnv3, detonation velocity: 1550 m
s-t; (b) soybean dust/40%/60% OZ/Nzmixture, dust concentration:
0.315 kg mW3, detonation velocity: 1650 m s-r; (c) oat dust/
oxygen mixture, dust concentration: 0.305 kg rnb3, detonation
velocity: 1680 m s-l.
suitable conditions for ignition. Experimental data and
detailed calculations confirm these expectations.
Numerical calculations also allow examination of some
details of particle ignition which cannot be measured
experimentally. It was found that during the ignition
period only a very thin (a few microns) surface particle
layer is heated to high temperature and reacts with
the gaseous oxidizer. Below this layer the temperature
remains nearly constant during the ignition. From this
analysis it can be inferred that particles of irregular
shape and of a porous nature will be more suitable
for ignition than spherical or dense particles.
Ignition data obtained from shock tube experiments
are very useful in evaluating the detonability of a dust
mixture, since one finds that dust particles which are
more readily ignited by incident shock waves are more
likely to support detonative combustion.
When the shock wave interacts with a dust mixture
different modes of combustion can be initiated. The
131
combustion front and shock wave can be separated,
and the shock wave will decay to an acoustic wave and
dust will bum in the deflagrative mode. A detonation
wave may be created, i.e. a shock wave closely coupled
with combustion. A quasi-detonation can be initiated,
i.e. a combustion wave separated by a constant and
distinct distance from the shock wave and traveling
with velocity equal to the speed of sound for the
combustion products. Finally, after initial separation of
the combustion front from the shock wave, the flame
may accelerate to a detonation (or quasi-detonation).
The quasi-detonation mode of dust mixture com-
bustion was discussed in the previous section. It need
only be added that quasi-detonation can be obtained
by the process of self-acceleration (as in the FAT) or
by shock wave initiation [19].
At the University of Michigan the vertical detonation
tube was used for the direct initiation of detonation
in mixtures containing dust, gases, and combinations
thereof, A schematic diagram of this tube is shown in
Fig. 23. The main tube is isolated from the initiator
and the atmosphere by mylar diaphragms. Before an
experiment the main tube can be evacuated and filled
with the required mixture. The dust feeder supplies
dust at a controllable rate, to produce the required
dust concentration in the mixture. The initiator tube
is filled with a gaseous explosive, usually a hydrogen/
oxygen/helium mixture. The initial pressure of the
mixture can be varied to obtain different initiation
energies. When the mixture in the main tube has been
established the ball valve is closed to protect the feeder,
the initiator is activated, and a blast wave is generated.
The shock tube is equipped with pressure transducers,
pressure switches, photodiodes, a microthermocouple,
and a four colour pyrometer. The tube is also equipped
with optical quality windows for direct and Schlieren
visualization of the process. Detailed descriptions of
this tube can be found elsewhere [13, 14, 46, 471.
Many experiments on direct initiation of a detonation
in different organic dust mixtures were carried out.
Typical Schlieren photographs of a dust detonation
obtained in this tube are shown in Figs. 24, 25. In this
tube, a detonation in air was obtained only for oats
and wheat dust, but the addition of oxygen to air or
running the experiments in pure oxygen allowed many
more dust mixtures to be detonated. Detonability limits
for air mixtures are obviously very narrow and very
close to stoichiometric conditions. Oats dust will det-
onate in the concentration range 0.22-0.275f0.02 kg
mm3 and wheat dust in the concentration range
0.25-0.305 f 0.03 kg mm3. Detonability limits for oats
dust in air and in oxygen enriched air as a function
of the stoichiometric coefficient are shown in Fig. 26.
In air the detonation velocity of wheat or oats dusts
ranges from 1470 m s-l to 1540 m s-l, with an average
132

Shock wave Dust agglomerations

front scattered region
\ ,
, .~..,.~.,.
: f.
,::,:::
:. . .
. E
1
. :,. : ..,. ;....* ,
In
,.. _...._. . ,. ..:
...-* ::
. ..,:. .. ..I.:.
I... . .
. . . ,.
. .
...
. I:...
..; . ..
..
.,
d
Combustion1 Combustion
Combustion region
mixture j__ _ --- Droducts

@)

Fig. 25. Detonation of oats dust/air mixture, dust concentration: 0.25 kg m -; detonation velocity: 1540 m s-l; (a) instantaneous
laser schlieren photograph; (b) diagram of the picture.

condition existing between the gas and solid phase.

Calculations of the one dimensional structure of a dust
detonation, despite some simplifications, give very rea-
--de7S Y sonable agreement with the limited experimental data
- 75<d < 106
(Fig. 27) and also make it possible to reveal nomnea-
surable features of such a process. Typical numerical
results for the temperature variation inside the particles
OOV during ignition and combustion are shown in Fig. 28.
It can be seen that significant nonuniformity of tem-
Fig. 26. Detonability range for the oats dust with different oxidizers.
perature results during the ignition and initial stages
of particle burning.
value of 1500 m s-l. In oxygen, the maximum detonation Research on the addition of combustible gas to the
velocity observed for organic dusts was close to 1800 dust mixture was also carried out [14,44]. It was found
m s-l. In air mixtures the maximum pressure at the that the addition of a small amount of a combustible
detonation front can reach 30 bars, and in oxygen up gaseous mixture can significantly modify the detonation
to 50 bars. For the region where the transition from limits, and can even make it possible to detonate a
deflagration to detonation occurs, a maximum pressure
of 60 bars has been observed. The temperature of the
burning dust particles in the detonation front was up
to 2800 K in air, and slightly more than 3200 K in DP=so)y1
oxygen. Schlieren visualizations as well as pressure and
temperature measurements allow evaluation of the basic
features of the detonation wave structure. It was found
for the given conditions that the detonation front thick-
ness varied from 0.1 m for some oxygen mixtures to
0.8 m for some dust air detonations. The strong evidence
of transverse waves was already reported in earlier
studies [13].
Detailed numerical calculations reveal the structure
m l3PEmENTR
and nature of dust detonations [47-49]. It was found
that, just behind the shock, local oxidizer enrichment
0.27
occurs which facilitates particle ignition. This is due
INITIAL PRRTIC~~L0ADIN3~~G,~3)
to the very rapid compression of the gaseous phase
and the relatively slow process of particle acceleration. Fig. 27. Variations of the calculated ignition delay time, time
for particles to reach C-J conditions, and the detonation velocity,
Due to this the maximum shock pressure appears at as a function of dust concentration for oats dust/air mixture.
a distinct distance behind the shock wave front. These For comparison, the experimentally measured detonation velocity
peculiar features are due to the highly nonequilibrium is indicated.

T tps1
RpfI)
Y I R-16.6 / / / I
I t-15
R-24.6
L dust combustion depends on the dust arrangement in
t.00 0.20 0.40 0.60 0.60
RROIRL POSITION IN PRRTICLEk/R)
Fig. 28. Numerically calculated temperature variations within the
burning particles in the detonation reaction zone.
mixture which could not be detonated without the
added component. It was also found that the addition
of nonreactive particles to the gaseous mixture can
widen the lean detonability limit [14, 46, 471, but the
addition of inert particles to the gaseous mixtures always
decreases the detonation velocity. The influence of
nonreactive particles is always proportional to the spe-
cific volumetric particle surface area. A more detailed
description of such effects can be found in [47] or in
the recent survey paper on dust-related detonations
WI.
Despite significant progress made here and in other
laboratories, more experimental and theoretical re-
search is necessary for better understanding of dust
detonation phenomena.
Conclusions
At the University of Michigan a comprehensive re-
search effort concerning dust explosions has been carried
out for almost two decades. It covers the most important
aspects of dust explosions, ranging from dust ignition,
through smoldering combustion and turbulent com-
133
bustion, to detonative combustion. The influence of the
addition of a combustible gas to the dust mixture and
of inert particles to combustible gaseous mixtures as
well as their effect on detonative combustion has been
studied.
Many commonly used dusts can be ignited relatively
easily, can smolder and glow, bum with relatively high
velocity, and even detonate. It has been found that
there is an optimum size of particles suitable for shock
wave ignition, and that a bimodal mixture can burn
more violently than a mixture containing only small or
large particles.
The combustion velocity of dust mixtures depends
very much on the turbulence intensity and scale, so
better understanding of the nature of turbulent com-
bustion is essential for modelling all dust explosion
phenomena. Recent experiments allow the dust burning
velocity to be related to the turbulence intensity and
scale, but research on turbulent dust combustion at
high turbulence intensity and large scales is still essential
for better understanding of fast dust flames and even
quasi-detonation processes.
A very thin layer of dust can support explosive
t-5 combustion in tubes. Flame acceleration due to layered
Ft.:
the layer but even more strongly on the dust properties.
Very fine and reactive dust can produce overpressures
of up to 26 bars, and even transition to quasi-detonative
combustion is possible.
Quasi-detonative and detonative combustion of some
dust-air mixtures can propagate with a velocity ranging
from 1000 m s- 1 (for quasi-detonation) to nearly 1550
m s- (for detonation), and with the addition of oxygen
or in pure oxygen the propagation velocity can be even
higher. The maximum pressure rise across the deto-
nation front is usually much higher than in a deflagrative
explosion, but the transition to detonation creates a
higher pressure rise than the detonation itself.
In hybrid mixtures or in gaseous mixtures with the
addition of nonreactive particles a widening of the
detonability range, and a more rapid acceleration to
detonation can be observed. However, the addition of
inert or slowly reacting particles to gaseous mixtures
decreases the detonation velocity, and they can even-
tually quench propagation.
Research on dust explosions carried out at the Uni-
versity of Michigan has improved understanding of many
fundamental aspects of dust ignition, smoldering com-
bustion, turbulent combustion, layered dust combustion,
and detonations. Dust explosions are a very complex
problem, and not all aspects of their initiation and
propagation are completely understood, so further and
more detailed experimental and theoretical studies of
this problem are in progress in our laboratories.
134

References 25 J. Nagy, J. W. Corm and H. C. Verakis, Rep. of Investigation

7279, U.S. Department of the Interior, Bureau of Mines,
Washington, DC, 1969.
D. J. Price and H. H. Brown, Dust Explosions, National Fire 26 J. Nagy, E. C. Seiler, J. W. Conn and H. C. Verakis, Rep.
Protection Association, Boston, MA, 1922. of Investigation 7507, U.S. Department of the Interior, Bureau
K. N. Palmer, Dusts Explosions and Fires, Chapman and Hall, of Mines, Washington, DC, 1971.
London, 1973. 27 W. Buksowicz and P. Wolanski, Biul. Inf Tech. KGSP, 3-#,
W. Cybulski, Coal Dust Explosions and Their Suppression, U.S. (1980) 87 (in Polish).
Dept. of Commerce, National Technical Information Service, 28 F. I. Tezok, Ph.D. Thesis, Dept. of Aerospace Engineering,
Springfield, VA, Warsaw, 1975. the University of Michigan, Ann Arbor, MI, 1985.
W. Bartknecht, Explosions, Course, Prevention, Protection, 29 G. E. Andrews and D. Bradley, Cornbust. Flame, 19 (1972)
Springer-Verlag, Berlin, Heidelberg, New York, 1981. 133.
J. Nagy and H. C. Verakis, Development and Control of Dust 30 D. Bradley and A. Michelson, Combust. Flame, 26 (1976)
Explosion, Marcel Dekker, New York and Basle, 1983. 201.
M. Hertzberg and K. L. Cashdollar, Industrial Dust Explosions, 31 Y. C. Li, Y. K. Pu, C. W. Kauffman, L. P. Bernal and V.
ASTM STP 958, Philadelphia, 1987, pp. 5-32. G. Johnston, Prog. Astronaut. Aeronaut., 132 (1991) 107.
C. W. Kauifrnan, Fuel-Air E#osions, University of Waterloo 32 S. R. Srinath, C. W. Kauffman, J. A. Nicholls and M. Sichel,
Press, Waterloo, Ont., 1982, pp. 305-347. Industrial Dust Explosions, ASTM STP 958, Philadelphia, 1987,
C. W. Kauffman, Indusrrial Dust Explosions, ASTM STP 958, pp. 90-106.
Philadelphia, 1987, pp. 243-264. 33 C. Bai and C. W. Kauffman,Annual TechnicalMeeting, Eastern
C. W. Kauffman and J. A. Nicholls, Fuel-Air Explosions, States Section of the Combustion Institute, Cornell University,
University of Waterloo Press, Waterloo, Ont., 1982, pp. Ithaca, NY, November 1991.
623-643. 34 S. R. Srinath, C. W. Kauffman, J. A. Nicholls and M. Sichel,
10 P. Wolanski, Proc. Int. Symp. on Grain Dust, Division of Prog. Astronaut. Aeronaut., IO5 (1986) 196.
Continuing Education, Kansas State University, Manhattan, 35 C. W. Kauffman, S. R. Srinath, F. I. Tezok, J. A. Nicholls
KS, 1979, pp. 422-446. and M. Sichel, Twentieth Int. Symp. on Cornbust., The Com-
11 P. Wolanski, First Specialists Meeting (International) of the bustion Institute, Pittsburgh, 1985, pp. 1701-1708.
Combustion Institute, American Institute of Aeronautics and 36 C. W. Kauffian, Third Int. Colloq. on DustExplosions,Institute
Astronautics, Bordeaux, 1981, pp. 497-502. of Heat Engineering of the Warsaw Technical University,
12 P. Wolanski, Fuel-AirExplosions, University of Waterloo Press, Szczyrk, Poland, 1988.
Waterloo, Ont., 1982, pp. 979-981. 37 D. N. Chi and H. E. Perlee, RI 7908, U.S. Department of
13 C. W. Kauffman, P. Wolanski, E. Ural, J. A. Nicholls and the Interior, Bureau of Mines, Pittsburgh, PA, 1974.
R. Van Dyk, Proc. Int. Symp. on Grain Dust, Division of 38 S. R. Srinath, Ph.D. Thesis, Department of Aerospace En-
Continuing Education, Kansas State University, Manhattan, gineering, the University of Michigan, Ann Arbor, MI, 1985.
KS, 1979, pp. 164-190. 39 H. Mirels, AIAA I., 22 (1984) 1582.
14 C. W. Kauffman, P. Wolanski, A. Arisoy, P. R. Adams, B. 40 D. P. Clark and L. D. Smoot, Combust. Flame, 62 (1985)
N. Maker and J. A. Nicholls, Prog. Astronaut. Aeronaut., 94 255.
(1984) 221. 41 T. W. Fox, C. W. Rackett and J. A. Nicholls, Proc. 11th Int.
15 B. R. Gardener, R. I. Winter and M. J. Moore, Twenty-First Symp. on Shock Tubes and Waves, The University of Wash-
Int. Symp. on Combust., The Combustion Institute, Pittsburgh, ington Press, Seattle, WA, 1977, pp. 262-268.
PA, Munich, 1986. 42 E. Ural, Ph.D. Thesis, the University of Michigan, Ann Arbor,
16 P. Wolanski, Proc. Shenyang Int. Symp. on Dust Explosions, MI, 1981 (also available from Xerox Corp., Ann Arbor, MI).
Shenyang, 1987, pp. 568-598. 43 M. Sichel, S. W. Baek, C. W. Kauffman, B. Maker, J. A.
17 A. A. Borisov, E. D. Saneev, S. V. Khomik, I. B. Fomin, B. Nicholls and P. Wolanski, AL4A I., 23 (1985) 1374.
A. Khasainov and B. Veyssiere, Fourth Int. Colloq. on Dust 44 C. W. KautIinan, J. A. Nicholls and P. Wolanski, Second
Explosions, Porabka-Kozubnik, Poland, 1990. National Workshop on Explosibili~ of Industrial Dusts, Czes-
18 R. Klemens, V. Johnston, C.Alexander,Y. Li, C. W. Kauffman tochowa, Poland, 10-12 November 1980, also Biul. Inf Tech.
and M. Sichel, Fourth Int. Colloq. on Dust Explosions, Porabka- KGSP, 34 (1980) 12 (in Polish).
Kozubnik, Poland, 1990. 45 F. P. Lee, C. W. Kauffman, M. Sichel and J. A. Nicholls,
19 P. Wolanski, W. Sacha and M. Zalesinski, Fourth Int. Colloq. ALAA J., 24 (1986) 1811.
on Dust Explosions, Porabka-Kozubnik, Poland, 1990. 46 M. Wolinski, P. Wolanski, R. M. Myers, C. W. Kauffman,
20 S. 0. Leisch, Ph.D. Thesis, Dept. of Aerospace Engineering, M. Sichel and J. A. Nicholls, Ninth Int. Pyrotechnics Semin.,
the University of Michigan, Ann Arbor, MI, 1983 (also Colorado Springs, CO, 1984, pp. 861-872.
available from Xerox Corp., Ann Arbor, MI). 47 P. Wolanski, J. C. Liu, C. W. Kauffman, J. A. Nicholls and
21 C. W. Kauffman, S. R. Srinath, F. I. Tezok, A. Arisoy and M. Sichel, Arch. Combustionis, 8 (1988) 15.
S. 0. Leisch, First Int. Colloq. on Exposibility of Ind. Dusts, 48 P. Wolanski, D. Lee, M. Sichel, C. W. Kauffman and J. A.
Baranow, Poland, 1984, p. 23. Nicholls, Prog. Astronaut. Aeronaut., 94 (1984) 241.
22 J. Fangrat, C. W. Kaulfman and M. Sichel, Fourth Int. Colloq. 49 D. Lee and M. Sichel, Prog. Astronaut. Aeronaut., 106 (1986)
on Dust Explosions, Porabka-Kozubnik, Poland, 1990. 505.
23 R. Klemens and P. Wolanski, Prog. Astronaut. Aeronaut., 105 50 B. C. Fan and M. Sichel, Twenty-Second Int. Symp. on Com-
(1986) 169. The Combustion Institute, 1988, pp. 1741-1750.
bustion,
24 I. Hartmann, A. R. Cooper and M. Jacobson, Rep. of In- 51 C. W. KautIman, M. Sichel and P. Wolanski, in A. A. Borisov
vestigation 4725, U.S. Department of the Interior, Bureau of (ed.), Dynamic Structure of Detonations in Gaseous and Dis-
Mines, Washington, DC, 1951. persed Media, Kluwer Academic, 1991, pp. 267-311.
Powder Technology, 71 (1992) 189-196 189

Theoretical analysis of dust explosions

S. Nomura
KureResearch Laboratov, Babcock Hitachi K L Takaramachi 3-36, Kiue, Hiroshima (Japan)

and T. Tanaka
Professor Emeritus of Hokkaido University, Sappom 060 (Japan)

Abstract

This paper reviews the theoretical work on dust explosions developed by the authors. The bases of the work
are a simplified model of the dust cloud suspension, the flame propagation process and the combustion of solid
particles. As a result, characteristics of the phenomena are expressed by combining reaction kinetics, heat transfer
theory and thermodynamics. Further, theoretical information on some protection methods is derived and shown
to be safe. It thus appears that, although this treatment includes some severe assumptions, it could provide a
basis for more rigorous theories for prediction of dust explosion and design of protection systems.

Introduction unburnt particles suspended at equal intervals in the

Many solid materials are combustible and handled
in various processes in -many industries. In these pro-
cesses, extreme care must be taken for fine powders
against dust explosions. Dust explosions are phenomena
of ignition and flame propagation between solid particles
suspended in air. The ease of occurrence (ignitability)
and the severity of explosion (explosrbility) are rep-
resented by ignition temperature, minimum ignition
energy, explosible limits of dust concentration, flame
propagation velocity, pressure developments in closed
vessels and so on. These explosion characteristics depend
upon the kind of dust material, the state of dust clouds,
the atmospheric conditions and also the type of explosion
apparatus.
Although a number of experiments have been con-
ducted in a variety of explosion equipment from the
fundamental and practical viewpoints [l, 21, theoretical
analysis remains difhcult and incomplete. There are
unsolved problems, particularly in accurate prediction
of explosion characteristics and in reliable protection
against explosions in various situations. This is due to
the extreme complexity of the many physical and chem-
ical factors and of the mechanisms of ignition, flame
propagation and solid combustion.
Under such circumstances, the authors [3-91 have
proposed a simple dust cloud model.aimed at theoretical
predictions of dust explosions. The model envisages
monosized dust particles dispersed uniformly in air.
Flame propagation occurs by successive ignitions of
direction of flame propagation.
Consequently, characteristics of dust explosions have
been analyzed and shown to be in reasonable agreement
with experiment, e.g. ignition temperature, flame prop-
agation velocity, upper and lower explosible limits and
rates of pressure rise in closed and vented vessels.
Further, our theoretical approach has generated some
important information on inerting and relief venting.
This paper reviews our published studies on dust ex-
plosions.
Bases for the theoretical analysis
Lhst cloud model
The uniform dispersion postulated by the authors
is illustrated in Fig. 1. The explosion begins with ignition
of the central particle 1 and the flame propagates
radially with spherical symmetry. If flat flame propa-
gation in pipes or ducts is considered, a one-dimensional
configuration of dust particles is appropriate [S].
Consider spherical surfaces centered at the origin
(shown by the dotted lines in Fig. l), numbered n in
the radial direction where n= 1 defines the centre.
Particles are distributed on each spherical surface with
one particle at the centre. Distances between two
neighbouring surfaces and between the centre and the
second spherical surface are all L, i.e., L is the rep-
resentative distance between two neighbouring particles,
burnt and unburnt, in the radial direction in this
spherical configuration. Assume that particle 1 in the
centre occupies a spherical space with diameter L. Since

0032-5910/92/$5.00 8 1992 - Elsevier Sequoia. All rights reserved

190

/
/
I

,- -o--_, \\d where m, is the mass of a particle, C, is the particle

specific heat, h is the heat transfer coefficient, T, is
,
Q \\\,
0 -.nr2 \n=3 \
T,(nL, 8), the gas temperature at distance nL from
the centre at time 0, F is the view factor, a, is the

la
I ,I- \ \
I / L \ particle absorptivity, lt is the emissivity of the flame,
I
/ Rb u is the Stefan-Boltzmann constant and Tf is the flame
front temperature. Integrating the above equation under
(40
\
\ \ \
6
; the initial condition of T, = T, at 0= 0,

1 (6)
\ Y&~~~ o,/ ,/

tieTg(nL, e) de+ z (tie--l)+T,

n
\ l__* _</ ,,,/ /
\
\
\
\
,
1 where A = 6h/(p&D,,,) and B = 3Fa,~~T~/(p,C,,D,,).
Next, the thermal diffusion equation with respect to
Fig. 1. Dust cloud model.
the spherical coordinates is applied to the temperature
profile of the ambient gas, T,(r, 0).
its dust concentration is C,, the relation between L
and C, becomes
i!!&K($+;!y
L = (PJCdY~p0 (1)
where ps and D,, are the density and diameter of a where K is the thermal ditfusivity. For the boundary
particle. condition of Tg= Tf at I= b and the initial condition
The dust concentration in the outer region is also of T,= T,(r) at 0= 0, the solution of eqn. (7) becomes
equal to C,. Therefore, the number of particles on the
nth spherical surface, N(n), must be
N(n) = 24n2 - 48n + 26 (2)
where n = 2, 3 . .. and N(1) is unity. x 1e - (r-+/4Ke _ (r-243 +i)2/4Kepr I
(8)
e -

Flame propagation mechanism
Flame propagation from burning particles to the next
unburnt ones is assumed to occur when the unburnt
particles are heated by convection and radiation from
the flame and the particle temperature reaches the
ignition point.
Suppose the flame has just propagated to particles
on the (n-1)th spherical surface, whose radius is (n - l)L.
Then, the flame front radius b is assumed to be
b=(n-l)L+R,
where Rb is the flame radius inherent to a single burning
particle. For liquid droplets, Miesse [lo] proposed
where k* is a material constant ranging from 1 to 10
cm for liquid. Since no data for solids have been
available, similar values of k* are assumed for solid
particles.
For the temperature of unburnt particles in the nth
surface at time 0, T,, a heat balance is taken on the
particles:
=rDP2[h(Tg-T,)+
In the case of Ti(r) =Ti:,,, a constant temperature, the
above solution is reduced to
Tg(r9
6)=(Tr To) ; (b)e
rfc(s)+L
Substituting eqn. (8) or (9) into eqn. (6), the particle
temperature T. can be calculated.
Let At,, be the time when T, reaches Ts2, the ignition
temperature. At, is the time interval required for the
(3) flame to propagate from the (n - 1)th spherical surface
to the nth one. Figure 2 shows At,, decreasing with
increasing n to reach a constant value. The trend can
be explained as follows. While heating particles in the
nth spherical surface, particles and gas in the outer
(4) region are warmed as well. Then, the initial temperatures
of particles and gas, T, and Ti(r), increase with in-
creasing n, accelerating the flame propagation at the
earlier stage of the process. On the other hand, the
acceleration shortens the time to warm the particles
and the gas in the outer region, resulting in an asymptotic
variation of At,,. Denoting the asymptote value as At,
At is related with dust concentration, C, in Fig. 3.
The flame propagation velocity is calculated to be
LlAt,, or LIAt. The case of one-dimensional flame prop-
i Fa,epTp] agation has been dealt with in our paper [5] taking
(5) account of the gas expansion due to increase in number
 191

6-
F ,0 Nornuraand McCarthy(1988)
Sakai and Saiio (1933)

c
0.1 0.2 0.3 0.4 , 5
Volatile matter, C v [ - ]

tmd 100
n (-1
200
Fig. 4. Relation between burning constant
content where C& is ash content.

where r is the time needed for complete combustion

and volatile matter

Fig. 2. Variation of At. with n. of a particle and.is proportional to the square of the
\
initial diameter as
r=Kd0DP2 (12)
where &,, is the burning constant. According to Es-
senhigh and Fells [12], K.+, is about 2 000 s cms2 for
solid particles on the average. For coals, an empirical
relation between KdO and the volatile matter content
C, as shown in Fig. 4 may be used [13].
When the flame propagates to the nth spherical
surface at time t after ignition at the centre at t = 0,
the total mass of particles burnt, M(t), is obtained as
the sum of the mass of particles burnt in each spherical
surface.

i-l I II (13)
3

M(t)=m, i N(i) l- l- r-(i:l) ht

0.5 I [
cd (g/l) Substituting eqn. (2) into eqn. (13) and approximating
Fig. 3. Relation between asymptotic time interval b and dust gives
concentration.
Al(t) =M(t)/M,= (8m,/Mo)(t/At)3 (14)
of gas moles and heat release as a result of particle Then,
combustion.

Cqmbustion rate of dust particles

(15)
Assuming that the rate of combustion of a particle where M, is the maximum mass of particles burnt in
is controlled by diffusion of oxygen through the reacted a closed vessel and is proportional to the vessel volume,
layer, it is reported [ll] that DP( e), the particle diameter
V. as
at time 6 after ignition at 8-0, is given approximately
bY M,=aV, (16)
e
-= l_D%g@ where a is the proportionality constant depending upon
7
(10) the dust and oxygen concentrations.
PO

and the mass of a particle burnt by time 0, m(0), is

then Ignition temperature

m(6)=m,
[
l-
(
1
-$I (11) Ignition temperature is a measure of the ignition
sensitivity of a combustible dust. Cassel and Liebman
192

[14] measured the ignition temperature as the minimum I ,

air temperature for which the flame due to combustion P 0 experiment

- 1,200 -
of the dust particles was just observed. They also
p!
discussed theoretically the ignition temperature using a 1,100 -
the rates of heat generation on a particle surface and 2

of heat loss, denoted as Z(T,, T,) and Q(T,, T,), F&00-

E
respectively, i.e., TBz and T,, are obtained by solving
$ %x3-
the two relationships Z= Q and dZ/dT,= dQ/dT,, as s
shown schematically in Fig. 5. Then, Tgz is the ignition Z
F 800-
temperature defined by these conditions. 0,
In the above derivation, T,, could be an ignition 0
point for the particle temperature as T, increases slowly
until it reaches T,, and then rises rapidly as the particle Fig. 6. Ignition temperature vs. particle size for various residence
bums (Fig. 5(a)). If particles can only stay for a limited times tI of coal particles.
time period in a hot environment, air temperatures
greater than TBzwould be required to ignite the particles Explosible of concentration
(or to raise the particle temperature to TJz within the
available time). Therefore, it is more general to define of
particle a particles by
the ignition temperature as the minimum air temper- dust only when definite
ature to raise the particle temperature to the solid is That in for flame propagate
based ignition point, Tsz [9]. This ignition temperature a cloud, dust must
is a function of the residence time of particles as well within explosible bounded upper
as the material properties. The case of an infinite lower These limits defined,
residence time is equivalent to that defined by Cassel as minimum maximum concentrations
and Liebman [14]. sustain propagation the cloud.
The effect of residence time on the ignition tem- the point view, powders
perature is shown in Fig. 6. For infinite residence time, be in within region
the theoretical ignition temperature decreases mono- these limits.
tonically with increasing particle size whereas a concave
curve is obtained for a limited residence time (e.g. Lower explosible limit
0.065 s). These trends were also observed experimen- Should ignition initiate from particle 1 in Fig. 1,
flame propagation to the next particle 2 at a distance
tally.
L occurs when the temperature of particle 2 rises to
its ignition point before particle 1 bums out. The lower
explosible limit is defined as the dust concentration
calculated from eqn. (1) by substituting a maximal
distance, L,,, which can be obtained from the condition
that the temperature of particle 2 reaches its ignition
point when particle 1 is just burnt out.
Tgl (Non Ignition)
Equations (6), (9) and (12) are used to determine
L IllaX,that is, T,(T)= T,, at r=L,_ and b =Rb in eqn.
t,. tr t1, time (9). Figure 7 shows the lower limit in relation to particle
a) Particle temperature - Time History size where the full lines are the calculations for several
k* values, ask* for these dusts are not clearly estimated.

Upper eqdosible limit

Near the upper limit, conditions are fuel-rich or
oxygen-lean. Under these conditions, flame propagation
is strongly influenced by the diffusion of oxygen. In
other words, propagation will cease if there is insufficient
oxygen supplied to burning particles or the flame front.
For simplicity, the total ambient oxygen in a dust cloud
b) Cassel and Liebmans Method
system is equally allocated to all of the particles in the
Fig. 5. Definition of ignition temperature. system and this is given by

urea rain vinyl butyml
0 xl0
D1A.h 1
Fig. 7. Lower explosible limits of dust concentration.
No2 = m,CglCd (17)
where No2 is the number of moles of oxygen allocated
to a single particle and C, is the oxygen concentration.
Assume that burning particle 1 consumes only the
amount of oxygen given by the above equation and
that the diffusion of oxygen allocated to other particles
is not taken into account. The flame propagation from
particle 1 to the neighboring particles should occur
before the quantity of oxygen allocated to particle 1
is entirely depleted. Then, the mass of particle 1 burnt
by consuming N,, moles of oxygen, ml is
m, = N,,M,IN, (18)
where MP is the molecular weight of the particles and
N1 is the number of moles of oxygen required for the
combustion of one mole of particle.
Combining eqns. (ll), (17) and (18), the time required
to exhaust N,, moles of oxygen by combustion of one
particle, 6,, can be obtained as
(19)
Thus, the upper limit at a given oxygen concentration
is calculated as a maximal dust concentration at which
T, at 8= 6, in eqn. (6) is equal to T,,, the ignition
point.
Although no exact measurements of the upper limit
for atmospheric oxygen content were made, values for
rather dilute oxygen contents have been obtained [16,
171 and are shown by the dotted lines in Fig. 8 where
the full lines are the calculated results. Oxygen contents
less than about 10% would support no flame propagation
and such critical oxygen contents should be the basis
for inerting.
Pressure development in closed and vented vessels
One of the important explosion characteristics is the
rate of pressure rise in a closed vessel. In particular,
193
0 500
0IA.b)
Aluminum
Terephthalic acid
DP = 40 Ir
Oxygencont. [ %I
Fig. 8. Upper and lower explosible limits of dust concentration
as functions of oxygen concentration.
the maximum rate of pressure rise provides information
for the safe design of relief vents.
Experimentally, using spherical vessels, Bartknecht
[18] obtained a scale-up relation between the maximum
rate of pressure rise [dpldt],, and the vessel volume
V,, called the cubic law as follows;
J/l
dF
-
0 =K3Z (20)
[
dt 1max
where P=P/P,,, the dimensionless pressure, PO is the
initial pressure and KGc is a constant. However, no
theoretical background was proposed for this empirical
law. In addition, for non-spherical vessels, no reliable
scale-up laws were found relating the vessel configu-
ration to the maximum rate of pressure rise and other
variables concerned.
In vented explosions, most of the data obtained were
plotted against the vent ratio, i.e. the area of vent per
unit volume of vessel. However, it is quite doubtful
that such a dimensional parameter can provide a basis
for designing vent areas required for various scales of
equipments.
This section deals with pressure development in closed
and vented vessels in relation to vessel scale.
194

Pressure development in closed vessels Figure 10 shows the comparison of the theory with
Adiabatic change is assumed in a closed vessel of the experimental data of Ishihama and Enomoto [19].
volume V,. After ignition at the vessel centre, the The predicted maximum pressures and rates of pressure
pressure rises due to the gas mole increase and the rise are greater than the experiments, as the theory is
heat released by combustion. The rate of pressure rise based upon the conservative assumption of adiabatic
at time t follows as change.
dP = ypl
- --l/yu(~mllw _ 1) Wt)
_
(21)
dt dt vessel
[ 1
configurations than sphere assumed
where F,,, = P,,/P,, with P,, =RT,-&n,IV,), TGOis the be as in 11 (Y the
maximum gas temperature, n, is the resultant total of length the D, aspect
number of gas moles and 3/uis the ratio of the specific Starting ignition a in vessel
heats of the gas. The above equation can be integrated at t=O, the flame propagates spherically until
to give t = tl, the time for the flame to reach the vessel side
1?= [(Imlm- 1)&f(t) + 117 wall, given by
(22)

t1= $t (25)
Spherical vessel
The relationship between P and t for spherical vessels
calculated from eqn. (22) using eqn. (14) is shown in The flame front at t > tl is assumed to expand spherically
Fig. 9. The maximum rate of pressure rise is obtained but to be partially cut off by the wall. This affects the
total mass of particles burnt at time t when t > tl. Thus,
at point (Pmax, t o) where to is the time when the total
mass of particles burnt it$to) becomes unity. From eqn. the change of pressure with time becomes as shown
in Fig. 12. The maximum rate of pressure rise in this
(14)P
case is obtained at point (Pl, tI). P, =P1/Po, the di-
mensionless pressure, is
(23)
Pl = [(&lYU- l)N+ 11% (26)
Thus, eqn. (21) with the aid of eqns. (14), (15), (22)
and (23) leads to the following important relation as
Starch dust

This equation proves the validity of the empirical cubic

law given by eqn. (20), where Koo is represented by
the right hand side of eqn. (24) and is constant for a
fixed dust concentration and particle size.

Fig. 10. Comparison of predicted maximum rate of pressure rise

with experiment for starch dust.

IO
Time (set)
Fig. 9. Variation of pressure with time in a closed spherical
vessel. Fig. 11. Cylindrical vessel.
 195

(VO= 1.2 1, [dpldt],, =1250atms-1)atC,=0.3g1-1.

&

I
__________ _________--

A---
The experimental value for Km/KG,-, is
K&KG,, = (5 000/1.2)1 (155/l 250) = 2.0
while our calculation using eqn. (28) gives
KGCIKGD
= 2.6
where N= (2 x 0.3)/(3 X 4.36 X 0.126) = 0.364 and F,,, =
10.78 are used. Thus, experiment and theory are actually
in reasonable agreement.

Vented dust explosions

time
Suppose the dust explosion occurs in a vented vessel
Fig. 12. Variation of pressure with time in a closed cylindrical
whose volume and vent area are V, and S. The
vessel.
macroscopic energy balance is applied to the system
and gives the following approximate equation for the
Cellulose acetate Cm&arch I
rate of pressure rise;
0 o( c4.36 - 0 01 z4.36

dF = rpl
_ - l/m(~m~/Yu- 1) Wt)
-

[
dt_($p-l/wpyu~)
+(:)I
1

(29)

The first term of the right hand side contributes to

cd WI)
Fig. 13. Comparison of predicted and experimental variation of
KGCJKGDwith dust concentration.
where N=2C,/(3azz). Similar to eqn. (24), with the aid
of eqns. (14), (XT), (25) and (26), eqn. (21) is reduced
to
xN[(l -&,-lvu)
xN+&l/Y]m-&KGD
Again, the cubic law is found to be applicable to
cylindrical vessels and KG,, becomes constant for given
material, state of dust cloud, and vessel configuration
represented by LYappearing in N.
Further, it is possible to convert data from spherical
apparatus to cylinders of any size and vice versa by
using the ratio of KGc to KG,, as
the pressure rise due to the combustion with the same
form as that of eqn. (21). The second term represents
cd -(ill) the reduction of the developed pressure by discharge
of burnt and unburnt gases through the vent, in which
the subscripts b and u denote burnt and unburnt,
respectively.
When the pressure becomes maximum in the vented
explosion, denoted as P,, the right hand side of eqn.
(29) is zero. Further, the most dangerous situation is
taken into account. The first term of the right hand
side of eqn. (29) could be replaced by the maximum
rate of pressure rise in a closed vessel as it has the
same form as for the closed case. Also, the second
term of eqn. (29) is approximated. Then,
(27)
(30)
where f(r,J and g(p) are given by
for P < 110.53,
l/2

KGCIKGD=N-2[(1_~~-ly)N+~,-1/~]1-- (28)
f(ru)=
@5) =p
h
(2 1

- l/Yp - lh _ 1)lR
Calculated KGC/KGDvalues are compared with exper-
iments of Nagy et al. [20] in Fig. 13 for two values of and for Is > 110.53,
cu, 4.36 and 1.55. As far as Fig. 13 is concerned, the
effect of configuration is less than expected. However,
Heimich and Kowall [21] tried unsuccessfully to
correlate data for a spherical vessel (VO= 5 000 1,
[dP/&],,=155 atm s-l) with results for a cylinder
196
0 Organic pigment
x Tris-2-hydroxyethyl
lsocynurate
Fig. 14. Comparison of predicted S/Vom with experiment.
Substitution of the cubic law into eqn. (30) leads to
(31)
where KG is KGc or KG,, depending upon the type of
vessel.
Thus, a new dimensionless scale up factor is proposed
for designing the area ofvent, instead of the conventional
vent ratio. S/Yom is estimated from KG value for a
given dust and P,, the predetermined maximum pressure
related directly to the strength of the plant.
With respect to vented explosions, many experiments
were carried out with various scales of test apparatus
[22-241. A set of the reported P, and KG values are
substituted into eqn. (31) to obtain the theoretical S/
V, which is plotted against the experimental SwO=
in Fig. 14. The calculated values are a little above the
corresponding straight line as the theory takes account
of the safe design in eqn. (30). In this sense, the overall
agreement with experiment is satisfactory not only in
the order of magnitude but also in the trend of the
straight line.
Conclusions
For such complicated phenomena as dust explosions,
our treatment seems to be rather idealized, simplifying
the state of dust clouds and the mechanism of flame
propagation as well as including some severe assump-
tions. Nevertheless the derived results are believed to
be quite acceptable as a first approximation. The overall
characteristics have been elucidated as a combination
of physical and chemical theories (reaction kinetics,
heat transfer theory and thermodynamics), environ-
mental conditions and material properties. Further, the
work has dealt with some protection methods and
provided important safety information.
It is also noted, however, that complete confirmation
of the theory against experiment has not been made,
because the present treatment includes some unknown
parameters or because of difficulties in obtaining rep-
resentative data. The extent to which these theoretical
and experimental problems should be solved depends
upon the accuracy required in future work. On the
whole, our approach could be a useful guide or the
first step to establishing sound theoretical bases for
the prediction of dust explosions.
References
1 K. N. Palmer, Dust Explosion and Fire, Chapman and Hall,
London, 1973.
2 D. F. Aldis and F. S. Lai, U.S. Dept. Agriculture, Misc. Pub.
No. 1375, 1979, prepared by Sci.&Edu. Administration.
3 R. Mitsui and T. Tanaka, Z&EC l+oc. Des. Dev., 12 (1973)
384.
4 S. Nomura and T. Tanaka, ffigaku Kogaku Ronbunshu, 4
(1978) 634.
5 S. Nomura and T. Tanaka, Kagaku Kogaku Ronbunshy 5
(1979) 47.
6 S. Nomura and T. Tanaka, I&EC Proc. Des. Dev., 19 (1980)
451.
7 S. Nomura and T. Tanaka, J. Chem. Eng. Jpn., 23 (1980)
309.
8 S. Nomura and T. Tanaka, Z&EC Proc. Des. Dev., 23 (1984)
420.
9 S. Nomura and T. G. Callcott, Powder Technol., 45 (1986)
145.
10 C. C. Miesse, 6th Symp. (Int.) Cornbust., New Haven, Cr,
1957, p. 732.
11 S. Yagi and D. Kunii, 5th Symp. (Znt.) Cornbust., Pittsburgh,
PA, 1955, p. 231.
12 R. H. Essenhigh and I. Fells, Dircuss Faraday Sot., 30 (1960)
208.
13 S. Nomura, Trans. Iron Steel Inst. Jpn., 27 (1987) 552.
14 H. M. Cassel and I. Liebman, Combust. Flame, 3 (1959) 467.
15 I. Hartmann, M. Jacobson and R. P. Williams, Invest. No.
5052, U.S. Bur. Mines Rep., 1954.
16 A. D. Craven and M. G. Foster, Combust. Flame, 11 (1967)
408.
17 W. Ishihama and H. Enomoto, Z5th Symp. (Znt.) Combust.,
Tokyo, 1974, p. 479.
18 W. Bartknecht, mZ Ber., I65 (1971) 24.
19 W. Ishihama and H. Enomoto, Anzen Kogaky 14 (1975) 243.
20 J. Nagy, E. C. Seiler, J. W. Corm and H. C. Verakis, Invest.
No. 7507, U;S. Bur. Mines Rep., 1971.
21 I-X.J. Heim-ich and R. Kowall, M>I Ber., 165 (1971) 53.
22 I. Hartmann and J. Nagy, Ind Eng. Chem., 49 (1957) 1743.
23 C. Donat, Loss Rev., II (1977) 87.
24 R. F. Schwab and D. F. Othmer, Chem. Process. Eng., April
(1964) 165.
520 C O M B U S T I O N A N D F L A M E 85:520-522 (1991)

BRIEF C O M M U N I C A T I O N
Effectiveness of Dust Explosion Suppression
by Carbonates and Phosphates
J. A M R O G O W I C Z and W. K O R D Y L E W S K I
Wroclaw Technical University, Institute of Power Engineering and Fluid Mechanics, Wroclaw, Poland

The mass concentrations of three inhibitors required to suppress dust explosion were measured in laboratory-scale
systems for four different dusts. The additives studied were commercially available--phosphate, bicarbonate, and
an inert powder. Two testing procedures were used: the LEL measuring method and the method of direct injection
of the inhibitor into the vessel. The resulting effectiveness of the inhibitors appeared to be sensitive to the method
of testing. The order of effectiveness for the LEL method was phosphate > bicarbonate > CaCO 3. For the method
of direct injection, it was bicarbonate > phosphate > CaCO 3.

The efficiency of three extinguishing powders
based on bicarbonate, phosphate, and C a C O 3 in
dust explosion suppression is compared. The early
(1957-1973) researches on the effectiveness of
extinguishing powders established the ranking of
effectiveness in which the phosphate was far be-
hind bicarbonates [1]. However, in 1984,
Hertzberg et al. [2] showed the ranking almost
reversed for suppression of methane and coal dust
explosions. These observations also were con-
firmed by Choi et al. [3] for the extinguishing of
coal flames. The high efficiency of the phosphate
as an explosion inhibitor is difficult to explain but
may result from the testing method in which the
lower explosion limit (LEL) of a powder-dust
mixture was measured [2]. These conditions are
different from those in which the explosion sup-
pression system operates [4]. Here a comparison
has been drawn between the inhibitor efficiencies
measured by the LEL method and the method of
direct injection of inhibitors by a miniature sup-
pression system.
EXPERIMENTAL DETAILS
The combustion vessel was a 22-L sphere used in
investigation of explosion parameters of dusts
[5]. Two types of experiments were performed:
1. measurements of the LEL against the extin-
guishing powder content in the powder-dust
mixture; and
2. measurements of the dust explosion overpres-
sures against the amount of extinguishing
Copyright 1991 by The Combustion Institute
Published by Elsevier Science Publishing Co., Inc.
powder injected into the combustion vessel
during explosion.
In experiment 1 the powder-dust mixture was
prepared and the standard procedure of LEL
evaluation was adopted [6]. In experiment 2 the
extinguishing powder was put into a 140-mm-
long, 16-mm-i.d. tube (Fig. 1). One end was
attached to the explosion sphere and had a nozzle.
At the other end was a 1.5-kJ chemical match that
was activated by the ignition delay unit. The
delay time was selected to give an explosion
pressure of 0.05 MPa in the sphere when the
extinguishing powder was dispersed from the tube
by combustion gases. In both experiments an
initiation of ignition was performed by the 1.5-1d
chemical match, and the explosion pressure was
recorded by a Kistler 601A transducer.
The following four explosion dusts were used:
melamine;
wood dust;
wheat flour;
coal dust So~nica.
All dusts wer sieved on a 80-#m mesh. The dusts
concentration was 0.5 kg/m 3 in experiment 2.
Commercially available extinguishing powders
representing phosphates, bicarbonates, and inert
inhibitors were investigated and are characterized
in Table 1. The results of the experiments are
presented against the inhibitor content in the sum-
0010-2180/91/$3.50
DUST EXPLOSION SUPPRESSION 521

(a) (b)
6OO
5OO
"~400
'-'~-'300
tu
..a 200
100
0
POWOER (c) (d),
6OO
500
~'~E~,~Oo
~ 300
200
Fig. 1. Installation for direct suppressant injection. 100

mary weight o f the inhibitor and dust introduced
into the vessel.
RESULTS
Figure 2 presents the L E L o f the p o w d e r - d u s t
mixture as a function o f the extinguishing powder
content in the p o w d e r - d u s t mixture. These mea-
surements have confirmed the effectiveness order
predicted earlier [2]. The most effective extin-
guishing powder in the suppression o f dust explo-
sions appeared to be the suppressant based on
N H 4 H 2 P O 4 (LB-3). Some o f these results should
be compared with the data in [2, 7] because the
size distributions for the present Sonica coal dust
and for the Pittsburgh bituminous coal Pgh-1
were similar. The inerting levels o f N H 4 H 2 P O 4
and C a C O 3 for coal dust were 1.5 times lower
than the corresponding levels found by Hertzberg
et al. [2]. Perhaps, these differences result from
the fact that So~nica coal was less explosive (28%
o
Fig. 2. Effect of added inhibitors on the flammability limits:
(a) melamine; (b) wood dust; (c) wheat flour; and (d) coal.
Vq, CaCO; A, LB-2, O, LB-3.
volatility, 11% ash, Q = 6400 k c a l / k g ) and a
weaker ignition source was used.
Figure 3 presents the L E L o f the explosion
overpressure as a function o f the content o f extin-
guishing powder which was directly injected into
the vessel during explosion. These measurements
contradict the results from experiment 1. O f the
four dusts investigated the most effective suppres-
sant appeared to be the extinguishing powder
base on N a H C O 3 - L B - 2 at the low and medium
( < 50%) content.of the inhibitor. However, for
higher inhibitor concentration the phosphate and
bicarbonate curves converge towards one an-
other.

TABLE 1

Characterization of Inhibitors.
Particle Size Bulk Density Surface Area Temperature of
Inhibitor Components Distribution [kg/m3] [cm2/g] Decomposition K
LB-2 NaHCO3 - 95% below 10#m - 12% 1000 4000 543
+talc + 10-60~m - 71%
chlorosilan 60-901zm- 17%

LB-3 NH4H2PO 4 - 97% below 10#m - 17% 800 4600 463

+talc + 10-60#m - 71%
chlorosilan 60-90#m- 12%

CaCO3 CaCO3 - 95% below 40/zm - 44% 1600 2500 825

40-63/~m - 42%
63-80/zm- 14%
522 J. AMROGOWlCZ AND W. KORDYLEWSKI

1 with earlier measurements of powder effective-

(a) (b)
0.6 ness in gaseous explosion quenching [8].
0.5 q
A possible explanation of the above discrep-
0.4 -._ l
ancy may be the different ability of NH,H 2PO4
-z
& 03 -. and NaHCO, to devolatilize as well as the dif-
a
02 \ ferent chemical kinetics of ignition and flame
0.1 propagation [2, 9, lo]. Perhaps, in the LEL
. -.- ___------
0.0 method NH,H, PO, more effectively prevents
(4 the development of an ignition kernel able to start
0.6

T 0.5
thermal wave propagating into the unburned mix-
g 0.4
ture. On the other hand, NaHCO, should be
T 0.3
more effective in terminating radicals in the flame
0.2
zone. Further work is needed to provide a more
0.1
detailed explanation of the inhibition mechanism
_ ____--- --- -. -
of powder extinguishers, especially phosphates.
0 20 10 60 20 LO 60

INHIBITOR
Fig. 3. Effect of inhibitors injected into the vessel on the
explosion overpressure; (a) melamine; (b) wood dust; (c)
wheat flour; (d) coal. 0, CaCO,; A, LB-2; 0, LB-3.
Jn both types of experiments, CaCO, appeared
to be the least effective suppressant, which sup-
ports the hypothesis on the chemical character of
explosion inhibition by phosphates and carbon-
ates.
CONCLUSIONS
It was found that the effectiveness of extinguish-
ing powders in dust explosion suppression de-
pends on the method of testing. We confirmed
that the LEL method strongly favors NH 4H 2PO,.
However, the method of direct injection of the
extinguishing powder led to different results. The
study provides evidence that at moderate concen-
trations of inhibitors NaHCO, is more effective
than NH,H, PO4 and at higher concentrations of
inhibitors the efficiencies of both extinguishers
converge. These observations are in agreement
CONTENT [ X]
REFERENCES
1. Hastie, J. W., High Temperature Vapors, Academic,
New York, 1975, pp. 344-347.
2. Hertzberg, M., et al., Twentieth Symposium (Interna-
tional) on Combustion, The Combustion Institute,
Pittsburgh, 1984, p. 1691.
3. Choi, T., Rahiman, S., and Essenhigh, E. G., Twenty-
First Symposium (International) on Combustion,
The Combustion Institute, Pittsburgh, 1986.
4. Moore, P. E., Shenyang International Symposium on
Dust Explosions, Shenyang, Republic of China, Septem-
her 14-16, 1987.
5. Jodkowski, W., et al., First International Colloquium on
Explosibility of Industrial Dusts, Baranow, November
8-10, 1984.
Polish norm PN-84 C-01200/7 (Polish Standard).
Cashdollar, K. L., and Hertzberg, M., Private commu-
nication.
Polak-Kuncerowicz, B., Ph.D. thesis (in Polish),
Krakow, 1987.
Hertzberg, M., Zlochower, I. A., and Cashdollar, K.
L., Twenty-First Symposium (International) on
Combustion. The Combustion Institute, Pittsburgh,
1988, pp. 325-333.
10. Sloane, T. M., Combust. Sci. Technol. 63:287 (1987).
Received 22 February 1990; revked 25 August 1990
 Sizing of dust explosion vents in the
process industries

Advances made during the 1980s

Rolf K. Eckhoff
Department of Science and Technology, Chr. Michelsen Institute, Fantoftvegen 38,
N-5036 Fantoft, Bergen, Norway

To prevent damage to humans and property caused by dust explosions in the process industries,
several kinds of precautions may be taken. Because of its favourable features, both technically
and economically, venting has become a widely used method for controlling dust explosions. It is
therefore unfortunate that the literature on dust explosion vent sizing is still in part contradictory.
It is generally appreciated that the vent area needed to keep the pressure in a dust explosion in a
given enclosure below a given limit. is determined by the rate of heat release in the explosion.
However, this rate depends not only on specific dust properties, such as chemistry and fineness,
but also on several properties of the dust cloud that are determined entirely by the industrial
process in which the cloud is generated and ignited. Taking these properties into account has
become a central issue during the 1980s. This paper reviews a number of realistic large-scale
experimental investigations conducted during recent years. The experimental results are
compared with the corresponding vent areas recommended by several different vent sizing
methods in current use. A differentiated method for vent sizing, based on data from experiments
conducted under realistic industrial conditions, is proposed. The continued need for conducting
further realistic exoeriments in various types of full-scale vented industrial enclosures is
emphasized.

(Keywords: vents; explosion; dust)

Several kinds of precautions may be taken to prevent
damage to humans and property by dust explosions in
industry. The importance of fighting potential ignition
sources has been stressed frequently, and this of course
still remains a major issue. However, it is normally
considered that trying to eliminate ignition sources
alone does not provide sufficient protection. It is also
necessary, in case an explosion should nevertheless be
initiated, to ensure that destructive effects are mini-
mized.
Venting has become the most widely used means
of additional protection, because of its favourable
technical and economical features. It is therefore
unfortunate that the technical literature on venting of
dust explosions is still to a considerable extent contra-
dictory. However, signs of a slowly developing clar-
ification are detectable. Useful summaries of existing
knowledge up to 1984 have been produced by
Schofield and Lunn*. Other sources of summarized
information are the books by Field3 and Bartknecht4J.
One particularly important achievement is the
increasing appreciation of the need for a more differ-
entiated approach to dust explosion vent sizing6. Such
Received I4 August I989
0960-4230/90/030268-12
@ 1990 Butterworth-Heinemann Ltd
an approach takes into account the industrial-process-
dependent variables that significantly influence the dust
cloud combustion rate, and the risk-analytical aspects.
The purpose of this paper is to analyse existing
evidence on dust explosion vent sizing and indicate a
general approach for the present and near future by
which adequate vent sizes may be assessed.
Nature of the dust explosion venting process
The maximum overpressure Pred in a vented enclosure
in which a dust explosion takes place is the nett result of
two simultaneous, competing processes: burning of the
dust, causing release of heat and thus pressure increase;
and flow of unburnt and burning dust cloud and
combustion products through the vent opening, causing
the pressure to decrease. In a given enclosure provided
with a given vent, the influence of the burning rate of
the dust on Prrd will be as illustrated in Figure I.
Clearly, predicting the rate of heat release that will
actually occur in a specific situation is the key issue.
However, adequate theories are lacking, and it is
necessary to rely on realistic experiments. The rate of
heat release depends on a number of factors:
l chemical composition of the dust, including mois-
ture;

268 J. Loss Prev. Process Ind., 1990, Vol3, July

 Sizing of dust explosion vents: R. K. Eckhoff

This method requires that Prcd c 0.14 bar(g) and that

the opening pressure and inertia of the vent cover are
small. The vent ratio is defined as vent area per unit of
enclosure volume. Originally a fixed vent ratio, deter-
mined by the maximum rate of pressure rise in the 1.2 I
Hartmann bomb, was specified for a specific dust.
However, as the enclosure volumes get larger, the
required vent area increases more than if geometrical
similarity is used for scaling. and unreasonably large
vents will result. For example, with a vent ratio of 1 mz:
6 m3, a 6 mJ spherical vessel would need only 6% of
the sphere surface for venting. whereas a huge sphere
of 24000 m3 volume would need the entire surface.
Because of this the vent ratio method has been
modified by reducing the required ratio as the enclo-
sure volume increases. It has not been possible to trace
the experimental basis for the vent ratio method. The
Time
method has, however, been widely used. e.g. in the
UK.
Figure 1 Illustration of the influence of the dust cloud burning
rata on the maximum pressure, P,.,, generated in a dust
explosion in a given enclosure provided with a given vent. P,,,, The nomogruph method (VW 367.3, NFPA 68)
is the static opening pressure of the vent cover This method rests partly on the result of cxtcnsivc
large-scale experimentation4,14 and partly on the
chemical composition, pressure and temperature of theoretical analysis by Heinrich 15. A self-contained
the gas phase; system for vent sizing has been developed. It consists of
distributions of particles sizes and shapes in the dust, a 1 m3 vessel Ih for determination of a maximum rate of
determining the specific surface area of the dust in pressure rise used as the characteristic of the explosion
the fully dispersed state; violence of a given dust (K,, value), and a series of
degree of dust dispersion, or of agglomeration of nomographs from which vent areas can be estimated,
dust particles, determining the effective specific using K,,, the enclosure volume, the maximum explo-
surface area available to the combustion process in sion pressure Prcd, and the static opening pressure of
the dust cloud in the actual industrial situation; the vent cover P,,,, as parameters. The relevance of the
distribution of dust concentration in the actual nomograph method is limited by the nature of the
cloud; large-scale experiments on which it rests. These experi-
distribution of initial turbulence in the actual cloud; ments were conducted with dust clouds generated by
possibility of in-situ generation of high turbulence blowing the dust into the experimental enclosures from
levels by rapid flow of still unburnt dust cloud, pressurized reservoirs through narrow nozzles, thus
induced by the expansion of the burnt part of the ensuring exceptionally uniform, well-dispersed, and
cloud. highly turbulent dust clouds. Consequently, the burn-
ing rate for a given dust was very high. in fact too high
The first three factors can usually be assessed reason-
to be representative of the dust clouds in most
ably accurately provided that representative dust sam-
industrial situations. In spite of this, the nomograph
ples are available, whereas the other factors are
method has gained in popularity and is widely used, for
determined by the actual industrial dust cloud genera-
example in the USA, where it has been incorporated in
tion process and the geometry of the enclosure. The
NFPA 68 (Ref. 17). NFPA 68 permits that the nomo-
last factor can be strongly influenced by the location of
graphs be extrapolated to Prctl = 0. I bar (8).
the ignition source.
Very recently, Lunn et al.ln extended the
For realistic vent sizing it is therefore necessary to
VDI 3673 nomographs to K,, values as low as
reconsider the common conception that explosion
10 harms-I, and Prcd values down to 0.05 bar(g).
violence is a specific dust property. In addition, the
However, the experiments on which this extension is
practical circumstances under which the dust cloud is
based were of the same kind as those forming the basis
genrated and ignited in any given case must be taken
of the original VDI 3673 nomographs (very high
into account7,H.
turbulence level and degree of dust dispersion). There-
fore the extended nomographs are subject to the same
Some existing methods for vent sizing
basic limitations as the original VDI nomographs.
During the last decade several authors, including
Schofield, Lunn2 and Field9, have compared different The Swedish method
vent sizing methods in current use. However, these In the Swedish methodl(, the explosion violence
overviews do not include the methods currently used in classification of the dust is by (dP/dt),,, from standard
Sweden and Norway10-t2. 1.2 1 Hartmann bomb tests, as in the case of the vent

J. Loss Prev. Process Ind., 1990, Vol3, July 269

Sizing of dust explosion vents: R. K. Eckhoff

ratio method. Dusts are classified in three groups: sive work that has been conducted during the last
decade*-z4 in larger vented silos of 500 rn and 236 m3
group 1: (dP/dl),,, s 300 bars-
volumes seems highly relevant. but is not mentioned
grou_p 2: 300 bars-l -C (dP/dl),,, -C 600 bars-
either by Bartknecht, or by Radandt?.
group3: (dP/dl),,, 2 600 bars-

The required vent area per unit enclosure volume is Other vent sizing methods
specified for each of the three groups. but the value Some further methods that have been suggested for
decreases with increasing enclosure volume. For sizing of dust explosion vents are discussed by
group 1 dusts and hinged vent panels of maximum Schofield I and Lunn?, and include the K-factor
20 kg m -2 mass, the tabular relationship between vent method, the Schwab and Othmer nomograph. the
area A (m*), volume V (m) and Prcd (bar(g)) can be equivalence coefficient method, and the Rust methodY.
approximated by the equation: the latter three being based on (dP/dt),,, from the
0.019 standard 1.2 I Hartmann bomb.
A=- V.hZS
p ,s
rd
Comparison of data from realistic full-scale
For group 2 dusts, and using hinged vent covers of
vented dust explosion experiments, and
maximum mass I2 kgm :, the Swedish method corres-
predictions by various vent sizing methods
ponds approximately to the equation:
Norwegian experiments in a 500 m3 silo of L/D = 4
(2) These experiments have been reported in detail
elsewhere2*. Pneumatic injection of dust through a
For enclosures of L,/l> 2 3. the enclosure must be 200 mm diameter pipe line was used for generating
divided into the minimum number of fictitious sub- explosible dust clouds in the silo. With wheat grain
volumes needed for all of these to have L,/II s 3. The dust, about 300 kg of dust was blown into the silo for
required vent area for the actual enclosure is then the each experiment. With maize starch the quantity was
sum of the areas calculated for all the fictitious somewhat less, = 20 kg. A larger quantity was re-
suhvolumcs. quired for the wheat grain dust because dispersion was
not complete due to fibrous particles, and lumps settled
The Norwegiun method (modiJied Donut method)
to the silo bottom before all the dust had been injected.
The method used in Norwayl, is a slightly modified
version of Donats methodI from 1971. based on
Hartmann bomb assessment of (dP/dt),;,,. The mod-
ification is the USC of continuous graphs, obtained by
interpolation and extrapolation of Donats tabulated
data IZ.I. For elongated enclosures having a length-to-
equivalent-diameter ratio exceeding 4. the enclosure
should he divided I? into the smallest number of
fictitious subvolumes that gives C,/fI of each
element 5 4. The vent area for each element is
assessed individually and the sum of all the element
areas is taken as the total area required for venting the
enclosure.

The Rudundt sculing luw ./or. vented silo explosions

In his recent book. Bartknecht indicates that the
equation derived by Radandt can be used for scaling
vent areas for silos. The equation was also presented by
0.05-
Radandt more recentlyZ.
l5obors- , \
P,,,, = (bV)/(A - aVL) (3) 6 \\

where A (m) is the vent area; P,,,, (bar(g)) is the 0 8

.
0.02 -
maximum explosion pressure in the vented silo; V (m)
is the silo vnlume; and a, b and c are empirical
constants depending on the K,, value of the dust. P,,,,, is
assumed to be equal to 0. I bar(g). and P re,l must not
0.0
0 5 IO I5
exceed 2 bar(g). For K,, = 200 harms-, i.e. the vent cwea of XXI d 6110(m*)
upper limit of the dust explosion class St 1, the
constants are a = 0.01 I, h = O.MY and c = 0.776. Figure 2 Results from vented maize starch and wheat grain dust
based on results from experiments in a 20 rnJ silo with explosions in a 500 m3 silo in Norway. Comparison with
direct injection of dust from a conventional pneumatic predicted P,,/vent area correlations by various vent sizing
methods in current uee. Data from Eckhoff and FuhreZZ. 0, data
transport line. It is not obvious, however. how the from 500 m3 silo, wheat grain dust, 1981; l , data from 500 m3
volume scaling constant has been obtained. The exten- silo. maize starch, 1982

270 J. Loss Prev. Process Ind., 1990, Vol3, July


The ignition point was close to the silo bottom. In all
experiments but one, dust injection was terminated a
few seconds before ignition, allowing the dust cloud to
become comparatively quiescent. The experimental
results for wheat grain dust as well as maize starch
explosions are shown in Figure 2, together with predic-
tions based on different vent sizing methods. The vent
cover used in the experiments was a sheet of plastic
with a low static opening pressure of the order
0.01-0.02 bar(g). The final, exceptional maize starch
explosion named turbulent jet was so violent that the
silo wall (nailed steel plates) ruptured at about
0.6 bar(g). In this way an additional vent of 50 m2 was
produced, preventing the pressure from rising even
further. In this exceptional experiment 300 kg of starch
had been charged to the dust feeding system, and dust
was still being injected at full rate when the ignition
source. was activated. This special configuration, blow-
ing dust upwards in the silo from a point about half-way
up on the silo axis, is probably not a realistid situation
in an industrial silo. However, it does demonstrate that
even the conservative VDI 3673 recommendations13
may for some very special situations specify vent
openings that are too small.
All the other experimental results in Figure2 show
that in the case of a large enclosure of 500 m3 and
L,fD = 4, VDI 3673 and NFPA 68 oversize the vent by
at least a factor of 2-3 if the K,, value (115 bar m s-l for
maize starch containing 11% moisture) is used, and at
least by a factor of 5 if the common St 1 nomograph is
applied. The Rust method9 agrees well with the wheat
grain dust experiments (150 bars-l in the Hartmann
bomb), but oversizes the vents for the maize starch
(300 bars-l) by a factor of 3. The vent ratio method
which simply requires half the silo top as vent for
keeping Prcdequal to maximum 0.14 bar(g), oversizes
the vent in this particular case by a factor of 2 to 3.
Radandts equation gives exceptionally large vents, but
L/D = 4 may be too small to make Radandts con-
stants for silos applicable.
Both the SwedishlO and the Norwegian11*12 vent
sizing methods are comparatively liberal, suggesting
vent sizes that are in fact smaller than the experimental
maximum values. The argument in defence of such a
liberal approach is that vent sizing involves risk-
analytical considerations, rather than being a fully
deterministic problem6. The 5QOm3 silo explosion
experiments attempted to create a worst-case situation
in terms of both dust concentration in the silo, and
location of the ignition point.
Experiments in slender silos of L/D = 6
Sizing of vents for large, slender silos of the kind
frequently used in the grain, feedstuffs and food
industries, is still an unclarified problem. However,
during the last decade, valuable information has been
gained through large-scale experiments. Two ex-
perimental programmes have been run in parallel, one
in Norway23T2Qand one in Switzerland and FRG20.25.
Both silos have L/D close to 6, but the volume of the
Sizing of dust explosion vents: R. K. Eckhoff
Swiss/German one is only 20 m3, whereas the Norwe-
gian silo has a volume of 236 m3. RadandtZo seems to
assume that effects observed in the comparatively small
20 m3 silo are also applicable to larger silos. However,
results from the 236 m3 silo experiments23*24 suggest
that this is not a valid assumption in general. A more
detailed discussion of this problem is given elsewhereX.
One set of results for which the vent areas are
fairly close to geometric similarity, is given in Figure 3.
The hatched envelope of all experimental data for the
236 m3 silo shows a dramatic decrease of the maximum
explosion pressure when the ignition point is moved
upwards along the silo axis. However, the hatched strip
enveloping the data for the 20 m3 silo, shows very little
of this effect. The discrepancy must be due mainly to
the different ways in which the dust clouds were
generated in the silos26 and to the difference between
the silo heights (10 m and 22 m).
In the case of the 236 m3 silo, the dust cloud was
essentially quiescent at the moment of ignition. There
is little doubt that the marked increase in the maximum
explosion pressure as the ignition point was shifted
downwards in the silo, is caused by flame acceleration
due to expansion-induced flow. This is in complete
accordance with previous results from numerous dust
and gas explosion experiments in one-end-open tubes,
ducts and galleries. In the case of the 20 m3 silo, the
strong, turbulent dust jet, extending from the top of the
silo and several metres downwards, was maintained
during the ignition and explosion process. This most
probably caused very rapid propagation of any flame
initiated at the top of the silo, to the central parts. This
would explain why top ignition and central ignition
gave almost the same explosion pressures. In the case
of bottom ignition, the initial flame propagation was
probably comparatively slow. As soon as the flame
front reached the turbulent dust zone in the central
parts of the silo, a much more rapid flame propagation
pattern, similar to that generated by central ignition,
probably developed.
Switzerland
= 1.0 L/D = 6 25,
vent orea= 0.5 mz
.Z
b
e
pe
0.5
vent area = 3.5 m2
0 I I
Silo Half Silo
bottall WOYup nwf
Location of ignition point m silo
Figure 3 Influence of location of the ignition point in the silo on
maximum vented explosion pressure. Comparison of trendsin a
20 m3 silo*O.zsand a 236 m3 sikP
J. Loss Prev. Process Ind., 1990, Vol3, July 271
Sizing of dust explosion vents: R. K. Eckhoff

In Figure 4, predictions made by Equation (3) are
compared with the results of the experiments in the
Norwegian 236 m3 silo. The dotted curves are obtained
by means of Equation (3) for 20 m3 and 236 m3
volumes, using the constants5 given for dust class St 1.
The upper limit of St 1 is K, = 200 bar m s-r, and it is
assumed that the dotted curves apply exactly to this K,
value. In the absence of further information, it is then
assumed that the ratios AA, between the vent areas for
K,, = 200 bar m s-l and K,, = 100 bar msl for the
20 m3 silo also apply to the 236 m3 silo (geometrical
similarity). This makes it possible to estimate theoretic-
al Radandt-predictions for 236 m3 and K, =
100 harms-, by shifting the Radandt curve for
200 bar ms-l and 236 m3 to the left by AA (which
varies with pressure).
The actual experimental*3 maximum explosion
pressures found in the 236 m3 silo of L/D = 6 are
partly considerably higher, and partly considerably
lower, than the estimated Radandt-value for
K,, = 100 harms-l. On the other hand, even the
highest experimental pressure of 1.2 bar(g) is signifi-
cantly lower than the Radandt value of 1.75 bar(g) for
K,, = 200 bar m s-l (St 1 dusts). It must be concluded
that so far the relevance of, and experimental and
theoretical basis for, the simple scaling law suggested
by Radandt has not been fully clarified.
From the start of the large-scale silo explosion
experiments in Norway*,*s, the dust clouds were
generated by pneumatic pipeline injection, in accord-
ance with typical industrial practice and the basic
philosophy adopted for these experiments27. Following
the first international presentation of results*s, Bart-
knecht and Radandts.2s replaced the traditional VDI
dust injection method 2g by pneumatic pipeline injec-
tion even in their 20 m3 silo experiments. This was a
significant decision, reflecting the need to conduct
experiments in accordance with reality in industry. In
fact, Bartknecht and Radandts,30 took a further re-
levant step, by adding experiments in which the dust
was not injected directly into the silo, but via a cyclone
at the silo top. In this way the dust cloud in the silo,
generated by discharge of dust from the cyclone bottom
via a rotary lock, was neither well dispersed, nor
turbulent. The published result.+30 using this method
are limited to one data point for maize starch, shown in
Figure 5. For a vent area of 1.3 m2, the maximum
explosion pressure generated by dropping the dust
from a cyclone at the silo top via a rotary lock, was only
0.2 bar(g). Direct pneumatic injection gave about
0.5 bar(g), and the traditional, artificial VDI-method
(discharge of dust from pressurized bottles) gave about
0.75 bar(g).
Figure 5 illustrates the need to apply a diierenti-
ated approach to sizing of dust explosion vents6 and for
further full-scale experiments under realistic industrial
conditions27. The pneumatic injection experiments in
the 20 m3 silo gave substantial scatter in the ex-
perimental maximum explosion pressure@. The P,&
data in Figure 5 are the highest values obtained for
each vent area, which means that the curve for direct
pneumatic injection is quite conservative. This kind of
scatter, which is also apparent in Figure 2, emphasizes
the relevance of applying risk-analytical considerations
in vent sizing6.
Recent realistic experiments in Switzerland in vessels
of small LjD
The German/Swiss school of dust explosion venting
research has now started to use industrial pneumatic
pipeline transport systems for generating dust clouds
even in vessels of small L/D. Recently, Siwek3r
discussed a series of explosion experiments in which the
dust clouds were generated in this way in vented
enclosures of 10, 25 and 250 m3 respectively. Some
results for maize starch are reproduced in Figure 6,
together with the current VDI 3673 recommend-
ations13. Experimental results for two quite high dust
cloud injection velocities are given. The dust entered
the vessel through a 90 mm diameter nozzle at the
\,

\ \

z
3
b
e
lqnitiin at silo bottom P
9

Free fall /-,;c -- -

V= 236 m= from cyclone
---
C)L
0 I 2

vent area (d)

Vent area Cm)

Figure 5 Results from vented maize starch explosions in a
20 m3 silo, demonstrating the marked influence of the mode of
Figure4 Comparison of Radandts scaling law for silo vent dust cloud generation on the maximum pressure. Data from
sizingazO and experiments in a 236 m3 vented silo in Norwayz3 Bartknechtm. K, = 226 harms-; P,,, = 0.1 bar(g)

272 J. Loss Prev. Process Ind., 1990, Vol3, July

 Sizing of dust explosion vents: R. K. Eckhoff

vessel, and thereby increasing the combustion rate,

IO m3 vessal, must necessarily decrease with vessel volume. In the
L/D= 1.0 experiments forming the basis of VDI 3673, the sys-
terns for dispersing the dust into the vessels were scaled
up with the vessel size until the desired, very high
turbulence level had been .reached even in large
volumes. Such experimental conditions are extremely
\ conservative and must lead to grossly oversized vents
\ VDI 3673 for large empty volumes of low and moderate L/D.
The systematic trend in Figure 6 of 78 ms-i
I-
yielding more violent explosions than 39 ms-* is
expected because both degree of turbulence and degree
of dust dispersion (de-agglomeration) increase with the
velocity of the dust cloud jet, The experimental data in
Figure 6 are by no means universally valid for pneuma-
tic injection of the maize starch used. Upward injection
from the bottom, or sideways injection, might give
different results. Furthermore, Siwek3i did not investi-
gate maize starch concentrations in the feeding pipe
any lower than 2-3 kg m-3. However, in the closed
1 m3 standard vesseli6, this particular starch gives the
most violent explosions with 0.6-0.7 kgmm3. It could
well be, therefore, that more violent explosions would
have resulted if the dust concentration in the feeding
ii- pipe had been in this range. Siwek31 also used wheat
\. 250 ma vessel flour (K,, = 75 harms-l) and Technocel (K,, =
L/D = 1.0 170 bar m s-l) in his experiments. He found a systema-
4-
tic increase of Pred with K,, for the three dusts, for the
78 Ins- . test conditions investigated.
3-

\ Y-- VDI 3673 Recent work in Japan on venting of dust explosions in

a 0.32 rn3 cyclone
2-
Earlier vented dust explosion experiments in a 1.2 m3
cyclone under realistic industrial conditions, were
reported by Tonkin and Berlemontsz, and by Palmer33.
As shown elsewhere6, the vent area requirements in
these realistic experiments were considerably smaller
(by a factor of five) than those prescribed by VDI 3673
(Ref. 13).
Figure6 Results from vented maize starch explosion experi- Similar cyclone explosion experiments have re-
ments (low moisture starch of K, = 226 bar m s-1 in vessels of cently been conducted in Japan by Hayashi and
various volumes. Pneumatic injection of dust into vessel Matsuda34. The volume of their cyclone was 0.32 m3,
through a 90 mm diameter nozzle located at the vessel apex,
pointing vertically downwards. Dust jet velocities of 78 and its total height was 1.8 m and the diameter of the upper
39 ms-. Dustconcentration in jet 12.5 kg m-J, P,,, = 0.1 bar(g). cylindrical part was 0.6 m. Dust clouds were blown into
Comparison with VDI 3673 predictions. Experimental data from the cyclone through a 150 mm diameter duct. The
Siwek3
desired dust concentration was acquired by indepen-
dent control of the air flow through the duct (suction
vessel top, which generated a strong dust jet vertically fan at downstream end of system), and the dust feeding
downwards into the vessels. The same dust injection rate into the air flow. The dust trapped in the cyclone
system was used for all three enclosures. For all three dropped into a 0.15 m3 dust collecting chamber bolted
volumes, the predicted VDI 3673 areas are substantial- to the bottom outlet. The exhaust duct (0.032 m2 cross
ly larger than those found experimentally for any given section and 3 m length) ended in a 0.73 m3 cubical
P rcd. The discrepancy increases systematically with quenching box fitted with two vents of 0.3 m2 and
increasing enclosure volume. For example, for 0.1 m2 respectively. The cyclone itself was vented
P red = 0.5 bar(g) the VDI 3673 vent area for the 10 m3 through the 0.032 m2 exhaust duct and the 10 m long
vessel is 3.7 times the experimental value, even for the 0.008 m2 dust feeding duct. During explosion experi-
highest injection velocity of 78 m s-l. For the 25 m3 ments, two water spraying nozzles for flame quenching
vessel the corresponding factor is 5.4 times, and for the were in operation in the exhaust duct to protect the fan
250 m3 vessel as large as 8.3 times. The reason for this just outside the quenching box. The ignition source was
trend could be as follows. The relative influence of the a 5 kJ chemical ignitor located in the dust feeding duct
dust jet with respect to inducing turbulence in the about 2 m upstream of the cyclone.

J. Loss Prev. Process Ind., 1990, Vol3, July 273

Sizing of dust explosion vents: R. K. Eckhoff
In addition to these realistic dynamic explosion
experiments, Hayashi and Matsuda34 conducted a maximum overpressure of about 2.5 bar(g), which is of
series of experiments with the same two dusts, using an
art&al static dust cloud generation method very
similar to that used in the VDI 3673 experiments4. The
dust feeding duct was then blocked at the entrance to
the cyclone, which reduced the effective vent area
slightly, to 0.032 m2. A system of two pressurized dust
reservoirs and perforated tube dispersion nozzles were
employed for generating the dust clouds. The S kJ
ignition source was located inside the cyclone, half way
up on the cyclone axis. The source was fired about
100 ms after onset of dust dispersion.
Envelopes embracing the results of both series of
experiments are given in Figure 7. Two polymer dusts
were used, namely an ABS resin dust of median
particle size 180 pm, and an ethylene-vinyl acetate
copolymer dust (EVA) of median particle size 40 pm.
The artificial static method of dust dispersion gave
much higher maximum explosion pressures in the
cyclone than the realistic dynamic method. This is in
accordance with the results of the earlier realistic
cyclone experiments 6,32. The static results in Figure 7
may be compared with predictions made using
VDI 3673 (Ref. 13). A slight extrapolation to 0.032 m2
I normal operation
r air-pulsed, self-cleaning type. Four different dusts were
0
;I
x to VDI 3673 (Ref. 13). The ignition source was located
2 was in the roof of the filter housing. Therefore, to get
C rized in Figure 8, together with the corresponding
0.5
Ncminol dust concmtmtion (kg ni?
Figure 7 Results from vented dust explosions in a 0.32 m3
cyclone using two different polymer dusts and two different
methods of dust cloud generation, and 0.03 and 0.04 m* open
vents with ducts. Data from Hayashi and MatsudaJ
274 J. Loss Prev. Process Ind., 1990, Wol3, July
vent area, assuming St 1 dusts, gives an expected
the same order as the highest pressures of 1.5 bar(g)
measured by the artifical static dust dispersion
method, and much higher than the pressures measured
in the realistic experiments.
However, the new NFPA 68 (Ref. 17) also in-
cludes an alternative nomograph which covers all St 1
dusts that do not yield higher P,, in standard closed
bomb tests than 9 bar(g). This nomograph gives
substantially lower Pled values than the standard
nomograph, especially for small volumes. In the case of
the 0.32 m3 cyclone with a 0.032 m* vent, the alterna-
tive nomograph gives Prcd equal to 0.50 bar(g), which
is in fact close to the realistic experimental values. This
alternative nomograph originates from Bartknecht5,
and represents a considerable liberalization, by a factor
of about two, of the vent area requirements for most
St 1 and St 2 dusts. However, the scientific and
technical basis for this liberalization does not seem to
have been fully disclosed in the literature, and the
situation appears somewhat confusing.
When applied to the 0.32 m3 cyclone, the method
presently used in Norway1pL2 gives a Prcd value of
0.2 bar(g), assuming (dP/dQmax in the Hartmann
bomb of 450 bars-l and P,,,, = 0.1 bar(g). This may
be taken as an indication of the Norwegian code being
quite liberal. On the other hand, risk analytical
considerations6 may justify the Norwegian practice,
since the Japanese realistic experiments were neverthe-
less worst case.
Realistic experiments in bag filters
Vented explosions in a 6.7 m3 industrial bag filter unit in
the UK. Lunn and Cairns35 reported on a series of
dust explosion experiments in a 6.7 m3 industrial bag
filter unit. The experiments were conducted during
of the filter, which was of the
used, and their K,, values were determined according
in the hopper below the filter bag section, and the vent
to the vent, the flame had to propagate all the way up
from the hopper and through the congested filter bag
section. The results from the experiments are summa-
VDI 3673 predictions.
Figure 8 shows that Pred values in the actual filter
explosions were mostly considerably lower than the
corresponding VDI 3673 predictions, and were close to
the theoretical minimum value 0.1 bar(g) at which the
vent cover ruptured. There is no sensible correlation
between the VDI 3673 ranking of expected pressures
according to the K,, values, and the ranking actually
obtained. Lunn and Cairns35 also reported on a series
of dust explosion experiments in a generously vented
8.6 m3 empty horizontal cylindrical vessel of L/D = 6.
The same dusts were used as in the filter experiments,
but the dust clouds were generated artificially by
injection from pressurized reservoirs as in the standard
 Sizing of dust explosion vents: R. K. Eckhoff

and peat dust, both having K,, = 115 bar ms-i and
polypropylene and silicon dusts, both having
I K,, = 125 bar m s-1. Considerable effort was made to
identify the worst-case conditions of dust concentration
/ and ignition timing. At these conditions, experimental
VDI 3673 correlations of vent area and Pred were determined for
---l
each dust, as illustrated in Figure 9 for the peat dust.
/ Figure 9 shows that Pred scattered considerably,
even when the nominal experimental conditions were
/
identical. This again illustrates the risk-analytical
x 0 aspect of the vent sizing problem6. Two extreme
predictions have also been included in Figure 9, that
/
from VDI 3673 (Ref. 13) and that from the vent sizing
0.1 -_----____-__ /
l- method currently used in Norwayll.i*. VDI 3673 is
definitely quite conservative, whereas it may again be

oL-
argued that the Norwegian vent sizing method is quite
liberal.
The 5.8 m3 filter results for all four dusts are
summarized in Figures 10 and 11 as functions of K,
and (dP/dt),,, in the Hartmann bomb. Predictions by
0 100 200
various vent sizing methods have also been included.
K (L+ from 20 L sphere (bar Ins-1 The data show poor correlation between the maximum
explosion pressures measured in the filter at a given
Figure8 P,_, measured in dust explosions in an industrial vent area, and the maximum rates of pressure rise
6.7 ms bag filter unit in normal operation. P,,, = 0.1 bar(g). Data determined in standard laboratory tests. Although the
from Lunn and Cairn@. Comparison with VDI 3673 (Ref. 13)
K,, values of the four dusts were very similar, ranging
from 115 to 125 harms-l, Pred (nitrocellulose flame
VDI 3673 method. In spite of the similarity between ignition) for the four dusts varied by a factor of two to
the dust dispersion method used and the VDI 3673 three.
dispersion methoda, there was no sensible correlation Figure 10 also shows that the explosion pressures
between Pred and K,, . measured were generally considerably lower than those
predicted by VDI 3673 if the ignition source was a
Vented explosions in a 5.8 m3 bag filter in Nor-
way. These experiments are reported in detail
elsewhere%. Dust explosions were initiated in the filter
0.2 m t ,/
during normal operation. A practical worst-case situa-
tion was realized by blowing dust suspensions of the
most explosible concentration into the filter at
Y
/I Si
I Ignition
35 m s-l, and igniting the cloud in the filter during
injection. Four dusts were used, namely maize starch

\ \
VDI 3673

0.2
,/,
I-c--
/ -
/
-- -
Peat ; t
-- - * ; - - -
I I I
0 50 100 150
-I
K,+ from 20 L sphere (bar rns )
I I I I I I
01
0 0.2 0.4 06
Figure 10 Maximum explosion pressures for four dusts in a
Vent area (m*)
vented 5.8 m3 filter at two vent areas, as functions of lG,
determined by the 20 I Siwek sphere: s, 0.2 m2 vent area,
Figure 9 Results from vented peat dust explosions in a 5.8 m3 nitrocellulose flame ignition; 0, 0.3 m2 vent area, nitrocellulose
filter at PM = 0.1 bar(g). Comparison with VDI 3673 flame ignition; f, silicon dust flame ignition of silicon dust.
predictions) and vent sizing method used in Nor~ay~~~~. P = 0.1 bar(g). Comparison with VDI 3673 predictionss for
Injected dust concentration 0.6 kg m-s: l , dusty section of filter; 3.?m3 volume (dusty section of filter). Data from Eckhoff, Alfert
o. clean section of filter. Data from Eckhoff, Alfert and Fuhre and Fuhre%

J. Loss Prev. Process ind., 1990, Vol3, July 275

Sizing of dust explosion vents; R. K. Eckhoff
100 200 300 400
(dP/dt jrnax in Hortmann bomb (bari)
Figure11 Maximum explosion pressures for four different
dusts in a vented 5.8 m3 filter at two vent areas, as functions of
determined by the Hartmann bomb: l , 0.2 m2 vent
(dP/dr),,
area, nitrocellulose flame ignition; 0, 0.3 m* vent area, nit-
rocellulose flame ignition; +, silicon dust flame ignition of
silicon dust. P,,, = 0.1 bar(g). Comparison with maximum
explosion pressures predicted for 3.8 m3 volume (dusty section
of filter) bv three different methods. Data from Eckhoff, Alfert
and Fuhre3
nitrocellulose flame. The result obtained for silicon
dust ignited by a silicon dust flame emphasizes the
different nature of initiation and propagation of metal
dust flames, as compared with flames of organic
dusts36.3. Figure 21 shows the same experimental filter
explosion data as Figure 10, but plotted as functions of
(dP/d%, measured in the Hartmann bomb. In this
case there is a weak positive correlation between P,,
and (dP/W,,, for nitrocellulose ignition, but it is by
no means convincing. Figure 11 also gives the corres-
ponding correlations predicted by three different vent
sizing methods based on Hartmann bomb tests. Both
the Swedish and the Norwegian methods are quite
liberal. The Rust method9 oversizes the vents for the
organic dusts excessively for (dP/dt)max > 1.50 bar s-l.
However, there is fair agreement with the data for
silicon dust ignited by a silicon dust flame.
Vent sizing procedure for the present and
near future
Basic approach and limitations
Realistic vented dust explosion experiments, mostly
conducted during the 198Os, have demonstrated that
none of the vent sizing codes in current use are fully
adequate. It is proposed, therefore, that for the present
and near future, sizing of dust explosion vents should
be primarily based on the available evidence from
realistic experiments. The following suggestions pre-
suppose that the initial pressure in the enclosure to be
276 J. Loss Prev. Process Ind., 1990, Vol3, July
vented is atmospheric. Furthermore, the vent covers
must open completely within times comparable with
the opening times of standard calibrated rupture
diaphragms. In the case of heavier, reversible vent
covers, such as hinged doors with counterweights, or
spring-loaded covers, additional considerations are
requiredss.39. The same applies to the use of vent
ducts4.Jr,QJ and the new, promising vent closure
concept that relieves the pressure, but retains the dust
and flame, thus rendering vent ducts superfluous41.
Large empty enclosures of L/D C 4
As shown in Figure2, a large empty enclosure of
volume 500 ms and L/D = 4 requires considerably
smaller vents than those specified by VDI 3673
(Ref. 13) or NFPA 68 (Ref. 17). This also applies to
the more liberal St 1 nomographsJ7 for P,, (
9 bar (g) As shown in Figure 6, an even more dramatic
reduction in vent area requirement was found recently
in a 250 m3 spherical vessel. In this case the vent area
actually needed was only 0.125 times that specified by
VDI 3673. When sizing vents for large enclosures of
L/D =Z 4, the exact vent area reduction factor, as
compared with VDI 3673, has to be decided in each
case, but it should certainly not be greater than 0.5. In
some cases it may be as small as 0.2 to 0.1.
Large, slender enclosures of L/D > 4
The only investigation of vented dust explosions in
large, slender silos of L/D > 4 and volumes greater
than 100 m3 that has been traced, is that conducted in
Norway23,24 for a St 1 dust in a 236 m3 silo of L/D = 6.
The strong influence of the location of the ignition
source on the explosion violence, as illustrated in
Figure 3, is a major problem. It is necessary in each
specific case to analyse carefully what kind of ignition
sources are likely to occur, and at what locations within
the silo volume ignition is probable42r43. If the explo-
sion in the silo cell can be assumed to be a secondary
event, initiated by an explosion elsewhere in the plant,
ignition will probably occur in the upper part of the silo
by flame transmission through dust extraction duets or
other openings near the silo top. In this case a vent of
moderate size can serve the purpose, even if L/D of
the silo is large. However, the analysis might reveal
that ignition in the lower part of the silo is probable, for
example because the dust has a great tendency to bum
or smoulder. In this case even the entire silo roof may
in some situations be insufficient for venting, and more
sophisticated measures may have to be taken to control
dust explosions in the silo.
Smaller, slender enclosures of L/D > 4
In this case the data of Bartknecht and Radandt5r25J9
from experiments in the 20 m3 silo can be used as one
point of departure.Further data for a 8.7 m3 vessel of
L,fD = 6 are found in the paper by Lunn and Cairn@.
However, it is necessary to pay adquate attention to the
way in which the dust clouds were generated in the
various experiments and select experimental conditions
that are as close as possible to the conditions prevailing

in the actual industrial enclosure (see Figure 5).
Intermediate (lo-25 m3) enclosures of small L/D
The experimental basis is the original VDI-3673
experiments4 with highly homogeneous, well-dispersed
and turbulent dust clouds, and the results for the recent
experiments31 with much less homogeneous and well-
dispersed clouds (Figure 6). The vent area require-
ments identified by these two sets of experiments differ
by a factor of up to 5. Adequate vent sizing therefore
requires that the conditions of turbulence, dust disper-
sion and level and homogeneity of dust concentration
for the actual enclosure be evaluated in each specific
case.
Cyclones
Two realistic investigations have been traced32,34, and
both suggest a significant vent area reduction, in
relation to VDI 3673. Thus, the early investigation by
Tonkin and Berlemont32, using a cyclone of 1.2 m3,
indicates a reduction factor of 0.2. The recent investiga-
tion by Hayashi and Matsuda, using a smaller cyclone
of 0.32 m3, indicates an area reduction factor of about
0.5. The situation for cyclones is similar to that for bag
filters. For organic dusts belonging to group St 1, there
seems to be room for a considerable vent area
reduction as compared with VDI 3673, in the range of
factors of 0.5-0.2. However, for metal dusts such as
silicon, although there is no direct evidence from
cyclone explosions with such dusts, VDI 3673 require-
ments should probably be followed as in the case of
filters.
Bag filters
The experimental basis is the evidence in Figures S-11
produced by Lunn and Cairns3 and Eckhoff, Alfert
and Fuhre%. If ignition inside the filter itself is the
probable scenario (no strong flame jet entering the
Xilter and no significant pressure piling prior to igni-
tion), the vent area requirements of VDI 3673 for St 1
dusts can be reduced by at least a factor of 0.5. If the
dust concentration in the feeding duct to the filter is
lower than the minimum explosible concentration, the
vent area may be reduced even more. However, in the
case of some metal dusts such as silicon, primary
ignition in the filter itself may be less probable and
ignition will be accomplished by a flame jet entering the
filter from elsewhere. In this case the vent area
requirements of VDI 3673 should be followed
(Figures 20 and II).
Mills
The level of turbulence and the degree of dust
dispersion in mills vary with the type of mill. The most
severe states of turbulence and dust dispersion prob-
ably occur in air jet mills. The experimental technique
for dust cloud generation used in the experiments on
which VDI 3673 is based, is likely to generate dust
clouds similar to those in an air jet mill. For this reason
it seems reasonable that VDI 3673 be used without
modifications for sizing vents for this type of mill. In
Sizing of dust explosion vents: R. K. Eckhoff
the case of mills generating dust clouds that are less
turbulent and less well dispersed, it should be possible
to ease the vent area requirements, depending on the
actual circumstances.
Elongated enclosures of very large L/D
This enclosure group includes galleries in large build-
ings, pneumatic transport pipes, dust extraction ducts
and bucket elevators. In such enclosures severe flame
acceleration can take place because of the turbulence
produced by expansion-generated flow in the dust
cloud ahead of the flame. In extreme cases transition to
detonation may occur. The generally accepted main
principles for venting of such systems should be
followed. Either the enclosure must be made sufflcient-
ly strong to be able to sustain even a detonation, and
furnished with vents at one or both ends, or a sufficient
number of vents have to be installed along the length of
the enclosure to prevent severe flame acceleration.
Further details can be found in the literature17.
Important new evidence of how dust explosions prop-
agate in long ducts under realistic process conditions
has recently been presented by RadandtU.
Scaling of vent areas to other enclosure volumes and
shapes, other Pred and other dusts
The number of realistic vented dust explosion experi-
ments is still limited. It may therefore be difficult to
find an experiment described in the literature that
corresponds exactly to the required case. A procedure
for scaling is therefore needed. NFPA 68 (Ref. 17)
presents the following simple equation intended for
scaling of vent areas for weak structures of
P,, 6 0.1 bar(g):
CA.
here A is the vent area, A, is the internal surface area
of the enclosure and Pred is the maximum explosion
pressure (gauge). C is an empirical constant expressing
the explosion violence, based on experimental evi-
dence. By using the internal surface area as the scaling
parameter for the enclosure size, the enclosure shape
is accounted for such that an elongated enclosure of a
given volume gets a larger vent than a sphere of the
same volume.
Equation (4) is intended for the low-pressure
regime only, but its form presents no such limitations.
This equation may be adopted even for P,, >
0.1 bar(g), and used for first approximation scaling of
vent areas from any specific realistic experiment, to
other enclosure sizes and shapes, other P,, and other
dusts. At the outset the constant C should be derived
from the result of the closest realistic experiment, from
which data are available. Subsequent adjustment of C
should be based on additional evidence/indications
concerning influence of dust type and turbulence. This
approach will usually imply extrapolation of ex-
perimental results, which is always associated with
uncertainty. Therefore, the efforts to conduct further
realistic experiments should be continued.
J. Loss Prev. Process Ind., 1990, Vol3, July 277
Sizing of dust explosion vents: R. K. Eckhoff
Conclusions
Over the last decade, our understanding of the dust
explosion venting process has increased considerably.
Unfortunately, however, this has not provided us with
a simple, coherent picture. On the contrary, new
experimental evidence gradually forces us to accept
that dust explosion venting is a very complex process.
What happens with a given dust under one set of
practical circumstances may be far apart from what may
happen in others. Therefore there is much to be said in
favour of the approach to vent sizing adopted in this
paper, the approach of differentiation and tailor-
making.
It is true that such an approach may require a
specialists experience and knowledge. The general
plant engineer may no longer be able to pick some
simple rule of thumb and design a vent in five minutes.
This may look like a step backwards, but in reality it is
how things have developed in most fields of engineer-
ing and technology. Increasing insight and knowledge
has revealed that apparently simple matters are in fact
complex, and need the attention of a specialist from
whom others can get advice and assistance.
For the present and the near future, data from
realistic full-scale experiments should be used as the
main basis for adequate vent sizing. Although further
realistic experiments are needed, the evidence from the
realistic experiments conducted during the last decade
constitutes a valuable starting point for practical vent
sizing work. The general situation is illustrated qualita-
tively in Figure 12 and reflects the basic nature of the
dust explosion venting process.
Figure 12 refers to one specific dust type in various
Higl
h-
Burning
rate in
dust
exolosion
r
---
1 High-speed/
air-jet mills
LO I J
:o, I Level of initial and explosion- iigh
induced turbulence, quality of
dust dispersion, homogeneity
of dust concentration in explosible
range
Figure 12 Qualitative illustration of correlation between degree
of dust dispersion, level of dust cloud turbulence and presence
of homogeneous explosible dust concentration for a given dust
in various industrial situations, and the burning rate of the dust
cloud
278 J. Loss Prev. Process Ind., 7990, Vol3, July
industrial situations. Clearly, the dust cloud burning
rate also depends strongly on the specific dust prop-
erties (chemistry, and particle size and shape), and it is
necessary to have some relative measure of that effect.
So far this has been done by measuring (dP/dt), in
closed-bomb tests7*8, such as the traditional Hartmann
bomb45*e, the 1 m3 ISO-vessel16 and the 20 1sphere5,46.
However, detailed differentiation of K,, values or
(dJ/dQ maxwithin a range of less than a factor of two
seems to make little sense. A classification in risk
groups appears to be a more rational approach, and the
widely used system of the three categories St 1, St 2 and
St 3 (VDI 3673 and NFPA 68) may serve the purpose.
Ideally the 1 m3 standard closed-bomb test should be
used for classification according to this system, but
sufficient accuracy can probably be obtained with
smaller apparatus, provided they are properly designed
and have been properly calibrated against the 1 m3
vessel. It would be useful if one could agree interna-
tionally on a selection of different standard dusts of
known K,, values that any laboratory could obtain for
calibration purposes.
References
1 Schofield, C. in Guide to Dust Explosion Prevention and
Protection-Part 1: Venting, IChemE, Rugby, UK, 1984
2 Lunn, G. A. in Venting Gas and Dust Explosions-A Review,
IChemE, Rugby, UK, 1984
3 Field, P. in Dust Explosions, Handbook of Powder Technology
Series, Volume 4, Elsevier, Amsterdam, The Netherlands, 1982
4 Bartknecht, W. in Explosionen, Springer-Verlag. Berlin, FRG,
1978
S Bartknecht, W. in Staubexplosionen, Springer-Verlag, Berlin,
FRG, 1987
6 Eckhoff, R. K. Bulk Solids Handling 1986,6(S)
7 Eckhoff, R. K. Fire Safery J. 1984, S(2), 159
8 Eckhoff, R. K. Report No. 863350-2, Chr. Michelsen Institute,
Bergen, Norway, 1986
9 Field, P. J. Har. Mat. 1984,8,2
10 Dammexplosioner, Arbetarskyddstyrelsens Fiirfattningsordning
AFS, 1981:s Lib&&lag, Stockholm, Sweden, 1981
11 Industrielle st@veksph%joner, Vemeregler nr.25 Directoratet
for Statens Arbeidstilsyn, Oslo, Norway, 1975
12 Eckhoff, R. K. Report No. 88/02501-3, Chr. Michelsen Institute,
Bergen, Norway, 1988
13 Druckentlastung van Staubexplosionen, VDI-Richtlinie 3673,
Verein Deutscher Ingenieure, Dilsseldorf, FRG, 1979
14 Donat, C. Staub-ReinhaltLuji 1971,31,154
15 Heimich, H. Arbeitsschutz 1974,11,314
16 IS0 6184/l: Explosion protection Systems--Part 1: Determlna-
tion of Explosion Indices of Combustible Dusts in Air, ISO, 1985
17 NFPA 68: Venting of Deflagrations, National Fire Pmtection
Association, USA, 1988
18 Lunn, G. A., Brookes, D. E. and Nicol, A. J. Loss Prev. Process
Ind. 1988,1,123
19 Eckhoff, R. K. Cereal Foods World 1978,23,650
20 Radandt, S. Einfluss van Betriebsparametem auf Explosionsab-
laufe in Silozellen. VDI-Berichte No. 701, Verein Deutscher
Ingenieure, Diisseldorf, FRG, 1989
21 Matusek, Z. and Stroch, V. Stub-Reinhalt Lufll980,40,503
22 Eckhoff, R. K. andFuhre, K. J. Oct. Act. 19846,229
23 Eckhoff, R. K., Fuhre, K. and Pedersen, G. H. J. Oct. Act.
1987,9,161
24 Eckhoff, R. K., Alfert, F., Fuhre, K. and Pedersen, G. H. J.
Loss Pm-v. Process Ind. 1988, 1,16
25 Radandt, S. Staubexplosionen in Silos. Untersuchungscrgebnis-
se, Tei13, Symposium No. 14, Berufsgenossenschaft Nahrungmit-
tel und Gaststitten, Mannheim, FRG, September 1985
26 Eckhoff, R. K. Report No. 88/02501-l, Chr. Michelsen Institute,
Bergen, Norway, 1988
27 Eckhoff, R. K. in Fuel-Air Explosions, (ed. J. H. S. Lee and C.

M. Guirao) University of Waterloo Press, 1982, pp. 657-678
28 Eckhoff, R. K. The Control and Prevention of Dust Explosions,
organized by OYEZ IBC, Basle, Switzerland, 16-17 September
1982
29 Radandt, S. Staubexplosionen in Silos, Untersuchungsergebnis-
se, Teil2, Symposium No. 12, Berufsgenosseschaft Nahrungsmit-
teland Gaststiitten, Mannheim, FRG, November 1983
30 Bartknecht, W. Bundesministerium fiir Forschung und Technolo-
gie, Abschlussbericht Project No. 01 HK 664 AO, 1988
31 Siwek, R. VDI-Bericht No. 701, Verein Deutscher Ingcnieure,
Dusseldorf, FRG, 1989
32 Tonkin. P. S. and Berlemont, F. J. Fire Research Note No. 942,
Fire Research Station, UK, July 1972
33 Palmer, K.N. in Dust Explosions and Fires, Chapman and Hail,
London, UK, 1973
34 Hayashi, T. and Matsuda, T. Archivum Combustionis submitted
for publication
35 Lunn, G. A. and Cairns, F. J. Haz. Mar. 19X5,12,87
36 Eckhoff, R. K., Alfert, F. and Fuhre, K. VDI-Berichte No. 701,
Verein Deutscher Ingenieure, Dusseldorf, FRG, 1989
Sizing of dust explosion vents: R. K. Eckhoff
37 Cassel, H. M. US Bureau of Mines, Rep. Inv. 6551,1964
38 Wiemann, W., Bauer, R. and Moller, F. VDI-Berichte No 701,
Verein Deutscher Ingenieure, Dusseldorf, FRG, 1989
39 Siwek, R. and Skov, 0. VDI-Berichte No. 701, Verein Deutscher
Ingenieure, Diisseldorf, FRG, 1989 .
40 Lunn, G. A. in Guide to Dust Explosion Prevention and
ProtectiowPart 3: Venting of Weak Enclosures and Effect of
Vent Ducts, IChemE, Rugby, UK, 1988
41 Alfert. F. and Fuhre, K. Report No. 89/25820-l, Chr. Michelsen
Institute, Bergen, Norway, 1989
42 Alfert, F., Eckhoff, R. K. and Fuhre, K. VDI-Berichte No. 701,
Verein Deutscher Ingenieure, Dusseldorf, FRG, 1989
43 Eckhoff, R. K. Special Techn. Publ. 958, Am. Sot. Test. Mat.,
USA. 1987, p. 265
44 Radandt, S. VDI-Berichte No 701, Verein Deutscher Ingenieure,
Dtisseldorf, FRG, 1989
45 Eckhoff, R. K. Staub-Reinhalt. Luft 1977,37,110
46 ASTM Designations: E789-86 (Hartmann bomb) and E1226-88
(20 I sphere), Annual Book of ASTM Standards, 1988, volume
14.02
J. Loss Prev. Process Ind., 1990, Vol3, July 279
96

Metal/Cyanide-Containing Wastes; Treatment Technologies, by S.A.K. Palmer,

M.A. Breton, T.J. Nunno, D.M. Sullivan and N.F. Suprenant, Noyes Data
Corp., Park Ridge, NJ, 1988, ISBN No. O-8155-1179-5,721 pp., $74.00.

The Resources Conservation and Recovery Act, as amended by the Hazard-

ous and Solid Waste Amendments of 1984, strictly regulates the land disposal
of numerous types of hazardous waste, among these being wastes containing
metals and cyanides of various concentrations.
The book provides a review of the regulatory background (complying with
the above laws) and a summary of the current hazardous waste management
data base. The introductory (regulation) background is followed by informa-
tion on waste minimization, techniques and an evaluation of a wide range of
treatment/recovery processes for metal- and cyanide-containing wastes.
Specific chapters include:
l Metal/cyanide waste sources and characterization
l Waste quantity management practices and treatment capacity
l Waste minimization processes and practices
l Membrane separation technologies for metal removal
l Extraction for metal removal
l Adsorption for metal removal
l Electrolytic processes
l Chemical treatment/removal program for metals
l Biological treatment for metal-containing waste
l Thermal destruction/recovery process
l Physical removal processes for cyanide
l Chemical destruction of cyanide
l Miscellaneous cyanide destruction processes
l Consideration for system selection
Because the book was written by a consulting firm as a government report,
i.e. by a U.S. Environmental Protection Agency contractor, it is not deep in its
treatment of the topic but is a comprehensive survey of the latest techniques
and the literature on the topic as well.
Indeed, the book is well-referenced. Those needing information on the topic
of cyanide/metal waste treatment will be well-advised to seek out this book. It
makes a good beginning.

GARYF.BENNETT

Dust Explosions, by W. Bartknecht, Springer-Verlag, Berlin, 1989, ISBN 3-

540-50100-2,270 pp., $80.00.

Dust represents the most hazardous form of solid matter, especially organic
and metal dusts. Dust may be considered a phase of its own with very special
 97

characteristics. This book summarizes present knowledge of the cause, course,

and consequences of dust explosions.
Since the windmill was introduced in 1752 for grinding cereal grains, dust
explosions have been reported. Although grain was involved in a quarter of
such explosions in the U.S. during 1900-1952, wood, feedstuff, flour, starch,
cork, sugar, plastics and other substances have been recorded. In recent years,
silos and bunkers, grinding plants, conveyors, separators and dryers have been
the scene of dust explosions. For an ignition source, mechanical sparks, smol-
dering particles, mechanic heating and static electricity are known sources.
Possibly the most useful chapter in the book is Chapter 5 on protective meas-
ures. The range of options is broad - from inerting, prevention of effective
ignition sources and hot surfaces, proper design of equipment, and process
venting. It is unfortunate that metal dusts, especially aluminum, magnesium,
and zirconium do not receive more editorial attention. The photographs in the
book are excellent, many in color. A total of 140 references are given, most of
which are to be found in the German literature.
This book is truly an excellent updated reference which should be useful to
anyone concerned with production, handling, storage or shipment of dusts in
any form.

HOWARD H. FAWCETT

Chemical Carcinogens. Some Guidelines for Handling and Disposal in the Lab-
oratory, by M. Castegnaro and E.B. Sansone, Springer-Verlag, New York,
NY, 1986, ISBN 0-387-16719-6.97pp., $17.00.

This little book appears to be a result of merging the National Institutes of

Health (NIH-USA) guidelines for laboratory work with chemical carcinogens
and similar guidelines published by the International Agency for Research on
Cancer (IARC ) . The introduction provides lists of the chemicals or processes
considered by IARC to be known, probably, or possibly carcinogenic in hu-
mans. This chapter also outlines the responsibilities of each person in a re-
search group which handles chemical carcinogens. Further discussion covers
supply, storage and transport of carcinogens, what to do in case of a spill, pro-
tection of staff, proper experimental procedures in laboratory and animal rooms,
and the design and outfitting of rooms where carcinogens are used. Methods
for treatment and disposal of waste from experiments with carcinogens are
given, as well as an evaluation of available methods for chemical deactivation
of various structural classes of carcinogens.
A list of almost 200 references is added. Unfortunately, older editions of
some of the main references are included when newer editions are now avail-
able. For future editions, more effort should be expended to update the book.
Fire Safety Journal, 14 (1989) 167 - 177
Safety and Fault Detection in Process Control Systems and Sensors
R. K. CHOHAN
124 Staines Road, Ilford, Essex IG1 2UY (U.K.)
B. R. UPADHYAYA
The University of Tennessee, Knoxville, TN 37996 (U.S.A.)
(Received May 1, 1988; in final form July 29, 1988)
SUMMARY
Recent severe accidents in the process
industry have demanded action and investi-
gation o f process systems for safety. The
present paper attempts to integrate various
aspects connected with safety o f systems and
faults in the e q u i p m e n t and sensors. E x c e p t
for dust explosions in grain stores and other
processes, most other serious accidents have
involved emissions from p o i n t failures such as
valves, seals and pipe connections. Weak links
k n o w n to have caused past accidents are
emphasized and some suggestions made for
accident prevention. Sensors have been identi-
fied as a weak element in a safety system and
various aspects are discussed. Fault detection,
especially early fault detection, is important
both for equipment, sensors and actuators.
Various procedures are reviewed. Early detec-
tion o f s y m p t o m s leading to an accident is
important, and therefore integrity o f all
elements in a detection system has to be
maintained.
1. INTRODUCTION
The present contribution attempts to inte-
grate various aspects of faults, failures and
safety issues in processes. Emphasis is then
directed to instrumentation, and fault detec-
tion in instruments and actuators.
The last t w o decades have witnessed major
process explosions and hazards. Among
typical and highly publicized incidents were
the release of 50 mg of cyclohexane to the
environment from a plant at Flixborough,
England, leading to the loss of 28 lives on
June l, 1974; and the explosion in a plant
manufacturing 2,4,5-trichlorophenol at
0379-7112/89/$3.50
167
Seveso, Italy, on July 10, 1976. One of the
most severe accidents happened very recently,
on November 19, 1984, in Mexico City at a
gas plant resulting in the loss of over 300 lives
and over 1500 injured. The most serious
accident appears to be the release of methyl
isocyanate gas at a plant in Bhopal, India, on
December 3, 1984. This incident claimed the
lives of more than 2000 people. All these
major incidents had one factor in c o m m o n -
the cause of each incident was n o t clear, at
least immediately after the incident. The most
severe nuclear accident at a U.S. commercial
power plant was the Three Mile Island
(TMI-2) accident in 1979. Some of these
incidents will be referred to further in the
discussion that follows.
In view of the above, it is n o t surprising
that there has been increased emphasis on
safety and reliability of process systems in
recent years. Publications have appeared on
the physical processes, such as explosions,
e.g., Bodurtha [1], on toxic and hazardous
waste, e.g., LaGrega and Hendrian [2], on
human detection and diagnosis of failures,
e.g., Rasmussen and Rouse [3], and on fault
detection and diagnosis, e.g., Himmelblau
[4l.
Various aspects of safety and fault analysis
tend to be emphasized from time to time.
For example, before TMI-2, a great deal more
attention was attached to utility of computer
support than the man-machine interface. It
is probable that at present, there is more
concern for human management and organiza-
tional failures than process hardware failures
[5]. Our purpose in the present paper is to
emphasize the integration of these multi-
disciplinary aspects. The o u t l o o k is one of
prevention rather than cure. Different aspects
of safety and accident prevention in chemical
Elsevier Sequoia/Printed in The Netherlands
168
processes are presented by Fawcett and Wood
[6].
After a discussion of integration of differ-
ent scenarios of hazardous and other undesir-
able events, we concentrate on aspects which
we feel have been somewhat neglected,
namely, sensors, measurement, actuators,
control, alarms and prediction of events.
Sensors are central to the provision of process
information for hazard prevention and con-
trol. Fault diagnosis of sensors, including that
of actuators, is discussed. Suggestions are put
forward for some new sensors and signal pro-
cessing for some known process designs.
2. GENERAL CONSIDERATIONS AND
SCENARIOS
Process systems in chemical, petrochemical,
food and nuclear technology tend to be com-
plex. This complexity is largely attributed to
the many interacting subsystems or processes.
Process systems usually have associated safety
systems which include alarm and protection
facilities. Besides considerations of equip-
ment, man-machine interactions, environ-
mental and legal considerations, safety
analysis should include instrumentation/con-
trol aspects as well.
Hazard prevention involves reliability of
both processes and instrumentation. Safety
systems can be improved by redundancy
{spares), upgrading components, control
structure/system modification, and perhaps
distributed control. While higher reliability
usually implies higher costs, the benefits may
be significant.
Most processes of significant complexity
have computer-based control and safety sys-
tems. Failure and fault analysis should include
the following considerations:
(a) hardware reliability
(b} software reliability
(c) system integrity
(d) maintenance
(e) sensor/instrumentation reliability
(f) actuator reliability
(g) man-machine interface including
operator training
(h) information flow to management.
Ideally, all the above should be fault-tolerant.
This may require some compensation for one
or more failed element, service or activity.
Commonly encountered failure events
include those involving mechanical device
failures due to extremes in process variables
{e.g., pressure, temperature, concentration),
control/safety system failure due to instru-
mentation failure and/or operator 'failure'.
Preventive maintenance must also receive
top priority. This implies inspection, repair,
replacement, and determining the probable
frequency of failure of equipment or a com-
ponent [7]. Timing and appropriate action
are important and early fault detection should
be encouraged [8]. Operator and management
;kills have to be utilized.
Whereas preventive maintenance is the
?hilosophy behind safety systems, post-
accident events also must receive attention in
a complete assessment to aid in the preven-
tion of future accidents. As stated earlier, we
shall emphasize the former.
Plamping and Andow [9] present a pro-
cedure for information flow for an alarm
system. This procedure is divided into nine
stages:
(1) determine the failure events of the
plant;
(2) assess their credibility;
(3) define the modes of operation;
(4) determine the response pattern of the
process variable;
(5) determine the corrective actions avail-
able to the operator;
(6) determine the level of diagnosis
required;
(7) analyse effect patterns and corrective
actions to group similar events;
(8) select the set of process variables;
(9) assess the alarms selected and if unsatis-
factory repeat stages 6 through 8.
This sequential procedure allows an appre-
ciation of the dependence of measurement
and states on failure and corrective action.
Various sets of states may be defined for
alarms given some measurable limits. These
limits may be dictated by rules such as the
following [9] :
(a) for those states defined to safeguard
e q u i p m e n t - use absolute limits;
(b) for those states defined to detect when
the process variable deviates b e y o n d a band
around a set-point value -- use deviation
limits;
(c) for those states defined to detect instru-
ment failure -- use zero or full-scale limits;

(d) for those states defined to detect pro-
cess variables -- use rate-of-change limits.
Measurement is essential to determine all
such limits. The multi-disciplinary aspects
of any safety procedure, such as that of
Plampling and Andow [9], must be recog-
nized.
Hazards are connected with substances
(gases, liquids, solids or mixtures) having
properties such as toxicity, corrosivity, flam-
mability, explosivity, reactivity with other
substances, and radioactivity. Most hazards
can be classified as:
(1) explosions including chemical and dust
explosions;
(2) fires;
(3) leaks, which eventually may lead to an
explosion or fire or a health hazard;
(4) radiation.
The discussion t h a t follows will be based on
these four major causes.
3. PROTECTION, CONTROL AND ALARM
Various relative time-scales can be identi-
fied with various hazard and other critical
situations. These can be conveniently classi-
fied according to the possible modes of
corrective action:
(a) short time-scale, i.e., situations where
catastrophe or accident can occur once initi-
ated:
(b) intermediate time-scale, i.e., catastro-
phe or accident can occur but can be stopped
if suitable action is taken rapidly.
(c) elemental time-scale, i.e., catastrophe or
accident can be avoided if proper control
action is followed.
It can safely be said that prediction of the
impending disaster is most urgent for case (a).
Most disaster predictions are best described as
combinations of art and science. Risk analysis
is part and parcel of these predictions. Morgan
[10] states that risk is due to the following
probabilistic factors:
(1) The values of significant variables are
not or cannot be known, and accurate predic-
tions cannot be made.
(2) The physical, chemical and biological
phenomena involved are n o t fully understood.
(3) The processes involved are inherently
stochastic, or at least complex enough to
impede the development of valid predictive
models.
169
Specific design details of a process or unit
are of significance n o t only with regard to the
general development of models but also pecu-
liarities as far as possibilities of leaks, exces-
sive pressure fluctuations, etc., are concerned.
Intermediate time-scales would involve
proper measurement followed by appropriate
alarm and action such as system shutdown.
This action must be rapid enough to minimize
the risk of a hazard or the extent of it. Ele-
mental time-scales involve appropriate mea-
surement and control. Both these time-scales
involve measurement sensors and prediction
may benefit from some measurement. There-
fore, we shall concentrate hereafter on mea-
surement and fault detection of instrumenta-
tion and actuators. Actuators, especially
involving moving parts, are subject to failure
and leaks.
4. SENSORS IN THE TECHNOLOGIES
Knowledge of operating conditions of a
process and its surrounding is essential for
analysis of potential hazardous situations and
their prevention. The required information
may include one or more of the following
variables: operating temperature and pressure,
flow rate, density, concentration of substance
or impurity, moisture content, particle size
and hardness, accumulation rates at local posi-
tions.
The availability, proper location and use of
appropriate sensors are all most important
factors. The post-accident diagnostics during
the accident at the Three-Mile Island Unit 2
reactor became complicated because of the
non-availability of some important process
instrumentation [11]. This incident also
demonstrated the importance of properly
placing instrument indicators on the control
panel, thus emphasizing the need for better
human factors considerations.
A brief consideration of sensors of impor-
tance to hazards and critical situations is in
order here. The parameters of interest are
pressure, temperature, toxic substance con-
centration, flow rate, and radiation levels
among others.
Pressure and temperature are the most
widely measured process variables. Technol-
ogy for these is highly advanced and well
described in textbooks and in symposia.
170

Recent discussions on high pressures and
temperatures are provided by Pegg [12], and
Anderson and KoUie [13], respectively.
Gas detection devices can be classified as
pellistors, infrared cells, electrochemical cells,
galvanic cells, semiconductor devices and
acoustic devices. Some of the problems
involved are:
(a) Flammable gas sensors such as pellistors
may be non-specific and the catalyst gets
poisoned. Semiconductive devices degrade to
the extent that they are not routinely used.
Further performance may be limited by asso-
ciated electronics and physical arrangement of
the sensing head. Long-term problems may
arise due to zero-drift and sensitivity fall due
to sensor degradation [ 14].
(b) Infrared sensors have m a n y virtues but
are not cheap.
(c) Electrochemical cells are not very sensi-
tive to low concentrations usually encoun-
One of these developments is based on laser
beams [19]. This can be online and responsive
to low gas concentrations, is fail-safe and can
be self-calibrated. These characteristics are
particularly important for some fugitive emis-
sions.
Hazardous substance detection and mea-
surement can be based on that of fugitive gas
detection systems [19]. Figure 1 shows a
representation of the important methods
available at present. Another representation
for fugitive gas detection is shown in Fig. 2.
The portable instruments enable detection at
REMOTESENSING.
METHODS
i. ELECIRO~ETRIC
SENSORS

tered for toxic gases and may not be specific. 2. SPECTROMETRIC

OR
The life of such sensors is also not appreciable ELECTROOPTICAL
SENSORS
[15]. 3. CHEMILUMINESCENT
SENSORS
Besides the above, a number of gas ana-
lyzers are used [16]. However, these are quite
time-consuming concentration determina-
tions. They also are prone to unrepresentative MANUAL
I J
CUMULATIVE
samples to be analysed. MANUALSAMPLING GRABSAMPLING
New sensor and device development is in ANDANALYSIS MID ANALYSS
progress. Some examples are [17] :
(a) zinc oxide surge arrestors to control the A) GAS ABSORPTION A) EVACUATED
FLASK
damaging effects of lightning on transmission
B) INTEGRATED B) DISPLACEMENT
equipment; SN~PLER SYSTEM

(b) optical boiler-flame scanners to prevent C) GAS ADSORPTION C) FLEXIBLE

BAG
boiler explosions;
(c) tagging compounds used in conjunction Fig. 1. Principle gas/liquid measurement systems.
with an ionization chamber to detect hot
spots in generators. I NSTRUI~NTATION
FOR
Recent developments in acoustic and opti-
FUGITIVEGAS~ETECTION
cal instrumentation are showing promise for
the provision of reliable sensors. Such sensors
may satisfy some desirable criteria as stated r NON-PORTABLEI
by Wangsmo and Riddervold [18]:
(1) elimination of the poisoning problem;
DETECTIONMETHOD
I
DETECTION~THOD
(2) automatic and immediate fault indica- l, INFRA-RED
~, FLAME IONIZATION
tion through self-test capability; DETECTION (FID)
2. LASER
(3) discrimination between hydrocarbon 2. CATALYTIC OXIDATION
(CAT OX) 3. FIBER OPTICS

and non-hydrocarbon gases or other contami- ] . INFRA-REDABSORPTION h. SAMPLING ANALYSIS

(It) ETC
nants, i.e., selectivity capability;
a. PHOTO-IO~IZATION (PID)
(4) fast response time;
(5) no ambiguity in response at high gas ~,~,~ IONIZE THE GAS, THEN
MEASURETHE DEGREEOF
IONIZATION
concentration;
(6) low maintenance at infrequent intervals. Fig. 2. Fugitive gas detection.

various locations such as valves and fans.
Valves are notorious for leaks of gases and
liquids [20] and it is n o t generally feasible to
have a sensor for each valve. However, this
should be seriously considered when highly
dangerous substances are involved.
All hardware, including sensors, degrade
and experience other functional problems.
Temperature and pressure sensors need cali-
bration checks and response time testing [21].
Some other problems with instrumentation
were pointed o u t above. Analytical instrum-
mentation has t w o major problems [22] :
(1) leaks in piping/sample lines;
(2) contamination of the measuring unit
during sampling.
Leaks in sample lines cause dilution, those
in the hydrogen supply system in a flame
ionization detector m a y n o t be dangerous b u t
may make the sensor non-functional.
Maintenance of processes including instru-
mentation, actuators and c o m p u t e r systems
must be present especially for high-risk
situations. System design should allow for
relatively easy maintenance. Analytical
instruments, for example, should have a few
indicators and allow easy checking of
plumbing and easy spotting of leaks. This
would enable simple techniques such as
'soaping' for leak detection for instruments
at r o o m temperatures. Soap bubbles from an
application o f a soap solution would indicate
a leak [23]. Valve packing should be period-
ically checked and if necessary, the packing
replaced [20].
Maintenance and proper design are neces-
sary for the prevention of hazardous accidents
and protection, b u t are not sufficient. Fault
detection and its analysis is also a most urgent
issue. A number of procedures and techniques
exist for the detection of faults in processes,
sensors and actuators. Since sensors and actu-
ators are central to the prevention, control of
and protection from accidents, their fault
detection is discussed in the next Section.
5. REVIEW OF INSTRUMENT AND ACTUATOR
FAULT DETECTION TECHNIQUES
5.1. Introduction
The methods used for monitoring and veri-
fication of process instrumentation and actua-
tion may be classified into "static" and
" d y n a m i c " [24]. Static methods are those
171
where the sensor performance is n o t corre-
lated with system dynamics. These include
like-sensor comparison, limit checking, instru-
ment-loop integrity checking, signal clipping,
calibration, etc. In the dynamic approach the
sensor validation is performed using signal
analysis in conjunction with a physical or an
empirical system model. The latter approach
is more versatile because of its application
during plant operation.
A noninvasive test, such as random data
analysis, can be used in some applications
w i t h o u t the need for a physical model. Static
and empirical methods are easy to implement
and are cost-effective. It is possible that most
of the sensors can be monitored b y this
approach. Often methods are developed that
are suitable for monitoring a specific detector.
In this philosophy a separate m e t h o d must be
available for each type of sensor. A third
approach for sensor validation is to use signals
that are acquired from dissimilar sensors. This
provides a diversity of measurement thus
overcoming the problem of common-cause
failures. In this Section we review dynamic
sensor validation methods and state a few
examples.
5.2. Statement o f sensor failure detection
problem
The methods developed for sensor fault
analysis must be able to distinguish between
sensor failure and changes in the system
behavior. For a general solution o f the prob-
lem, a system dynamic equation and a mea-
surement equation are necessary. The linear
form of these equations is given by
~((t) = A X ( t ) + BU(t) + W(t) (continuous time)
(5.1)
Y(k) = H X ( k ) + V(k) + SF(k) (discrete time)
(5.2)
where
X is an n-dimensional state vector
U is a p-dimensional perturbation vector
Y is an m-dimensional measurement vector
SF is a vector of sensor errors
k denotes the current sampled data index
W and V are process and measurement noise
components
Sensor failures can be modeled as changes
in the measurement matrix H, a bias in the
172
sensor output, and additional sensor noise.
Other fault estimation must be performed by
first isolating the sensor and using appropri-
ate fault estimation algorithms. Empirical and
physical model descriptions must also be
considered.
5.3. Dynamic methods for sensor fault detec-
tion and estimation
The fundamental considerations in this
approach are:
(1)generate hardware and/or analytical
redundancy of the measurements;
(2) a comparison logic for isolating the
sensor error from system behavior;
(3) quantitative estimation of sensor errors.
Analytical redundancy methods using
Kalman filtering techniques [25] or dedicated
observers [26] have been applied in the aero-
space industry [27] and in nuclear power
plant models [28]. The Kalman filter is appli-
cable to both steady-state and transient
conditions, and for linear and nonlinear
dynamics.
A generalized likelihood ratio method of
comparing various fault conditions was
applied to a pressurized water reactor (PWR)
pressurizer level detection system [29]. This
approach is computationally slow. A more
general method of isolating faulty sensors
using a parity-space representation was used
for a boiling water reactor (BWR) suppression
pool signal validation [30]. All these algo-
rithms make use of physical models requiring
validation. Such models may fail in a transient
situation.
A data-dependent modeling strategy and
estimation of sensor errors by appropriate
measurement model has recently been devel-
oped [31] and applied for faulty estimation
of core-exit thermocouples in the loss-of-fluid
test (LOFT) reactor. A subsystem of the
reactor (in this case the reactor core) was con-
sidered for analysis a n d a dual-hypotheses
testing procedure was applied to the model
generated from the in-core neutron detector
and core-exit thermocouple signals. Sensor
errors such as bias, excess noise, changes in
response time can be estimated using proper
measurement models.
Utilization of a pair of signals not related
to the sensor under study could provide infor-
mation about sensor validation. Such a tech-
nique for monitoring fluid flow velocities has
been reported. A pair of thermocouples [32]
or a pair of 16N detectors [33] may be corre-
lated to determine the fluid transit time and
hence the velocity. This provides an indepen-
dent m e t h o d of verifying flow meters and
thus provides diversity against common mode
failure of redundant sensors.
The choice of a particular approach for sig-
nal validation and instrument fault detection
depends on the implementational aspect of
the m e t h o d and the available measurements.
Distinction must also be made between slow
degradation and transient failures.
The above discussion on fault detection
and identification concentrated on signal pro-
cessing. However, other procedures are also
to be encouraged such as physical/chemical
effects on component aging, use of mechan-
istic models for system behavior and historical
records of components.
6. P O T E N T I A L P R O C E D U R E S F O R P R E V E N T I O N
OF ACCIDENTS
Prevention of and protection from acci-
dents can be visualized as decisions and
actions made at various hierarchical levels.
The highest of these levels would be at the
national level (such as the EPA in the United
States) and lowest level would be a plant
operator. Generally speaking, as one ascends
such a hierarchy, data becomes somewhat
imprecise, infrequent and incomplete result-
ing in a greater or lesser degree of risk associ-
ated with decisions and actions.
A serial procedure for accident prevention
is shown in Fig. 3. The first step in Fig. 3
attempts to classify a system, region or space
involving a hazardous material (solid, liquid,
gas or mixed phase). Factors which are to be
considered in such a grading of flammability
and toxicity include the following [34] : com-
ponents present in the space, volume of space,
relative quantity of each component present,
phase of component, pressure, temperature.
Some substances can easily be classified as
having a high hazard rating, for example,
methyl isocyanate.
Suitable action(s) will be required for sub-
stances having a high hazard rating. Further
measures or quantification may be carried
out (e.g., see the Appendix in ref. 34).
A literature search of all known physical,
 173

ILON F~ILIIY~ONICITY Threshold conditions for autoignition or

ORoNs-
I~- intolerable fugitive emissions are desirable and
~ssl MO~BATE
Em,~
' IU~Y~OXICITY
S
' TORES') ~ some procedures are available as pointed o u t
~HION F~ILIIY/TOXIEITY earlier (see also ref. 1). If the process condi-
tions are such that there is a fair probability
2. LOCATEPOSITIONS IN SYSTEMMOSTLIKELY TO of exceeding such a threshold, suitable and
(A) ACCUMULATE timely action is required. Process management
(0) ESCAPE
(C) HAVECONDITIONSCOHDUCIVETO AUTOIGINHIONOR OTHEROPERATIONS
is centrally involved and pre,accident/post
(SUCH AS WASHINGOIL TANRERCARGOSWITH WATERJETS (JONES & BOND accident actions are shown in Fig. 3.
(198L~)). A number of disasters, except dust explo-
sions, appear to be due to emissions to the
3. ~I.COLAIE~STIi"IATE THRESHOLDCONDITIONS(E.G. ELECTROSTATIC[HRESHOLD) IN atmosphere and can be caused by:
A C,RAIN STOREOR OIL TANKERSCARGO)BASEDON WORSTCASEASSUMPTIONSSUCH
AS i'IINIM~ IGINIT]ON ENERGY,FLAI,It~AI~LEVAPORCONCENTRATION,ETC.
(a) relief valve operation;
(b) equipment failure such as pipe rupture;
~. ASSESSIF LIKELY/POSSIO~CONDITIONSEXCEEDTHOSEOF THE
(c) leaks in valves, fans, pump seals, etc.
ESTI~TED/EALCULATED THRESHOLDSFROilSTEP (3). The disaster in Bhopal involved the operation
of a relief valve due to internal over-pressure
5. M~J~AGERIALAETION
caused when water contaminated methyl iso-
MANAGEMENT cyanate. The Seveso incident also involved
I I liquid escape from a relief valve caused b y a
PRE-ACCIDENT ACTION POST-ACCIOENT ACTION
I runaway reaction. Valves and other units with
B) EXPLOSION/FIRE
CONTAINMENT moving parts have to be present in a process.
C) INSURANCE
LOCATION CONSIDERATION Failures of such components is also possible
and safety/protection systems including
I I I I management should allow for this.
SENSORS ACTUATORS, DESIGN OPERATION
r FLANGES
OTHER MOV NG PARTS 4=
i In view of the above, any components with
I
I moving parts or weak points such as seals
I
I I 1
must receive attention. Early detection of
PREVENTIVE CONTROL
ACTION
ALARM'4- . . . . . . . LIMITING CONDITION emissions is highly desirable followed b y
(AUTOMATED/OTHERWISE)
p r o m p t action. Appropriate sensors together
Fig. 3. Procedure for accident prevention. with alarm/control/protection action is abso-
lutely necessary where highly hazardous
substances, such as methyl isocyanate, are
chemical, and toxic properties must be carried involved. This and the degree of redundancy
o u t using references such as Sax [35]. These of sensors, actuators, etc., should reflect the
would account for the chemical hazards. It is risk and hazardous nature of the substances
important to consider the physical and equip- involved and process conditions.
ment hazards such as high temperature, pres- Automation is desirable as it takes negli-
sure, construction material, equipment age, gible time for specified actions, as w e l l as
maintenance, reliability of instruments/actu- avoiding human errors. However, the need for
ator]control systems and human error. The h u m a n - c o m p u t e r interaction is there. As Lees
release of toxic/flammable substance from a [38] points out, there is little that a computer
valve, leaks from pipe flanges and pump seals, can do to compensate for failures and mal-
failure of process vessels due to corrosion or functions in plant equipment valves and
excess pressure are examples of equipment sensors. Operator/management action is
hazards. Equipment hazards tend to be site- required and should be timely, suitable and
specific. Strictly speaking, the analysis would sufficient. Computer support would certainly
require reference to a particular plant. How- enhance reliability and accident prevention
ever, some guidelines are available for a fre- but it should be recognized that it is only one
quently occurring equipment hazard, namely, subsystem comprising the whole system (the
leaks from valves [36, 37]. Decisions based process plant). Interactions between these is
on models such as leak detection and repair an important consideration. Lees [38] points
(LDAR) [36] and regular maintenance are o u t the importance of the operator in recog-
recommended. nizing patterns, diagnosing faults, decision-
174
making and input c om m a nds to a computer.
Information presentation to operators, either
verbal or from instruments or c o m p u t e r out-
put, should be simple and minimal.
We describe below some possible post-
design procedures in specific cases. In the
manufacture o f carbon m onoxi de, hydrogen
is a by-product. Because hydrogen has a wide
flammable range in air and a low ignition
energy of 0.002 mJ, it is i m p o r t a n t t hat it is
separated from carbon m o n o x i d e [34]. If
such a mixture flows as a stream near person-
nel or a residential district, then an arrange-
m e n t shown in Fig. 4 may be employed.
Because the thermal conductivities o f hydro-
gen and carbon m o n o x i d e are easily differ-
entiable, a thermal conductivity gas analysis
can be e m p l o y e d for hydrogen gas det ect i on
[39, 40]. Regular tests should be carried o u t
on this instrument, such as soap-bubble leak
d e t ectio n using a soap solution. If a Wheat-
stonebridge is em pl oye d, t hen a step current
input may be used to analyse the instrument
in the same way that immersion-type thermo-
meters are analysed [21]. Alarms should be
sufficiently audible and be coupled with ot her
protective measures. Control action should
allow for enhanced removal of hydrogen
either immediately downstream o f the mea-
surement location or at an upstream point.
Worst-case situation should be used to design
and implement the p r o t e c t i o n system
including personnel training and awareness.
Leaks f r o m seals and valves should be assessed
regularly using a portable instrument.
F o r highly hazardous substances such as
m e t h y l isocyanate, there are at least two
CONTROL AND
PROTECTO
I N SYSTEM
~ SAMPLE LINE
I CARBON
MONOXI
CONTAI G DSTREAM
E IMPURITY
NINHYDROGEN
P
Z
Fig. 4. Prevention and protection for a carbon
monoxide stream containing hydrogen as impurity.
levels of post-design decision-making. One is
the national level, and the o t h e r is the plant
operational level. The national level m ay pos-
sess the power to shutdown a plant but such
a decision is usually a pre-design one. Should
the plant concerned be operational, then the
following should be located and possible
provision of sensors, positioned in their
proximity, be considered:
(a) control valves;
(b) relief/other valves;
(c) seals, connections, other weak loca-
tions.
E q u i p m e n t design usually allows for a
factor of t w o or more times the highest pres-
sures and temperatures attainable. This has
proved to be desirable as the Seveso incident
has shown t hat u n e x p e c t e d pressures and tem-
peratures can be attained. C o m p o n e n t s with
moving parts and potential weak spots such as
pipe connections must be isolated for atten-
tion. Possible actions may be depicted as
shown in Fig. 5. A d e t e c t o r should be located
near the potential opening area of a valve
or seal. A gas absorber unit surrounding the
valve should be provided, and should be easily
removable. Alternatively, the emitted gases
can be vented to a flare as suggested by Hust-
vedt [41]. Absorption and flaring may not
be suitable for all gases. Other control devices
include those based on thermal oxidation, the
use of emissions as fuels, carbon adsorption
and condensation [42].
Sensors are to be located at various areas
around the plant with signals activating, upon
SPATIALLY ~ a
LOCATED GAS
TOXIC/FLA~ABLE GAS
ABSORBER ORFLARE
VA,~r
.... ~ ~
I- ~TOXIC/F~MMABLE
DETECTOR
Fig. 5. Provision for leak detection, containment and
protection.

reading threshold values, alarms for personnel
and surrounding residential areas. All sensors
should be regularly analysed for faults either
by methods described in Section 5 of b y
direct testing. Site personnel should carry por-
table gas detectors. The frequency of inspec-
tion of valves, seals and connections for immi-
nent failures and leaks should reflect the
design, k n o w n failure rates of the c o m p o n e n t
and damage due to process operational char-
acteristics or internal forces. Further, this fre-
quency should increase with age of the com-
ponent. Plant shutdown does n o t imply that
the alarm, control and protection systems are
to be disabled. Seveso should be a lesson that
an emission can take place after a plant
shutdown.
Processes with highly flammable substances
must also be provided with fire and explosion
protection. These must respond quickly. New
technology should be assessed and be utilized
depending u p o n the costs and benefits. Micro-
computer-based fire and security systems have
the following benefits [43] ;
(a) enhanced response, such as voice mes-
sages tailored to the location of fire and zones
of safety;
(b) relative reduction in hardware;
(c) relative reduction in system cost;
(d) easily upgraded.
Dust explosions have also received atten-
tion due to increased incidence, especially in
connection with food dusts and polymeric
powders [44, 45]. Such explosions involve
dust particles o f appropriate sizes, air or oxy-
gen environment, and an ignition source.
Minimizing the incidence requires consider-
ations such as:
(a) equipment design
(b) careful operation such as avoiding any
naked flame.
As for the other potential accident situa-
tions, protection must be provided. Methods
are available and discussed b y Palmer [45].
These are based on minimizing cloud forma-
tion, containment, separation of plant, vent-
ing, inerting and automatic suppression. As
Palmer points out, it may be wise to use these
methods in combination rather than singly.
7. CONCLUSION
The present paper has attempted to inte-
grate the various aspects related to safety in
175
process, f o o d and other systems. The follow-
ing factors aid in making difficult the task of
accident prevention:
(a) incomplete, and imprecise knowledge of
processes involved;
(b) human errors;
(c) c o m p o n e n t failure;
(d) lack and/or absence of sensors at poten-
tial sources of accidents;
(e) unsatisfactory man-machine interfaces;
(f) lack of proper/effective communication
and action within a plant, and between the
plant and the public.
There is room for progress in all these
factors. Post-accidents must be used to pro-
vide information for analysis and improve-
ment of safety systems. Post-accident studies
such as that of Theofanous [46] are of value
and should be used for further analysis and/or
decision-making in connection with accident
prevention and protection.
The most severe accidents have shown that
the following were involved: c o m p o n e n t fail-
ure or valve opening due to excess pressure,
lack of early detection due to absence or non-
availability of appropriate sensors, and
absence of alarms and appropriate action.
Disasters involve short time-scales as discussed
in Section 3. Proper system and c o m p o n e n t
design, personnel training, early fault/leak
detection followed by prompt, suitable action
is therefore most important. Reliability of
various components in a safety system,
human errors and communication is also
important. The e c o n o m y of microprocessors
does allow hardware redundancy of computer
support thus providing standby processors in
emergency situations.
Recommendations which can be provided
are:
(1) computerized control system with
manual override;
(2) components should have a manual over-
ride due to, for example, power failure or loss
of compressed air to pneumatic valves;
(3) fixed fugitive gas sensors, with hard-
ware redundancy in critical situations, at
probable locations of gas in case of release.
Periodic testing to be provided;
(4) alarm systems which should be suffi-
cient but not give intolerable false alarms;
(5) system desig n should include intrinsic
safety capabilities [47]. For highly explosive
media, this should also apply to power-
176
consuming devices; design of solenoid valves
for intrinsically safe systems is discussed by
Johnson and Colony [48].
All processes involving highly hazardous
materials should be flexible enough for tech-
nology updating. This could involve, for
example, improved design valve installation
and software addition to the computer
support. Action for implementation during
and immediately after a potential accident
must be thoroughly examined and period-
ically assessed. All preventive and protective
measures should be considered for safety of
the plant, personnel and the public.
REFERENCES
1 F. T. Bodurtha, Industrial Explosion Prevention
and Protection, McGraw-Hill, 1980.
2 M. D. LaGrega and L. K. Hendrian (eds.), Toxic
and Hazardous Waste, Butterworths, Boston,
1983.
3 J. Rasmussen and W. B. Rouse (eds.), Human
Detection and Diagnosis o f System Failures,
Plenum Press, New York, 1981.
4 D. M. Himmelblau, Fault Detection and Diagnosis
in Chemical and Petrochemical Processes, Else-
vier, Amsterdam, 1978.
5 A. C. Barrell and D. C. Thomas, Assessing Man-
agement in Major Hazard Operations, (Jubilee)
Symp. Series No. 73, Inst. Chem. Eng., 1982,
pp. B9 - B17.
6 H. H. Fawcett and W. S. Wood, Safety and Acci-
dent Prevention in Chemical Operations, Wiley,
2nd edn., 1982.
7 D. T. Smith, Maintenance of chemical equip-
ment, in H. H. Fawcett and W. S. Wood (eds.),
Safety and Accident Prevention in Chemical
Operations, Wiley, 2nd edn., 1982, Ch. 13.
8 R. Isermann, Process fault detection based on
modeling and estimation methods, Automatica,
20 (1984) 387 - 405.
9 K. Plamping and P. K. Andow, The design of
process alarm systems, Trans. Inst. Mess. Control,
5 (1983) 161 - 168.
10 M. G. Morgan, Probing the question of technol-
ogy-induced risk, IEEE Spectrum, 18 (Nov.)
(1981) 5 8 - 64.
11 Three Mile Island and the future of nuclear
power, IEEE Spectrum, 16 (Special Issue) (11)
(Nov.) (1979).
12 G. N. Pegg (ed.), High Pressure Measurement
Techniques, Applied Science Publ., Barking, U.K.,
1983.
13 R. L. Anderson and T. G. Kollie, Problems in high
temperature thermometry (Critical Review),
Anal. Chem., 6 (1976) 171 - 221.
14 L. R. Cooper, M. J. Gibson and M. J. Willett,
Institute meetings report on workshop on new
sensors for gas detection, Measur. Control., 16
(1983) 3 5 6 - 357.
15 Gas monitoring, Chem. Indust., (July) (1984) 27.
16 J. N. Driscoll, Flue Gas Monitoring Techniques,
Ann Arbor Science Publishers, 1974.
17 J. Fishetti, Report in IEEE Spectrum, (Jan.)
(1983).
18 T. Wangsmo and H. Riddervild, Improved safety
in fire and gas]ESD systems, Proc. ISA Conf.
Advances in Instrumentation, Vol. 38, part 2,
pp. 1053 - 64, 1983.
19 R. W. Gerstle, Source sampling, analysis and
monitoring of gaseous emissions, in W. Strauss
(ed.), Air Pollution Control, part 3, Wiley, 1978,
Ch. 1.
20 A. L. Jones, Fugitive emissions of volatile hydro-
carbons, Chem. Eng., (Aug./Sept.)(1984).
21 R. K. Chohan and T. W. Kerlin, Temperature
sensors in process control: the importance of
dynamic performance and testing, Proc. Industrial
Temperature Measurement Symposium, Knoxville,
TN, September, 1984, University of Tennessee,
Knoxville, paper 20.01.
22 G. Flanagan, Selecting a volatile organic chemical
detector, Chem. Eng. Prog., (Sept.) (1984) 3 7 -
39.
23 T. W. Rhoads, D. G. Cole and R. L. Norton, Opti-
mizing fugitive VOC emission control, Chem.
Eng. Prog., (Sept.) (1984) 39 - 44.
24 B. R. Upadhyaya, Sensor fault detection and esti-
mation, to appear in Nucl. Safety (1985).
25 A. Gelb (ed.), Applied Optimal Estimation, The
MIT Press, Cambridge, MA, 1974.
26 R. N. Clark, Instrument fault detection, IEEE
Trans. Aerosp. Electron. Syst., Aes-14 (1978)
456 - 465.
27 J. C. Deckert et al., F-8 DFBW sensor failure
identification using analytic redundancy, IEEE
Trans. Auto. Cont., AC-22 (1977) 795 - 803.
28 R. N. Clark and B. Campbell, Instrument fault
detection in a pressurized water reactor pres-
surizer, Nucl. Tech., 56 (1982) 23.
29 J. L. Tylee, A generalized likelihood ratio
approach to detecting and identifying failures in
pressurizer instrumentation, Nucl. Tech., 56
(1982) 489 - 492.
30 Demonstration o f a PWR Suppression Pool Signal
Validation Program, Report NP-3641, EPRI,
California, July, 1984.
31 B. R. Upadhyaya and M. Skorska, Sensor fault
analysis using decision theory and data-driven
modeling of pressurized water reactor subsystem,
Nucl. Tech., 64 (1984) 70 - 77.
32 T. Tsunoda, Application of temperature fluctua-
tion to anomaly detection in fuel assemblies, J.
Nucl. Sci. and Tech. (Japan), 13 ( 1 9 7 6 ) 1 0 3 -
110.
33 W. Reimche and D. Stegemann, Velocity profile
and flow measurements in liquid sodium by signal
correlation of fast intrinsic thermocouples and
electromagnetic flowmeters, Proc. SMORN-IV,
Reactor Noise, Dijon, France, October, 1984.
34 A. S. Goldfarb et al., Organic Chemicals Manufac-
turing Hazards, Ann Arbor Science Publishers,
Michigan, 1981.

35 N. I. Sax, Dangerous Properties of Industrial
Materials, Van Nostrand, 4th edn., 1975.
36 J. L. Hess and T. A. Kettleman, Predicting volatile
emissions, Chem. Eng. Prog., 79 ( N o v . ) ( 1 9 8 3 )
40 - 42.
37 R. G. Wetherold, G. E. Harris, J. A. Steinmetz
and J. W. Kamas, Economics of controlling
fugitive emissions, Chem. Eng. Prog., 79 (Nov.)
(1983) 4 3 - 48.
38 F. P. Lees, The improvement of plant integrity
using a process control computer, Measur. Con-
trol, 5 ( A u g . ) ( 1 9 7 2 ) 3 2 2 - 325.
39 D. J. Huskins, Quality Measuring Instruments in
On-Line Process Analysis, Ellis Harwood, Chiches-
ter, U.K., 1982.
40 D. J. David, Gas Chromatographic Detectors,
Wiley, New York, 1974.
41 K. C. Hustvedt, Control strategies for VOC equip-
ment leaks, Environ. Prog., 2 (4) (1983) 2 6 2 -
264.
177
42 J. W. Blackburn, Organic emission control
devices, Environ. Prog., i (3) (1982) 182 - 189.
43 A. Saigal, Computerized fire and security system
performance, Heat. Pip. Air Condit., (Nov.)
(1984) 145 - 148.
44 J. A. Radley (ed.), Starch Production Technol-
ogy, Applied Science Publishers, U.K., 1976,
Ch. 5.
45 K. N. Palmer, Dust Explosion Hazards in the Bulk
Handling o f Polymers, Symp. Series No. 73, Inst.
Chem. Eng., 1982, pp. C39 - C57.
46 T. G. Theofanous, The physicochemical origins of
the Seveso accident -I, Chem. Eng. Sci., 38
(1983) 1615 - 1629.
47 R. Garside, Intrinsic safety: prevention not cure,
Adv. in Instrum., ISA, 37 (3) (1982) 1645 - 1654.
48 T. L. Johnson and R. J. Colony, The design and
application of solenoid valves for intrinsically safe
systems, Adv. in Instrum., ISA, 37 (1982) 1263 -
71.
 Methods for sizing dust explosion vent
areas: a comparison when reduced
explosion pressures are low

G. A. Lunn
Explosion and Flame Laboratory, Health and Safety Executive, Harpur Hill, Buxton,
Derbyshire SK1 7 9JN, UK

Various methods that are available for estimating the explosion venting requirements of weak
dust handling equipment are compared. The regression analysis that Simpson has performed on
the well known K,, nomographs is shown to extend to Kst values less than 50 bar m s- and to
reduced explosion pressures between 1 .l bar a and 1.2 bar a.

(Keywords: explosion; vents; regression analysis)

The dust from many materials processed in the increasing explosibility. Much dust handling equip-
materials handling and other industries is explosible. ment, especially in the UK and USA, is relatively
One method of protection commonly used in industry weak, and the reduced explosion pressure must be kept
to minimize the potential damage from a dust explosion low. Unfortunately, the nomographs in VDI 3673
is explosion venting. By allowing the explosion to issue cannot be used if the reduced explosion pressure, Pred,
from vent openings in the equipment the internal needs to be less than 0.2 bar g. The vent ratio
explosion pressure is kept low. Totally enclosed explo- approach3 has been used for vent sizing for such weak
sions can generate overpressures as high as 10 or equipment, but this is an empirical method which can
11 bar g; vented explosions need only generate explo- lead to gross oversizing. While other methods have
sion overpressures of 0.2 bar g or lower. been developed for estimating venting requirements at
One method of estimating the requisite vent area is low overpressures, e.g. the Rust method4 and the Hart
the well known K,, nomograph approach described in methods, for one reason or another they have not had
VDI 3673 (Ref. 1). The information needed to apply wide use.
this technique is: the opening pressure of the vent cover Recently, however, additional approaches for vent
(P,,,), the volume of the equipment (V), the internal area prediction at low reduced explosion pressure have
overpressure appropriate to the strength of the equip- become available. These are: an extension of the K,
ment, i.e. the reduced explosion pressure (P&, and nomographs6; a simple to use equation in the new
the explosibility of the dust (Kst). The K,, value is NFPA 68 (Ref. 7); and there is always the possibility of
measured under standard conditions in a test vessel extending the regression analyses that have already
usually of either 1 m3 or 20 1 capacityz. It is essentially a been performed on the original K,, nomographs*. This
measure of the rate of pressure rise and is calculated paper discusses a comparison of these methods.
from

K=dP
Discussion
s, (1)
( dt 1,,x~~ Extended nomographs and NFPA 68 equation
where (dP/dt),, (bar s-1) is the maximum rate of Figure I is from a paper by Grolmes and Fauskeg. It
pressure rise and V(m3) the vessel volume. The K,, demonstrates a relation between the dimensionless
value (bar ms-1) is considered to be volume- vent area, A .,/W/J, and the reduced explosion pressure
independent. for a K,, = 300 bar ms-1 dust, i.e. a dust at the top of
Depending on the Kst value, dusts can be allocated the St2 group. A v is the vent area, usually in m2.
to one of four groups: group StO, where Kst = 0; group Above a reduced explosion pressure of 10 psig
Stl, where 0 > K,, > 200; group St2, where 200 > (0.7 bar g), a high pressure sonic venting model de-
K,, > 300; and group St3, where K,, > 300. Group St0 veloped by Epstein et al. O has been used to calculate
contains non-explosible dusts; groups Stl-St3 have A v/Vzk
A, B[P,,,/y - Predh][PmBxh - l]
- =
Received 7June I989 V/3 p redqpmx /I, - I] - Pmax/YIP,,ly - l] (2)
099&4230/89/04020&09$3.00
0 1989 Butternorth & Co. (Publishers) Ltd

200 J. Loss Prev. Process Ind., 7333, Vol2, October

 Methods for sizing dust explosion vent areas: G. A. Lunn

where P,,, is the maximum explosion pressure mea- when 6C = 0.25 Equation (3) gives results close to the
sured in an enclosed explosion. For the purposes of this K,, = 300 barms-1 nomograph predictions. as shown
equation all the pressures have been normalized by in Figure I, from the paper by Grolmes and Fauske.
reference to PO, the initial pressure, and are absolute The nomograph predictions follow a slightly curved
pressures rather than gauge pressures; y is the ratio of line, whereas Equation (3) produces a straight line.
specific heats, C,/C,; and the constant B = An extension of the K,, nomograph method to
61S,p,/CdG includes combustion and vent flow terms; reduced explosion pressures of 0.05 barg has been
S, is the fundamental burning velocity (the model producedl2. Figure 1 and Figure 4 from Ref. 12 are
having originally, been developed for gas explosion combined in Figure 2. As the K,, value changes, a
venting); pU is the density of the unburnt combustible series of approximately parallel lines are produced. The
mixture; r3 a turbulence factor; Cd a flow discharge values of 6C that produce agreement between the
coefficient; and G a mass rate of discharge per unit results of Equation (3) and the extended nomographs
vent area. The authors set l/y as unity and consider the are given in Table 1. There is an approximately linear
uncertainties introduced by this approximation to be relation between 6C and the K,, value. Equation (3) is
acceptable considering the uncertainties in the calcula- included in the new NFPA 68 (Ref. 7) in the form:
tion of B.
A, = C~As/(Pred - P,,)2 (4)
5 where A, is the total internal surface area of the
enclosure and C, has units of (pressure)lp. Equation
(4) is considered applicable to any shape of enclosure
while Equation (3) is limited to compact enclosures, in
which A I z 6VU.

0 011
. 0.2 0.5 1 5 10 50

Maximum pressure Pred(psig)

Figure 1 Dimensionless vent area versus reduced explosion

pressure calculations from vent models by Swift and Epstein
(P, = 0.7P,J Kay: -, K = 300 dust nomogram; --, Equa-
tion (3) 6C = 0.25; - - -, Equation (2) B = 0.066. (Reproduced with I I III1 I t I ,I,,(
0.011 ' "'111'
permission from Grolmes and Fauske (Ref. 9)) 0.2 0.5 1 5 10 50

Maximum pressure Pred(pslg)

When the reduced explosion pressure is above
15 psig (1 barg) Equation (2) follows the expected Figure 2 Comparison of vent models
behaviour of vented explosions much better than the
K,, nomographs. The dimensionless vent area falls to
Table 1 Relation between K, value and 6C from Equation (3)
relatively low values as the allowable Prcd increases and and C, from Equation (4)
approaches P,,,. The model also demonstrates that
decreasing the vent opening pressure, P,,,,, to very low K,, (bar ms-I) 6C C, [psi)1/2
values has only a limited effect on the vent size.
10 0.008 0.005
For reduced explosion pressures below 3 psig 0.016 0.010
(0.2 barg), Grolmes and FauskeY use an equation g 0.024 0.015
40 0.033 0.021
derived from a subsonic venting model derived by Swift
50 0.042 0.027
and Epstein I I : 75 0.064 0.041
A P 00 l/2 100 0.066 0.055
-
v/2/3
= 6C
t-1P
150
200
250
0.130
0.163
0.196
0.064
0.105
0.127
where P,is the ambient. atmospheric pressure and 300 0.253 0.163
reduced explosion overpressure, 406 0.326 0.210
with pressures in absolute units. The 500 0.364 0.246
600 0.466 0.300
dimensionless constant 6C = B [( Pmax/Po) I/Y 7 11, and

J. Loss Prev. Process Ind., 1989, Vol2, October 201

Methods for sizing dust explosion vent areas: G. A. Lunn

A comparison of Equations (3) and (4) shows that Table 2 Comparison between regression analysis, Equation (5).
6C = 6CIfi;. With P,, = 15 psi, Cl equals 0.105 and extended K., nomographs (figures l-3 in Ref. 12) for K,
(psi)/* when K,, = 200 barms-1, 0.16 (psi)1/2 when values below 50 bar m s-l
K,, = 300 barms-1 and 0.30 (psi)/2 when
.%PP
K,, = 600 bar m s -I. A previous comparison6 between K,
the extended nomographs and Equation (4) suggested (bar ms-1 Lzg, i:rgI Regression Nomographs
that C, = 0.1 (psi)/2 for St1 dusts, CI = 0.14 (psi)@ Equation (5) (figs l-3 in
Ref. 12)
for St2 dusts, and CI = 0.25 (psi)/* for St3 dusts gave
reasonable agreement. The values of Cl recommended 40 0.1 0.2 0.077 0.072
in the new NFPA 68 (Ref. 7) are: 0.10 (psi)l for St1 0.6 0.035 0.036
1.0 0.025 0.024
dusts, 0.12 (psi)l/z for St2 dusts and 0.20 (psi)/* for St3 1.4 0.020 0.019
dusts. 1.8 0.016 0.015
Swift has suggested a value of 0.20 (psi)ilr for 2.0 0.015 0.014

highly turbulent explosions of St1 and St2 dusts13. 40 0.2 0.3 0.069 0.070
0.6 0.042 0.044
With the values of 6C given in Table I, Equation
1 .o 0.029 0.030
(3) can be used to size vent areas for compact 1.4 0.023 0.022
enclosures over a range of K,, values when reduced 1.8 0.019 0.018
2.0 0.018 0.017
explosion pressures are less than 0.2 bar g. A value for
C, can readily be calculated (Table I) when a value of 40 0.5 0.6 0.071 0.065
1.0 0.048 0.045
6C is ktown and the use of Equation (4) can thus be 1.4 0.037 0.034
extended over the range of K,, values for non-compact 1.8 0.030 0.026
enclosures. The qualifications as to the use of this 2.0 0.027 0.024
equation and siting of vent openings that Swift details 30 0.1 0.2 0.068 0.055
should be borne in mind. 0.8 0.027 0.027
1.0 0.019 0.018
The vent opening pressure, P,,,,, is not included as 1.4 0.015 0.014
a variable in Equation (3). The data in Figure I 1.8 0.012 0.011
suppose that P,,,, = 0.7 Pred (bar g) and the extended 2.0 0.012 0.010
nomographs in Ref. 12 that P,,,, = OSP,d (barg), but 30 0.2 0.3 0.053 0.050
these are only recommendations and do not influence 0.6 0.032 0.032
1.0 0.022
the calculations in any way. 1.4 0.017 0.016
1.8 0.015 0.014
Extended nomographs and regression anaIyses 2.0 0.013 0.013
In recent years, equations have been fitted to the 30 0.5 0.6 0.054 0.049
original K,, nomographs published in VDI 3673. These 1.0 0.036 0.034
1.4 0.028 0.025
are not theoretical derivations, but are regression 1.8 0.023 0.020
analyses, and the equations can be used for rapid 2.0 0.021 0.018
calculation of vent areas using computer programs. 20 0.1 0.2 0.039 0.036
An equation developed by Simpson* is recom- 0.6 0.018 0.018
mended in the new NFPA 68: 1.0 0.013 0.013
1.4 0.010 0.0095
A/VZ!J = aKs,bPred 1.8 0.008 0.0075
(5)
2.0 0.008 0.007
where 20 0.2 0.3 0.036 0.034
a = 0.000571 exp[2P,,,,]; 0.6 0.022 0.021
b = 0.978exp[- O.l05P,,,,]; 1.0 0.015 0.015
c = - 0.687exp[0.226PS,,,]. 1.4 0.012 0.011
1.8 0.010 0.008
P,,,, and Pred are bar g, K,, is bar m s-1, V is m3 and A v 2.0 0.009 0.008
is m*. The correlation limits are the same as the limits 20 0.5 0.6 0.037 0.031
to the original K,, nomographs: 1.0 0.025 0.022
1.4 0.019 0.017
2 barg > Pred > P,,,, + 0.1 barg 1.8 0.016 0.013
2.0 0.015 0.012
0.5 bar g > P,,,, > 0.1 bar g
10 0.1 0.2 0.020 0.018
6OO>K,,>50barms-1 0.6 0.0092 0.009
1.0 0.0065 0.006
1000m2>V>1m~ 1.4 0.0051 0.0045
1.8 0.0043 0.0036
2.0 0.0040 0.0034
K,, values below 50 barms-1. The nomographs ex-
tended to K,, values below 50 barms-1 have been 10 0.2 0.3 0.018 0.017
0.6 0.011 0.011
shown previouslylz. A comparison between the regres- 1 .o 0.0077 0.0075
sion Equation (5) and predictions from these nomo- 1.4 0.0061 0.0055
graphs is given in Table 2. The values are close enough 0.0051 0.0045
0.0047 0.004
to indicate that the regression curves can be used for

202 J. Loss Prev. Process Ind., 1989, Vol2, October

 Methods for sizing dust explosion vent areas: G. A. Lunn

Table 2 kont) Table 3 (cont.1

AlflB AIV?P
P
karms-) 2::) kg, Regression Nomographs kr m s-j) fdG g) Regression Nomographs
Equation (5) (figs l-3 in Equation (5) (fig 18, Ref. 121
Ref. 12)
75 0.05 0.32 0.275
10 0.5 0.0 0.019 0.018
0.075 0.25 0.22
1.0 0.013 0.011
0.10 0.21 0.19
1.4 0.010 0.008
0.125 0.18 0.176
1.8 0.0084 0.007
0.15 0.185 0.18
2.0 0.0077 0.0065
0.175 0.15 0.15
0.20 0.14 0.14
KSt values between 50 barm s-1 and 10 barm s-1 with
100 0.05 0.42 0.38
confidence. The other correlation limits should remain
0.075 0.33 0.32
as listed earlier. 0.10 0.28 0.27
0.125 0.24 0.24
Reduced explosion pressures below 0.2 burg. A com- 0.15 0.22 0.22
0.175 0.20 0.21
parison between Equation (5) and predictions from the 0.20 0.19 0.19
extended nomographn is given in Table 3. PStat (bar g)
has been taken as 0.5 Pred (bar g) for the regression 150 0.05 0.83 0.58
calculation. The comparison indicates that there is 0.075 0.49 0.45
0.10 0.41 0.39
generally a reasonable agreement at reduced explosion 0.125 0.38 0.35
pressures of 0.1 barg and above, but below 0.1 bar g 0.15 0.32 0.32
the regression curve gives a high value compared with 0.175 0.30 0.29
0.20 0.28 0.27
the nomograph.
200 0.05 0.84 0.72
Table 3 Comparison between Regression Analysis, Equation 0.075 0.85 0.58
(5). and extended K, nomograph (figure 18, Ref. 12). for 0.10 0.54 0.51
Reduced Explosion Pressures below 0.2 bar g: Pst,, (bar g) = 0.5 0.125 0.47 0.45
Prsd(bar g) 0.15 0.43 0.42
0.175 0.39 0.38
0.20 0.38 0.38
A/~~
KSl
ibarms 1 PdeargI Regression Nomographs 250 0.05 1.04 0.91
Equation (5) (fig 18, Ref. 12) 0.075 0.80 0.74
0.10 0.87 0.84
10 0.05 0.045 0.034 0.125 0.59 0.58
0.075 0.035 0.029 0.15 0.53 0.51
0.10 0.029 0.025 0.175 0.49 0.47
0.125 0.028 0.022 0.20 0.45 0.44
0.15 0.023 0.020
0.175 0.021 0.019 300 0.05 1.24 1.09
0.075 0.98 0.88
20 0.05 0.088 0.075 0.10 0.80 0.76
0.075 0.088 0.082
0.125 0.70 0.88
0.10 0.058 0.054 0.15 0.83 0.81
0.125 0.051 0.047
0.175 0.58 0.58
0.150 0.046 0.043 0.20 0.54 0.52
0.175 0.042 0.039
0.20 0.039 0.038
400 0.05 1.84 1.40
30 0.05 0.131 0.108 0.075 1.27 1.12
0.075 0.102 0.090 0.10 1.08 0.95
0.10 0.085 0.079 0.125 0.93 0.87
0.125 0.075 0.070 0.15 0.84 0.80
0.150 0.088 0.084 0.175 0.77 0.74
0.175 0.082 0.059 0.20 0.71 0.70
0.20 0.058 0.054
500 0.05 2.04 1.83
40 0.05 0.174 0.150 0.075 1.58 1.38
0.075 0.135 0.123 0.10 1.32 1.19
0.10 0.113 0.110 0.125 1.15 1 .Ol
0.125 0.099 0.097 0.15 1.04 0.94
0.150 0.089 0.088 0.175 0.95 0.88
0.175 0.082 0.081 0.20 0.88 0.83
0.20 0.077 0.078
50 0.05 0.218 0.195 800 0.05 2.43 1.90
0.075 0.187 0.182 0.075 1.88 1.55
0.10 0.141 0.142 0.10 1.58 1.34
0.125 0.123 0.128 0.125 1.38 1.20
0.150 0.111 0.115 0.15 1.24 1.10
0.175 0.102 0.102 0.175 1.14 1.00
0.20 0.095 0.097 0.20 1.08 0.93

J. Loss Prev. Process Ind., 1999, Vol 2, October 203

Methods for sizing dust explosion vent areas: G. A. Lunn

Extended nomographs and the vent ratio approach the venting process developed by Heinrichir. Accord-
The vent ratio approach requires measurements in the ing to this model, the vent area is given by
Hartmann apparatus3 a dust explosibility apparatus
developed in the USA and now generally superseded
by the 1 m3 and 20 I test vessels. The difficulty in using
the vent ratio approach when only the K,, value of the (6)
dust is known is that rates of pressure rise measurement
in the Hartmann apparatus cannot be satisfactorily
correlated with K,, values. Data glveni4 suggest that where VL is a test vessel volume, (dP/dt)Prrd_VL is the
the vent ratio approach is applicable to low K,, dusts rate of pressure rise measured at P,,, in the test vessel,
only.
P, is the atmospheric pressure, and Mw the molecular
The vent ratios applicable to equipment volumes weight, taken to be that of air. Equation (6) can be
up to 30 m3 are given in Table 4. The rate of pressure modified to incorporate the K,, value and a factor,
rise refers to Hartmann bomb data. Above 30 m3 derived from experiment, which relates (dP/dt) P,,,V,. to
volume, the vent ratio decreases linearly to a value of
(dP/dt) maxas measured in an enclosed vesselb.
l/25 m-i at 300 m3 for the dusts of lowest explosibility.

Harts equation. Hart has derived an equation for

Table 4 Ratios for the vent area method for equipment volume vent areas that explicitly includes Plta,s. Hart begins
up to 30 rn3 with an equation for the explosion pressure in a vented
container:
Maximum rate of pressure Vent ratio (m2/m3 = m-l)
rise (dP/dt),., (bar s-l)
WR 711
4 350 f f6.1 (7)
350-700 1 f4.6 p= v- Vr(l-2)
< 700 113.1

where P is the internal pressure at any time during the

Table 5 gives some comparisons of predictions explosion, W is the mass of gas present, To the initial
temperature, V the volume of the vessel, VF the
from the extended nomographsi* and the vent ratio
volume of the fireball, PO the initial pressure and P,,,
approach. Strictly, the vent ratio approach acts to limit
the maximum explosion pressure in a closed container.
the reduced explosion pressure to 0.14 barg. For
When VF = V the flame completely fills the vented
approximate agreement when the vent ratio equals
vessel and
l/6.1 m-i, the K,, is around loo-150 barms-*, but it
has been suggested3 that the true comparison is with
K,, = 50 bar m s -1. Other comparisons in Table 5 show p = $P_.%
that, for approximately like conditions, the nomograph
gives lower vent areas than the vent ratio approach. where W,, is the mass of gas originally present. If it is
assumed that the reduced explosion pressure is mea-
Comparison of the nomograph approach and the sured at the moment flame fills the vessel and the
methods of Hart and Rust burning rate is at a maximum, then P = Pred in this
The K,, nomographs are based on a simple model of equation. Equation (7) can be differentiated to give:

Table 5 Comparisons of predictions from extended nomograph (figure 18 in Ref. 12) with vent ratio approach (P,d = 1 .I4 bar a)

Volume (m3) 30 70 100 200 300 500 700 1000

Vent area (m*)

Vent ratio l/6.1 mm1 4.9 6.3 7.3 10.0 12.0 20.0 28.0 40.0
at 30 m3, decreasing
linearly to
l/25 m-l at300 m3
Ref. 12:
K, = 50 bar m 5-l 1.16 2.04 2.58 4.1 5.4 7.55 9.46 12.0
K, = 100 bar m s-l 2.25 3.95 5.0 8.0 10.44 14.7 18.4 23.3
K, = 150 bar m .sl 3.28 5.77 7.32 11.6 15.23 21.4 26.8 34.0
K, = 200 bar m 5-l 4.25 7.47 9.48 15.0 19.71 27.7 34.7 44.0
Vent area (m2)
Vent ratio l/4.6 m- 6.52
Ref. 12:
K, = 100 bar ms- 2.25
Vent ratio l/3.1 m-l 9.68
Ref. 12:
K, = 200 bar m s-l 4.25

204 J. Loss Prev. Process Ind., 1989, Vol2, October

 Methods for sizing dust explosion vent areas: G. A. Lunn

-1
avF

at w
(16)

Substituting Equation (16) into Equation (13) gives:

aw
+v- (-1
VFF+J at t.5
A,=

VI RTdCd
P,Cd.
T0.S P SIX

where 0.85 has been substituted for

(17)
The similarities between Equations (17) and (6) are
(1-G). clear. (The molecular weight is included in the R term
At the reduced explosion pressure, dP/dt = 0, and so in Equation (17), and although T and To are in
0.85W actuality different, they could be taken as equal without
much loss in accuracy. P,,,, could be taken as equal to
v_F(l PO ) (%I, =-Fz).. (9)
P&). The equations are not exactly the same, howev-
P max
er; if P,,, is taken to be equal to Prcd, there is an extra
At the moment the vent opens, from Equation (7) P&O. in the denominator of Equation (l7), although
WR T,, Pred = unity at low values of P&.
(10) However, the main differences between Equations
V- VF(, _Z!_) = ps1,1 (17) and (6) are merely due to different vent outflow
equations. Thus, if Heinrichs equation for (- dP/dt)
Hart assumes that Equation (10) holds while the is modified to calculate - dW/dt:
pressure rises from P,,,, to Pred, i.e. the fireball size
does not grow appreciably in the time taken for the
pressure to accomplish this rise. It is by this assumption
that P,,,, is introduced into the final equation. How- P
= CdA.-$$~2RT(P,.ci(P,.d - P,))
ever, the same assumption could be applied in reverse,
i.e. replacing P,,,, in Equation (10) by Pred, and with and this equation replaces Equation (11) in the
more justification. However, Harts argument will be derivation of Equation (17), then
maintained and P,,,, used here.
Hart calculates d W/dr by
dW
- = Js(Pr.,, - P,)o.5 (Pred < 2 bar a) (11) P Stat
dt
X
where J is a constant, Cd the discharge coefficient, A, P
RToCd +hT(Pre&d - pa))
the vent area, P, the atmospheric pressure, and T is
the temperature of the gas being vented. (18)
Hart calculates (dVF/dr)w by
The temperature term Tin Equation (18) strictly refers
to the temperature of the vented gas, but it is assumed
(12) unburnt material is being vented and so To/T = 1;
where U is a burning velocity, The constant, 30.71, similarly, P,@, = Pred. Thus, Equations (18) and (6) are
includes an expansion factor of 6.35. By substitution of almost identical, remembering that the molecular
Equations (IO)-(12) into Equation (9) and replacing weight is included in R the gas constant, in Equation
VF).h7 by f.67 W7, where f = ~/F/V,,: (18).
A = (0.85)(30.71)T-P,,,J;b-h7V.h7U Hart presents his equations in the form:
(13) A = K(PstadV0~67
JCd(Pred - P,).RT,, (P& < 2 bar a) (19)
where 7 = fat the maximum burning rate. In a closed
(Pred - l).5
container, Hart shows where K is an experimental constant, specific to each
material, concentration, degree of turbulence and size
(14) of ignition source.
When Pred > 2 bar a, the equation becomes:
where VL is the volume of the closed test vessel and P*
is the internal pressure when (dP/dt),,, occurs. In the (20)
vented container Equation (14) may equally well be
written: The factor K is defined by Hart as:
K = (0.85)(30.71) T05fC0.67 U/JRToCd (21)
V.33(30.71)(0.85)P,dz (15)
and substitution of Equation (15) into Equation (21)
and by the cube root law: gives

J. Loss Prev. Process Ind., 1999, Vol2, October -205

Methods for sizing dust explosion vent areas: G. A. Lonn

12
K = To.s(dP/dt)red.VV0.33/JRTgCdPred2

= TO-sK,,/JRToCdP,,d2 (22) \
11 \
i.e. K 0~ Ksr/(Pred)*_ \
To test the proportionality of K to K,, and Prd and \
to assess the variation of K with P,,,,, Equations (19) \
10
and (20) have been re-arranged to give K in terms of \
A,, V, Pred and P,,,,. Values of K can then be \
calculated using the K,, nomograph predictions. These \
equations and the results of the calculations are given in 9
\
Table 6. The K values calculated in this way are similar \
to the K values for the St groups given in Harts papers. \
Generally, to within f lo%, K is independent of a
volume (up to 1000 m3) and is independent of P,,,, for ..
l
I..._. .
l
. . . . .
I:
P,,,, values of 1.1 bar a and 1.2 bar a. K does vary with l
P rcd, but at a given Predr K is proportional to K,,, up to 7 l
K,, = 400 bar m s-1. This proportionality is given at b 0
z
P red values of 1.2 bar a to 1.3 bar a by K = h
7.52 X K&0000, at P,d = 2.0 bara by K = 6
5.81 X K,,/lOOOO, and at P,d = 3.0 bar a by K =
4.88 x K,,,/lOOOO. The proportionality factor is plotted
against P& in Figure 3.
Over the ranges of the K,, nomographs where \

these proportionality factors apply, substitution into

Harts vent area Equations (19) and (20) gives good
agreement with the K,, nomographs. It should be noted
that K,, values from 50 bar ms-L down to 10 barm s-i
have been incorporated in these calculations without
problems; it is extrapolation to P,,,, and P& values less
than 1.1 bar a and 1.2 bar a, respectively, that presents
a difficulty because of the unknown behaviour of the
proportionality factor. One possible extrapolation is 2-
0.01 0.1 1 10
shown as a dashed line in Figure3, i.e. a linear
extrapolation to low values of Pred, but apart from P redoverpressure(barg)
simplicity there is no justification for this extrapolation.
Figure 3 Factor relating K from Harts equation to K, value as a
Alternatively, the extended nomographsl2 can be used function of P,
to provide the proportionality factors at low Pred by
substituting the various values of P& and a P,,,, equal process yields an approximately constant value of the
to 1.1 bar a into the equations given with Table 6. This proportionality factor below a P& of 1.2-1.3 bara.

Table 6 Calculations of K by Harts equations from the & nomographs (Pw~ in bar al

K, P La, = 1.1 bara P** = ,.*Laara P,,n= 1Sbara

(bar ms-I
P,@=l.* ,-,_,=2 ~,~=3 P,ti=,.2Pw=2 PM=3 P,d=,.3P.d=2 P.d=3 P,ad=l.3Prd=2 P.e,,=3 P,e.,=1.6Pre.d=2 P&=3 Pr..~1.6P,w,=Z P,ti=3

V- V= v= V= V= V= V= V= V= V= v= v= v= V= V= v= . V- V=
1 ma Id 1 m3 1000 1000 1000 ld lrd lm3 1000 IWO 1000 1m3 lm3 lm3 lOiJ0 lOOa 1000
m3 m0 m3 m m m3 m3 m3 m3

10 0.00732 0.00732 0.00554 00046140.00776 0.00767 o.oo608 0.0512 0.00852 0.00740 0.00774 0.0073 o.w59
20 0.0146 0.0188 0.0144 0.0113 0.0095 0.0148 0.0121 0.0146 0.0121 0.0100 0.01&a 0.0140 0.0122 0.0154 0.01467 0.0117
30 0.0220 0.0173 0.0224 0.0168 0.0139 0.0233 0.0192 0.0160 0.0228 0.0175 0.0155 0.0248 0.0217 0.0190 O.Ql238 0.022 0.018
40 0.0301 0.0235 0.0195 0.0289 0.0227 0.019, 0.029, 0.0246 0.02W O.wo6 0.0250 0.0210 0.0330 0.0287 0.0225 0.01315 0.0293 0.0235
50 0.0x34 0.0309 0.0259 0.0378 0.0295 0.0248 0.0383 0.0312 0.0254 0.0374 0.0309 0.0282 0.0413 0.0360 0.0315 0.0392 0.0367 0.0295
75 0.0569 0.0445 0.0375 0.0569 0.0445 0.0388 0.0548 0.045a 0.0375 0.054S 0.0450 0.0375 0.0645 0.0567 0.050 0.0609 0.0567 0.046
100 0.0772 0.0800 0.0511 0.0752 0.0591 0.0491 0.0730 0.0808 0.0500 0.0717 0.0583 0.0487 0.0557 0.0733 0.066 0.0745 0.074 0.060
150 0.1100 0.0864 0.0723 0.1057 O.w),S 0.0845 0.1118 O.DSl7 0.0782 0.10%5 O.SWO 0.0750 0.1265 0.110 0.035 0.123s 0.1133 0.092
200 0.1504 0.1182 o.oSS5 0.1423 0.1118 0.0954 0.1438 0.1192 0.0987 o.w3S 0.,,87 O.lWa 0.1652 0.1487 0.130 0.1576 0.1467 0.120
250 O.la# 0.14!55 0.1214 0.1788 0.136 0.1172 0.1825 0.1500 0.1250 0.1825 0.15W 0.1250 0.2169 0.1866 0.166 0.2014 0.1866 0.152
300 0.2561 0.1727 0.1432 0.2073 0.1636 0.1384 0.2145 0.175 0.1475 0.2100 0.1750 0.1475 0.258 0.2267 0.200 0.2375 0.2200 0.180
400 0.2887 0.2273 0.10 0.2805 0.2136 0.1814 0.2830 0.2333 0.1937 0.2739 0.2250 0.1900 0.3150 0.2667 0.22 0.2&u) 0.2667 0.215
600 0.3374 0.2727 0.2318 0.3252 0.2545 0.2166 6.3515 0.2317 0.2375 0.3286 0.2708 0.2312 0.3408 0.3WO 0.23 0.3oSS 0.2933 0.240
So0 0.3882 0.3091 0.2591 0.3653 0.2908 0.2455 0.4016 0.3330 0.2750 0.5880 0.3167 0.2887 0.3504 0.3133 0.24 0.3873 0.3133 0.255

For P,, < 2 bar a, K = A(P,~_,).S/(P,,)V0.61; Pd > 2 bar a, K = A(P,d)/2(P,)vD-67

206 J. Loss Prev. Process Ind., 1989, Vol2, October

 Methods for sizing dust explosion vent areas: G. A. Lonn

Thus, if Pred is below 1.2 bar a, a value of vent area can

be first obtained from the extended K,, nomograph.
Because this vent area strictly applies when P,,, equals
then, the final equation for the vent area, A , assuming
1.1 bar a, a correction to the vent area for the lower
spherical combustion, is identical to Heinrichs equa-
value of P,,., can be made by multiplying the vent area
tion multiplied by a factor (P&P,,) (J. A. Abbott,
by the factor (P&/1.1, where P,,, is in bar a. This
private communication). This version of Rusts equa-
procedure could essentially fine-tune the predictions
tion:
obtainedIT, though as the minimum value of the ratio is
(1.02/1.1) = 0.927, it is doubtful whether such a
procedure is worthwhile.
A,=

Rust4 has developed an equation cd (Pred(Pred - pa))

Rusts equation.
which has been recommended for mild explosions. In
the derivation, a spherically expanding flame front is will always give higher vent areas than those predicted
assumed, at least initially, and equations for the volume by the K,, nomographs by the ratio (P&/P,I). At
of burnt gases and the internal pressure are given in reduced explosion pressures below 0.2 bar g, this factor
terms of time from the moment of ignition: will be relatively small.
The derivation discussed here is merely to show
(23) that the thinking behind the Rust and Heinrich venting
models is so similar that at very low reduced explosion
where I is time after ignition, U is the burning velocity pressures there should be very little difference between
and Vt the volume of expanded gases from a spherical them. Rust uses somewhat different discharge equa-
flame if they remain at a pressure of PO. P,,, is the tions in his own derivation, leading to an equation:
maximum explosion pressure in an enclosed vessel. The
C~(P,,,V)wC/
pressure at any time during the explosion is given by: A,= (31)
- P,)W
(Pred

(24) where C2, incorporating a shape factor, is a constant

and K is a dimensional factor equal to (4/3)nS,IP,,,.
Rusts condition for venting is that, at Predrthe volume Due to a confusion of burning velocity, S,, and flame
outflow from the vessel, Q, equals the rate of volume speed16 this equation overestimates either vent area or
expansion due to combustion, dV,/dt. Differentiating reduced explosion pressure for vented gas explosions 11.
Equation (23) gives: For dusts

(25) (dP/dt) = ls (32)

and differentiating Equation (24) gives and replacing K in Equation (31) leads to
A,=>
P max 213 ~213
(33)
3 P2p(P,d - P,)p
Therefore Rust uses Hartmann data to evaluate K, but if
Equation (33) is assumed to apply when 20 1 data is
T=(z),&)= F(!$),, (27) available, then P equals P,,, and
C2V2/3VLL/3(dP/dt)v,
A,= (34)
In a vented vessel, (dP/dt)p,d is strictly zero, but in 3(pre_d - pa)
Equation (27) (dP/dt)pred refers to a measurement in a
If (dfb) VL is taken to be (dP/dt),, p,, and an
closed vessel and, as in Heinrichs derivation, is the rate appropriate value of Cr introduced, then Equation (34)
of pressure rise due to the explosion. By the cube root
differs from Equation (6) only by a factor (Pred)12,
law, Equation (27) becomes: which at low values of P& (less than 1.2 bar a) will not
exceed 1.095.
(28)

where VL is the volume of the test vessel: Thus, as with Conclusions

Heinrichs equation some relation is needed to give
The various methods for calculating vent areas for dust
(dP/dl)PrEd,VL in terms of (dP/dt)max,vL. The final form
explosions when the reduced explosion pressure is
of the Rust equation then depends on the expression
below 0.2 bar g have been shown to give generally good
for Q.
agreement with K,, nomograph extensions by means of
If the expression used by Heinrich for pressure loss
Heinrichs equation. The regression analysis equation
due to venting is modified thus:
of the K,, nomographs recommended by NFPA 68 can
be used with confidence for K, values below
50 bar m s-l and for reduced explosion pressures down

J. Loss Prev. Process Ind., 1989, Vol2, October 207

Methods for sizing dust explosion vent areas: G. A. Lunn

to 0.1 bar g. (P,,, (bar g) should be set equal to 0.5 P,,J
(bar g) in this range). With the appropriate constants
detailed in this paper, the simple venting equation
A, = CAs/AP recommended in NFPA, and the ver-
sion for cubical enclosures presented by Grolmes and
Fauske, A /V*l3 = 6C[P,,/AP]@, can be used to
calculate vent areas that are in close agreement with the
extended K,, nomographs. Venting models by Rust and
Hart have been shown to differ from Heinrichs model
by only small amounts when Pred is less than 0.2 bar g.
References
1 VDI 3673. Guidelines on Venting of Dust Explosions. Verein
Deutscher Ingenieure, Dusseldorf, FRG, 1979
2 Bartknecht, W. Explosions. Course, Prevention, Protection.
Springer-Verlag, Berlin, FRG, 1981
3 Schofield. C. Guide to Dust Explosion Prevention and Protec-
tion. Part 1, Venting, IChemE, Rugby, UK, 1984
Rust, E. A. Chem. Eng., 1979, 102
Hart, R. L. Plant/Operations Progress 1985,4, 1
Lunn, G. A., Brookes, D. E. and Nicol, A. J. Loss Prev.
Process Ind. 1988,1,123
National Fire Protection Association, Guide for Venting of
Deflagrations NFPA 68 Quincy, MA, USA, 1988
Simpson, L. L. Plant/Operation Progress 1986,5,49
Grolmes, M. A. and Fauske, H. K. Dust Explosion Technology
Applications, IBC Flammable Dust Explosion Conference, 2-4
November 1988. IBC, St Louis, MO, USA
Epstein, M. eral. Combustion and Flame 1986.66, 1
Swift, I. and Epstein. M. Plant/Operation Progress 1987.6.98
Lunn, G. A:, Guide to Dust Explosion Prevention and
Protection. Part 3. IChemE, Rugby, UK, 1988
Swift, 1. Chem. Eng. 1988.63 -
Field. P. and Abrahamsen, A. R. Dust Explosions - The
resoective roles of the Hartmann bomb and 20-litre sphere ___ __.
in
prescribing the size of explosion relief vents - a preliminary
studv. Buildine Research Establishment Note N81/81,198t
15 Heinrich. H. J: Chemie-lng. Techn. 1966.38, 1125
16 Chappell, W. G. Pressure/Time Diagram for Explosion Vented
Space, AIChE Loss Prevention Symposium 1977, p. 76
rQCrown Copyright

208 J. Loss Prev. Process Ind., 1999, Vol2, October

 Review

The effect of vent ducts on the reduced

explosion pressures of vented dust
explosions

Geoff Lunn, David Crowhurst* and Michael Hey

Explosion and Flame Laboratory, Health and Safety Executive, Harpur Hill, Buxton,
Derbyshire, UK
*Fire Research Station, Boreham wood, Hertfordshire, UK

An investigation into the effects of vent ducts on reduced explosion pressures is described.
Experiments were made using an 18.5 m3 explosion vessel and a modified 20 I sphere, with dusts
having K.r values ranging from 144 barms~ to 630 barms. The vent area/vessel volume ratio,
bursting pressure of the vent cover, and the length to diameter ratio of the vent duct have been
varied. Straight vent ducts, and ducts containing sharp 45O and SO bends have been used. A
simple model to describe the effect of vent ducts on the reduced explosion pressure has been
derived and compared with the experimental results. Agreement is shown to be satisfactory in
nearly all cases. A comparison between the experimental results and guidance on the effect of vent
ducts already available in the literature is discussed.

(Keywords: vents; explosion; modelling)

Many materials, when dispersed in air in a finely divided rise, (dP/dt)mar (bar s-i) is measured over a range of
form, at a suitable concentration and in the presence of dust concentrations in air, using a standard technique.
an effective ignition source, are explosible. In enclosed The highest value of the maximum rate of pressure rise
vessels the pressures generated in these explosions may is used to calculate the K,, value of the dust:
reach 10 bar for organic dusts, and still higher values for
metal dusts such as aluminium. Such pressures cannot
be contained by most industrial dust handling equip-
ment, and measures need to be taken either to prevent The K,, value is essentially defined as the maximum rate
the explosion or to protect plant against the effects. of pressure rise measured under standard conditions in a
A method of protection used widely in industry is 1 m3 vessel, and is used to characterize the explosibility
explosion relief venting. In this technique, panels or of the dust by reference to four groups:
membranes covering openings in the walls of the equip-
KM : Grout St& Non-exolosible
ment open at low overpressures in the event of an
0 c K,, < 200 : Group St1
explosion, and thus the pressure is released. The Increasing
200 < K,, < 300: Group St2
highest pressure attained in a vented explosion is known explosibility
300 < K,, : Group St3 I
as the reduced explosion pressure, Pred. and is much less
than the explosion pressure generated in an enclosed The equation for K,, is referred to as the cube root law
vessel. When the explosion relief is properly designed, and expresses the relation between the rate of pressure
Prcd is not high enough to cause damage to the plant. rise and the volume of the vessel when standard condi-
The nesessary vent area depends on several factors: the tions are maintained. A theoretical analysis of the
volume of the equipment, V (m3); the opening pressure venting process in compact vessels by Heinrich3 has led
of the vent covering, Pstat (bar a); the strength of the to a series of nomographs that allow estimates of the
equipment or the reduced explosion pressure, which vent area when P,,,,, P& V and the K,, value are
must not be exceeded, P,d (bar a) and the explosibility known. These nomographs are published in VDI 3673
of the dust as characterized by the K,, value (bar m s-i) (Ref. 4). The results obtained using this method are
(Ref. 2). usually conservative, and for a given set of conditions
The K,, value of the dust is measured in the standard the calculated vent areas can be up to three times
20 1 sphere apparatus. The maximum rate of pressure experimentally determined values. This is because the
K,, nomographs apply to dusts with maximum
explosion pressures in enclosed explosions, Pmax, of
Received I March 1988; revised 13 June I988 11 bar a when the K,, value is less than 300 bar m s-
0950-4230p8/040182- 15$3.00
Q 1988 Butterworth & Co. (Publishers) Ltd
182 J. Loss Prev. Process lnd., 7988, Vol 1, October
 The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

and 13 bar a when the K,, value is greater; most dusts were conducted in an 18.3 m3 explosion vessel at the
have P,=, values less than these5. Nevertheless. it can be Explosion and Flame Laboratory, Buxton. A drawing
argued that this is a valuable safety factor, especially of the vessel showing the dimensions of the explosion
when conditions are somewhat different to those in the chamber has been published6. The length to diameter
standard test, and the nomograph method has found (L/D) ratio of the explosion chamber is approximately
wide favour. 1.6, and it thus qualifies as a compact enclosure to
The nomograph method is designed for unobstructed which the VDI venting nomographs apply.
vents. Vented dust explosions, however, are usually of The explosion chamber is designed with a working
large volume, and the burning material that ejects from pressure of 14 bar. Various ports in the chamber walls
dust handling equipment during a vented dust explosion were utilized as entry points for the ignition source
cannot be allowed merely to disperse into the work connections, and for attaching the dust injection cylin-
place. The danger to personnel that the cloud of hot gas ders. One end of the vessel contained a centrally posi-
and dust poses is an obvious one. In practice the tioned circular vent opening, diameter 1.1 m. A
combustion products are guided to a safe place through blanking plate was fitted over this opening to seal the
a duct fitted to the vent opening. chamber when enclosed explosions needed to be carried
The presence of the duct, however, alters the out.
explosion characteristics, and changing the system in The dust injection system comprised three pressure
this way is bound to affect the reduced explosion vessels with pepper pot nozzles. The vessels were spaced
pressure. Thus fitting a vent duct may result in along the top of the explosion chamber, and their
explosion pressures that unwittingly exceed the strength positions have been shown6. The injection vessels were
of the vessel. connected to pepper pot-type dispersion nozzles via ball
There are no truly satisfactory methods for predicting valves. These injection devices are different to the ring
the effects of vent ducts. A rule of thumb used in the UK system employed in the standard 20 1 sphere, but have
is that the duct should not exceed 3 m in length, should the advantage that the scale of the dust injection process
be straight, and should have an area not less than the is larger and thus conforms, approximately, to the
area of the vent itself. VDI 3673 contains some guide- larger scale of the explosion chamber. In any event, a
lines4 but these are only for straight ducts and give only series of calibration tests carried out with enclosed
limited help on the one hand for ducts 3 m in length and explosions minimized the effects of differences in injec-
on the other for those more than 3 m in length. tion techniques.
This paper describes a project in which the effect of Prior to an experiment the dust was divided equally
vent ducts on reduced explosion pressures has been between the three injector cylinders and stored under
studied. The results are discussed, and a method of the required injection pressure. This pressure was
calculation is described. This project was carried out for determined in the series of enclosed explosion calibra-
the British Materials Handling Board, which has tion tests. The dust was released into the explosion
brought together some 30 sponsors from industry and chamber through the ball valves operated by air cylin-
government to finance the project. The experiments ders. Because of the large volume of the explosion
were carried out at the Explosion and Flame Laboratory chamber, the dust injection did not contribute signific-
of the Research and Laboratory Services Division antly to the explosion pressure.
(RLSD) of the Health and Safety Executive at Buxton The ignition source was positioned at the centre of the
and at the Fire Research Station, Borehamwood. This explosion chamber and comprised a 30 g charge of
paper is a shortened version of the report provided to black powder ignited by an electrically fired powder
sponsors of the project. fuse. The ignition source did not contribute to the
explosion overpressure or to the maximum rate of
pressure rise. The ignition delay between initiation of
Experimental
the dust injection and firing of the powder fuse was
determined in the series of enclosed explosion calibra-
The dusts tion tests.
Three dusts were originally chosen to cover the range of A circular vent opening of 1.1 m diameter was posi-
explosibilities likely to be found in industrial dust tioned centrally in one end of the explosion chamber.
handling processes, and to be representative of the three Orifice plates were fitted over this opening when smaller
groups of dust explosibility. The St 1 dust was coal dust; vent diameters were required. These plates had bolt
the St2 dust was aspirin dust; the St3 dust was alumin- holes around the central opening to which both the
ium flake. explosion relief bursting discs and ducting could be
Two other dusts, toner dust and polyethylene, were fastened. The details of the vent opening are given in
used in a short series of tests to check that the behaviour Table 1. The bursting discs were designed for three
of the three main dusts was typical of the St-Groups. static bursting pressures: 1.1 bar a, 1.2 bar a and 1.5 bar
a. These bursting pressures conform to those for which
The explosion vessels the VDI nomographs are applicable4. A bursting disc in
position is shown in Ref. 6. Some experiments using
The 18.5 m explosion vessel. The large scale tests open vents have also been carried out.

J. loss Prev. Process lnd., 1988, Vol 1, October 183

The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

Table 1 Details of the vent openings in the 18.5 m3 vessel Table 2 Values of K., and Pm,. for test materials at the concen-
tration used in small scale vented explosions
Vent K-factor
diameter Vent area, A, K-Factor (cross section
(m) (rn? fVU3/A.) of chamber/A,) Concentra-
tion KS, P
1.1 0.950 7.38 4.93 Powder (kgm-) fbarms-1 fba:)
0.9 0.636 11.0 7.35
0.7 0.385 18.18 12.15 Silkstone coal 0.6 144 8.4
0.5 0.196 35.71 23.87 Polyethylene 0.25 193 8.8
Aspirin* 1 .o 220 8.6
Aspirin 1.5 254 8.7
Toner powder 0.25 236
Aluminium flake 0.5 272 :::
Aluminium flakeb 1 .o 333 9.3

* The concentration used was not that which gave maximum Ka

* The concentration used in a limited number of experiments

Igniter
/
Vent timer/
pressure tran

Figure 1 Explosion chamber and ducting

The vent ducts were of circular cross-section with

diameters and areas equal to the four vent openings. For
each vent diameter four straight lengths of duct were support frame
prepared, three of 5 m length and one of 1 m length;
one sharp 90 bend and two sharp 45 bends were also
prepared for each duct size. In Figure I, 16 m of 0.5 m
Figure 2 Modified 20 litre sphere
diameter duct is shown fitted to the explosion vessel.

The 20 1 explosion vessel. The 20 1 spherical vessel was ignition source upon activation from approximately
used to carry out the initial classification of materials. 1 bar (approximately S-10 times the vent bursting
This vessel, when operated in accordance with the pressure) to a negligible level thereby preventing burst-
specified test procedure**, is acepted internation- ing of the vent cover as a result of the ignition source.
ally 4*8~9as a suitable instrument for the determination Classification tests on each of the powders to be used
of the maximum pressure (P,,,) and rate of pressure showed that reducing the ignition energy had no effect
rise (dP/dt), and hence values of K,,. All the materials on the value of K,r obtained. For all vented explosions
used were classified according to the standard test all other parameters (ignition delay, pre-dispersion
procedure and the results are summarized in Table 2. pressure in vessel, dispersion pressure in the dust
To carry out explosions vented through relief ducts, hopper) were the same as for the standard test.
the standard 20 1sphere had to be modified to accept the Three vent diameters were investigated; 76 mm,
vent cover and ducting. The top igniter support 100 mm and 128 mm. To seal the vent simple vent
assembly and top flange were removed and replaced by covers, made from thin (50pm) polythene film lightly
a vent support plate and ducting adaptor. Two new glued to a cardboard frame to form a diaphragm, were
support arms for the igniters were made and mounted used for all experiments. The external diameter of the
through spare 14 mm holes in the small flange plate used frame was 140 mm while its internal diameter varied to
to mount the sphere pressure transducer. The whole match the diameter of the vent. The diaphragm was
assembly was then mounted onto a new stand, which sandwiched between a vent cover support grid and the
allowed the sphere to be rotated through 90 so that the ducting adaptor flange and the complete assembly
ducting when fitted lay in a horizontal position. The bolted to the sphere (see Figure 4).
modified sphere is shown in Figure 2. The support grid was made from 220 mm diameter
For vented explosions the energy of the ignition 1.5 mm stainless steel plate with a hole cut from the
source was reduced from the standard 10 kJ to I kJ. centre to match the diameter of the duct, into which was
This was done to reduce the pressure generated by the welded a 2.5 mm wire mesh with a grid size of approxi-

184 J. Loss Prev. Process lnd., ?988, Vol 7. October

 The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et.al.

mately 25 mm. This grid supported the diaphragm The injection conditions altered from dust to dust.
against the evacuated pressure of 0.4 bar absolute However, for the three dusts that did not coagulate
(-0.6 bar gauge) during the pre-dispersion phase of (coal, toner and aluminium) the conditions were the
each test. The presence of the grid reduced the effective same. More rigorous conditions have been necessary for
area available for venting by 15-18%, thus the vent aspirin dust, perhaps because it coagulates and the large
coefficient, normally given by injection ports in the 18.5 m3 vessel do not break it up
as well as the ring system in the 20 1 sphere. The highest
K= V23jA,
K,, value obtained for the polyethylene dust was
where A, equals the vent area, was increased for each 106 harms-t (compared with 190 bar m s- in the 20 1
vent by a corresponding amount, i.e. is given by sphere), again because the dust coagulates readily.

Vented tests without ducts

where AfV is the free vent area. Values of the vent
Comparison with nomograph predictions. As a pre-
coefficient, for each of the vent diameters used are
liminary to the main work, a series of tests was carried
presented in Table 3.
out without the fitting of vent ducts. The reduced
All the ducting was made from galvanized heavy
pressures measured in these tests have been compared
weight steel pipe to BS 1387, fitted with galvanized steel
with predictions from the VDI venting nomographs.
flanges to BS 10 Table E. Up to four lengths of pipe
The reduced explosion pressures cover a wide range of
could be connected to form the duct; these had nominal
lengths of 100 mm (adaptor flange), 400 mm, 900 mm values from as low as 1.12 bar a to as high as 8.2 bar a.
and 1900 mm allowing various duct lengths up to a The results of the tests are given in Table 5. In all
cases other than aluminium flake, the measured
maximum of approximately 3400 mm to be investi-
explosion pressure is below the predicted value. The
gated .
predicted overpressure can be up to three times the
measured overpressure. These results are in keeping
Enclosed explosion tests
with expected comparisons between nomograph predic-
The Kst value was chosen as the indicator of dust
tions and measurements with dusts of P,,, around
explosibility in both series of tests, so that comparisons
between the results from the 20 1 sphere test investiga-
tion and the 18.5 m3 vessel investigation could be made.
Table 5 Results of vented explosion Tests in the 18.5 m3 vessel,
A preliminary series of enclosed explosions was con- and comparison with reduced explosion overpressures predicted
ducted in the large scale vessel. Changes were made to by the VDI venting nomographs
the ignition delay and dust injection pressure until the
Vent area (ml Vent P Pred bar gl Pred (bar g)
maximum rate of pressure rise corresponding to the K,, K factor diameter (ml (barStat
g) measured oredicted
value and the cube root law was obtained, at the same
dust concentration. In terms of the rate of combustion Vented explosions: coal dust IK,, = 144 barms )
0.95 1.1 0.1 0.21 0.50
in an enclosed explosion as measured by the maximum 7.36 i.l 0.2 0.25 0.56
rate of pressure rise, this procedure effectively calib- 1.1 0.5 0.61 1.04
rated the vessel for the dusts used in the tests5. The 0.9 0.1 0.24 0.84
injection pressures and time delays for each dust are 0.636 0.9 0.2 0.64 1.03

given in Table 4. 11.00 0.9 0.5 1.24 1.64

0.385 0.7 0.1 0.58 1.52

18.10 0.7 0.2 1.56 1.83
0.7 0.5 1.38 2.81
Table 3 Vent areas used in small scale experiments
0.196 0.5 0.1 2.44 3.13
35.7 0.5 0.2 2.58 3.68
Vent
0.5 0.5 3.77 5.47
Vent diameter Total vent area, Free vent area, coefficient
(mm) A, (mm) At, (mm1 K = pi3/Ati Vented explosions: aspirin dust (K,, = 254 bar m s )
0.95 1.1 0.1 0.71 1.04
7.36 1.1 0.2 0.94 1.26
76 4536 3720 19.8
1.1 0.5 1.04 1.97
100 7854 6519 11.3
128 12868 10552 7.0 0.385 .0.7 0.1 2.45 2.80
18.18 0.7 0.2 2.96 3.28
0.7 0.5 3.10 4.93

Vented explosions: aluminium flake (K,, = 630 bar m s 1

Table 4 Details of dust injection and ignition procedure
0.95 0.1 4.15 2.81
7.36 1.1 0.2 3.85 3.32
InjectIon Ignition 18.5 Ill3
pleSSW delay K,-value Cogcentrstion 0.5 4.83 4.95
pnl.x
DUST (bar a) Imsl lbarms-I lkgm 1 (bar al 0.5 7.12
0.5 7.23
Coal 20 760 144 0.5 8.5
0.636 0.1 3.10 4.21
Aspirin 30 540 254 1 .o 8.3
236 0.25 8.8 11.00 0.9 0.2 5.0 4.93
TOflW 20 760
Polyethylene 30 540 106 0.5 7.9 0.5 4.76 7.25
Aluminum 20 760 630 0.25 11.0
Dust concentration = 0.5 kg m 3

J. Loss Prev. Process Ind., 1988, Vol 7, October 185

The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

9 bar a (Ref. 5). Generally, the nomographs give conser- Experimental tests are useful for measuring the effect
vative results for St1 and St2 dusts, as has been pointed of vent ducts in particular circumstances, but more
out by Eckhoffg. It is clear from work by Eckhoff general guidelines are needed for users of dust handling
that in most industrial processes the dust clouds produce equipment. A simple empirical model has been derived
less violent explosions than those to which the nomogra- that predicts the reduced explosion pressure that can be
phs apply. However, the nomograph method is a well expected in a vented vessel, when a vent duct of given
used approach to calculating venting requirements and L/D ratio is fitted. The model does not attempt to
the conditions in the dust cloud to which the nomogra- describe the complicated processes of combustion that
phs apply (central ignition and a state of turbulence must take place in a vented vessel/vent duct system. In
producing a combustion rate generating the rate of outline, the model considers the explosion to take place
pressure rise expected from the dusts K,, value and the in two parts-the explosion in the main vessel, followed
cube root law) have been used in the present tests so that by a secondary explosion in the duct. The explosion in
comparisons can be made with previously published the main vessel is assumed to generate a pressure
explosion data obtained under the same conditions. throughout the system, and the explosion in the vent
In the case of aluminium flake, the comparison duct causes a further increase in this pressure.
between estimation and measurement of the reduced In reality, the explosion process will be a continuous
pressures suggestions that the Kst value can be estimated one, although much of the increase in explosion
only approximately. The results suggest that in some of pressure caused by the vent duct will occur because the
the tests the Kst value was above 630 bar m s-l. Tests volume of the system has been increased while the vent
done with a concentration of 0.5 kgme3 indicate a KS1 area has remained essentially the same. The empirical
value well in excess of 630 harms-. nature of the model arises from the specification of a
combustion rate in the explosion in the vent duct. This
Vented tests with ducts has been done as simply as possible and the information
required to use the model consists of the duct configura-
A method for the prediction of the eflect of vent ducts tion (L/D ratio, number and type of bend), the K,, value
on the reduced explosion pressure. The measured of the dust and the reduced explosion pressure in the
explosion pressures are tabulated in Tables 6-9. The vessel when the duct is absent. This latter value may
ducts were empty of dust prior to the experiment, and either be measured or estimated from the Kst nomogra-
the only dust taking part in the experiment was that phs in VDI 3673 (Ref. 4). Thus apart from the duct
injected into the main vessel. Some examples of the parameters, the information required for the vent duct
measured effects of vent ducts on reduced explosion model is the same as is needed to use the Kst nomogra-
pressures are shown in Figures 3-10. The lines in these phs. Comparisons between predictions and experimen-
graphs represent the results of calculations using a tal results are shown in Figures 5-12.
simple model of the effect of vent ducts on the reduced
explosion pressure. Vented explosions when the vent is unobstructed. The

Table 6 Experiments in 18.5 m3 vessel with a straight duct

Vent
Vent area diameter P stat Reduced explosion pressure (bar al
Dust Im2) lm) (bar a) and duct (L/D) ratio

Coal 0.95 1.1 Open vent LID 0 0.90 5.5 10.0 14.5
PW 1.12 1 .l 1.20 2.0 2.3
Coal 0.95 1.1 1.2 LID 0 0.90 5.5 10.0 14.5
PWd 1.20 1.20 1.55 2.6
Aspirin 0.95 1.1 open vent LID 0 0.90 5.5 10.0 14.5
PW, 1.60 2.45 2.55 3.00
Coal 0.196 0.5 open vent LID 0 2 12 22 32
PWJ 2.55 2.35 2.4 4.0 4.0
Coal 0.196 0.5 1.5 LID 0 2 12 22 32
P red 4.8 4.5 4.8 5.4
Coal 0.385 0.7 Open vent LID 0 1.4 8.6 15.7 22.8
PM 1.8 1.55 2.4 2.4 3.7
Aspirin 0.385 0.7 1.1 LID 0 1.4 8.6 15.7 22.8
3.4 3.4 4.0 5.8
Coal 0.636 0.9 Open vent I$ 0 1.1 6.7 12.2 17.8
PWd 1.25 1.15 1.3 2.03 1 .a5
Aspirin 0.636 0.9 Open vent LID 0 1 .l 6.7 12.2 17.8
P,X 2.1 2.4 2.8 3.97 4.3
Toner 0.636 0.9 Open vent LID 0 1.1 6.7 12.2 17.8
2.05 2.2 3.3 3.6 4.1
Aluminium flake 0.95 1.1 1.5 :jL; 0 0.9 5.5 10.0 14.5
(0.25 kgm ? P reo 5.7 6.5 7.8 8.1 7.9
Aluminium flake 0.95 1.1 1.5 LID 0 0.9 5.5 10.0 14.5
(0.5 kgm ? Pr.0 8.1 a.4 8.8 9.5 9.1

186 J. Loss Prev. Process Ind., 7988, Vol I, October

 The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

Table 7 Experiments in 20 I vessel with straight ducts

Vent
Vent area diameter Maximum reduced explosion pressure
Dust (m*f tm) (bar a) and duct (L/D) ratio

Coal 0.00454 0.076 1 .o LID 1.2 6.6 13.0 17.3 26.2 31.8 38.4 43.6
pm, 1.37 1.97 2.5 2.9 3.25 3.55 3.62 3.8
Coal 0.00788 0.1 1 .o LID 0.9 5.1 10.3 14.3 20.5 24.3 30.0 33.7
Pred 1.15 1.26 1.30 1.75 2.20 2.02 3.0 2.6
Coal 0.0129 0.128 1 .o LID 0.8 3.9 7.8 11.1 18.9 25.8
P,*, 1.15 1.22 1.25 1.32 2.06 2.25
Polyethylene 0.0129 0.128 1 .o LID 0.8 3.9 7.8 11.1 18.9 25.6
P <a 1.25 1.9 2.1. 2.75 2.85
Aspirin 0.00454 0.076 1 .o LID 1.2 6.6 13.0 17.3 26.2 31.8 38.4 43.6
P red 2.8 2.9 3.52 3.40 3.70 3.67 3.75 3.57
Aspirin 0.00786 0.1 1 .o LID 0.9 5.1 10.3 14.3 20.5 24.3 30.0 33.7
PM 1.32 1.42 2.32 2.57 2.57 1.85 2.35 3.20
Aspirin 0.0129 0.128 1 .o LID 0.8 3.9 7.8 11.1 18.9 25.8
pm 1.08 1.35 1.73 2.22
Toner 0.0129 0.126 1 .o LID 0.8 3.9 7.8 1 T.l 18.9 25.8
P red 1.35 2.25 2.90 3.05 3.2
Aluminium flake 0.0129 0.128 1 .o LID 0.8 3.9 7.8 11.1 18.9 25.8
Pm.3 2.67 5.3 5.15 5.52 5.05 5.25

Table 8 Experiments in 18.5 m3 explosron vessel with 45O bend in vent duct

Vent
Vent area diameter Pstat Reduced explosion pressure (bar a)
Dust (m*) fm) (bar a) Vent duct L/D ratio and number and positions of the bends

Coal 0.95 1.1 Open vent LID 0 2.7 7.3 11.9 16.4
Bend 1 m from 1 m from 1 m from 1 from
duct exit duct exit duct exit duct exit
P,& 1.12 1.25 1.75 2.1 2.3
Coal 0.95 1.1 1.2 LID 0 2.7 7.3 11.9 16.4 13.6
Bend 1 m from 1 m from 6 m from 2 bends
duct exit duct exit duct exit
P,a, 1.25 1.25 1.80 2.20 3.0
Aspirin 0.95 1.1 Open vent LID 0 2.7 7.3 11.9 16.4
Bend 1 m from 1 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
Pred 1.58 2.25 3.6 3.1 3.9
Coal 0.196 0.5 Open vent LID 0 6.0 16.0 26.0 36.0 30.0
Bend 1 m from 1 m from 1 m from 6 m from 2 bends
duct exit duct exit duct exit duct exit
P red 2.55 2.7 3.85 3.8 4.05 4.90
Coal 0.196 0.5 1.5 LID 0 6.0 16.0 26.0 36.0 30.0
Bend 1 m from 1 m from 1 m from 6 m from 2 bends
duct exit duct exit duct exit duct exit
PWd 4.75 4.50 5.20 4.75 5.8 5.3
Coal 0.365 0.7 Open vent LID 0 4.3 11.4 18.5 25.7 28.5
Bend 1 m from 1 m from 6 m from 10 m from 2 bends
duct exit duct exit duct exit duct exit
1.5 2.9 3.15 3.4 4.05
Aspirin 0.385 0.7 1.1 4.3 11.4 18.5 25.7
Bend 1 m from 1 m from 6 m from 10 m from
duct exit duct exit duct exit duct exit
3.40 3.9 5.2 5.85 5.9
Coal 0.636 0.9 Open vent 0 3.33 8.9 14.4 20.0
1 m from 1 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
PIed 1.25 1.35 1.65 1.90 2.10
Toner 0.636 0.9 Open vent 0 3.33 8.9 14.4 20.0
$Z 1 m from 1 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
Bend 2.0 3.0 4.35 4.55 4.15

J. Loss Prev. Process lnd., 1988, Vol I, October 187

The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

Table 9 Experiments in 18.5 m3 vessel with 90 bend in vent duct

vent
Vent area diameter P,,,, Reduced explosion pressure (bar a)
Dust (rn) (ml (bar a) Vent duct L/D ratio and number and positions of the bends

Coal 0.95 1.1 Open vent LID 0 2.8 7.3 11.8 16.4
Bend 1 m from 1 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
P red 1.12 1.25 1.90 2.15 2.75

Coal 0.95 1.1 1.2 LID 0 2.8 7.3 11.8 16.4

Bend 1 m from 1 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
P red 1.25 1.45 1.90 2.50 2.90

Aspirin 0.95 1.1 Open vent LID 0 2.8 7.3 11.8 16.4
Bend 1 m from 1 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
P red 1.58 2.5 4.3 3.1 4.0

Coal 0.196 0.5 Open vent LID 0 6 16 26 36

1 m from 1 m from 6 m from 1 m from
duct exit duct exit duct exit duct exit
PTd 2.55 3.65 4.0 4.6 4.9

Coal 0.196 0.5 1.5 LID 0 6 16 26 36

Bend 1 m from 1 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
P red 4.75 5.05 5.90 5.50 6.15

Coal 0.385 0.7 Open vent LID 0 4.3 11.4 18.5 25.7
Bend 1 em from 6 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
p,ed 1.8 2.2 3.0 3.0 4.0

Aspirin 0.385 0.7 1 .l LID 0 4.3 11.4 18.5 25.7

Bend 1 m from 1 m from 1 m from 6 m from
duct exit duct exit duct exit duct exit
P.ed 3.40 4.50 5.54 6.05 5.8

Coal 0.636 0.9 Open vent LID 0 3.33 8.9 14.4 20.0
Bend 1 m from 1 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
P& 1.25 1.40 1.8 2.05 2.00

Toner 0.636 0.9 Open vent LID 0 3.33 8.9 14.4 20.0
Bend 1 m from 1 m from 1 m from 1 m from
duct exit duct exit duct exit duct exit
P& 2.02 4.05 3.3 5.0 4.9

model uses as a starting point the conditions expected in which can be written
the vented vessel when the vent duct is absent. For a
1
compact vessel, the reduced explosion pressure can be 105(( Pred)O- pO)=zpu2 (1)
obtained from the Kst nomographs in VDI 3673 (Ref. 1)
when the vessel volume, v(m3), the vent area, A, (m), where ( Pred)e has already been defined as the reduced
the dust Kst value (bar m s-i), and the bursting pressure explosion pressure (bar a) in a vessel with an unobstruc-
of the vent cover, PSt,, (bar a), are known. This value of ted vent; PO is the atmospheric pressure when the orifice
the reduced pressure is usually a conservative estimate, flow is subsonic (taken as I bar a in these calculations);
and a measured value would be lower. Nevertheless, p is the density (kgmm3); u is the velocity at the vent
whether either measured or estimated, a value or Pred opening (ms-); and C, is an orifice contraction
for an unobstructed vent (i.e. one without a vent duct) coefficient which has a value of 0.61 for a vent area
can be obtained, (Pred)& small in comparison with other flow areas in the system.
The Kst nomographs are developed from an equation This is assumed to be the case for a vent discharging into
derived by Heinrich3. The basis of this equation is that the open air.
at the reduced pressure the rate of pressure rise due to Thus
the combustion equals the rate of pressure fall due to
lo5 (( &d)O - PO) = 2.687(; pu2)
outflow of material through the vent. A simple equation
can be derived to calculate the rate of mass outflow if The rate of mass discharge, kgs-, is given by
the pressure in the vessel is known.
A standard orifice flow equation equates the pressure dm
dt = WA,
loss to a number of velocity heads:

where A, is the area of the vent (m).

Substitution of Equation (3) into Equation (2) gives,

188 J. Loss Prev. Process lnd., 1988, Vol I, October

 The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

m a
k
5
-E
PL
Vent duct LID ratio 0 4 8 12 16 20 24 28 32 36
Vent duct LID ratio
4, I

1L
0

2
I
4
I
6
Vent duct
I
6
L/D
10
I

r&o
12
I

14
1 1

16
I
18 f&q--i
0 4 8 12 16 20 24 28 32 36
Vent duct L/D ratio

31 1 I I I I I , I I
0 4 8 12 16 20 24 28 32 36
Vent duct LID ratio

0 4 a 12 16 20 24 28
Vent duct LID ratio

Figure 5 Effect of vent duct on reduced explosion pressure

vessel volume = 20 I, straight duct: a, coal dust; vent dia-
meter = 0.1 m; area = 0.00706 m2 P,,=, E 1 .O bar a; b, aspirin
I I I f I dust; vent diameter = 0.1 m; area = 0.00786 m2 P_, z 1 .O bar a;
2
0 4 8 12 16 20 24 C. toner dust; vent diameter = 0.129 m; area = 0.129 m
Vent duct L/D ratio P,,,, G 1 .O bar a

Figure 3 Effect of vent duct on reduced explosion pressure

vessel volume = 18.5 m, straight duct: a, coal dust; vent dia-
meter = 1 .l m; area = 0.95 m*; PSI,, = 1.2 bar a; b, aspirin dust;
vent diameter = 1 .l m; area = 0.95 m2; open vent; c, coal dust;
vent diameter = 0.5 m area = 0.196 m2 P,,* = 1.5 bar a; d. aspirin
dust; vent diameter = 0.7 m; area 0.385 m2 P,,,, = 1.1 bar a

0 4 8 12 16 20 24
Vent duct L/D ratio

0 4 8 12 16 20 24 28
Vent duct L/D ratio

Vent duct L/D ratio

.
2*
.
1 I I I I 1 I
0 2 4 6 8 10 12 14 16 16 1 I I I I I 1
Vent duct LID ratio 0 4 a 12 16 20 24
Vent duct L/D ratio
Figure 4 Effect of vent duct on reduced explosion pressure
vessel volume = 19.5 ms, straight duct: a, coal dust; vent die Figure 6 Effect of vent duct on reduced explosion pressure
meter = 0.9 m: area = 0.636 m* open vent; b, toner dust; vent vessel volume = 20 I. straight duct: a, polyethylene dust; vent
diameter = 0.9 m; area = 0.636 m2 open vent; c, aluminium flake; diameter = 0.128 m; area = 0.0129 m*; P,,., I 1 .O bar a; b,
vent diameter = 1 .l m; area = 0.95 m* Pst,, = 1.5 bar a. 0, aluminium flake; vent diameter = 0.128 m; area = 0.0129 mz;
0.25 kgm ; q , 0.5 kgm~3 P,,,, - 1 .O bar a

J. loss Rev. Process Ind., 1988, Vol 1, October 189

The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

0 2 4 6 8 10 12 14 16 18
Vent duct L/D ratio 31 1 I I I I I
0 4 8 12 16 20 24
Vent duct L/D ratio

0 2 4 6 8 10 12 14 16 18
Vent duct L/D ratio Vent duct L/D ratio

Vent duct L/D ratio

11 I I I I I I I I I Figure 9 Effect of vent duct on reduced explosion pressure

0 2 4 6 8 10 12 14 16 la vessel volume = 18.5 m3 a, aspirin dust; vent diameter = 0.7 m;
Vent duct LID ratio area = 0.385 m2; P,, = 1.1 bar a; duct = 90; b. coal dust; vent
diameter = 0.9 m; area = 0.636 m2, open vent, duct = 45O; c,
Figure 7 Effect of vent duct on reduced explosion pressure coal dusti vent diameter = 1.1 m; area = 0.95 m*, open vent,
vessel volume = 18.5 m3: a, coal dust; vent diameter = 1.1 m; duct = 45 A, single bend 1 m from duct exit; A, single bend
area = 0.95 mp; P,r,? = 1.2 bar a; duct = 45O; b, coal dust; vent From duct exit
diameter = 1 .l m; area = 0.95 m; PS,.t = 1.2 bar a; duct = 90; c,
aspirin dust; vent diameter = 1.1 m; area = 0.95 mz open vent,
duct = 45O. q , single bend 1 m from duct exit; n , single bend
6 m from duct exit; l , two bends

ii
0 2 4 6 a 10 12 14 16 18
Vent duct LID ratio

i$; 5
0 4 8 12
Vent
16
duct
20
L/D
24
ratio
28 32 36

n n
0 4 8 12 16 20 24 28 32 36
C
Vent duct L/D ratio

ZG; 21
0
I
4
I
a
1
12 16
Vent duct
I
20
L/D
I
24
ratio
I I
28
I
32
1
36

0 4 6 12 16 20 24 28 32 36
Vent duct LID ratio

0 4 8 12 16 20 24 28
11 1 Vent duct L/D ratio
0 4 a 12 16 20 24 28
Vent duct L/D ratio Figure 10 Effect of vent duct on reduced explosion pressure
vessel volume = 18.5 m3 8, aspirin dust; vent diameter = 0.9 m;
Figure 8 Effect of vent duct on reduced explosion pressure area = 0.636 m, open vent, duct = straight; b, coal dust; vent
vessel volume= 18.5 m3 a, coal dust; vent diameter =0.5 m; diameter = 0.5 m; area = 0.196 m, open vent, duct = 45O; c,
area = 0.196 m; P,,,, = 1.5 bar a; duct = 45O. b. coal dust: vent coal dust; vent diameter = 0.5 m; area = 0.196 m*, open vent,
diameter = 0.5 m: area = 0.196 mz; PSt,r = 1.5 bar a; duct = 90; duct = 90; d. coal dust: vent diameter = 0.7 m;
c, coal dust; vent diameter = 0.7 m: area = 0.385 m2 open vent, area = 0.385 m2, open vent, duct = 45O. d, single band 1 m from
duct = 90. q , single band 1 m from duct exit; n , single band duct exit; I, single bend 6 m from duct exit; l , two bends; A,
6 m from duct exit; 0, two bends single bend 10 m from duct exit

190 J. Loss Prev. Process lnd., 7988, Vol I, October

 The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

after some algebra exit. When 4 > 1.89Po, the flow is choked and
2 x 10s I* dm P2A,
g = Ad&d)0 - pO)L2P2
(. >
2687 (4)
-=-
dt
T,/2 5776

Substituting p = (P,d)oM,/RT, where M, is taken as To estimate the new reduced pressure caused by
the molecular weight of air (0.029 kg mol- ),
explosion in the vessel some comparison has to be made
R = 82.0552 x lo-$ m3 bar/Kmole and T (K) is the with the. situation when the vent duct is absent. It is
temperature assumed to be obtained by adiabatic com- assumed here that the rate of mas outflow remains the
pression from ambient (T = 298(( P,,~)o/Po)~286), into same as at (Pred)~ and as calculated by either Equation
Equation (4) gives (5) or Equation (7). Knowing the value of (dm/dt)
means that the pressure 9 can be calculated by either
dm A,
dt = ~/2 (&d)Ci*((&dh - fo)*5128 (9 Equation (9) or Equation (10).
The pressure loss along the duct depends on the
Equation (5) is suitable only when the flow through length to diameter (L/O) ratio of the duct and the
the vent is subsonic. When (Pred)O > 1.89P0. the flow is friction factor. The general equation is:
choked, and
105(p, - Pz) = 8f $ $ (11)
AP = (P&O - $!$ = 0.47( Prcd jo (. >
where PI is the pressure inside the duct at the entrance
Thus, for choked flow (bar a); f is the friction factor; p is the density (kg me3);
and u is the velocity in the duct at the entrance (m s-l).
10 X 0.47(Pred)O = 2.687(;pu2) (6)
A satisfactory engineering approximation is
and, by the same procedure as for subsonic flow: 8 f = 0.025 (Ref. 12) and this value has been used in all
the calculations. Substituting dm/dt = puA, and the
dm
-=- A,
T,,2 (p&o3516 (7) equation of state into Equation (11) gives
dt

The rate of mass outflow can thus be calculated if PI (PI - P2) = y (5) & (!$)*(0.0283 x 10-6)
(Pred)O and A, are known.
(12)
The changes to jlow resistance caused by fitting a duct.
where T is calculated by assuming adiabatic compres-
Fitting a duct to the outside of a vent opening changes
sion from ambient (298K) to PI. PI can be calculated
the resistance to flow. The pressure loss is now made up
from Equation (12) by trial and error.
of three components: the loss at the duct exit, the
The pressure loss at the entrance to the duct from the
frictional loss along the duct and the loss at the duct
vessel depends on the configuration of the entrance. The
entrance.
general equation for this pressure loss is
For the moment any effect of an explosion in the duct
is neglected, and a procedure is described for estimating 105((P,,d), - P,) = F* fpU2 (13)
the new reduced pressure resulting from an explosion in
the vessel under the new conditions for flow discharge. where (Prcd)r is the new reduced pressure for the vessel
The area of the duct is assumed equal to the vent area, explosion (bar a) and F is a factor, which depends on
and of the same shape. the configuration of the junction between the vessel and
Data* has been used to calculate the pressure losses the duct.
at various pipework configurations in terms of velocity When the cross-section of the vent and the cross-
heads. The pressure loss at the end of a duct discharging section of the duct are the same for area, shape and
to atmosphere is equal to one velocity head: possible, Fdepends on the area ratio of the duct and the
vessel. Data is given in Ref. 12.
AP=fpu Substituting dmldt = pA,u and the equation of state
Thus in Equation (13) gives:
2
105(P2 - PO) = ;&Iu* (8) lo((Pred), - 8) =; $7 RT (14)
( ) PI MwA,Z
where P2 (bar a) is the pressure inside the duct at the
exit, and u is the flow velocity (ms-) inside the duct. and so
Substituting for u in Equation (8) gives an expression
for dmldt: p,((p,d),-~,)=E; $ 2(O.O293xlO-6)
( >
(15)
$ = (Pz (Pz - Po))lf2 +2 8407 (9) where T is obtained assuming adiabatic compression
from ambient to PI. (p&)1 can be calculated by
T, the temperature (K), is calculated by assuming Equation (15) using trial and error.
adiabatic compression from ambient (298K) to P2. The pressures throughout the vessel-duct system can
Equation (9) is suitable for subsonic flow at the duct now be calculated. Unless the duct length is very long

J. Loss Rev. Process lnd., 1988, Vol I, October 191

The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

(P&)1 will be less than (Prcd)O because the resistance to account the overpressure in the duct, the right hand side
flow is less than the duct is fitted than for the orifice of Equation (16) is multiplied by the density ratio
alone. be/pn) to give the volume of the dust cloud at the
When the duct contains bends, the restriction to conditions in the duct. pD is the density assuming
outflow is increased. adiabatic compression from ambient pressure to the
The following has been presented for sharp bends IL: duct pressure, PI.
The volume of dust taking part in the explosion in the
AP = 1.2(1 - cos a)$pu2 for circular ducting,
duct depends on the relative values of the ejected
where CYis the angle of the bend. When several bends are volume of dust and the volume of the duct, VD,, where
present, the pressure loss can sometimes be obtained by vu = AVL. If VE > VD, then the volume of dust to be
simple summation. burnt equals vu, if VE < VDu, then the volume of dust
Thus the equation for the pressure loss along the duct to be burnt equals VE.
when bends are present is The net rate of volume generation depends on the rate
of combustion of this dust cloud compared with the rate
of outflow from the duct to the open air. However, the
rate of combustion depends on many factors that make
where N is the number of bends of a given angle; FF is an accurate assessment difficult.
the relevant loss term; and the summation sign implies A simple empirical model utilizing the K,, value of the
bends of different angles. dust and the cube-root law has been used throughout
Using dmldt = pAu and the equation of state gives a these calculations. A simple formulation for the rate of
variation of Equation (12) expansion due to combustion, Rr, is

lO(P, - P2)Pl = 0.025 $+EN* FF) : s V2d,3

RE=-!$.-=K,,_
VDu
VDu PO PO
(17)

-$ (0.0283 x 10-6) Equation (17) is the cube root law applied to the duct

volume using the K pf value measured by a standard
and thus PI can be found by the usual means of trial method. Normally, scaling the rate of pressure rise by
and error. the K,, value and the cube root law is only applicable to
vessels with L/D ratio less than 5 : 1, and its use here is
The explosion in the duct. This model considers the not intended to suggest any similarity between the
explosion in the vessel-duct system to take place in two advancing flamefront that would be generated by an
stages. Heinrichs model assumes that at (Pred)O, com- ignition source in the duct and the combustion resulting
bustion inside the vessel is essentially complete. Simi- from an explosion surging into the duct from the main
larly, the present model assumes that at (Pr&)[, vessel. Equation (17) is used only as an empirical
combustion inside the vessel is ended, and that any formula and makes no statement about the actual rate
further increase in explosion pressure due to the duct is of pressure rise taking place in the duct volume; it is
caused by the flame entering the duct and burning the assumed to describe the number of cubic metres at
dust/air mixture therein. It is the net creation of volume atmospheric pressure generated per second by combus-
from this secondary explosion that is assumed to cause tion. The model does not calculate pressure changes
the increase in the explosion pressure. throughout the system with either time or position; it
The conditions of the start of this secondary dust calculates a final reduced explosion pressure by assum-
explosion are first, the pressure in the explosion vessel ing the explosion volume increase is evenly distributed.
equals (PI&)], and second, the pressure in the duct is Equation (17) does not imply any consideration of the
taken to be constant along the duct and equal to PI. The method of combustion; however, in the initial surge of
volume of dust, VE, ejected from the vessel through the the explosion into the duct, combustion can be expected
duct, is given by to be dispersed throughout a length of the duct. Thus,
when the L/D ratio is short, combustion is likely to take
VE = vl _ y, - PO)
((Predh (16) place throughout the entire duct volume, but for longer
(Pm - PO) L/D ratios the initial surge is likely to extend only part
where VI is the vessel volume, and P,,,,, is the maximum way along the duct, the rest of the dust being consumed
explosion pressure generated in an enclosed explosion. as the explosion propagates. It is thus unlikely that the
P,,,,, can be measured either in the vessel under con- rate of expansion will continue to increase indefinitely
sideration or obtained from tests in the Standard 20 1 with duct volume as the L/D ratio increases. To take
sphere test. The value of Pmax should represent the this into account, when the L/D ratio > 10. the rate of
energy release of the dust explosion only; when applied expansion is assumed to stay constant at the value
to the 20 1 sphere results, 1 bar has been subtracted from appropriate for L/D = 10.
the measured value of Pm,, to compensate for the The progress of the duct explosion is followed by
energy imparted by the igniter. VE is the volume of dust calculating the amounts of volume created and lost
cloud measured at 1 bar pressure, and to take into during small time steps until the dust initially in the duct

192 J. Loss Prev. Process lnd., 1988, Vol I, October

 The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

has been either burnt or ejected. The time step, dt, is combustion rate compared with that in a straight duct.
chosen so that the final pressure is independent of dt. When the bend is close to the vent opening, however, its
The volume expansion in the time step, is given by presence is likely to accelerate the combustion rate.
The net volume expansion is given by:
Pl
VEX=RE- dt (18)
PO VN= VEX- V, (23)
The ratio (PI/PC,) has been introduced to take into The increase in pressure is obtained by dividing VN by
account the effect of initial pressure on the expansion the appropriate volume. At the start of the duct
due to combustion. This ratio changes as the duct explosion the pressure in the duct is lower than in the
explosion progresses. vessel, and so initially the pressure is allowed to rise only
The amount of dust air mixture burnt in a time step to in the duct, and thus the appropriate volume is the duct
give this expansion is given in volume terms by volume. When the pressure is uniform throughout the
vessel-duct system, the appropriate volume is the sum of
vB = RE dt
I(pma;;po)
where (P,,,,, - Po)/Po is the expansion factor of the
(19) the vessel plus duct volumes.
The amount of dust left is obtained from the follow-
ing equation:
combustion.
Vn is assumed to remain constant throughout the VosT=(vnST)r-C va+ t5.p (24)
i( PI > i
duct explosion; its value is subtracted from the volume
of dust initially in the duct at the end of each time step. where i is the number of the time step and ( VDUST)Iis the
The volume lost from the duct through the open end volume of dust initially present in the duct. pi is the
is given by density in the duct at the start of the duct explosion. The
density ratio in Equation (24) is present to convert the
v
I.
=dmdt volume lost at the conditions within the duct at time step i,
dt P to the volume at conditions initially present in the duct.
The rate of outflow from the duct, dm/dt, is calculated When the amount of dust left is zero, the final explosion
assuming that there is a constant pressure, PI, throughout pressure is reached. A simple calculation after the last time
the duct. As the explosion continues, this pressure rises. step (which overshoots) brings the final pressure back to
The equation for rate of mass outflow includes pressure the actual amount of dust present.
losses due to friction along the pipe and at bends, and at
the duct exit. For subsonic flow the equation is Comparison with experiment
The model predictions have been compared with the
reduced explosion pressures measured in the explosion
vessel. Generally the agreement between prediction and
experiment is satisfactory. The poorest agreement occurs
+CN* FF with explosions in the 18.5 m3 vessel and with open vents,
>
when in some sets of results the measured pressures
where FF is a restriction factor for bends and N is the regularly exceed the predicted values as the L/D ratio
number of bends. increases. In some of the open vent tests agreement
For choked flow between prediction and measurement is good, but ducts
5764
added to an open vent sometimes produced pressures
dm _ PIA,
dt-T/ZX (22) ranging from 0.3 to 1 bar higher than expected, especially
1 + 0.025 LID
when ( PXd)O exceeded 2.0 bar a. In themselves however,
The rate of volume outflow is calculated by dividing the open vent results are consistent; the addition of bends
Equations (21) and (22) by the density. Both density and to the duct increases the reduced explosion pressure by the
temperature T are obtained by assuming adiabatic expected amount, and the trend of the reduced pressure
compression from ambient to PI. No account is taken with increasing L/D ratio is satisfactorily calculated by the
of hot gas; the outflow is assumed to be always of model, as Figure 12 shows.
unburnt material. The discrepancies between prediction and experiment
The outflow equations used during the duct explosion are greatest with coal dust, at the smallest diameter of vent
incorporate the loss terms for the bends. The length of and when the opening overpressure of the vent is low or
duct incorporated in the L/D term is shown on the non-existent. It may be that when dust is ejected into a
appropriate figures. Generally, it is taken as the shortest narrow duct well in advance of the flames arrival,
length of duct, and terms are added to the usual L x A, increased turbulence in the dust cloud can raise the
method of calculating the duct volume to take into combustion rate enough to show an effect on the
account the volume of the corners. explosion pressure, especially with dusts in the St1 group
Essentially, this procedure assumes that for a given but not when the dust has a higher K,, value. When Psratis
duct volume, the effect of a bend on the reduced relatively high, the flame follows close behind the entry of
explosion pressure is due to the influence of the dust into the duct, and does not propagate into dust which
increased resistance and not to any increase of the has increased turbulence.

J. Loss Prev. Process Ind., 1988, Vol I, October 193

The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

Straight vent ducfs. Some examples of the comparisons bend. Some tests were repeated with the bend closer to the
between the maximum reduced explosion pressures explosion vessel. Usually the closer the bend was to the
measured inside the explosion vessel and the explosion vent opening the higher the pressure measured, but in one
pressures calculated by the model are shown in Figures 5 case the higher pressure was measured when the bend was
and 6. at the end of the duct. Generally the position of the bend
Generally, both the trend of the explosion pressure as did not have a large influence on the explosion pressure,
the L/D ratio increased and the numerical value of the only about 0.3 bar, although tests were not done with
explosion pressure are well predicted. Figure 6b shows bends closer than 2 m to the explosion vessel. The effect of
some results with an open vent. Although the trend of the two 45 bends in the vent duct is shown in some of the
changes in explosion pressure as the L/D ratio increases is figures. These results demonstrated that in some cases the
satisfactorily predicted by the calculations, the experimen- effect of multiple bends on the explosion pressure was
tal points all lie on one side of the calculated line. An greater than expected from calculations assuming bend
attempt to calculate a line passing through the points, by restrictions to be simply additive.
choosing a slightly higher (Prcd)O would have been suc- Some measurements using a single curved bend rather
cessful in this example. than a sharp bend in the 20 1 sphere tests showed that
The results of the aluminium flake tests are shown in explosion pressures were not noticeably different to those
Figure 6~. Approximate values only for the K, and P,,,,, when the duct was straight, and of equal (LID) ratio.
are available; these have been taken as 630 bar m s-i and
11 bar a respectively. In any event the experimental The eflect of duct closure and other auxiiiary
reduced pressures are much less than predicted by the experirnen ts
model described earlier. The effects of obstructing the end of the vent duct with
However, if Equation (24) is modified to give: either a grid or a membrane are demonstrated in Figure
1I. The conditions for these tests were a Pstat of 1.1 bar a,
&ST=(h3ST,I-~ (vB+ & (25) a vent diameter of 0.5 m and coal as the explosible dust.
( PI > i
In these experiments, the membrane increased the reduced
and the model used as before, the calculated reduced explosion pressure by about 0.3 bar, whereas the grid has
pressures fall closer to the measured values. Considering little effect. Figure Ii also shows two examples of a vent
the uncertainties typical of the high K,, value tests the explosion with the ignition source at the back end of the
agreement is probably reasonable. explosion vessel and not at the centre. In the test con-
Some results from the 20 1 explosion vessels tests are ducted without a vent duct, the reduced explosion
compared with calculations in Figures 7 and 8. No pressure was increased by approximately 0.5 bar, whereas
measurements at L/D = 0 were taken in these tests, so the with a short vent duct present moving the position of the
comparison is used to demonstrate that the trend of the ignition source had no effect on the reduced explosion
results as L/D ratio increases is satisfactorily predicted. pressure.
Generally, the trend is calculated satisfactorily. How-
ever, the agreement between calculation and experiment Comparisons with data from other sources
for the aspirin explosions in the 20 1 sphere is not good, Some tests on the effect of vent ducts have been carried
except in the example shown (Figure 7b). Usually the out in the LJSAi2. The conditions of dust injection and
measured pressures are much lower than predicted. When ignition are different to those in the present tests and the
the calculations are compared with toner dust results, K,, values of the dusts are not known. However, a
however, under similar conditions and a similar value of comparison between calculated and measured reduced
K,,, the agreement is satisfactory as can be seen in Figure explosion pressures was made for some coal dust tests. A
7c. This result suggests that the 20 1 sphere apparatus had K,, value of 144 bar m s- was used in the calculations. In
not given true values in some of the aspirin dust tests. The general, the agreement between calculation and experi-
trend in the aluminium flake explosions is calculated, as ment was reasonable, considering all the unknowns, and it
shown in Figure 8b, although the Ksr value is much lower
than in the 18.5 m3 aluminium flake tests.

Vent ducts with bends. In Figuresg-II, calculations for 5 8

a single 45 and a single 90 bend in the duct are
4 t
compared with some of the values of the explosion
pressure measured in the explosion vessel. As for the 3 8
linear ducts, the trend of reduced explosion pressure with I
2L I I I * I I I I I
the L/D ratio is generally satisfactory, and in many cases 0 4 6 12 16 20 24 28 32 36
the numerical value of pressure has been satisfactorily Vent duct LID ratio
calculated also. On some of the figures, two measure- Figure 11 Effect of vent duct closures on reduced explosion
ments for pressure are shown at some values of L/D ratio. pressure; vessel volume = 18.5 m3; COSI dust
These results demonstrate the effect of position of the (K,=144barms-): vent area=0.136m2; P,,=l.l bar a;
straight duct; vent diameter = 0.5 m; 0, 0.5 kg me3 central igni-
bend along the pipe. The bend was usually positioned at tion; q , 0.5 kg mm3 back ignition: 0, mesh fitted to duct end; A,
the end of the ducting, with a 1 m length of duct after the membrane fitted to duct end

194 J. Loss Prev. Process lnd., 1988, Vo/ I, October

 The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

v
,
4 PO1
I t
0.01 0.1 1.0 10 1 IO
P red overpressure (bargl without vent duct P red overpressure (bar g) without vent duct

Figure 12 Comparison of results with guidance from Ref. 14: A, Figure 13 Comparison of prediction with guidance from Ref. 14.
18.5 msvessel. ~3 m duct; 0, 18.5 m3 vessel. 3 m duct; n , 20 I Vent duct length > 3 m. c, KS,= 300 barme, P,,,, = 1.5 bar 7;
sphere. 3 m duct; _, guidance Ref. 12; ----, equal reduced 7 K,,=300barms ,PS,,,=1.2bar a; 0, K,,=300barms ,
pressures P Sfaf= 1 .l bar a; K,,=lOObarms~, P,,,=1.5bar a; l,
K,,=lOObarms~. PS,,,=1.2bar a; o, K,,=lOObarms ,
P 5,at= 1 .l bar a
can be concluded that the American tests show that the
effect of vent ducts on the reduced explosion pressure is of
4.0
approximately the same degree as found in the present b

FL
tests.
The results of the present tests are compared in Figure 3.0

12 with the vent duct guidance from VDI 3673 (and given
in Ref. 13). Reasonable correspondence is found when the
vent duct is greater than 3 m in length. The recommend-
ations are slightly higher than the measurements made in
the 18.5 m3vessel, probably because the recommendations
are an envelope of experimental measurements. When the
vent duct length is 3 m the measured pressures in the
0
18.5 m3 vessel are lower than the recommendations. In all 0 200 400 600 800 1
the present results for straight ducts, the rise in reduced Time fms)

pressure caused by a vent duct either starts off gradually 4.0 4.0

or decreases at low vent duct lengths. The graphs of

reduced explosion pressure against vent duct length given
in Ref. 13 do not show this behaviour, and it does not
appear to have been taken into account in deriving the
recommendations. The present results at low duct level,
would therefore be expected to fall below the recommend-
ations. The results from the 20 1 sphere tests are also
shown in Figure 12 for a duct length of 3 m, and do not
coincide with the guidance.
One further comparison can be made with the present -
0 200 400 600 800 1000
o- 0 200 400 600 800 11
results and the recommendations given in Ref. 13. If the Time (msl Time fms)
vessel volume, the K,, value of the dust, the bursting
Figure 14 Pressure time traces: toner dust: vent dia-
pressure of the vent cover and the vent area are known, meter = 0.9 m: open vent; straight vent duct 11 m long. a;
then (P&)0 can be obtained from the nomographs and explosion vessel; b. 0.48 m from duct entrance; c, 6.1 m from
the model described earlier applied. When the K,, value is duct entrance; d, 11 .O m from duct entrance

equal to 300 bar m s-i (the top of the St2 group) and P,,
equals 11 bar a, the results given by the model for a vent Pressure measurements in the duct
duct greater than 3 m in length coincide with the guidance A typical example of pressure time traces in the vessel duct
in Ref. 13, as Figure I3 shows. When the K,, value is system is shown in Figure 14. The empirical model does
100 barms -I, however, the increase in pressure caused by not attempt to reproduce pressure changes with time or
the vent duct is not as great. This comparison suggests position, only to calculate the final reduced explosion
that current guidance is a conservative estimate applicable pressure in the explosion vessel. However, pressure
to the top of the St2 group, and that at lower Kst values measurements along the duct generally showed maximum
that effect of vent ducts on the reduced explosion pressure pressures somewhat less than the pressure measured in the
will not be as great. explosion vessel. Often the pressure trace at a point in the

J. Loss Rev. Process lnd., 1988, Vol I, October 195

The effect of vent ducts on the reduced explosion pressures of vented dust explosions: G. Lunn et al.

duct was double peaked and on some occasions the first Brookes, A. Nicol, B. Quince, W. Wardle and H. Good-
peak did register a higher pressure than measured in the win are acknowledged.
explosion vessel. However, this result was rare and these
results support the general conclusion that the ducts
should be made as strong as the explosion vessel.

Conclusions
The effect of vent ducts on the reduced explosion
pressures of vented dust explosions has been investigated
and a simple empirical model has been shown to calculate
adequately the pressure increases that result, except in
some examples with vents of low bursting pressure or
open vents. The obvious use of this data and model is to
provide practical guidance for industrial dust handling
equipment. Because of the experimental conditions used
in this series of tests, this guidance could be directly
associated with the K,, nomograph method for estimating
venting requirements. It has been demonstrated that such
guidance would agree with that already available at the
top of the st2 dust explosibility Group
(IL = 300 bar m s- ), but would enable users to take into
account a wide range of variables and so design for lower
explosion pressures than at present for K,, values less than
300 harms-.
Acknowledgements
The extensive contributions to this project
References
1 Schofield, C., in Guide to Dust Explosion Prevention and Protec-
tion, Part l-Venting. IChemE Rugby, UK. 1985
2 Bartknecht, W., in Explosions Course Prevention Protection,
Springer Verlag, Berlin, 1981
3 Heinrich, H. .I. Chemie-lng-Tech. 1966, 38, 1125
4 VDI 3673, Pressure Release of Dot Explosions, Verein Deutscher
Ingeneiure, 1983
5 Bartknecht, W. Plant/Operations Progress 19%. 5 (4), 1%
6 Lunn, C., J. Loss Prev. Process Ind. 1988, 1, 123
7 Operating Manual for 20 I Spherical Explosion Vessel
8 British Standards Institute, BS 6713, Part 1, Explosion Protection
Systems, London, 1986
9 National Fire Protection Association. Guide for Explosion
Venting, NFPA 68 Boston, MA, USA, Draft 1987
10 Eckhoff, R. K. Bulk Solids Handling 1986, 6 (5), 913
I1 Eckhoff, R. K., and Fuhre, K. Dust Explosion Experiments in a
vented 500 m3 vessel, Fourth International Symposium on Loss
Prevention and Safety Promotion in the Processing Industries,
1983, IChemE Symposium Hazards Series No 82
I2 The Efficient Use of Fuel, HMSO, 1958
13 Nagy, J.. and Verakis, H. C. in Development and Control of
Dust Explosions, Marcel Dekker Inc. New York, 1983
14 Siwek, R. Additional Considerations Concerning an Effective
Pressure Venting: paper presented at the Europex Seminar Course
1986. Dust Explosion Venting, Amsterdam, 17-19 November,
of D. E. 1986

196 J. Loss Prev. Process lnd., 1988, Vol 1, October

 RWkW

Safety considerations in conveying of bulk

solids and powders

Stanley S. Grossel
HofSmann-La Roche Inc., Corporate Engineering, 340 Kingsland Street, Building 105,
Nutley, NJ 07110, USA

This review covers a wide range of safety factors which need to be considered when handling bulk
solids and powders. The physical dangers of the mechanical equipment are well covered by
published standards and codes, as are noise levels and the use of electrical equipment. Dust
explosions caused both by static electricity and other ignition sources, are more complex. The plant
requires careful investigation to ensure that explosion hazards are kept to a minimum and suitable
protective measures installed. Different types of conveyers, e.g. belt, pneumatic and bucket
elevators, each pose their own hazard and the system chosen should be considered carefully in the
light of the material to be handled.

(Keywords: bulk solids handling: safety factors; dust explosions)

There are a wide range of safety factors to be considered

Table 1 Permissible noise exposures
in the handling of bulk solids, from the safety of the
machinery used to the toxicity and explosibility of fine Sound level
dBA slow
powders. This review looks at these different safety Duration Per day fhl ~l3SpOSl3
aspects.
8 90
6 92
Conveyor personnel protection considerations 4 95
3 97
2 100
Machinery guarding 1: 102
All bulk solids/powders conveyors have a large number 1 105
: 110
of moving parts, including power transmission machin-
: or less 115
ery and equipment (shafts, pulleys, couplings, speed
reducers, etc). In accordance with OSHA and ANSI When the daily noise exposure is two or more periods of noise
requirements guards must be provided to protect per- exposure of different levels their combined effect should be
considered, rather than the individual effect of each. If the sum of
sonnel and state and local codes and regulations must the following fractions: Cl/T, + G/T2 G/T, exceeds unity, then,
also be satisfied. the mixed exposure should be considered to exceed the limit
value. Cn indicates the total time of exposure at a specified noise
The OSHA and ANSI standards listed below should
level, and Tn indicates the total time of exposure permitted at that
be consulted for details. level. Exposure to impulsive or impact noise should not exceed
140 d6 peak sound pressure level.
0 OSHA Safety and Health Standards (29CFR 1910):
General Industry, Paragraph 1910.219 (1981) -
Mechanical Power - Transmission Apparatus.
machinery cannot be obtained with noise levels comply-
0 ANSI B15.1 (1972) - Safety Standard for Mechan-
ing with OSHA standards, then engineering controls,
ical Power Transmission Apparatus.
such as sound mufllers, must be installed. If this does
0 ANSI/ASME B20.1 (1984) - Safety Standard for
not bring the sound levels within the limits listed in
Conveyors and Related Equipment. Table I, then suitable protective equipment must be
provided for the operators.
Noise exposure
Equipment and machinery noise levels must be kept
Electrical equipment
below the values given in OSHA Safety and Health
All electrical equipment (e.g. motors, switchgear,
Standards (29CFR 1910), Paragraph 1910.95 - Occupa-
wiring,) must be designed and installed to comply with
tional Noise Exposure (see Table I). If equipment/
the latest edition of the National Electrical Code (NEC),
and paragraphs 1910.301 to .399 of OSHA Safety and
Received19 January 1988 Health Standards (29CFR 1910).
OSSO-4230/%%/020062-13S3.00
0 1988 Butterworth & Co. (Publishers) Ltd
62 J. Loss Prev. Process Jnd., 1988, Vol 1, April
 Safety considerations in conveying bulk solids and powders: S. Grossel

When electrical equipment needs to be repaired or hazards. Palmer3 lists minimum ignition energies for
maintained, it must be locked out and tagged to prevent many bulk solids and powders. These values are
injury to personnel. There is currently no OSHA stand- listed in Table 2.
ard, but legislation has been proposed and submitted
for review, entitled Control of Hazardous Energy
Dust explosion hazards
Sources (Lockout/Tagout). There is also an ANSI
The subject of dust explosions is too large and com-
consensus standard 2244.1 (1982).
plicated to cover in depth in this review, but certain
aspects of it will be discussed below to present some
Static electricity and dust explosion hazards
fundamentals and background materials. For further
reading on the subject, the following books are recom-
Static electricity hazards
mended: Cross and Farrer, Palmer, Bartknecht4,
When conveying bulk solids and powders, especially
Field, and Nagy and Verakis6.
organic ones, static electrical charges can develop.
A dust explosion results when finely divided combus-
These charges arise from contacts made between sur-
tible matter is dispersed in an atmosphere containing
faces during the movement of the particles. The charge
sufficient oxygen to permit combustion and a source
on a powder particle is governed by three factors: the
of ignition of appropriate energy is present. Dust
charge production rate; the charge leakage rate when
explosions have certain similarities to gas explosions,
the particle is in contact with a ground; and the
especially with regard to the chemical processes involved
electrical breakdown of air initiated by the high field
and in cases where the particle size of the dust is less
around the charged particle.
than 5 pm. However, there are significant differences
There are five fundamental quantities in an under-
standing of electrostatics. The most basic is the electric which make the study of dust explosions extremely
difficult.
charge that is transferred to a material, usually by
For a dust explosion to occur there must be a degree
friction. When an object is charged it exerts a force on
of turbulence, if only to disperse the dust into a
any other charged object, and is then said to have an
suspension. Gas explosions can occur when the gas is in
electric potential or voltage, k. The rate of change of
a quiescent state, the mixture being homogeneous and
voltage with distance or potential gradient is the electric
consisting of molecular-size particles. The suspensions
held, E. The potential reached by an object having a
of dusts encountered in dust explosions are, however,
charge q depends on its electrical capacity, C. The
unlikely to be homogeneous, and would normally con-
higher the capacity the more charge is needed to achieve
tain a range of concentrations of particles which are
a given potential.
many orders of magnitude larger and heavier than gas
The rate at which charge dissipates depends primarily
molecules and which settle out of suspension due to
on the electrical resistance R between the stored charge
and ground. An electrostatic spark occurs when an gravity.
A dust explosion involves such a high rate of combus-
isolated charged object is suddenly grounded. The
accumulation of static electricity on an object produces tion that individual particles and agglomerates are either
consumed or oxidized. The combustion of carbon in
an electric field around it and a spark will occur if the
organic material produces gaseous products which in
field strength exceeds the breakdown value of the sur-
themselves take up more space than the solids of the
rounding atmosphere. For air, this is approximately
parent material. An expanding flame front will also
3000 kV/m.
Electrostatic sparks can cause dust explosions if they result from the ignition of flammable gases produced by
achieve a minimum ignition energy and the dust cloud in the decomposition of the dust. A dust explosion there-
the air is within the explosive concentration range. The fore requires more space because of the expansion of the
minimum ignition energy is the energy which just ignites hot gaseous products.
the most easily ignited mixture and is usually measured In industrial plant, the heat released during a dust
with a capacitor discharge by varying the charge quan- explosion is likely to exceed the natural rate of cooling
tity, the capacitance and the electrode separation at and consequently an explosion would be accompanied
standard temperature and pressure conditions (1 bar, by significant, and, in some cases, uncontrolled expan-
2OC), or when applicable, in saturated vapour. sion effects. In an unconfined situation, there would be
The minimum ignition energy of a dust-air mixture mainly localized flames and pressure effects. However,
(with the exception of explosives and other reactive in the confined situations commonly found in plant
materials is lo-100 mJ and is therefore 50 to 1000 times handling particulate matter, the expansion effects are
greater than those of gas-air or vapour-air mixtures. likely to be sufficient to burst through the confines of the
Determining the minimum ignition energy is, with other plant equipment and/or piping.
data, an important aid in quantifying the potential The following conditions must exist for a dust
hazard of electrostatic charges. explosion to occur:
Cross and Farrer present an excellent discussion of l The dust must be combustible.
static electricity phenomena and measurement tech- l The dust must be in suspension in the atmosphere
niques for minimum ignition energy. Also, the book by which must contain sufficient oxygen to support
Haase is a good source of information on electrostatic combustion.

J. Loss Prev. Process lnd., 1988, Vol I, April 63

Safety considerations in come ying bulk solids and powders: S. Grossel

Table 2 Dust explosion parameters

Maximum
oxygen
Minimum coce-
ignition Minimum Maximum tration
temperature explosible Minimum Maximum rate of to prevent
fC) COCB- ignition explosion pressure ignition
tration energy pressure rise 1% by
Dust cloud layer fgl I (mJ) lib I ? (lb in >sl volume) References Notes

Acetamide 560 - _ - _ _ Group (bl dust

Aceto acetanilide 560 ~ 0.030 20 90 4 800
Acetoacet-p-phenetedtde 560 - 0.030 10 87 >10000 _
Acetoecet-o-toluidine 710 - _ _ - -
2 Acetylamino-5-nitro 450 450 0.160 40 137 9 000 _
thiazole
Acetyl-p-nitro-o- 450 - - - _ _ 8
toluidine
Adipic acid 550 - 0.035 60 95 4 000 4.8
Alfalfa 460 200 0.100 320 88 1 100 _ 5
Almond shell 440 200 0.065 80 101 1 400 _ 5
Aluminium, atomized 650 760 0.045 50 84 > 20 000 - 3
Aluminium, flake 610 320 0.045 10 127 > 20 000 _ 3
Aluminium-cobalt alloy 950 570 0.180 100 92 11 000 - 3
Aluminium-copper alloy ~ 830 0.100 100 95 4 000 3
Aluminium-iron alloy 550 450 - _ 36 300 3
Aluminium-lithium alloy 470 400 <O.l 140 96 6000 - 3
Aluminium-magnesium 430 480 0.020 80 86 10000 _ 3
alloy
Aluminium-nickel alloy 950 540 0.190 80 96 10000 _ 3
Aluminium-silicon alloy 670 - 0.040 60 85 7 500 _ 3
Aluminium acetate 560 640 - _ 59 950 1 Guncotton ignitton
source in pressure
test
Aluminium octoate 460 - - - - 8
Aluminium stearate 400 380 0.015 10 86 >lOOOO _ 1
2 Amino-5-nitrothiazole 460 460 0.075 30 110 5 600 - 1
Anthracene 505 Melts _ 68 700 _ 2.8
Anthranilic acid 580 - 0.030 35 84 8 500 1
Anthraquinone 670 ~ _ _ - 8
Antimony 420 330 0.420 1 920 28 300 3
Antipyrin 405 Melts 53 - - 2
Asphalt 510 500 0.025 25 94 4 800 _ 7
Aspirin 550 Melts 0.015 16 87 7 700 _ 8
Azelaic acid 610 ~ 0.025 25 76 4 700 4
awl Azo isobutyronitrile 430 350 0.015 25 134 8 000 _ 4
Barley 370 - _ _ 8
Benzethonium chloride 380 410 0.020 60 91 6 700 _ 1
Benzoic acid 600 Melts 0.011 12 95 10 300 _ 8
Benzotriazole 440 - 0.030 30 103 9 200 _ 1
Benzoyl peroxide _ 21 _ 8
Beryllium 910 540 _ - Did not Ignite 3 Contained 8 per
cent oxide
Beryllium acetate, basic 620 - 0.080 100 87 2 200 15 1 Inert gas carbon
dioxide
Bis (2hydroxy-5- 570 - 0.040 60 70 2 000 13 1 Inert gas carbon
chlorophenyl)-methane dioxide
Bis (2-hydroxy-3. 5,6,- Did not 450 _ _ _ - 1
trichlorophenyl).methane ignite
Bone meal 490 230 11 100 _ 6 Guncotton ignition
source in pressure
test
Boron 730 390 Did not _ 41 200 3 Guncotton ignition
ignite source in pressure
test
Bread 450 - _ _ - _ 8
Brunswick green 360 - - _ 8
P-t-butyl benzoic acid 560 - 0.020 25 88 6 500 - 4
Cadmium 570 250 _ 4oocl 7 100 _ 3
Cadmium yellow 390 _ _ - 8
Calcium carbide 555 325 _ _ 13 - _ 2
Calcium citrate 470 - _ _ - _ 8 Group (bl dust
Calcium gluconate 550 - _ - _ 8 Group (bl dust
Calcium DL pantothenate 520 - 0.050 80 105 4 600 _ 1
Calcium propionate 530 _ _ 90 1 900 8
Calcium silicide 540 540 0.060 150 86 20 000 _ 3
Calcium stearate 400 - 0.025 15 97 >lOOOO _ 1

64 J. Loss Prev. Process Ind., 1988, Vol 7, April

 Safety considerations in conveying bulk solids and powders: S. Grossel

Table 2 (continued)

Maximum
oxygen
Minimum coce-
ignition Minimum Maximum nation
temperature explosible Minimum Maximum rate of to prevent
(OC) concen- ignition explosion pressure ignition
tration energy pressure rise (% by
Dust cloud layer (gl 1 ImJl (lb in ) (lb in * s.1 volume) References Notes

Caprolactam 43b - 0.07 60 79 1 700 8 8

Carbon, activated 660 270 0.100 92 1 700 _ 7 Guncotton ignition
source in min. expl.
cont. and max.
expl. pressure tests
Carbon, black 510 - _ _ _ _ _ 7
Carboxy methyl cellulose 460 310 0.060 140 130 5 000 _ 4
Carboxy methyl hydroxy 380 - 0.200 960 83 800 _ 4
ethyl cellulose
Carboxy polymethylene 520 ~ 0.115 640 76 1 200 _ 4
Casein 460 - - _ 89 1 200 _ 8
Cellulose 410 300 0.045 40 117 8 000 _ 4
Cellulose acetate 340 - 0.035 20 114 6 500 5 4.8 Inert gas nitrogen
Cellulose acetate butyrate 370 - 0.025 30 81 2 700 7 4
Cellulose proprionate 460 - 0.025 60 105 4 700 _ 4
Cellulose triacetate 390 ~ 0.035 30 107 4 300 4,8
Cellulose tripropionate 460 ~ 0.025 45 88 4 000 _ 4
Charcoal 530 180 0.140 20 100 1 800 _ 7
Chloramine-T 540 150 _ _ 7 150 _ 1 Guncotton lgnitlon
source in pressure
test
o-Chlorobenzmalono nitrile 0.025 90 >10000 1
o-Chloroaceto acetanilide 64; ~ 0.035 30 94 3 900 1
p-Chloroaceto acetanilide 650 ~ 0.035 20 85 5500 1
Chloro amino toluene 650 - _ _ _ _ _ 8
sulphonic acid
4-Chloro-2 nitro aniline 590 120 co.750 140 123 3500 _ 1
p-Chloro o-toluidine 650 - _ _ - _ _ 8
hydrochloride
Chocolate crumb 340 - _ _ _ _
Chromium 580 400 0.23; 1 40 56 5000
Cinnamon 440 230 0.060 30 121 3900 _
Cttrus peel 500 330 0.060 100 51 1 200 _
Coal, brown 485 230 _ _ _ - _ See also Lignite
Coal, 8 per cent volatiles 730 - _ _ _ _ _
Coal, 12 per cent volatiles 670 240 _ _ _ _ _
Coal, 25 per cent volatiles 605 210 0.120 1 20 62 400 _
Coal, 37 per cent volatiles 610 170 0.055 60 90 2300 Standard Pittsburgh
coal
Coal, 43 per cent volatiles 575 180 0.050 50 92 2 000 _
Cobalt 760 370 - _ _ _ _ ;
Cocoa 500 200 0.065 120 69 1 200 _ 5
coconut 450 280 _ _ _ _ 2
Coconut shell 470 220 0.035 60 115 4 200 _ 5
Coffee 360 270 0.085 160 38 150 10 5.8 Inert gas carbon
dioxide
Coffee, extract 600 - _ _ 47 _ _ 2
Coffee, instant 410 350 0.280 Did not 68 500 _ 5
ignite
Coke >750 430 _ _ _ _ 2
Coke, petroleum, 13 per 670 ~ 1 .oo 36 200 _ 7 Guncotton ignition
cent volatiles source in min. expl.
cow. and max.
expl. pressure tests
Colophony 325 Melts _ - - 2
Copal 330 Melts _ 68 _ _ 2 See also Gum
manila
Copper 700 - _ Did not Did not Did not _ 3
ignite ignite
Copper-zinc, gold bronze 370 190 1 .oo 44 1 300 _
Cork 460 210 0.035 96 7 500 _
Corn cob 450 240 0.045 127 3 700 _
Corn dextrine 410 390 0.040 124 7 000 _
Cornflour 390 -
Cornstarch 390 0.040 30 145 9 500
Cotton flock 470 - 0.050 25 94 6 000 _

icontinoedl

J. Loss Prev. Process Ind., 1988, Vol 1, April 65

 Safety considerations in conveying bulk solids and powders: S. Grosset

Table 2 (continued)

Maximum
oxygen
Minimum COCB-
ignition Minimum Maximum tration
temperature explosible Minimum Maximum rate of to prevent
fC) concen- ignition explosion pressure ignition
tration energy pressure rise 1% by
Dust cloud layer fgl II fmJ) (lb in 1 (lb in-*s) volume) References Notes

Cotton linters 520 - 0.50 1920 73 400 5 5

Cottonseed meal 530 200 0.055 80 89 2 200 - 5
Coumarone-indene resin 550 - 0.015 10 93 11000 11 4
Crystal violet 475 Melts _ _ _ _ _ 2
Cyclohexanone peroxide - _ _ 21 84 5 600 _ 8
Dehydroacetic acid 430 - 0.030 15 87 8 000 _ 1
Dextrin 410 440 0.050 40 99 9 000 6
Dextrose monohydrate 350 - - - _ 8
Diallyl phthalate 480 - 0.030 20 90 8 500 _ 1
Diamino stilbene 550 - _ _ _ _ _ 8 Group fbf dust
drsulphonic acid
Diazo aminobenzene 550 - 0.015 20 114 >10000 - 1
Di-t-butyl-p-cresol 420 - 0.015 15 79 13000 9 4
Dibutyl tin maleate 600 - _ _ _ _ 8
Dibutyl tin oxide 530 - - _ _ - 8
Dichlorophene 770 - _ _ 72 3 000 - 1
2.4.Dichlorophenoxy ethyl 540 ~ 0.045 60 84 2 200 _ 1
benzoate
Dicyclopentadiene dioxide 420 - 0.015 30 89 9 500 _ 4
Dihydrostreptomycrn 600 230 0.520 _ 42 200 7 1
sulphate
3-3 Dimethoxy 4-4 0.030 _ 82 >10000 _ 1
diamino diphenyl
Dimethylacridan 540 _ _ _ _ -
Dimethyl drphenyl urea 490 ~ _ _ _ -
Dimethyl isophthalate 580 - 0.025 15 84 8 000
Dimethyl terephthalate 570 - 0.030 20 105 12 000 s
S-S-Dimethyl xanthogene- 400 - 0.300 3200 84 1 500
thylene bis dithiocarba-
mate
Dinitro aniline 470 ~ _ _ _ - 8
3, 5-Dinitrobenzamide 500 Melts 0.040 45 163 6 500 - 1
3. 5-Dinitrobenzoic acid 460 - 0.050 45 139 4300 _ 1
Dinitrobenzoyl chloride 380 - - 8
Drnitrocresol 340 Melts 0.030 - _ _ _ 1.2
4. 4.Drnitro-sym-diphenyl 550 - 0.095 60 102 2 500 _ 1
urea
Drnitro stilbene disulphonic 450 ~ - _ _ 8
acid
Dinitrotoluamide 500 - 0.050 15 153 >lOOOO _ 6
Diphenyl 630 - 0.015 20 82 3 700 _ 1
4,4-Diphenyl di 590 140 0.065 30 143 5 500 _ 1
sulphonylazide
Drphenylol propane 570 ~ 0.012 11 81 11 800 5 4.8 Inert gas nitrogen
(BisphenoCA)
Egg white 610 - 0.14 640 58 500 _ 5
Epoxy resin 490 - 0.015 9 94 8 500 4.8
Esparto grass _ 94 7 300 8
Ethyl cellulose 340 330 0.025 15 112 7 000 4
Ethylene diamine tetra 450 - 0.075 50 106 3 000 _ 1
acetic acid
Ethyl hydroxyethyl 390 - 0.020 30 94 2 200 _ 6
cellulose
Ferric ammonium 390 210 1.500 _ 17 100 1
ferrocyanide
Ferrrc dimethyl dithio 280 150 0.055 25 86 6 300 _ 1
carbamate
Ferric ferrocyanide 370 - _ 82 1 000 - 1
Ferrochromium 790 670 2.00 _ _ 3
Ferromanganese 450 290 0.130 80 62 5oDO 3
Ferrosilicon 640 - _ _ _ _ _ 2
(45 per cent Sif
Ferrosilicon Did not 980 0.240 1280 113 3 500 _ 3
(90 per cent Si) ignite
Ferrotitanium 370 400 0.140 80 55 9 500 3
Ferrous ferrocyanide 380 190 0.400 _ _ _ _ 1
Ferrovanadrum 440 400 1.300 400 _ _ _ 3

66 J. Loss Prev. Process Ind., 1988, Vol I, April

 Safety considerations in conveying bulk solids and powders: S. Grossel

Maximum
oxygen
Minimum concan-
ignition Minimum Maximum tration
temperature explosible Mintmum Maximum rate of to prevent
(OC) concen- ignition explosion pressure Ignition
tration energy pressure rise (% by
Dust cloud layer (gl I (mJ) (Ibin 2, (lb in 2 s) volume) References Notes

Fish meal 485 ~ _ _ _ _ _ 2

Fumaric acid 520 - 0.085 35 103 3 000 _ 4
Garlic 360 ~ 0.10 240 57 1 300 _ 5
Gelatin, dried 620 480 co.5 78 1 200 1
Gilsonite 580 500 0.020 25 78 4 500 7
Graphite 730 580 _ - _ - _ 7
Grass _ _ 56 400 _ 8
Gum arabic 500 260 0.060 100 117 3 000 _ 4
Gum Karaya 520 240 0.100 180 116 2 500 _ 4
Gum manila (copal) 360 390 0.030 30 89 6 000 _ 4
Gum tragacanth 490 260 0.040 45 123 5 000 4
Hexa methylene tetramlne 410 - 0.015 10 98 11000 11 4
Horseradish <0.100 _ 96 1 600 _ 6
Hydrazine acid tartrate 570 - 0.175 460 30 200 _ 1
p-Hydroxy benzoic acid 620 - 0.040 _ 37 - _ 1
Hydroxyethyl cellulose 410 - 0.025 40 106 2 600 _ 6
Hydroxyethyl methyl 410 - _ _ _ _ 8
cellulose
Hydroxy propyl cellulose 400 ~ 0.020 30 96 2 900 6
Iron 430 240 _ _ _ 2
Iron, carbonyl 420 230 0.105 100 47 8 000 _ 3
Iron pyrites 380 280 1 .oo 8 200 5 100 _ 3
lsatoic anhydride 700 - 0.035 25 80 4 900 _
lsinglass 520 - _ - Nil Nil _ 8
lsophthalic acid 700 ~ 0.035 25 78 3 100 4
Kelp 570 220 Did not ionite 19 200 5
Lactalbumin 570 240 0.040 50 97 3 500 4
Lampblack 730 - _ _ _ 7
Lauryl peroxide _ 12 90 6 400 8
Lead 790 290 _ Did not 3 100 _ 3 Flame ignition
ignite source in pressure
test
Leather 390 ~ _ _ _ 8
Lignin 450 - 0.040 20 102 5 000 7 4
Lignite 450 200 0.030 30 94 8 000 7
Lycopodium 480 310 0.025 40 75 3 100 9 5
Magnesium 560 430 0.030 40 116 15000 _ 3
Maize husk 430 - _ _ 75 700 _ 8
Maize starch 410 - _ - _ _ _ 2
Maleic anhydride 600 Melts _ _ 2
Malt barley 400 260 0.055 35 95 4 400 5
Manganese 460 240 0.125 305 53 4 900 _ 3
Manganese ethylene bis 270 - 0.07 35 - _ _ 8
dithio carbamate
Maniac 430 - _ _ - _ _ 8
Mannitol 460 - 0.065 40 97 2 800 1
Melamine formaldehyde 410 ~ 0.02 50 93 1 800 _ 4.8
resin
DL Methionine 370 360 0.025 35 119 5 700 7
1 -Methylamino 830 Melts 0.055 50 71 3 300 _
anthraquinone
Methyl cellulose 360 340 0.030 20 133 6 000 _ 4
2. 2-Methylene bis-4-ethyl- 310 - _ _ 76 7 300 _ 8
6-tmbutyl phenol
Milk 440 ~ _ 8
Milk, skimmed 490 200 0.050 50 95 2 300 5
Milk sugar 450 Melts _ _ 31 _ _ 2
Molybdenum 720 360 _ _ _ _ _ 3
Molybdenum disulphide 570 290 _ _ _ - _ 6
Monochlorecetic acid 620 ~ _ _ _ _ 8
Monosodium salt of 540 ~ _ _ 8 Group lb) dust
trichloro ethyl phosphate
Moss, Irish 530 230 Did not ignite 21 300 - 5
Naphthalene 575 Melts _ _ 87 - - 2
P-Naphthalene-azo- 510 Melts 0.020 50 70 2 300 _
dimethyl aniline
P-Naphthol 670 ~ _
Naphthol yellow 415 395 _ - _

J. Loss Prev. Process Ind., 7988, Vol 1, April 67

Safety considerations in conveying bulk solids and powders: S. Grossel

Table 2 (continued1

Maximum
oxygen
Minimum concen-
ignition Minimum Maximum nation
temperature explosible Minimum Maximum rate of to prevent
(OC) concen- ignition explosion pressure ignition
tration energy pressure. rise (o/o by
Dust cloud layer (gl II (mJ) (lb in *) (lb I sl volume) References Notes

Nigrosine hydrochloride 630 - _ _ 8

p-Nitro-o-anisidene 400 -- _ _ _ _ 8
pmN$y-benzene arsenic 360 280 0.195 480 77 900 _ 1

Nitrocellulose _ 30 > 256 > 20 900 _ 8

Nitro diphenylamine 480 ~ ._ _ - 8
Nitro furfural semi 240 - _ >143 8 600 _ 8
carbazone
Nitropyridone 430 Melts 0.045 35 111 >10000 _ 1
p-Nitro-o-toluidine 470 - _ _ _ _ - 8
m-Nitro-p-toluidine 470 - _ _ _ _ _ 8
Nylon 600 430 0.030 20 95 4 000 6 4
Oilcake meal 470 285 _ _ - _ 2
Onion, dehydrated 410 - 0.130 Did not 35 500 _ 5
ignite
Paper 440 270 0.055 60 96 3 600
Para formaldehyde 410 - 0.040 20 133 13000 _
Peanut hull 460 210 0.045 50 116 8 000 _
Peat 420 295 _ _ _ _
Peat, sphagnum 460 240 0.045 50 104 2 200 _
Pectin 410 200 0.075 35 132 8 000 _
Penicillin, N-ethyl 310 - _ - _
piperidine salt of
Penta erythritol 450 - 0.030 10 90 9 500 7 4
Phenol formaldehyde 460 ~ 0.015 10 107 6 500 _ 8.4
Phenol furfural resin 530 - 0.025 10 88 8 500 - 4
Phenothiazine 540 - 0.030 56 3 000 - 1
p-Phenylene diamine 620 - 0.025 30 94 11 000 4
Phosphorus, red 360 305 _ _ _ ._ 2
Phosphorus pentasulphide 280 270 0.050 15 64 >10000 _ 1
Phthalic acid 650 Melts - _ 62 _ _ 2
Phthalic anhydride 605 Melts 0.015 15 72 4 200 11 2,4
Phthalimide 630 - 0.030 50 89 4 800 _ 1
Phthalodinitrile 2700 Melts _ 43 _ _ 2
Phytosterol 330 Melts 0.025 10 76 > 10000 _ 1
Piperazine 480 - _ - 72 1 400 1
Pitch 710 - 0.035 20 88 6 000 _ 7
Polyacetal 440 ~ 0.035 20 113 4 100 _ 4
Polyacrylamide 410 240 0.040 30 85 2500 _ 4
Polyacrylonitrile 500 460 0.025 20 89 11 000 _ 4
Polycarbonate 710 - 0.025 25 96 4 700 _ 4
Polyethylene 390 - 0.020 10 80 7 500 4.8
Polyethylene oxide 350 - 0.030 30 106 2 100 5 4
Polyethylene terephthalate 500 - 0.040 35 98 5500 _ 4
Poly isobutyl methacrylate 500 280 0.020 40 74 2800 _ 4
Poly methacrylic acid 450 290 0.045 100 97 1800 _ 4
Polymethyl methacrylate 440 - 0.020 15 101 1 800 7 4
Polymonochlorotrifluoro 600 720 Did not ignite _ 4
ethylene
Polypropylene 420 - 0.020 30 76 5 500 _
Polystyrene 600 500 0.020 15 100 7 000
Polytetrafluoro ethylene 670 570 Did not ignite _
Polyurethane foam 510 440 0.030 20 87 3 700 _
Polyurethane foam, fire 550 390 0.025 15 96 3 700 _
retardant
Polyvinylacetate 450 - 0.040 160 69 1 000 11 4.8 Inert gas carbon
dioxide
Polywnyl alcohol 450 Melts 78 _ _ 2
Polyvinyl butyral 390 - 0.020 10 84 2 000 5 4
Polyvinyl chloride 670 - Did not ignite 38 500 _ 4 Flame ignition
source
Polyvinylidene chloride 670 ~ _ _ _ _ _ 8 Group lb) dust
Polyvinyl pyrrolidone 465 Melts _ _ 15 _ - 2
Potassium hydrogen 520 - _ _ - 1
tartrate
Potassium sorbate 380 180 0.120 60 79 9 500 1
Potato, dried 450 ~ _ _ 97 1 000 _ 8

66 J. Loss Prev. Process Ind., 1988, Vol 7, April

 Safety considerations in conveying bulk solids and powders: S. Grossel

Table 2 (conrinued)

Maximum
oxygen
Minimum concen-
ignition Minimum Maximum fration
temperature explosible Minimum Maximum rate of to prevent
CC) concen- ignition explosion pESSU,e ignition
tration energy pressure rise (% by
Dust cloud layer fgl I (mJI (lb in ) (lb in s) volume1 References Notes

Potato starch 430 - - _ _ _ _ 2

Provender 370 - _ _ 93 1 400 a
Pyrethrum 460 210 0.100 80 95 1 500 1
Quillaia bark 450 ~ _ _ _ _ a
Rape seed meal 465 ~ _ _ _ _ _ 2
Tayon. viscose 420 - _ - _ _ _ 8
Rayon, flock 0.03 _ _ _ _ 8
Rice 440 240 0.050 50 105 2 700 _ 5
Rosin 390 - 0.015 10 87 12000 4
Rubber 380 - _ 8
Rubber, crude, hard 350 ~ 0.025 50 80 3 800 13 4
Rubber, crumb 440 ~ _ 84 3 300 8
Rubber, vulcanized 360 - _ _ 40 _ _ 2
Rye flour 415 325 _ 35 _ 2
Saccharin 690 - _ _ _ _ _ 1
Salicylanilide 610 Melts 0.040 20 73 4 800 _ 1
Salicylic acid 590 - 0.025 84 8 800 4.6
Sawdust 430 ~ 97 2 000 8
Sebacic acid _ 74 400 - 8
Sea 440 - 0.010 105 49 300 8
Shellac 400 - 0.020 10 73 3 600 9 4
Silicon Did not 760 <O.lO 80 94 13000 _ 3
ignite
Soap 430 600 0.085 100 77 2 800
Sodium acetate 590 ~ 0.030 35 90 4 600
Sodium amatol 580 Melts 0.140 _ 65 800
Sodium benzoate 560 680 0.050 80 91 3 700 _
Sodium carboxymethyl 320 - 1.10 440 49 400 5
cellulose
Sodium 2-chloro-5 nitro- 550 440 _ _ _ _ _ 1
benzene sulphonate
Sodium 2,2-dichloro 500 ~ 0.260 220 68 500 1
propionate
Sodium dihydroxy 510 - _ _ _ _ _ a Group lb1 dust
naphthalene dlsulphonate
Sodium glucaspaldrate 600 - _ _ _ _ _ 8
Sodium glucoheptonate 600 - _ _ _ _ _ a
Sodium monochloracetate 550 - _ _ _ _ _ 8
Sodium m-nitrobenzene _ _ 92 400 _ 1
sulphonate
Sodium m-nitrobenzoate _ 87 2 900 _ 1
Sodium pentachlorophenate Did not 360 _ _ Drd not ignite _ 1
ignite
Sodium propionate 479 - _ _ 70 700 _ 8
Sodium secobarbital 520 - 0.100 960 76 800 _ 1
Sodium sorbate 400 140 0.050 30 87 6 500 1
Sodium thiosulphate 510 330 11 <lOO _ 1 Guncotton ignition
source in pressure
test
Sodium toluene sulphonate 530 - _ _ _ _ _ 8
Sodium xylene sulphonate 490 - _ _ _ _ _ 8
soot >690 535 _ _ Did not ignite _ 2
Sorbic acid 440 460 0.020 15 106 >lOOOO 5 i,a Inert gas nitrogen
L-Sorbose 370 ~ 0.065 80 76 4 700 1
Soya flour 550 340 0.060 100 94 800 9 5
Soya protein 540 - 0.050 60 98 6 500 9 5
Starch 470 - _ _ - _ - 8
Starch, cold water 490 - - _ 8
Stearic acid 290 - _ 25 80 8 500 _ 1
Steel 450 ~ 8
Streptomycin sulphate 700 - _ _ - - - 8
Sucrose 420 Melts 0.045 40 86 5 500 _ 1
Sugar 370 400 0.045 30 109 5 000 - 5
Sulphur 190 220 0.035 15 78 4 700 _ 1
Tantalum 630 300 < 0.20 120 55 4 400 3
Tartaric acid 350 ~ _ _ _ _ _ 8
Tea 500 - _ _ 93 1 700 _ 8
Tea, instant 580 340 Did not ignite 48 400 _ 5 lcontinuedl

J. Loss Prev. Process lnd., 1988, Vol 7, April 69

Safety considerations in conveving bulk solids and powders: S. Grossel

Maximum
oxygen
Minimum concen-
ignition Minimum Maximum tration
temperature explosible Minimum Maximum rate of to prevent
(C) concen- ignition explosion pressure ignition
tration energy pressure rise f% by
Dust cloud layer fgl 1 fmJ1 (lb in *) (lb in *s) volume) References Notes

Tellurium 550 340 - _ _ 3

Terephthalic acid 680 - 0.050 20 84 8 000 _ 4
Tetranitro carbazole 395 Melts _ - _ - _ 2
Thiourea 420 Melts _ 29 100 _
Thorium 270 280 0.075 5 79 5 500 _ 3
Thorium hydride 260 20 0.080 3 81 12000 _ 3
Tin 630 430 0.190 80 48 1 700 3
Titanium 375 290 0.045 15 85 11 000 Ignites 2.3
in
carbon
droxide
Titanium hydride 480 540 0.070 60 121 12000 3
Tobacco 485 290 _ _
Tobacco, dried 320 - _ _ 85 1 000 -
Tobacco, stem 420 230 Did not ignite 53 400
Tribromosalicyl anilide 880 Melts _ - 63 2 100_ _
Trinitro toluene 0.070 75 _
s-Trioxane 480 ~ 0.143 _ 85 600 _
cv, cx-Trithiobis (N, N- 280 230 0.060 35 96 6 000 _
dimethyl-thioformamide)
Tung 540 240 0.070 240 74 1 900 5
Tungsten 730 470 _ - Did not ignite _ 3
Uranium 20 100 0.060 45 69 5 000 _ 3
Uranium hydride 20 20 0.060 5 74 9 000 - 3
Urea 900 -, Did not ignite 4.8 Group (b) dust
Urea formaldehyde 460 - 0.085 80 89 3 600 9 4
moulding powder
Urea formaldehyde resin 430 - 0.02 34 110 1 600 8
Vanadium 500 490 0.220 60 57 1 000 10 3
Vitamin Bl mononitrate 380 190 0.035 35 120 9 000 - 1
Vitamin C 460 280 0.070 60 88 4 800 _ 1
Walnut shell 420 210 0.035 60 121 5 500 _ 5
Wax. accra 260 - _ _ _ 8
Wax. carnauba 340 - _ _ 8
Wax, paraffin 340 - _ _ - _ 8
Wheat, flour 380 360 0.050 50 109 3 700 5
Wheat, grain dust 420 290 _ _ 43 _ 2
Wheat starch 430 ~ 0.045 25 100 6 500 5
Wood 360 - 90 5 700 8
Wood, bark 450 250 0.020 60 103 7 500 4
Wood, flour 430 - 0.050 20 94 8 500 4
Wood, hard 420 315 _ _ 66 2
Wood, soft 440 325 _ _ 63 _ 2
Yeast 520 260 0.050 50 123 3 500 _ 5
Zinc 680 460 0.500 960 70 1 800 3
Zinc ethylene dithio- 480 180 45 300 _ 1
carbamate
Zinc stearate 315 Melts 0.020 10 80 >10000 1.2
Zirconium 20 220 0.045 5 75 11 000 Ignites
in 3
carbon
dioxide
Zirconium hydride 350 270 0.085 60 90 9 500 3 3

Reprinted by permission of Chapman and Hall Lrd

The dust must have a particle size distribution that volume and characteristics of the vessel or chamber
will propagate flame. containing the dust suspension, the dispersion and
The dust concentration in the suspension must be concentration of the dust suspension and the degree of
within the explosible range. turbulence in the vessel.
The dust suspension must be in contact with an The explosibility of a dust can be determined by tests
ignition source of sufficient energy. which are described by Field. The tests specific to dust
explosion venting are described by Field and
Under these conditions the hazard from a dust Schofield .
explosion depends upon the explosibility of the dust, the Generally speaking, the explosibility of a combustible

70 J. Loss Prev. Process Ind., 1988, Vol 7, April

 Safety considerations in conveying bulk solids and powders: S. Grossel

dust is greater the smaller the particle size. The Where possible select less dusty alternatives for
minimum ignition energy is reduced and the maximum materials and minimize attrition.
explosion pressure and rate of pressure rise are Minimize handling of dusty materials and design the
increased with a decrease in particle size. In addition, handling system to minimize dust generation and the
fine particles more readily stay in suspension increasing size of dust clouds.
the probability of producing an explosible concen- Avoid the accumulation of dust (which can be dis-
tration. Particles greater than 2500 pm diameter are turbed to form a dust cloud) by the detailed design of
unlikely to cause dust explosions, although the possi- equipment, building and working practices.
bility of coarser materials producing fine dust by Anticipate possible ignition sources and eliminate
attrition during handling must be anticipated. them, as far as is reasonably practicable, by appro-
Minimum explosible concentrations in air are typ- priate equipment design, grounding, maintenance
ically in the range lo-500 gme3 and some values are and working practices.
given in Table 2 together with other explosion param- Take appropriate additional precautions, where
eters. Explosible concentrations are much higher than practicable, such as inerting, containment, venting
those associated with toxic hazards or nuisance prob- or suppression.
lems (which might range from = 1 to 10 mgme3) and Isolate vulnerable plant where appropriate.
such explosible concentrations would most likely occur
very close to a dust source or within an enclosed space Prevention and protective measures wi!l be described in
where the dust cloud cannot spread. Indeed, enclosure detail below.
and confinement of dust sources to solve occupational
hygiene or nuisance problems may well concentrate the
dust cloud increasing the risk of explosion. This danger Pneumatic conveying systems
is often overlooked by environmental protection engi-
neers. Pneumatic conveying systems cause the highest risk of
Ignition sources of sufficient energy to cause a dust dust explosions and fires for the following reasons:
explosion are many but the main ones are listed below:

Flames Static electricity is generated by contact between

Hot surfaces particles themselves and between particles and the
Incandescent material pipewall.
Spontaneous heating Dust concentrations within the explosible range can
Welding or cutting operations arise at the delivery point where the dust is separated
Friction heating or sparks from the air (silos, cyclones, baghouses).
Impact sparks Heated particles which are created during grinding or
Electrical sparks drying may be carried in a pneumatic transport
Electrostatic discharge sparks system and fanned to a glow by the high air velocity.
These particles can then cause an ignition in the
Precautions against dust explosions may be for preven- storage or collection system at the end of the
tion or protection and are summarized in Table 3. pneumatic transport. Tramp metal in pneumatic
The following general approach to dust explosion systems may also cause frictional heating or sparks
precautions is recommended: as it passes through the system.

Table 3 Summary of dust explosion precautions

Method Comments

Prevention (of an explosion occurring)

Exclusion of dust cloud
Exclusion of ignition sources
Exclusion of oxygen (inerting)
using Nz, CO2 or other suitable
gas.
Diluent dust addition - to
reduce explosibility of dust
Containment
Venting
Suppression
Material can be rendered less dusty and handling system designed to minimize dust. Impossible to
guarantee total dust free environment short of changing to wet process.
All practical measures must be taken to exclude ignition sources but because sources are often
unknown It is difficult to guarantee so other precautions usually taken.
Reduces oxygen content below minimum necessary to support combustion (typically c6-15%).
Requires continuous monitoring of oxygen content. Usually requires closed system to conserve
inert gas. Expensive.
Non-combustible diluent, well mixed with dust, acts as heat sink thus reducing explosibility of
dust. Limited application because of contamination and expense. (Typically the amount of inert dust
exceeds 50%).
Vessel and associated pipework etc built sufficiently strong to withstand the maximum explosion
pressure. Expensive in all but the smallest systems.
Vents provided in walls of vessel to allow escape of dust and combustion products to limit pressure
nse to an acceptable level. Widely used.
Start of explosion detected by instruments which trigger release of fire suppressants. Useful where
venting is unacceptable or impracticable e.g. when the dust is toxic.

J. Loss Prev. Process Ind., 1988, Vol 1, April 71

Safety considerations in conveying bulk solids and powders: S. Grosset

Design considerations to minimize hazards piping and even cyclones for this is feasible. Usually,
Pneumatic conveying systems should possess the large-volume baghouses are not designed to contain
following general characteristics: explosions, for economic reasons. In recent years, how-
ever, several baghouse vendors have been offering
They should be air tight, to prevent the escape of
small-to-medium volume units designed for up to
dust from the system where it might present a fire,
150 psig, which can contain dust explosions.
explosion or health hazard. If operating under a
Explosion suppression is a technique which has been
negative pressure the system should be air tight to
in use for many years and is highly regarded. The shock
prevent pulling in air or other contaminants.
wave is detected by pressure transducers, located strate-
They should be strong enough to remain intact and
gically in and on the plant item where an explosion is
air tight under normal operating conditions, includ-
anticipated. The pressure wave is subsequently attacked
ing vibration; and, in some cases, to withstand or
by a very fast injection of a high concentration of Halon
contain explosive pressures.
or other suppressant. The vaporization of this Halon
Static electrical charges should be minimized by
breaks down the flame front and leaves behind an inert
grounding, including bonding across joints where
atmosphere, which prevents a secondary fire being
necessary. Eletrically isolated metallic objects within
started by any residual glowing particles within the
the system may accumulate dangerous static charges.
vessel. Plant vibrations and pressure .fluctuations may
Wire braid within rubber-covered transfer hose may
disturb this sort of system. It is also quite expensive,
act as a static accumulator.
especially when spurious trips occur and the container
Electrically conductive bags may be used in bag filters has to be refilled. A good discussion of explosion
handling extremely static-sensitive materials. Care must suppression systems is presented in NFPA6gY.
be taken to ensure that the conductive bags do not Another protection system has recently been
become ungrounded. developed which can detect glowing particles at temp-
eratures down to 400C (radiation detector operating in
0 An electrical installation must meet the electrical
the wavelength 1.5-3.0 cm). The detector activates a
classification imposed by the conveyed materials as
fire water injection system which extinguishes the glow-
well as the surrounding environment.
ing particles lo.
0 Suitable construction materials compatible with the
Several other good discussions of fire and explosion
materials handled and the surrounding environment
protection of pneumatic conveying systems and their
should be used.
components have been published3V-3.
0 Screens, magnets and metal detectors should be
installed for the detection or removal of any foreign
material which might create hazards in the system. Mechanical conveyors
a Where appropriate special materials should be used,
such as non-ferrous metals, to minimize mechanical Screw conveyors
sparking in the event of misalignment or failure of Screw conveyors have minimal free volumes so that dust
moving parts with the process stream. suspensions cannot form and thus dust explosions are
0 An adequate programme of maintenance and inspec- not usually a problem. However, screw conveyors may
tion must be instigated to ensure proper alignment of occasionally act as a source of ignition by generating
drives, proper clearances, dust tightness, electrical friction, which can lead to subsequent overheating and
bonding and grounding and control of ignition sour- ignition of the conveyed material. This hazard can be
ces. minimized by installing an overload trip on the motor
a Joints and openings should be suitably located to driving the screw.
facilitate cleaning or unplugging. High velocities
(3000 to 4000 fpm) will minimize settling and there- Belt conveyors
fore reduce frequency of cleaning. A belt conveyor system presents two principal fire
hazards: first, those of the belt itself, and second, those
Fire and explosion protective measures of the material being conveyed. Belts made of rubber or
All potentially dangerous pneumatic conveying systems synthetic products are combustible. Their combus-
which transport bulk solids should be designed to tibility and the extent to which the heat is released from
contain an explosion or provided with explosion vents. the belt can cause additional damage to housing struc-
Two alternative protective measures are: to use an inert tures covering the conveyors.
gas, usually nitrogen, to transport the solids; and to Some belts are less easily ignited than others, but they
provide explosion suppression systems. are still essentially combustible. Those made of poly-
The sizing and design of explosion vents is covered in vinyl chloride are one example. Those meeting the fire
detail in NFPA 68, and this document should be retardancy standards of the US Bureau of Mines or the
consulted when explosion venting is to be provided for British and Canadian equivalents are specific examples.
the conveying piping and/or downstream collection or Tests in the Factory Mutual material testing calorimeter
storage facility (cyclone, baghouse, silo). With respect indicate extremely high heat release rates once ignited.
to explosion containment, design of the conveying Several of the more promising types of fire retardant

72 J. Loss Prev. Process Ind., 7988, Vol I, April

 Safety considerations
belting, meeting the specifications of the US Bureau of
Mines, etc., were ignited in tests conducted at FM
Research Corporation and, once ignited, burned to
completion with high heat release rate. However, use of
these will reduce fire frequency because they are more
difficult to ignite.
Belt conveyors can also overheat either because of a
jammed idler roller or, if the belt jams, as a result of
drive rollers continuing to run. The belt can also
generate static electricity, and should therefore be of
anti-static material.
Belt conveyors used for dusty systems should be
enclosed. The free volume within the enclosure is likely
to be much greater than with a drag link conveyor and
the explosion hazard is thereby increased. The drag link
type should be used in preference. If a belt conveyor is
chosen, then full protection against explosion should be
provided in accordance with NFPA 68.
Factory Mutual recommends a number of protective
measures for belt conveyors4: Automatic sprinklers or
water spray protection should be installed for all com-
bustible conveyor belts. The belt drive should be inter-
locked to shut down on sprinkler water flow.
0 The system should be hydraulically designed for the
operation of ten automatic sprinklers and two small
hose streams (sprinkler spacing 100 ft per head).
The system should be designed for a flow pressure of
10 psi on the end sprinkler. (This will result in about
200 gpm for sprinklers and the same for small hose.)
In a conveyor enclosure less than 15 ft wide, this can
usually be accomplished by installing a single con-
tinuous supply line with an unlimited number of
sprinklers.
0 Water supplies should be adequate for one hour of
use.
l High initial water pressures should be used to over-
come friction and elevation loss. A constant dis-
charge pressure system would be one method of
providing such initial high pressure.
A small hose or equivalent should be provided at
suitable intervals with sufficient FM-approved li in hose
at all locations to reach any part of the conveyor system.
Small hose connections can be made to the sprinkler
piping. A suitable alternative to this is the provision of
hydrants and easy access to the conveyor by fire fighting
equipment. All weeds, brush and trees should be cleared
from underneath and at least 25 ft from both sides of
conveyor supports. Unprotected combustible buildings
and similar exposures should be removed. Each con-
veyor belt system should be provided with tamperproof
devices arranged to automatically shut off driving
power in the event of greater than 20% belt slowdown
or misalignment of belts. In addition, interlocking
devices should be arranged to shut off power on all
contributing conveyors. Where conveyor belts are
critical to plant operations, spare belts in quantity
consistent with expected fire damage (with recom-
mended protection in service) should be kept on hand.
in conveying bulk solids and powders: S. Grosset
Some damage will occur even with recommended pro-
tection in service.
Bucket elevators
Bucket elevators are often subject to explosions and
fires, and numerous examples have occurred in the
past, particularly with vegetable dusts, for which these
elevators have customarily been used3. The design of
these elevators lead to dust clouds being continuously
present during working, particularly at the head and the
boot of the elevator. The buckets are also regularly
subject to impact and the belt supporting the buckets
can slip on the pulleys and generate frictional heat. As
a result, a source of ignition and a dust suspension can
be present simultaneously, causing explosion or fire.
Modern high capacity elevators, with separate delivery
and return legs, have a reduced risk because of the
reduced volume per unit weight of dust conveyed. In the
general case, with explosible dusts, other types of
elevator are preferable, and are particularly necessary
for dusts of more severe explosibility, e.g., those giving
maximum rates of pressure rise in excess of
5000 lb in- s (35000 kN m-s) in small scale test appa-
ratus. Use of elevators should also be avoided for
dusts known to be readily ignited by friction, e.g.,
sulphur.
Where the use of bucket elevators is unavoidable,
their positioning should be carefully considered and
regular maintenance is essential. The elevators should
be mounted outside the building, e.g., supported by an
outside wall, and the intake and delivery points should
preferably be isolated from the rest of the dust handling
plant by means of chokes. The elevator casing should be
a fire resistant construction, sufficient to retain a fire,
dust-tight and of sufficient strength not to rupture in the
event of an explosion. To meet the strength require-
ments, the casing should be provided either with auto-
matic suppression or with explosion relief at the head
and the boot, with vent areas calculated per NFPA 68.
Long elevators, say more than 6 m (20 ft), may require
additional relief at intervals along the casing to ensure
that no point is too remote from a vent.
Particular care should be taken to ensure that flame
burning dust, etc., discharged from the vents during an
explosion cannot injure operators or damage nearby
plant. Provision of ducting or deflectors over the vents
may be required. Where it is unavoidable to site a
bucket elevator inside a building it is desirable for the
internal pressure to be slightly below atmospheric to
minimize leakage of dust. Discharge of combustion
products from vents to the outside of the building is
essential, and the design requirements of ducting from
vents should be in accordance with NFPA 68. The need
for dust tightness and adequate casing strength and fire
resistance should be rigorously met.
Steps should also be taken when designing bucket
elevators to minimize the generation of ignition sources.
These steps may include the provision of strong fixing
for the buckets and strong bearings for all shafts,
external to the casing, provided with detectors for
J. Loss Prev. Process lnd., 1988, Vol I, April 73
Safety considerations in conveying bulk solids and powders: S. Grossel

overheating. The main drive to the elevator should be research should shed new knowledge on how to vent
external to the casing. Belt slip within the casing can be duct-like volumes containing obstructions and shows
detected by belt speed meters, and anit-runback devices that supression systems can be effectively used to stop
can also be installed. Development of friction within the explosions.
casing can thereby be reduced.
Exhaust ventilation can be applied to the casing of En Masse Conveyors
bucket elevators which assists the removal of suspension The explosion hazard situation in en masse conveyors is
of dust in air. These suspensions contain the dust similar to that in bucket elevators but more complex
fractions of smaller particle size, and ventilation would especially when operated vertically. In the elevating leg
also give a slight negative pressure relative to atmos- the channel will be full of material, probably above the
phere, within the casing. The ventilation system should upper explosive limit, but during start-up, or when the
be conducted to a dust collection unit in a safe area, inlet is starved, an explosible concentration could occur.
and should be provided with explosion protection on a It is impracticable to vent the elevating leg so this should
similar basis to that in the elevator casing itself. Care be strengthened and vented at the top.
should be taken to ensure that the air flow is sufficient to Exposion vents should be provided on the empty
prevent deposition of dust in the ducting of the collec- return leg in the same way as for bucket elevators.
tion unit, and also sufficient to permit the collector to
function efficiently. Dust collection systems are con-
sidered in more detail in the literature3*. References
Automatic explosion suppression can be used to
protect bucket elevators when explosion relief is not I Cross, J. and Farrer, D. Dust Explosions, Plenum Press, New
York. 1982
practicable, often because of the situation of the el- 2 Haase, H., Electrostatic Hazards: Their Evaluation and Control,
evator being located within a building. The design of the Verlag Chemie, Weinheim, West Germany and New York, 1977,
suppression system should be done by a qualified English translation by M. Wald
3 Palmer, K. N., Dust Explosions and Fires, Chapman and Hall,
vendor. London. UK. 1973
Gillis and Fishlock conducted experimental investiga- 4 Bartknecht, W., Explosions-Course. Prevention, Protection.
tions on bucket elevator explosions to determine ways to Springer Verlag, Berlin, Germany, and New York, 1981, English
translation
limit their effects, or to extinguish the explosion5. The 5 Field, P., Dust Explosion, Handbook of Powder Technology,
purpose of these tests was twofold: first, to obtain data Volume 4, Elsevier, Amsterdam, The Netherlands, 1982
and information on how bucket elevators may be vented 6 Nagy, J. and Verakis, H. C., Development and Control of Dust
Explosions, Marcel Dekker, Inc., New York, 1983
to eliminate physical damage and to minimize explosion 7 Schofield, C., Guide to Dust Explosion Prevention and Protec-
propagation; and second, to demonstrate the effect- tion, Part 1 - Venting, IChemE. Rugby, UK, 1984
iveness of an explosion suppression system in detecting 8 NFPA 68 Guide for Venting of Deflagrations, National Fire
Protection Association, Quincy, MA, USA, 1988
and extinguishing a dust explosion within the bucket 9 NFPA 69, Standard on Explosion Prevention Systems, National
elevator. Fire Protection Association. Ouincv. MA. USA. 1986
Halon suppression systems were used successfully to IO Forsyth, V. G., Dust Explosi& Prc%ection in Pneumatic Convey-
ing Processes, Fire Prevention, No. 135, March 1980, pp. 25-30
isolate flames from propagating in the head and boot II Bennett, N., Explosion Protection for Fabric Dust Collectors,
section during explosion vent tests. A specially designed Specifying Engineer, September 1982, pp. 81-83
flame retarding distributor, supplied by Union Iron 12 soo. s. L.. J. PiDPlimY 1981. I. 57
13 NFPA 650. Pneimatic Conveying Systems for Handling Com-
Works, was tested and found effective in limiting flame bustible Materials, National Fire Protection Association, Quincy,
propagation to a single path when all but one outlet was MA, USA, 1984
sealed. Various types of optical detectors were tested as 14 Factory Mutual Loss Prevention Data Sheet 7-1 I, Fire Protection
for Belt Conveyors, August 1972
means of detecting the initiation of an explosion and 15 Gillis, J. P. and Fishlock, F. H., J. Powder ond Bulk Solids
compared with a pressure detector. Results from this Technology 1982, 6(2), 5

74 J. Loss Prev. Process Ind., 7988, Vol 1, April

 Full Ripers

Maize starch explosions in a 236 m3

experimental silo with vents in the silo wall
Rolf K. Eckhoff, Franz Alfert, Kjell Fuhre and Geir H. Pedersen
Chr. Michelsen Institute, N-5036 Fantoft, Bergen, Norway

Large-scale wall venting experiments were conducted in the same steel silo of 22 m height, 3.7 m
diameter and 236 m3 volume used in previously published roof venting experiments. Maize starch
test dust (125 kg), was blown into the silo at the top through a conventional pneumatic pipeline.
The dust injection process lasted for 28 s, whereafter the air flow in the pipeline was terminated and
the ignition source activated. The dust concentration in the silo was measured during dust injection.
Pressure as a function of time and flame speed at different locations in the silo were recorded during
each explosion experiment. In the case of an uncovered vent of 4.4 m2 in the cylindrical silo wall
close to the silo top, the maximum explosion pressures in the silo were generally considerably lower
than with a 3.6 m* roof vent, and even somewhat lower than with a roof vent of 5.7 m2. Lower
turbulence, and hence combustion rate in the dust cloud in the silo, due to more restricted flow out
of the vent, may be the reason for this. On the condition that adequate precautions are taken to
prevent destructive effects of reaction forces, venting through the silo wall may therefore be
preferential to roof venting even from the point of view of minimizing the explosion pressure.

(Keywords: explosions: silo; vents)

A comprehensive series of vented dust explosion experi- of the 236 m3 silo is shown in Figure 2. The silo is 22 m
ments in the new 236 m3 steel silo in Norway were in height and has a diameter of 3.7 m. The main silo
conducted in 1985 . The vent openings were located in body is made of 8 mm thick steel plate, all-welded, and
the silo roof. An abbreviated account was published can withstand at least 5 bar internal overpressure. It is
later2* 3. The 1985 experiments showed that the welded to a horizontal steel ring flush with the ground
maximum explosion pressure was influenced strongly by level, which is fixed to the solid rock underneath by 15
the location of the ignition point within the silo, and for bolts. Sufficient strength for withstanding strong wind
this reason the main emphasis was put on quantifying forces has been provided. A winding staircase is pro-
this effect, which is of great practical significance. vided alongside the silo. The staircase allows easy access
However, it was also of considerable practical interest to any desired level. Every 1.5 m, the silo wall is
to investigate the feasibility of venting dust explosions furnished with a 3 threaded hole, accessible from the
in silos through vents in the cylindrical silo wall, rather staircase, and permitting ignition sources and various
than in the roof. This is because the entire roof of silos types of diagnostic probes to be mounted in desired
with external walls will then be available for other uses. locations.
The main objective of the present experimental pro- A strong steel grid has been constructed across the top
gramme, conducted during the first half of 1986 and of the silo, allowing any desired part of the roof to be
reported in detail elsewhere4, was to investigate the blocked by bolting a number of 0.25 m2 steel plates to
feasibility of venting dust explosions in large silos the grid. Thus, varying vent opening areas and shapes in
through an opening in the cylindrical silo wall close to the roof is easy. The size and shape of the double vent
the silo top, as compared with venting through the silo used in comparative roof venting experiments in 1986, is
roof. However, the strong lateral reaction forces caused shown in Figure 3, together with the 3.4 m roof vent
by the venting of a violent dust explosion through an used in 1985 . The free area of each grid square is
opening in the wall of a slender silo, close to the top, 0.23 m2.
may cause the silo to overturn or break. Therefore the Figure 4 shows a section through the silo at the level
reaction force problem had to be given serious con- of the vent system. The total vent area of 4.4 m2 is
sideration, and adequate precautions taken. distributed on eight identical slots of height 1.23 m and
width 0.45 m, separated by vertical steel beams of
width = 10 cm. A fenced gangway accessible from the
The 236 m3 experimental silo
silo roof by a ladder was constructed alongside the
Figure I shows a sketch of the experimental site located entire vent arrangement, as also shown in Figures 2
on the island of Sotra, west of Bergen. A vertical section and 4.
The reaction forces generated by vented dust and gas
Received 24 Augusr 1987 explosions have been investigated both theoretically and
0950-4230/88/010016-09 $3.00
0 1988 Butterworth & Co. (Publishers) Ltd
16
 Maize starch explosions in a 236 m3 experimental silo with vents in the silo wall: R. K. Eckhoff et al.

Figure 1 Sketch of experimental site with 236 m3 silo equipped with vent in the silo well

. experimentally by several workers5-. The equation:

Pressure probe P3_ Eight rectangular F R.max = O.J2APm,,
, vent openings.
Dust injection point -
Tot=
covers most published data. FR,,,~~ is the maximum
is 4.4 Y
m area
Highest ignition point - reaction in MN; A is the vent area in m2; and P,,,,, is the
maximum explosion pressure in bar (g). On the basis of
Dust concentration the results of the previous roof venting experimentslm3,
probe c3 it was assumed that the maximum explosion pressure
would not exceed 1.5 bar(g), corresponding to a
maximum lateral reaction force at the silo top of about
0.7 MN. Based on these considerations, a system of
three pre-stressed wire ropes, anchoring the silo to the
nearby rock, was adopted, as indicated in Figure 4.
Dust concentration Wire ropes (20 mm thick) were found to serve the
probe c2 purpose. One end of each rope was fixed to specially
made attachment eyes welded to the top rim of the silo,
whereas the other was anchored to eye bolts molded into
Pressure pmbe P2 - the nearby solid rock. Each of the two main wire ropes
were pre-stressed to 0.02 MN (2 tonnes) and the third
rope on the opposite side of the silo to 0.04 MN (4
Dust concentration tonnes).
pm& Ct

Ignition, diagnostic and timing system

The ignition source was the same as that used in the
Pressure probe P , -
previous experiments. It consisted of about 50 g of
0 1 2 3 dried nitrocellulose powder contained in a plastic bag and
Lowest ignition point - L I I ,
meta fired by a pair of electrically ignited 100 J Ce-Mg fuse
heads. The nitrocellulose flame was fully developed to
50- 100 cm diameter approximately 0.1 s after firing the
Figure 2 Vertical section of 236 m3 silo fuse heads, and it maintained its full size for approxi-

17
Maize starch explosions in a 236 m3 experimental silo with vents in the silo wall: R. K. Eckhoff et al.

Wire rope
t

Net vent area : 3.6 m* (1986) Figure 4 Section of silo through the eight vent slots, each of
height 123 cm and width 45 cm, and gangway providing access
to the vents

the dust concentration development at three different

points in the silo during some of the experiments.
Details of the probe design and performance are pro-
vided in Ref. 1. A set of flame front detectors, con-
sisting essentially of the receiver part of the dust concen-
tration probes, were constructed and mounted at
various altitudes along the silo wall. Finally, 25 frames
per second video recordings w&e made of every
explosion experiment.
A new, integrated data acquisition and control system
was used. The system includes a 13 bit high-speed
voltmeter (100 kHz), a 24-channel high-speed FET
multiplexer with thermocouple compensation, and a
16channel general purpose switch. All timing and
measuring operations were automatically controlled by
a computer combined with this system. More detailed
information about the system is given elsewhere.

The maize starch used in the experiments

Net vent area : 3.4 m* (1985) The native maize starch used was of the same type as
used in 1985 . Samples were taken from a number of
Figure 3 The 3.6 m2 roof vent arrangement used in the present
experiments and the 3.4 m2 roof vent used in 1985 (Ref. 1)
the bags immediately before emptying the bags into the
small storage hopper, from which the dust was dischar-
ged into the dust injection system. The samples were
mately 1 .O s. The total energy liberated by the ignition kept in sealed plastic jars and brought to the CM1 lab at
source was about 200 kJ. Fantoft for determination of moisture content and
Figure 2 shows the locations of the various diag- explosiveness. The moisture content in relation to wet
nostics. Three Kistler No. 7261 pressure sensors, one dust was determined for samples from nine different
located 3 m above the silo bottom, one 9 m above the 25 kg bags of starch, and it varied between 10.7 and
bottom, and one in the silo roof, were used for measur- 11.9 mass 070,with a mean value of 11.2%. This is close
ing the development with time of the explosion pressure to the mean value of 11.1% found in corresponding
inside the silo. Three of CMIs simplified version of the samples in 1985 .
USBM light attenuation probe were used for monitoring Previous particle size analyses using an Alpine air-jet

18
 Maize starch explosions in a 236 m3 experimental silo with vents in the silo wall: R. K. Edhoff et al.

Roots blower. With no starch in the pipeline, the

Table 1 Results in Hartmann bomb tests of maize starch
upstream static pressure was very close to atmospheric
Tested by Year of test P max (dP/dtl,., and the mean linear air velocity in the pipeline was
Ibarlg)) (bar(s)) 38.3 m s- (Reynolds number = 300 000). During dust
injection, the static pressure would increase to typically
CMI 1982 6.3 282
CM1 1985 5.2 220 0.2 bar (g) with peak values approaching 0.5 bar (g),
CMI 1986 7.6 180 and the air flow would drop by up to 15-20070.
The maize starch was fed into the pipeline from a
1.5 m3 hopper a few metres downstream of the blower
by means of a rotary lock. The steady-state feeding rate
was about 5 kgs-, which corresponds to an average
Table 2 Results of 20 litre-sphere test of maize starch dust concentration in the pipeline of about 8 kgme3.
Tested by Year of test P nlax (dP/drl,.x
When running an experiment, the blower was first
(barlg)l (bar(s)) started and air was blown through the system in order to
FRS 1982 7.3 425
dry up and clean the line. Then the air supply to the line
CMI 1985 7.2 363 was shut off to allow the desired quantity of maize
Ciba-Geigy 1986 9.5 542 starch to be charged into the 1.5 m3 hopper at ambient
CMI 1986 7.3 370
pressure. The hopper was subsequently sealed, and the
air from the blower was again allowed to flow freely
through the pipeline. Then the rotary lock was started
and run for the time required to just feed all the dust in
sieve showed that nearly 80% by weight fell in the size
the hopper into the pipeline, plus a few seconds to
range 10 to 20 ,um. This is confirmed by SEM-pictures of partially clear the line. Then the air flow was again shut
the starch and conforms with the fact that maize starch
off to allow the turbulence in the dust cloud to dissipate
is essentially composed of nearly monosized and nearly
for a few seconds before the ignition source was acti-
spherical native grains of mean diameter about 12 Bm.
vated _
Table 2 gives the mean values of Hartmann bomb
tests of native maize starch carried out at CM1 in 1982,
1985 and 1986. Table 2 gives the mean results of 20 1 Results from vented maize starch explosions
sphere tests carried out at CM1 in 1985 and 1986, at FRS in the 236 m3 silo
in 1982 and at Ciba-Geigy AG in 1985. The higher P,,,
obtained in the Hartmann bomb test in 1986 as com-
pared with the results of earlier tests can be explained Control experiments using vents in the silo roof
when looking more closely at the test results from The previous investigation was concerned with roof
1985 I. In the earlier tests the highest dust concentration venting only. The overall objective of the present
used was 1000gmm3, whereas the 1986 tests revealed investigation was to study the influence of moving the
that the highest explosion pressures in the Hartmann vent from the silo roof to the wall near the silo top.
bomb are in fact generated at 1500 g rne3. To establish a sound basis for comparison, a few
The 20 1 sphere results show that Ciba-Geigy found experiments with roof venting were carried out even in
considerably higher values for P,,, and (dP/dt),,, than the present programme. The majority of the experi-
the two other laboratories. This is likely to be due to a ments were conducted with 125 kg maize starch, a dust
change in moisture content of the particular sample of injection period of 28 s, and a 2 s delay between shut-off
starch tested by Ciba-Geigy. The samples tested at CM1 of air flow and ignition. If the entire quantity of 125 kg
had mean values of 11.1% (1985) and 11.2% (1986), of starch had been dispersed evenly throughout the
and the samples tested at FRS about 10%. However, 236 m 3 silo volume, the average theoretical dust concen-
Ciba-Geigy found the moisture content of the sample tration in the silo would have been 530 g rne3. In reality
tested by them to be only about 6%. Therefore, the the average dust concentration was somewhat lower,
(dP/dt)max value to be related to the present silo experi- because some starch remained in the injection pipe after
ments is the one measured at CMI, corresponding to a termination of the 28 s injection period, and some had
&-value (W. German VDI 3673-system) of about settled on the silo floor. Furthermore, the dust concen-
100 bar ms-. tration distribution was not uniform. However, this
procedure of dust cloud generation and ignition seemed
to give the most violently exploding dust clouds that
Dust injection system
could be produced by the present pneumatic injection
As indicated in Figure 2, explosive clouds of maize system.
starch were generated in the silo by blowing the starch The development of the dust concentration in the silo
into it through a conventional pneumatic conveying line during dust injection was followed in some of the
of i.d. 155 mm. The starch entered the silo asa horizon- experiments by means of three calibrated light attenu-
tal jet from the semi-tangential pipe exit close to the silo ation probes. The probes were located as indicated in
top. The air for conveying the dust was supplied by a Figure 2 with the measurement heads approximately

19
Maize starch exdosions in a 236 m3 experimental silo with vents in the silo wall: R. K. Eckhoff et al.

roof venting experiments, which were all conducted at

the worst-case conditions of the dust cloud generation
system used, have been plotted in Figure 6 as a function
of the location of the ignition source above the silo
bottom. The P,,,=, values given are those measured by
the lowest pressure transducer (PI in Figure 2). The
maximum pressure measured by the intermediate trans-
ducer P2 was generally only slightly lower than that
measured by the PI, whereas transducer 4 in the silo
roof generally gave lower values than the two others.
Figure 6 also gives the maximum pressures found in
the previous experiments using a 3.4 m2 vent in the silo
roof and worst-case dust clouds. It is seen that the
results from the present investigation using the 3.6 m2
roof vent arrangement (Figure 3) are essentially
identical with the results from the previous programme
Figure 5 Local dust concentration in 236 m3 silo close to silo using the 3.4 m2 roof vent arrangement (Figure3). This
axis during dust injection, at 6 m and 12 m above silo bottom
is assumed to imply that the concentration distributions
of the experimental dust clouds used in 1985 and 1986
1 m from the silo wall. Details of the probes and their were essentially the same. It then follows that the results
calibration are given elsewhere. Figure 5 shows an from using a 5.7 m2 roof vent in the previous program-
example of the dust concentration as a function of time, me are also directly comparable with these results.
measured at 6 m and 12 m above the silo bottom
respectively. As can be seen, the concentration at the Experiments using 4.4 m 2 vent area in the silo wall close
moment of ignition (30 s) was about 400 g rnT3 at 6 m to the silo top
and slightly lower, about 350 g m - 3 at 12 m. The results of the 16 main experiments using 4.4 m2
The maximum explosion pressures generated in the vent area in the silo wall close to the silo top are plotted

1300

1200

1100

loo0

900

600

500

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Location of ignition sourca above silo bottom (m)

Figure 6 Maximum explosion pressures in vented explosions in the 236 m3 silo. 0, 3.4 m2 (1985) roof vent; 0, 3.6 m2 I1 986) roof vent;
hatched area, envelope embracing all 3.4 rn and 3.6 m* roof venting results

20
 Maize starch explosions in a 236 m3 experimental silo with vents in the silo wail: R. K. Eckhoff et al.

lzzl 3.4 m2 and 3.6 m2 roof vent

6.7 m* roof vent

ISI

I 4.4 m* wall vent

700

606

500

400

300

200

100

I I I 1 I I I I I I I I

0 1 2 3 4 5 6 7 8 9 10 11 12
_ 13_ 14 15
- lfi
. 17 1R 163 20

Location of ignition source above silo bottom (m)

Figure 7 Maximum explosion pressures in the 236 ms silo with a 4.4 m* vent in the silo wall. Comparison with envelopes embracing all
experiments with 3.4, 3.6 m2 and 5.7 m* vents in the silo roof

in Figure 7. In all the experiments the dust clouds were

generated in the same way as in the experiments with 0.040

roof vent, i.e. 125 kg of maize starch was injected at the

silo top over a period of 28 s and the ignition source was
activated 2 s after termination of dust injection. Figure
s 0.020

7 also shows the envelope from Figure 6 and the 5

corresponding envelope from the previous 5.7 m2 roof
vent experiments. It seems clear that the 4.4 m2 wall
t% O
venting pressures are generally considerably lower than
-0.020
those from the 3.4-3.6 m2 roof venting experiments, I- I-.
and even somewhat lower than those generated with the In 3 n 1 E n 11 1. * 11 1

larger 5.7 m2 roof vent. The pressure versus time histor-
ies generated by the explosions vented through the silo
wall were, apart from a general reduction of the peak
pressures, similar to those found with roof vents.
Figure 8 shows a typical pressure development in an
experiment with ignition at 13.5 m above the silo
bottom. The standing-wave oscillations of frequencies
3-6 Hz (depending on gas temperature) are very similar
to those observed in roof venting experiments with
ignition at 13.5 rnlm3. The frequency corresponds to the
first harmonic of a standing wave in a one-end-open
pipe of length 22 m.
Figure 9 shows a comparison of two pressure versus
time histories generated in the wall vented silo in two
apparently identical experiments, with ignition at 4.5 m
above the silo bottom. The upper trace exhibits a
distinct spiky, oscillatory pattern after the first main
0 1 2 3
Time (s)
Figure 8 Pressure ~ersos time history of explosion in the 236 ms
silo with a 4.4 m2 vent in the silo wall. Ignition 13.5 m above silo
bottom. Pressure transducer located 3 m above silo bottom
pressure rise. This corresponds closely to the 17-18 Hz
oscillations observed previously with ignition at 4.5 m
above the silo bottom using a roof vent. The exact
nature of these oscillations has still not been resolved.
However, as indicated previously2S 3, one possible
explanation could be generation of some kind of
Helmholtz oscillations across the vent opening during a
short phase of the explosion, when the dust cloud
outside the vent explodes. This will occur just after the
flame front inside the silo has reached the vent opening,
i.e. when the total pressure pulse reaches its peak.

21
 Maize starch explosions in a 236 m3 experimental silo with vents in the silo wall: R. U. Eckhoff et al.

outside the silo, with the firing of the ignition source

0.50 inside. As in the previous experiments-, a close
correlation was found between the moment of arrival of
= 0.40 the flame at the vent and the moment of the main pres-
c
$ 0.30
sure peak. The pressure peak generally occurred slightly
P later than the flame arrival, by about 0.05-o. 1 s.
I
E 0.20 An average axial flame front speed was estimated by
dividing the distance from the ignition point to the vent
b 0.10
by the time from ignition till the first appearance of
0 flame at the vent (video). In the case of ignition at the
bottom of the silo, explosions with the 3.6 m2 vent in
0.50 0.80 1.10 the silo roof gave about 26 ms-. whereas with the
Time (s)

0.40

0.30 -
z
2
jJ 0.20 -

j 0.10 -

o;--- ----- -----

-
._.-

t I I I I I I I 1 1 I I I I L
b.40 0.60 0.80 1.00 1.2

Time 0)

Figure 9 Comparison of different pressure versus time histories

from two explosion experiments in the 236 m3 silo, conducted at
apparently identical conditions, with 4.4 mz vent area in the silo
wall and ignition 4.5 m above silo bottom: -, PI;-.-., P2;----, P3

The pressure versus time history in the upper half of

Figure 9 shows that the main pressure peak was reached
at 1.04 s after ignition, whereas the video recordings
showed that the first sign of flame reached the vent
opening some time between 0.92 and 0.96 s after ig-
nition. During the short duration of the unconfined
explosion just outside the vent the over-pressure in this
region may be close to the pressure inside the silo, and
Helmholtz oscillations may take place. The fact that the
oscillations only occur occasionally, as illustrated by the
smooth, bell-shaped pressure versus time history in the
lower half of Figure 9, suggests that rather unstable con-
ditions must be fulfilled for the oscillations to appear.
The trace in the upper half of Figure 9 is from
transducer PI (Figure 2), whereas the traces from all
three pressure transducers are shown in the lower half.
As can be seen, PI and P2 gave almost identical results,
whereas P3, located in the silo roof, gave a slightly
lower peak pressure. This pattern was observed consis-
tently, using the 4.4 m2 vent in the silo wall.

Flame front propagation in silo during vented

explosions @
The arrival times of the flame front at the vent opening
at the silo top were determined from the video record- ClOSd Vented
ings of the explosions. The moment of ignition of the Exaggerated schematic illustration of flame front pro-
Figure 10
dust cloud inside the silo was made visible on the video pagation during dust explosions in elongated vessels with and
by synchronizing the firing of a flash lamp mounted without venting

22
 Maize starch explosions in a 236 m3 experimental silo with vents in the silo wall: R. K. Eckhoff et al.

4.4 m2 wall vent they gave significantly lower values, in of the burnt gas, which in its turn increases yet further
the range 19-21 ms-. the flow rate of the unburnt gas ahead of the flame and
The vertical propagation of the flame front along the so on. In other words, there is a self-accelerating
silo wall was measured by means of a number of explosion process, whose key mechanism is the gener-
infrared radiation detectors mounted 10 cm from the ation of flow-induced turbulence in the gas ahead of the
silo wall. The detectors, developed at CMI, are flame.
described elsewhere. In some experiments the flame Extensive experimentation in several countries with
arrival was detected simultaneously at different flame coal dust explosions in long pipes, ducts and large-scale
probes up to several metres apart. Figure 10 gives an galleries has confirmed that this rather multifarious
exaggerated illustration of a possible reason for this picture indeed also applies to dust explosions. For this
observation. If ignition takes place at the silo bottom reason venting of dust explosions in elongated enclos-
and the explosion is vented at the silo top, the strong ures of L/D > 4 is a very complex process. The tur-
axial flow towards the vent opening will cause a marked bulence generated ahead of the flame due to expansion-
axial stretch of the flame. Therefore, as the flame front induced flow, has a strong enhancing influence on the
approaches the vessel wall, it will become nearly parallel combustion process. This is why the location of the
with the wall over quite long distances. ignition point with respect to the vent, is of such crucial
It is of interest to compare the average axial flame importance for the pressure build-up during dust
speeds estimated from the video recordings with those explosions in the 236 m3 silo. If ignition occurs in the
found in comparable roof venting experiments (ignition upper part of the silo, close to the vent, there is no
at silo bottom, worst-case dust concentration) con- possibility of generating high-velocity flow and resulting
ducted in 1985 . With 3.4 mz vent area average flame high turbulence levels in the unburnt cloud further
speeds of 25-30 ms- were measured, in complete down.
agreement with the value of 26 ms- for the 3.6 m2 It seems that this overall conception can in fact offer
roof vent in the present investigation. Further, the a qualitative explanation of why in the present experi-
previous experiments showed that a larger roof vent, of ments in the 236 m3 silo the wall vent gave considerably
5.7 m2, gave higher average flame speeds, in the range lower explosion pressures silo than the roof vent of
of 30-35 ms-. This is most probably due to a larger similar size. Venting through the silo wall forces the
vent allowing a higher flow rate at a given pressure drop dust cloud to change flow direction by about 90 degrees
than a smaller vent. Following this line of thought it to pass through the vent. This necessarily increases the
may at the first glance seem somewhat strange that a pressure drop across the vent, i.e. the resistance to
4.4 m2 vent gave a significantly lower average flame vertical flow in the silo. Had the rate of combustion of
speed, about 20 m s- , than the 25-30 m s- found with the dust cloud been independent of the flow-induced
the smaller roof vent of 3.6-3.4 m2. turbulence, the reduced flow rate through the wall vent
would clearly have resulted in higher maximum
explosion pressures than venting through a roof vent of
similar size. However, as discussed above, the measured
Why does a vent in the silo wall give less
average axial flame front speed of 20 m s- for a 4.4 m2
violent explosions than a roof vent of the same
vent in the silo wall, is considerably lower than the speed
size?
of about 30ms- estimated for a roof vent of the same
Earlier this century, gas explosion experiments in long size. This would imply that the overall combustion rate
pipes have shown that as the explosion propagated was higher with the roof vent and probably more than
along the pipe, it could undergo a dramatic change from compensated for the initially lower pressure drop across
harmless laminar flames via a wide spectrum of tur- the vent. The net result would be a higher maximum
bulent explosions, to detonation. It was also observed explosion pressure with the roof vent.
that this change of explosion violence was greatest when As discussed elsewhere, computer simulation of the
the pipe was closed at one end and open at the other, propagation of dust explosions is the probable long-
and ignition took place at the closed end. term solution for quantifying the influence of flow-
The secret of the dramatic flame acceleration has also induced turbulence on the combustion rate. However,
been resolved on a qualitative basis. After ignition, the developing fully comprehensive computer codes for
first combustion phase occurs rather slowly as an dust explosions is a very demanding task and will take
approximately laminar flame spread. However, as a time. An intermediate solution for the time being could
result of the heat produced by the combustion, the be to use the advanced computer codes for gas explosion
combustion products expand, and this leads to the still simulations that have recently been developed at CMI,
unburnt gas ahead of the flame being pushed towards taking the dust cloud as being equivalent to some
the open end of the pipe. Depending on pipe diameter, explosive gas mixture. Such simulations would most
flow rate, and the roughness of the pipe walls, the probably indicate whether the above suggestion of why
flowing unburnt gas will be rendered more or less wall venting gives lower maximum explosion pressures
turbulent. When the combustion zone approaches these than roof venting, is valid. This type of simulation
turbulent regions further along the pipe, the combustion seems to offer a promising possibility of studying the
rate will increase. This leads in turn to faster expansion influence of geometry on gas and dust explosions in a

23
Maize starch explosions in a 236 m3 experimental silo with vents in the silo wall: R. K. Eckhoff et al.

variety of different configurations, such as head houses and Mr Norman K. Larson, Chairman of the NGFA
and bottom tunnels in grain storage facilities. Explosion Suppression and Venting Committee, should
be especially mentioned. Sincere thanks are also due to
Mr James D. Mills, former research assistant at CMI,
Conclusions for skilful and enthusiastic participation in the experi-
In the case of a 4.4 mz vent in the silo wall, the ments.
maximum explosion pressures generated in the 236 m3
silo were generally significantly lower than with a roof References
vent of similar size. Ignition just above the silo floor still
gave quite high pressures of 0.7-0.8 bar(g), but a roof Eckhoff, R. K., Fuhre. K. and Pedersen, G. H., Vented Maize
Starch Explosions in a 236 m3 Experimental Silo. Fire and
vent of similar size would give about 1.0 bar(g). With Explosion Research Report No. ESV-86-070. prepared for NGFA
ignition at 7.5 m above the silo floor, the maximum by CMI. Also obtainable from CM1 as CMI-Report No. 843307-2,
pressures were 0.2-0.3 bar(g), as opposed to 1985
2 Eckhoff, R. K., Fuhre, K. and Pedersen, G. H.. Dust Explosion
0.4-0.6 bar(g) estimated for the vent in the roof. Ig- Experiments in a vented 236 m3 Silo Cell. 5th Int. Symp. on Loss
nition at 18 m, i.e. close to the silo top, gave very low Prevention and Safety Promotion in the Process Industries,
pressures, of the order of 0.01 bar(g). The probable SocietC de Chimie Industrielle Cannes, Sept. 15-19, 1986
3 Eckhoff, R. K., Fuhre. K. and Pedersen, G. H., Dust Explosion
reason for this somewhat unexpected effect-of locating Experiments in a Vented 236 m Silo Cell. J. Oct. Accidents 1987,
the vent in the wall, is that the higher pressure drop submitted for publication
caused by a wall vent, compared with a roof vent of 4 Eckhoff, R. K., Alfert. F.. Fuhre. K.. Mills. J. D. and Pedersen.
G. H., Maize Starch Explosions fin a 236 A3 Experimental Silo
similar area, restricts the vertical flow of unburnt dust with Vents in the Silo Wall. Fire and Explosion Research Report
cloud in the silo during the explosion. This in turn No. ESV-87-071, prepared for NGFA by CMI. Also available
restricts the build-up of flow-induced turbulence, and from CM1 as CMI-Report No. 863307-l. 1986
Hattwig, M., Investigations on the Recoil of Vented Vessels.
limits the combustion rate of the dust cloud and the Paper given at 3rd Int. Symp. on Loss Prevention and Safety
resulting rate of pressure build-up in the silo. Promotion in the Process Industries, Basle, Sept., 1980, Preprint
The strong influence of the location of the ignition collection p. 968
Faber, M., Reaction Forces in Explosion Pressure Relief. Paper
point on the maximum pressure in the wall-vented presented at the Symposium on the Control and Prevention of
236 m3 silo explosions is in agreement with the results of Dust Explosions, Basle. Nov. 16 and 17, organized by Scientific
previous roof-venting experiments. This re-emphasizes and Technical Studies/OYES-IBC, 1982
Hattwig, M. and Faber, M.. Riickstosskrlfte bei der Explosion-
the importance of identifying likely ignition source sentlastung. VDI-Berichte Nr. 494, 1984, pp. 219-226.
locations in silos in practice. Alfert, F., Eckhoff, R. K., Fuhre, K., Pedersen, G. H. and Mills,
J., Initiation of Dust Explosions in Silo Cells by Smouldering
Nests. CMI-Report No. 863380-1, December, 1986
Eckhoff. R. K.. Fuhre, K., Guirao, C. M. and Lee. J. H. S.,
Acknowledgements Venting of Turbulent Gas Explosions in a 50 m Chamber. Fire
sapy Journal 1984, 7, 191-197
CM1 thanks the National Grain and Feed Association, Eckhoff. R. K., Measurement of Explosion violence of Dust
Washington DC for the substantial financial support Clouds. Proceedings of International Symposium on the
Explosion Hazard Classification of Vapours, Gases and Dusts.
rendering this project possible, and for unfailing con- Nat. Materials Advisory Board, Washington DC, July 15-18,
structive advice and support throughout. The keen 1986. Publication NMAB-447, National Academy Press, 1987
personal interest and support of Mr Mark G. Goedde, Hjertager, B. H., Three-Dimensional Modelling of Flow, Heat
Transfer and Combustion in the Handbook of Heat and Mass
previous Director of Engineering Services, NGFA, Transfer, Gulf Publishing Company, Houston, 1986, Chapter 14,
Washington DC, his successor Mr Kerry J. Goforth, pp. 1303-1350

24
Fire Safety Journal, 12 (1987) 97 - 102
Measurements of the Explosibility of Peat Dust
HENRY WECKMAN
Technical Research Centre of Finland, Fire Technology Laboratory, SF-02150 Espoo (Finland)
(Received December 11, 1986; in revised form January 26, 1987)
SUMMARY
Results from dust explosion tests carried
out with peat dust samples are reported. The
m a x i m u m explosion pressure and the Kst
value have been measured, and especially
their dependences on the moisture content
and mean particle size have been investigated.
It is concluded that the maximum explosion
pressure and the Kst value o f homogeneous
dry (moisture content <35% W/W) and fine
(mean particle size ~ 2 0 0 I~m) peat dust can
be estimated with a reasonable accuracy from
the moisture content and the mean particle
size. For moist (moisture content >35% W/W)
and~or coarse (mean particle size >200 ~m)
peat dust, it is often possible only to predict
whether the dust is explosibie or not. For
heterogeneous peat mixtures composed o f
dusts with very different moisture contents
and~or mean particle sizes, it is not possible
in a reliable way to estimate the maximum
explosion pressure or the Kst value based on
the mean moisture content and the mean par-
ticle size o f the mixture. It is not always even
possible to predict whether the mixture is
explosible or not.
INTRODUCTION
During the last few years the use of peat
for energy production has increased rapidly
in several countries. This development has
shown that peat dust is a reactive substance,
which relatively easily can lead to fires and
explosions in peat power stations. According
to an investigation, there occur 2 - 3 fires and
explosions every year in each peat power sta-
tion in Finland [1]. The total number of fires
and explosions recorded during a five-year
period was 115. Of these incidents only five
were explosions.
0379-7112/87/$3.50
97
In order to avoid the occurrence of the
usually costly fires and explosions at the
power stations, as much as possible of the fire
and explosion properties of peat dust should
be known. In this paper we discuss how the
m a x i m u m explosion pressure and the Kst
value of peat dust explosions depend on the
mean particle size and moisture content of
the dust.
A comprehensive study of the explosibility
of peat dust is published by Scholl [2]. Among
the many results there, it is worth mention-
ing the findings that a peat sample with a
mean particle size of 2650 pm (moisture con-
tent 11.5% W/W) was explosible and that
another sample with a moisture content of
41% W/W (mean particle size 39 pm) was also
explosible, while a sample with a moisture
content of 51% W/W (mean particle size 60
/~m) was not explosible.
EXPERIMENTAL METHODS
The determination of the m a x i m u m explo-
sion pressure and the Kst value of the samples
was made in accordance with the International
Standard ISO 6184/1 [3]. The tests were car-
ried out with a commercially available appa-
ratus known as the '20 litre sphere', in which
a weighed sample of dust is injected with com-
pressed air through a perforated ring-nozzle
into a partially evacuated spherical pressure
vessel with a volume of 20 litres. The dust is
ignited in the middle of the sphere with two
chemical igniters with a total energy o u t p u t
of 10 kJ, 60 ms after the beginning of the
dust injection. The pressure in the sphere is
measured with two piezo-electric pressure
transducers. To find the maximum explosion
pressure and the m a x i m u m rate of pressure
rise for each dust, the tests are repeated over
a large dust concentration range, e.g., 250 -
Elsevier Sequoia/Printed in The Netherlands
98

2000 g/m 3. The appropriate Kst value is cal- ticles may be reduced by the dispersion pro-
culated from the 'cube root law', which states cess. To check if this is the case the particle
that the Kst value is equal to the product of size distributions of two different peat dust
the maximum of the maximum rates of pres- samples were measured both before and after
sure rise and the cube root of the volume of dispersion without ignition. No clear differ-
the test equipment. The standard deviation ences between the particle size distributions
for the maximum explosion pressure is re- were, however, observed.
ported to be about 4% and for the Kst value The moisture content of the dust samples
15 - 20%, in the appropriate range. was determined by drying to constant weight
It should be noticed that the standard ISO at 105 -+ 2 C and the particle size distribution
6184/1 [3] states that the volume of the test by dry sieving.
apparatus is 1 m 3. Other test equipment can
be used providing it has been proven that they
give results directly, or by calculation, that RESULTS
are commensurate (+20%) with the results
obtained using the 1-m 3 apparatus. Siwek In the first phase, peat dust samples were
[4] has shown that the '20 litre sphere' meets either collected from several power stations
this requirement. in Finland (samples nos. 1 - 10) or prepared
The standard further states that generally from sod peat of different origins (samples
the 1-m 3 test apparatus is suitable for evalua- nos. 11 - 18). The samples were air<tried be-
tion of maximum explosion pressures and K s t fore the dust explosion tests. The results are
values of combustible dusts which have a par- given in Table 1.
ticle size not exceeding 63 pm and a moisture In the second phase the influences of the
content not exceeding 10% WfW. Dusts with moisture content and the particle size of the
larger particle size and/or higher moisture peat dust on the maximum explosion pressure
contents can be characterized, provided that and Kst value were investigated. From one
such dust can be dispersed effectively in the sample of milled peat, 12 dust samples with
explosion chamber. In this study peat dusts different moisture contents and particles sizes
with particle sizes up to 500 pm and moisture
contents up to 59.5% were tested. There are
means to observe the efficiency of the disper- TABLE 1
sion of the dust in the 20-1itre sphere. The The m a x i m u m e x p l o s i o n pressure a n d K s t value for
measuring and recording unit of the apparatus air-dried peat dust samples
contains routines which measure the effective
ignition delay time or the time between the Sample Moisture Particles Maximum K s t value
start of the dust injection and the m o m e n t no. content <63/~m explosion (barm/s)
when atmospheric pressure has been reached (% W/W) (% W/W) pressure
(bar)
within the sphere and the ignition is activated.
The nominal effective ignition delay time is 1 11.0 47.7 7.9 133
60 ms. If the dust does not flow freely, but, 2 l 1.1 34.6 8.3 141
for example, tends to clog up the perforations 3 8.2 39.8 8.1 138
of the ring-nozzle, the effective ignition delay 4 7.3 34.0 8.2 131
5 7.5 28.4 7.8 102
time will be affected. The measuring and re-
6 12.3 20.4 7.5 116
cording unit accepts only results where the 7 8.5 36.8 8.5 133
ignition delay time is 55 - 65 ms. After each 8 8.7 74.9 8.4 151
test the equipment is cleaned and the end 9 8.4 30.6 8.1 137
caps of the ring-nozzle are removed and the 10 8.5 26.1 8.1 124
11 8.3 19.2 8.2 142
nozzle is inspected for any residues of undis- 12 8.6 28.5 8.2 130
persed dust. During the test series no signs of 13 7.6 33.3 8.1 154
problems with the dispersion of the dust were 14 10.4 24.1 7.7 97
observed. 15 9.6 27.9 7.9 121
It is known that the ring-nozzle of the test 16 12.5 23.5 7.5 94
17 12.6 27.0 7.9 126
apparatus may have a grinding effect on the 18 13.2 20.7 7.7 106
dust being tested, i.e., the size of the dust par-
 99

TABLE 2 mean particle size and moisture c o n t e n t in the

The m a x i m u m explosion pressure and Kst value for investigated range.
prepared peat dust samples with different moisture In the third phase the effects on the maxi-
contents and particle sizes m u m explosion pressure and Kst value of
mixing peat samples with very different mois-
Sample Moisture Particle size Maximum Kstvalue ture contents and particle sizes were investi-
no. content fraction explosion (bar m/s)
gated. These measurements were considered
(% W/W) (~m) pressure
(bar) necessary to find o u t the results of adding
small amounts of strongly explosible peat
100 0.0 0...500 8.2 165 dust to non-expl0sible peat dust. Bartknecht
101 0.0 0... 74 8.4 183 [5] reports that mixing 5% of a fine explo-
102 0.0 74...125 8.3 146
sible polyethylene dust into a coarse non-
103 0.0 125...250 8.1 133
explosible polyethylene dust gives an explo-
200 14.0 0...500 7.9 143
sible mixture. The m a x i m u m explosion pres-
201 14.1 0... 74 8.4 139
202 13.0 74...125 7.8 95 sure and the K s t value remained almost con-
203 12.9 125...250 7.3 71 stant when the a m o u n t of coarse dust was
300 33.6 0...500 6.8 43 0 - 95%. The finding by Scholl [2] that a peat
301 32.1 0 . . . 74 7.7 80 dust sample with a mean particle size of 2.65
302 33.6 74...125 7.0 53 mm was explosible could perhaps be under-
303 33.8 125...250 6.8 42 stood if it contained small amounts of fine
explosible peat dust.
From a new sample of milled peat the fol-
were prepared. The results of the dust explo- lowing three basic samples were prepared:
sion tests are given in Table 2. -- 'dry/fine', moisture c o n t e n t 0% W/W, par-
The results in Table 2 are sketched in Fig. 1. ticle size fraction < 100 #m.
From Fig. 1 it can be seen t h a t both the mean - - ' d r y / c o a r s e ' , moisture content 0% W/W,
particle size and moisture content have a particle size fraction 250 - 500/zm.
stronger effect on the Kst value than on the - - ' m o i s t / c o a r s e ' , moisture content 59.5%
m a x i m u m explosion pressure. Both the maxi- W/W, particle size fraction 250 - 500 pm.
m u m explosion pressure and the Kst value From these basic samples the mixtures
decrease almost linearly with increase of the listed in Table 3 were prepared. Table 3

IIHIfllF IIIIII I III I _ J

iiii rtl I]Z ; oo
. iLlll,il LIHIHlillllil, ' _I "'00

100 (/j, rfl) 1o0 ( iL/,r'lq)

- ~ o ~ , ~ %0 /porticle
~ size ~ o 0 porticle size
Fig. 1. The dependence of the moisture content and mean particle size on the maximum explosion pressure and
Kst value of a peat sample.
 100

TABLE 3 DISCUSSION
The maximum explosion pressure and Kst value of
peat dust mixtures F r o m a practical p o i n t o f view, it w o u l d be
very useful to k n o w the borderline b e t w e e n
Composition of the mixture Maximum Kst value explosible and non-explosible peat dust with
explosion (bar m/s) respect to the m o i s t u r e c o n t e n t and mean
Dry/fine Dry/ Moist/ pressure
(% W/W) coarse coarse (bar) particle size. T o find this, several regression
(% w/w) (% w/w) surfaces were c o m p u t e d f r o m the measure-
m e n t s o f Tables 1 and 2, b u t n o n e o f t h e m was
100 -- -- 9.0 169 considered to be satisfactory for m o i s t u r e
75 25 -- 8.9 158
50 50 - 8.7 125 c o n t e n t s over 35% W/W and m e a n particle
25 75 -- 8.2 102 sizes over 2 0 0 / a m . T h e reason for this is t h a t
10 90 -- 7.0 59 there are o n l y very few m e a s u r e m e n t s in this
5 95 -- 6.4 38 region.
- 100 -- 6.5 27 F r o m the 30 m e a s u r e m e n t s o f Table 1 and
- 75 25 4.9 10
.... 50 50 2.1 4 2 the following regression surfaces, valid for
-- 25 75 0 0 m o i s t u r e c o n t e n t s less than 35% W/W and
-- -- 100 0 0 m e a n particle sizes b e t w e e n 30 /am and 2 0 0
25 -- 75 2.9 6 /am, can be d e t e r m i n e d with multiple regres-
50 -- 50 7.4 49 sion analysis:
75 -- 25 8.4 117
Pm~x = 11.1 -- 0.031W -- 0.63 In (~m (1)
gives also the results f r o m the d u s t e x p l o s i o n Kst = 341 -- 2.58W -- 42.0 In ~bm (2)
tests.
where
T h e results o f Table 3 are sketched in Fig. 2.
Pmax is the m a x i m u m e x p l o s i o n pressure (bar)
O n e n o t e w o r t h y result is t h a t the m i x t u r e
Kst is the Kst value (bar m/s)
c o n t a i n i n g 75% ' m o i s t / c o a r s e ' and 25% ' d r y /
W is the m o i s t u r e c o n t e n t (% W/W)
fine' was explosible ( m e a n m o i s t u r e c o n t e n t
~m is the m e a n particle size (/am)
44.6%), while t h e m i x t u r e with t h e same m e a n
moisture c o n t e n t c o n t a i n i n g 75% ' m o i s t / T h e f o r m for eqns. (1) and (2) is indirectly
coarse' and 25% ' d r y / c o a r s e ' was not, t h o u g h suggested by Scholl [ 2 ] , w h o s h o w e d t h a t
the mean particle sizes o f the m i x t u r e s did Pmax and Kst d e p e n d a l m o s t linearly on the
n o t differ very m u c h ( 3 2 0 / a m and 3 9 0 / a m ) . m o i s t u r e c o n t e n t o f the peat dust in the range

..... -.
_.. " \\

10 J f ~ J 200 i

. f ](}

5 ~'0 100 -\\

~,~ , "\
f

"\

fioe p0rt~cle size ~

rio ~oFtlCle size
Fig. 2. The dependence of the composition of a peat dust mixture on the maximum explosion pressure and Kst
value of the mixture.
 101

15 - 35% W/W. Similarly, Pmax and Kst depend unsafe direction, t h a t is, eqns. (1) and (2)
almost linearly on the logarithm of the mean underestimate the strength of the explosion.
particle size of the peat dust in the range 50 - As a regression surface valid for a larger
400 pm. area could not be determined, the measure-
The coefficients of determination (multiple ments of Tables 1 and 2 and also those of
R-square, R 2) of eqns. (1) and (2) are 89.1% Scholl [2] were plotted in the particle size-
and 90.9%, that is, the independent variables moisture c o n t e n t plane of Fig. 3. On the basis
W and Cm explain 89.1% and 90.9% o f the va- of these results, the plane can be divided into
riation in the dependent variables Pma, and the following parts:
K s t . The 30 measurements of Tables 1 and 2 - - a n upper part, where a dust explosion is
for P~ax fall within +0.46 bar from the values very unlikely,
given by eqn. (1) and for K s t within +19.6 bar -- a lower part, where a dust explosion is very
m/s from the values given by eqn. (2). Using likely, and
surfaces of higher degree, higher values of R 2 - - a middle part, where a dust explosion is
can be obtained*. uncertain.
The eqns. (1) and (2) do n o t apply equally The results of the measurements with the
well for the heterogeneous mixtures of Table peat dust mixtures are also plotted in Fig. 3.
3. For example the calculated values for the Figure 3 shows that it is not always possible
last mixture in Table 3 are p ~ , = 7.4 bar and to predict whether a mixture is explosible or
K s t = 88 bar m/s, while the measured values not on the basis of the mean particle size and
were 8.4 bar and 117 bar m/s. The differences moisture c o n t e n t of the mixture. The mixture
are thus 1.0 bar and 29 bar m/s and are larger marked A is explosible while mixture B is not
than for the samples of Tables 1 and 2. It is explosible, though the difference in the mean
also to be noted that the calculations err in an particle size is small.

* T h e c o r r e s p o n d i n g linear e q u a t i o n s are: CONCLUSIONS

Pmax = 8.9 -- 0 . 0 2 9 W -- 0.0064@ra (1') From the observations made it can be con-
Kst = 193 -- 2.46W-- 0.43~m (2') cluded that,
For (1') R 2 = 89.2%, max. deviation is +-0.42 bar. --for homogeneous dry (moisture con-
For (2') R 2 = 91.4%, max. deviation is -+22.8 bar m/s. tent <35% W/W) and fine (mean particle size

70

~" 6O 0
Non-exploslble peot dust
x
5O X

OO
c 40 - - _ +~ ~ "~ x x
Q)
c- \
O 50

ID x
~ 2O +
[xplostble p e o t dumt
~+
+ + e;t
0 10 + +
+ ~ + 4.+ +
2~
0 i , ,zl ,i--I I , I I - - , I I i i i I i i i i i i i i

10 100 1000 10000

Mean particle size (p,m)
Fig. 3. T h e division o f t h e m e a n particle size--moisture c o n t e n t p l a n e i n t o areas w h e r e a peat d u s t e x p l o s i o n is
likely a n d unlikely. + e x p l o s i o n o b s e r v e d ; x n o e x p l o s i o n o b s e r v e d , e x p l o s i o n for m i x t u r e o b s e r v e d ; o n o explo-
sion for m i x t u r e o b s e r v e d .
102
< 2 0 0 pm) peat dust, the maximum explosion
pressure and the g s t value can be estimated
with a reasonable accuracy from the moisture
content and mean particle size (with, for ex-
ample, eqns. (1) and (2) or similar);
- - f o r homogeneous moist (moisture content
>35% W/W) and/or coarse (mean particle
size ~ 2 0 0 gm) peat dust, it is often possible
to predict whether the dust is explosible or
not based on the moisture content and mean
particle size (with, for example, Fig. 3 or sim-
ilar), and
- - f o r peat mixtures composed of dusts with
very different moisture contents and/or mean
particle sizes, it is not possible to estimate the
maximum explosion pressure and the K s t
value reliably, or even to predict whether the
mixture is explosible or not, from the mean
moisture c o n t e n t and the mean particle size
of the mixture.
The explosion properties of peat dust are
also affected by other properties than the
moisture content and mean particle size. Such
properties are, for instance, the degree of hu-
mification, the botanical composition and
the morphology of the particles. The influ-
ence of these is, however, small compared
with the moisture content and mean particle
size. For eqns. (1) and (2) it was found that
R 2 was 89.1% and 90.9%. If the equations
are computed for the twelve samples of
Table 2 prepared from one sample of milled
peat, it is found that R 2 is 92.2% and 93.2%.
REFERENCES
1 K. Mutanen et al., Safe Production and Use o f
Domestic Fuels, Part 1, A Preliminary Study,
Research Notes 323, Technical Research Centre
of Finland, Espoo, 1984, 118 pp. (in Finnish).
2 E.-W. Scholl, Explosionsverhalten yon Torf, Torf-
forschung GmbH, Bad Zwischenahn, 1978, 21 pp.
3 ISO 6184/1, Explosion Protection Systems. Part
1: Determination o f Explosion Indices o f Com-
bustible Dusts in Air, International Organization
for Standardization, Geneva, 1985, 5 pp.
4 R. Siwek, Development o f a 20-1itre Laboratory
Apparatus and its Application for the Investiga-
tion o f Combustible Dusts, Ciba-Geigy Ltd.,
Basle, 1985, 189 pp.
5 W. Bartknecht, Explosionen, Springer, Berlin,
2nd edn., 1980, p. 33.
Journal of ~ccu~a~~o~u~Accidents, 9 (1987) 161-175 161
Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands

Dust Explosion experiments iu a Vented 236 m3

Silo Cell*

R. K. ECKHOFF, K. FUHRE and G. H. PEDERSEN

Chr. ~~c~e~en Zns~i~ute,Bergen (Norway)
(Received 25 February 1987; accepted 7 April 1987)

ABSTRACT

Eckhoff, R. K., Fuhre, K. and Pedersen, G. H., 1987. Dust explosion experiments in a vented
236 m3 silo cell. Journal of Occupational Accidents, 9: 161-175.

The influence of dust concentration and location of the ignition point, on flame propagation
and pressure development, has been studied in a series of vented maize starch explosions in a new
236 m3 steel silo, constructed specially for dust explosion experiments. The investigation consti-
tutes a continuation of the experiments in a 500 m3 silo, reported previously. However, the new
silo of height 22 m has a diameter of 3.7 m, and thus L/l?=6, whereas the 500 rn' silo had L/D= 4.
Explosible clouds of maize starch were generated in the silo by conventional pneumatic injection,
either at the top or at the bottom of the silo. Two vent areas of either 3.4 or 5.7 m, were used.
The most explosible dust concentration was found to be 400-500 g/m. The average minimum
explosible concentration was about 100 g/m3.
The experiments confirmed that for explosions in large, slim silos, vented at the top, the max-
imum explosion pressure depends strongly on the vertical distance between the vent and the ig-
nition point further down. Therefore no straight-forward correlation between vent area and
maximum explosion pressure exists. Ignition close to the silo bottom generated pressures exceed-
ing 1 bar (g) and flame speeds exceeding 100 m/s, whereas ignition close to the vent at the silo
top gave only marginal pressures, of the order of lo-20 mbar (g), and flame speeds of a few m/s.
Realistic sizing of vents for slim silos must imply identification of likely ignition source loca-
tions and be performed within an overall context of risk assessment.

REALISTIC FULL-SCALE VENTED DUST EXPLOSION EXPERIMENTS ARE NEEDED

From a theoretical point of view, dust explosion venting is a very complex

process, and no adequate theory exists. Realistic full-scale experimentation
therefore still seems to be the most beneficial means of obtaining further in-
sight of practical value to industry (Eckhoff, 1982),
In coal mine explosion research, experiments under realistic dust cloud gen-

*Paper presented at the 5th International Symposium on Loss Prevention and Safety Promotion
in the Process Industries, Cannes, France, September 15-19,1986.

0376-6349/87/$03.50 0 1987 Elsevier Science Publishers B.V.

162

eration conditions have been conducted in full-scale mine galleries for many
years ( Cybulski, 1975).
Dust explosion experiments conducted in a vented 500 rn+silo some years
ago f Eckhoff and Fuhre, 1984) confirmed the need for realistic experiments.
W&h initially quiescent dust clouds, significantly smaller vent areas were
required to reduce the explosion pressure to a given value, than those specified
in the VDI-3673 and NFPA-68 codes (VDI, 1979; NFPA, 1978). However,
initial turbulence increased the pressure by at least one order of magnitude,
and even beyond the VDI code prediction.
An interesting investigation of vented dust explosions in a cyclone plant,
working under conditions of realistic dust cloud generation, has been described
by Palmer (1973 ) and Tonkin and Berlemont ( 1972). The experiments revealed
that. t,he vent areas required for not exceeding a given maximum pressure, were
substantially smaller, by a factor of five or more, than those prescribed by VDI-
3673.
In a recent, similar investigation, Lunn and Cairns (1985) studied the pres-
sure build-up during vented dust explosions in a bag filter unit, run as in nor-
mal industrial operation. Even in the worst case, with ignition in the bottom
hopper section and venting at the top of the filter housing, the explosion pres-
sures were significantly lowere than those predicted by VDI-3673. When the
vent was located in the rear wall of the housing, close to the point. of ignition,
very much lower explosion pressures were generated.
In t,he present contribution attention is focused at explosion of dust clouds,
generated by conventional pneumatic conveying, in a large, slim silo of volume
236 rn and L/D=6, and with a vent in the top surface. A detailed account of
the investigation is given elsewhere (Eckhoff et al., 1985).

THE EXPERIMENTAL SILO INSTALLATION

Thesilo

The cylindrical silo of height 22 m and diameter 3.7 m is shown in Fig. 1.

The main body, made of 8 mm thick steel plate, all-welded, can withstand at
least 5 bar internal overpressure. The silo wall is provided with a number of
threaded holes of diameter 85 mm, accessible from a winding staircase, and
permitt,ing ignition sources and various types of diagnostic probes to be mounted
at desired levels above the silo bottom. A steel grid across the entire silo top,
allows any desired part of the top to be blocked by bolting a number of 0.25 m
steel plates to the grid. Thus, varying the vent opening area and shape is easy.
In the present investigation two alternative vent openings were used, one square
of 5.7 m2 open area, and one rectangular of 3.4 m2 open area.
 163

Fig. 1. Vertical section of 236 m3 experimental silo.

The dust injection system

Explosible clouds of maize starch were generated in the silo by blowing the
starch into it through a conventional pneumatic conveying line of i.d. 155 mm.
Three alternative dust entrance configurations were used:
a. Upwards injection from the bottom, with the conveying pipe exit about 5.5 m
above the silo bottom, as shown in Fig. 1.
b. Upwards injection from the bottom, but with the 5 m vertical pipe removed,
i.e. with the conveying pipe exit only 0.5 m above the silo bottom.
C. Semi-tangential, horizontal injection at the silo top, with the conveyingpipe
exit flush with the silo wall.
The injection system consisted of a Roots blower driven by a 55 kW electric
motor. The starch was fed into the pipeline from a 1.5 m3 hopper a few metres
downstream of the blower. A conventional rotary lock was used for the feeding
process. During feeding, the hopper was sealed and pressurized to balance the
overpressure in the conveying line. The steady-state feeding rate of maize starch
164

into the pipe line was about 8-9 kg/s. With no dust in the pipeline the blower
yielded 2600 m/h (standard state). During injection of maize starch the air
flow normally dropped to about 2150 m/h (standard state). Typical average
concentrations of dust in the conveying pipe during injection were lo-13 kg/m
( standard state ) .
In most of the tests the air flow was maintained for the time required to feed
all the dust in the hopper into the pipeline, plus a few seconds to clear the
pipeline of remaining starch. Then the turbulence in the dust cloud was allowed
to dissipate for a few seconds, before activating the ignition source. However,
in a few tests the ignition source was activated earlier, whilst dust was still
being injected.

The ignition source

The ignition source was the same as that used in the previous 500 rn exper-
iments (Eckhoff and Fuhre, 1984). It consisted of about 50 g of dried nitro-
cellulose powder contained in a plastic bag and fired by a pair of electrically
ignited Ce-Mg 100 J fuse heads. The special quality of nitrocellulose used
(12.3% N ) was chosen so as to prevent any significant blast effects that could
disturb the dust cloud, while at the same time the burning was sufficiently fast
to produce a momentary and sufficiently large and efficient flame source
(approx. 50-100 cm diameter). The flame was fully developed approximately
0.1 s after firing the fuse heads and maintained its full size for approximately
1.0 s. The total energy liberated by the ignition source was about 200,000 J.
Due to continued burning of the plastic bag, the ignition source, after the initial
main energy release, kept on burning for up to half a minute. Consequently, if
no ignition of the maize starch cloud was obtained, and the ignition source had
been activated sufficiently early, temporary presence of explosible dust con-
centrations in the ignition source region at any instant during the experiment
could, due to the prolonged burning of the source, be excluded.

Diagnostic system

Four different types of conventional measurements and observations were

made during an explosion experiment:
a. Three Kistler No. 7261 pressure transducers, one located 3 m above the silo
bottom, one 9 m above the bottom, and one 2.5 m below the top, were used
for measuring the development with time of the explosion pressure inside
the silo.
b. Six light attenuation probes, constructed by CM1 on the basis of work done
at USBM (Liebman et al., 1977; Conti et al., 1982) and CERL (Gardner
and Moore, 1983 ) , were used for continuous monitoring of the dust concen-
tration in six different locations in the silo during dust injection. The probes
 165

were calibrated in the laboratory by exposing them to clouds of known con-

centrations of the same quality of maize starch as used in the silo explosion
experiments.
c. Optical transducers (photo-diodes) were employed for measuring flame
arrival times at up to six different locations in the silo. In fact, these trans-
ducers were identical with the receiver part of the light attenuation probes
used for measuring the dust concentration (see b. above).
d. A 16 mm, 64 frame/s camera, remotely operated, was positioned on a steep
hillside about 40 m from the silo top, to provide high quality live pictures of
selected experiments. Visual recordings of the explosions were also under-
taken manually by means of a conventional 25 frame/s video camera. One
important objective of this was to determine the delay between the appear-
ance of the light flash of a conventional flash bulb mounted outside the silo,
synchronised with the firing of the ignition source, and the arrival of the
flame front at the vent opening at the top of the silo.

THE MAIZE STARCH USED IN THE EXPERIMENTS

The native maize starch used was of the same batch as that used in the
previous 500 m3silo experiments (Eckhoff and Fuhre, 1984). From 1982 to
1985 the starch was stored indoors at CMI.
Gross samples of about 1 1 were taken from a number of bags immediately
before emptying the bags into the small storage hopper, from which the dust
was discharged into the dust injection system of the silo. The samples were
kept in sealed plastic jars and brought to the CM1 laboratory for analysis.
The moisture content, related to weight before drying, and determined by
drying for 4 h at about 105 C, varied between 9.7 and 11.8%, with a mean value
of 11.1%. This is slightly higher than the mean value of about 10% found in
corresponding samples in 1982 (Eckhoff and Fuhre, 1984).
Size analyses by means of the Alpine air jet sieve revealed that nearly 80%
by weight fell in the size range lo-20 ,um, in agreement with the fact that the
maize starch is essentially composed of nearly monosized and spherical native
grains of mean diameter about 12 pm.
Some maize starch samples were subjected to explosion violence tests in the
Hartmann bomb and the 20 1 sphere (Bartknecht, 1978). The gross mean
values of (dP/dt) max and P,,,, were 220 bar/s and 5.15 bar (g) in the Hart-
mann bomb and 365 bar/s and 7.2 bar (g) in the 20 1 sphere. These values are
slightly lower than those measured previously (Eckhoff and Fuhre, 1984))
probably in part due to the slight increase of the moisture content in the dust.
166

RESULTS AND DISCUSSION

Dust concentration distribution

Different dust quantities, ranging from 25 kg to 300 kg, were injected into
the silo by all three methods described above. The delay between abrupt ter-
mination of dust injection, and activation of the ignition source, was varied
from 11.6 s down to negative values, in the case of which the explosion started
while dust was still being injected.
In the first 29 experiments, the dust was introduced from the bottom, through
the 5 m vertical pipe, i.e. with the injection point about 5.5 m above the silo
bottom. The average dust concentration in the jet entering the silo was esti-
mated to about 13 kg/m (standard state).
In the next 18 experiments, the dust was injected through the pipe exit at
the top of the silo. With this configuration significant masses of dust accu-
mulated in the 21 m vertical pipe and became injected at a later stage. However,
if prolonged injection times were used to clear the line completely, the main
part of the dust already injected would settle out at the silo bottom to a larger
extent than new dust was blown in, and the average concentration would
decrease. Therefore it had to be accepted that with dust injection at the silo
top, an appreciable fraction of the dust fed into the pneumatic transport line
would not take part in the explosion.
Complete series of six measurements of dust concentration by the light
attenuation probes were, due to breakdown of the probes by heavy thermal
loads during some of the explosions and subsequent fires, only obtained in a
limited number of the experiments. As expected, bottom injection through the
5 m vertical pipe tended to give higher dust concentrations in the lower half of
the silo than in the upper half. Top injection, as also expected, seemed to gen-
erate a more axially even dust concentration distribution.

Explosion pressure development

The three pressure traces, recorded at 3 m, 9 m and 19.5 m above the silo
bottom, were normally quite similar, although slightly higher pressure peaks
were usually recorded in the bottom part of the silo than further up. On average
the peak pressure at 9 m was about 94%, and at 19.5 m about 87%, of the peak
at 3 m.
The pressure traces obtained can be grouped in three main categories. The
first was those exhibiting a smooth, bell-shaped main pulse, followed by 4-6
Hz oscillations, of considerably smaller amplitudes. Such traces were encoun-
tered in the case of bottom ignited explosions of moderate strength. However,
occasionally a smooth main pulse was found even in the case of quite strong
explosions, as shown in Fig. 2.
 167

t
IGNITION 0 2 04 06 0.8 10 1.2 (5)

Fig. 2. Example of smooth, bell-shaped pressure pulse followed by oscillations of smaller ampli-
tudes. Ignition close to silo bottom.

1
500 mbar

A Pl

IGNITION 0 2 014 0:6 018 1.0 1.2 1.4 (5)

Fig. 3. Example of jagged pressure pulse resulting from ignition 3-6 m above silo bottom.

The second main category of pressure traces, shown in Fig. 3, occurred in

the case of the strongest explosions, of peak pressures of 0.8-1.2 bar (g). In
these cases a first phase of smooth pressure rise would be abruptly followed by
a phase of much steeper rise, leading into a more or less systematic sequence
of sharp pressure spikes. In the particular example shown in Fig. 3, the oscil-
latory pattern of the spikes is quite regular, making it possible to identify a
typical frequency of 17-18 Hz. The physical nature of these rapid oscillations
168

I
I ,
1
8 I
I
I

IGNITION 0.5 1.0 15 20 25 (5)

Fig. 4. Example of oscillatory pressure development caused by ignition in upper half of silo (13.5
m above bottom). Oscillations persisted for about 5 s.

has not been fully disclosed. A form of Helmholtz oscillation may be one guess.
No doubt this phase of the explosion is due to violent turbulent combustion,
starting at about the same time as the flame edge reaches the region of the vent
opening. The mechanism for creating the high turbulence level is probably
complex. Turbulence will be generated both in the shear layer along the silo
wall, when unburnt dust is pushed towards the vent opening by the expanding
burning cloud further down in the silo, and in the cloud just below the vent,
due to the baffle effect of the edges of the vent opening. The process is likely
to be of the same nature as the strong turbulence-induced combustion accel-
eration observed in large-scale gas explosions (Eckhoff et al., 1984; Moen et
al., 1982).
The third main category of pressure development encountered is shown in
Fig. 4. It is characterised by a long train of oscillations, often with a negative
phase of the same magnitude as the positive one. The oscillations, which could
have peak values approaching 100 mbar, show a characteristic frequency of
4-7 Hz. This is the basic standing wave frequency of a one-end-open silo. The
moderate increase of the frequency with time is probably due to the increase
of the average gas temperature as the explosion develops. This exaggerated
oscillatory pressure development occurred only when the ignition source was
located in the middle and upper part of the silo. The oscillatory nature of the
explosion could be clearly seen on the video recording. Packets of flame were
ejected at a frequency matching exactly that exhibited by the pressure traces.
 169

The significance of this type of oscillatory pressure development with respect

to the structural response of silos, may have to be discussed.
In the previous study of vented dust explosions in a 500 m3 silo (Eckhoff
and Fuhre, 1984) it was found that a substantial increase in the explosion
violence resulted when the dust cloud was ignited whilst the dust injection was
still taking place. It was of interest to investigate whether a similar effect could
be observed in the new silo. Two replicate experiments had already been per-
formed by injecting 100 kg of starch from the silo bottom through the 5 m
vertcial pipe. The injection period was 14 s, followed by automatic activation
of the ignition source located 1.5 m above the silo bottom, 2.6 s after. It was
therefore decided to undertake a similar experiment, the only difference being
that 125 kg was fed into the system during 18 s, instead of 100 kg during 14 s.
Ignition at 1.5 m above the silo bottom was, however, performed at 16.6 s as
normal. This means that the turbulent dust jet would persist until 1.4 s after
ignition. However, no significant violence-promoting influence of the jet could
be observed. One partial reason for this could be that the explosion-induced
turbulence generated in the long, slim silo in the case of bottom ignition, masks
the influence of the turbulent jet. Another likely reason is the high concentra-
tion of dust in the injection pipe, of about 13 kg/m3. The regions of high tur-
bulence close to the pipe exit therefore probably contained too high dust
concentrations for very rapid combustion to take place.
As far as reasonable estimates of average dust concentrations could be
obtained, it was found that the most explosible concentration was 400-500
g/m3, although quite violent explosions were also generated at higher concen-
trations, of up to 600-700 g/m. On the lean side the explosion pressure seemed
to drop relatively sharply with decreasing concentration to an average mini-
mum explosible limit of about 100 g/m.

Influence of ignition source location on maximum explosionpressure

As the experiments progressed, it was confirmed that the location of the

ignition source was a decisive parameter with respect to the maximum explo-
sion pressures generated in the vented silo. The pressure was reduced by two
orders of magnitude, from about 1 bar, to only 10 mbar when the ignition point
was moved from the bottom of the silo to close to the vent opening at the top.
The likely main reason for the much higher pressures generated by bottom
ignition is generation of turbulence, as discussed above. Video recordings
revealed that the time for the flame to reach the vent in fact decreased as the
ignition source was moved downwards, in spite of the fact that the distance
that the flame had to travel to get to the vent increased.
The previous experiments with venting of dust explosions in 500 m3 silo cells
of L/D = 4 (Eckhoff and Fuhre, 1984) revealed a reasonably systematic decrease
of explosion pressure with increasing vent area, provided the dust cloud was
 lgnltlon at 21 m

2 3 4 5 6 7

vent ac?a Of 236 d stio inI')

Fig. 5. Influence of ignition source location and vent area on maximum explosion pressure. Worst-
case dust clouds from all three different modes of dust cloud generation. Comparison with VDI
3673 (VDI, 1979).

video
- WALL

0 Q,l Q,2 Q,3 0.4 0.5 0.6 0.7

TIME AFTER IGNITION (sl

Fig. 6. Vertical displacement of flame front as a function of time after ignition at 1.5 m above silo
bottom 5.7 m2 vent area.
 171

initially non-turbulent. Only bottom ignition was used in these experiments,

but it is felt that due to the smaller L/D, the influence of the ignition source
location would be less pronounced than in the present experiments.
Figure 5 summarizes the most important result of the present work with
respect to industrial practice: a quantification of the strong influence of the
location of the ignition point. Experience from gas and dust explosions in mine
galleries and long pipelines, conducted in the past, suggests qualitatively that
the kind of relationship shown in Fig. 5 would exists. However, a quantification
was needed. In fact, the influence of the ignition source location is so strong
that the vent area, over the range considered, only constitutes a second order
parameter.
This confirms the general need for a new approach to vent sizing in the
process industry. In future, simple nomographs and formulae will have to be
replaced by methods tailoring the vent arrangement to suit each specific appli-
cation, within an overall context of risk assessment (Eckhoff, 1986, 1987).

- WALL

Max flame spew

= 20 m/s

r
I I 7 n
0 0.2 0.L 0.6 0.8 1.0 1.2

TIME AFTER IGNITION (sl

Fig. 7. Vertical displacement of flame front as a function of time after ignition at 13.5 m above silo
bottom; 5.7 m2 vent opening.
Note: The oscillatory nature of the explosion is not resolved by these measurements.
172

Flame front displacement

Figure 6 shows an example of the measured vertical position of the flame

front as a function of time. The time of flame-arrival at the vent opening, as
obtained from the video recording, fits well into the picture and provides a most
valuable additional data point. Because of the limited resolution, the precision
of the time derived from the video recording is 2 40 ms, as indicated. As indi-
cated in Fig. 1, the supporting rods of three of the flame detection probes were
0.5 m long, whereas the other three were 1.5 m. This made it possible to record
the vertical flame front displacement both near the axis and near the wall of
the silo. In general, the flame front near the wall lagged somewhat behind that
near the axis. The maximum flame speed estimated from the gradient of the
displacement curve in Fig. 6, is about 70 m/s. The highest flame speed meas-
ured in the investigation altogether was 110 m/s.
Figure 7 shows the flame development following ignition at 13.5 m above the
silo bottom. In this case the flame propagated simultaneously upwards and
downwards. In the particular case shown, the ignition source was activated

TIME FROM IGNITION TO FLAME AT VENT [ms]

Fig. 8. Coincidence of moments of arrival of flame front at the vent opening, and occurrence of
pressure peak.
 173

1.4 s before terminating dust injection, rather than 2.6 s after, which was the
standard procedure. Consequently, heavy dust suspension being still injected
at the silo top, dropped rapidly from the injection point towards the silo bottom
during the explosion, and gave rise to a comparatively high downwards flame
speed of about 20 m/s in the lower part of the silo. As expected, a somewhat
stronger explosion resulted, than when quiescent clouds were ignited at 13.5
m. The explosion pressure exhibited oscillations of 6-7 Hz and maximum posi-
tive amplitudes of about 100 mbar, similar to the course in Fig. 4.

Coincidence of flame-arrival at silo top, and occurrence of pressure peak

In the previous investigation of explosions in a 500 m3 silo (Eckhoff and

Fuhre, 1984)) it was found that the moment of arrival of the flame at the vent
at the silo top, and the moment of the occurrence of the pressure peak coin-
cided. It was of interest to check whether the same coincidence existed even in
the present experiments. In the case of the oscillatory explosions a single, dis-
tinct pressure peak was not easy to define. In others the video recordings were,
for various reasons, not obtained. However, for 30 of the experiments, both
time intervals were available, and are plotted in Fig. 8. Although there is a
tendency of the pressure peak occurring slighly later than the first sign of flame
outside the vent, the correlation is conspicuous. The onset of venting of hot
combustion gas is one likely reason for the drop in pressure following the arrival
of the flame front at the vent.

CONCLUSIONS

1. The study of maize starch explosions in a specially constructed top-vented

steel silo of volume 236 m3 and L/D = 6, revealed that the most explosible con-
centration was 400-500 g/m3, although violent explosions were also obtained
at higher concentrations of 600-700 g/m3. The estimated average minimum
explosible concentration in the silo was of the order of 100 g/m.
2. Varying the location of the ignition source along the silo axis, from the
bottom to the top, had a substantial influence on the explosion pressure. With
a vent area of 3.4 m* ignition in the lower third of the silo, resulted in over-
pressures of up to 1.2 bar, whereas ignition in the upper half gave less than 0.1
bar. Ignition very close to the vent at the silo top gave only marginal overpres-
sures, of the order of 0.01 bar.
3. Contrary to what was found in the earlier 500 m3 silo experiments (Eck-
hoff and Fuhre, 1984), ignition of the dust cloud while dust was still being
blown into the silo by a conventional pneumatic conveying system, did not
cause any significant rise of the maximum explosion pressure. High dust load-
ing in the turbulent dust jet of more than 10 kg/m3, i.e. well above the upper
explosible concentration, is a likely main reason for this.
174

4. The maximum vertical flame front speed during an explosion in the silo
ranged from a few m/s in weak explosions, to more than 100 m/s in the strong-
est ones.
5. A differentiated approach to dust explosion vent sizing is needed, which
accounts for the risk assessment aspects of the problem. It is necessary to
discuss which dust cloud formation and ignition processes and associated com-
bustion rates should, under various circumstances in industry, be regarded as
the likely worst cases that the vents to be designed should be able to
accommodate.

ACKNOWLEDGEMENTS

The authors are greatly indebted to the National Grain and Feed Associa-
tion, Washington, DC, for the substantial financial contribution rendering this
work possible, and for unfailing constructive technical advice and support
throughout. Sincere thanks are also due to the Swedish Fire Research Board,
Brandforsk, Stockholm, for granting a valuable partical financial contribution.

REFERENCES

Bartknecht, W., 1978. Explosionen - Ablauf und Schutzmassnahmen. Springer, Berlin, pp. 43-45.
Conti, R. S., Cashdollar, K. L. and Liebman, I., 1982. Improved optical probe for monitoring dust
explosions. Rev. Sci. Instrum., 53 (3) : 311.
Cybulski, W., 1975. Coal Dust Explosions and their Suppression. English translation published
by Foreign Scientific Publ. Dept. of National Center for Scientific, Technical and Economical
Information, Warsaw, Poland.
Eckhoff, R. K., 1982. Current dust explosion research at CMI. In: J. H. S. Lee and C. M. Guirao
(Eds.) , Fuel-Air Explosions. University of Waterloo Press, p. 657.
Eckhoff, R. K., 1986. Sizing dust explosion vents - The need for a new approach based on risk
assessment. Bulk Solids Handling, 6 (5 )
Eckhoff, R. K., 1987. Identifying likely ignition source locations - A key to sizing dust explosion
vents. Symposium on Industrial Dust Explosions, Pittsburgh, USA, June 1986. ASTM Special
Technical Publication.
Eckhoff, R. K. and Fuhre, K., 1984. Dust explosion experiments in a vented 500 m silo cell. J.
Occup. Accid., 6: 229.
Eckhoff, R. K., Fuhre, K., Guirao, C. M. and Lee, J. H. S., 1984. Venting of turbulent gas explo-
sions in a 50 m3 chamber. Fire Saf. J., 7: 191.
Eckhoff, R. K., Fuhre, K. and Pedersen, G. H., 1985. Vented maize starch explosions in a 236 m
experimental silo. Fire and Explosion Research Report No. ESV-86-070, prepared by Chr.
Michelsen Institute for National Grain and Feed Association, Washington, DC, Report No.
843307-2, Chr. Michelsen Institute, Bergen, Norway.
Gardner, B. R. and Moore, M. J., 1983. Personal communication to R. K. Eckhoff at Central
Electricity Research Laboratories, Surrey, UK.
Liebman, I., Conti, R. S. and Cashdollar, K. L., 1977. Dust cloud concentration probe. Rev. Sci.
Instrum., 48 (10): 1314.
 175

Lunn, G. A. and Cairns, F., 1985. The venting of dust exlosions in a dust collector. J. Hazardous
Materials, 12: 87-107.
Moen, I. O., Lee, J. H. S., Hjertager, B. H., Fuhre, R. K. and Eckhoff, R. K., 1982. Pressure
development due to turbulent flame propagation in large-scale methane-air explosions. Comb.
Flame, 47: 31.
NFPA, 1978. NFPA-68. Guide for Explosion Venting. National Fire Protection Assocation, Bos-
ton.
Palmer, K. N., 1973. Dust Explosions and Fires. Chapman and Hall, London.
Tonkin, P. S. and Berlemont, F. J., 1972. Dust explosions in a large-scale cyclone plant. Fire
Research Note No. 942, Fire Research Station.
VDI, 1979. VDI-Richtlinie 3673. Druckentlastung von Staubexplosionen. Verein Deutscher
Ingenieure, Dusseldorf.
Fire Safety Journal, 12 (1987) 153 - 164
Initiation of G r a i n D u s t E x p l o s i o n s b y H e a t G e n e r a t e d d u r i n g Single Impact
between Solid Bodies
GEIR H. PEDERSEN and ROLF K. ECKHOFF
Chr. Michelsen Institute, Department of Science and Technology, 5036 Fantoft, Bergen (Norway)
(Received March 3, 1987; in final form May 25, 1987)
SUMMARY
Several independent investigations o f past
dust explosions in the grain, feed and flour
industries in the U.S.A. and Europe arrive at
the conclusion that 'friction sparks' m a y have
been the initiator o f a substantial fraction, up
to 50%, o f all the explosions recorded. How-
ever, this conclusion has to a large e x t e n t been
based on indirect evidence, and essential de-
tails o f the suspected ignition process most
often remained unconfirmed.
In the present investigation a comprehen-
sive series o f single-impact ignition experi-
ments were carried out, by generating explo-
sible dust clouds o f dried maize starch, in the
region o f tangential impact between a moving
body and a stagnant horizontal anvil.
The net energy o f the impact was defined
as the loss o f kinetic energy o f the moving
body during impact. The influence o f the
impact energy on the frequency o f ignition
was determined. It was f o u n d that much
higher energies were required for ignition at
high velocities o f impact than at low ones.
The reason is that any moving object colliding
with a solid surface, reduces the ignition sensi-
tivity o f the dust cloud in the vicinity o f the
impact p o i n t by inducing turbulence.
Details o f the impact and subsequent igni-
tion o f dust clouds were studied using a high-
speed movie camera. The temperatures o f the
metal sparks generated from a range o f differ-
ent metals were measured by a four-wave-
length optical pyrometer. Typical spark tem-
peratures were ~ 1500 - 2 7 0 0 C for mild steel
and ~ 2 0 0 0 - 2 9 0 0 C for titanium.
The overall conclusion from the present
investigation is as follows: steel sparks pro-
duced by single impacts o f net energies o f up
to 20 J, between steel and concrete, steel or
rusty steel, are unlikely to ignite clouds o f
0379-7112/87/$3.50
153
dried maize starch. In the case o f maize starch
o f 10% moisture, ignition is unlikely even
with titanium sparks. Maize starch is amongst
the easiest ignitable dusts encountered in the
grain, feed and flour industry. Also, most
dusts f o u n d in practice will contain some
moisture. Therefore, it seems unlikely that
dust explosions involving dusts o f grain, feed
or flour can be initiated by heat from acciden-
tal single impacts between tramp metal, and
anvils o f metal, corroded metal, stone or con-
crete, unless the net impact energies are much
higher than 20 J.
1. BACKGROUND AND MOTIVATION FOR THE
PRESENT WORK
Whether or not metal sparks or hot spots
from single accidental impacts between solid
bodies can initiate dust explosions, has re-
mained a controversial issue for a long time.
Several attemps [1 - 4] have been made at
resolving the puzzle with reference to the
grain, feed and flour industry by analysing
past accidents with the objective to identify
the ignition sources. A summary is given in
Table 1.
As can be seen, 'friction sparks' are claimed
to play a significant part. If it is further taken
into account that it is often tacitly implied
t h a t a substantial part of the 'unknowns' may
have been initiated by some untraceable
sources such as metal sparks and electrostatic
discharges, the friction spark becomes the
most suspect of all the potential ignition
sources.
The situations in which metal sparks can be
generated in an industrial process plant fall
into two main categories. The first is grinding
and cutting operations, by which continuous,
dense showers of sparks are produced. The
second is single accidental impacts.
Elsevier Sequoia/Printed in The Netherlands
154
TABLE ]
Percentage of dust explosions in the grain, feed and food industry assumed t,) be initiated i)y 'friction sparks" or
unknown sources
Ref. Number of % ignited by
explosions friction sparks
1 535 20
1 128 17
2 91 54
3 137 9 62
4 83 28
The ability of metal sparks or hot spots
from grinding and cutting to ignite dust clouds
has been demonstrated by several researchers.
The experiments by Leuschke and Zehr [5],
published in 1962, are probably the first ones
in which dust clouds were ignited by grinding
wheel metal sparks. However, no clouds of
organic dusts ignited. Zuzuki et al. [6] ignited
different coal dusts using both sparks and hot
spots from a piece of steel in contact with a
grinding wheel rotating at 23 - 47 m/s peri-
pheral velocity, Later Allen and Calcote [7]
conducted similar experiments in which metal
sparks were generated by pressing a steel rod
against a rotating grinding wheel. By retarding
and focusing the spark stream, it was possible
to ignite clouds of natural organic dusts such
as maize starch and wheat grain dusts. An-
other recent, and very comprehensive study is
Ritter's [8] investigation. This work confirm-
ed that metal sparks generated by grinding,
rubbing and repeated impacts can ignite
clouds of several dusts, including cellulose and
lycopodium.
The second situation, in which metal sparks
may be generated in industry, is accidental
single impacts between solid bodies. In this
case relevant evidence of the ignition poten-
tial has been lacking. However, as long as the
necessary conditions for initiation of dust
explosions by such impacts remain unidenti-
fied, one is forced to maintain the hypothesis
that such sparks may be hazardous in general.
This in turn forces industry to take precau-
tions that may be superfluous, and causes fear
that may be unnecessary.
This unsatisfactory state of affairs created
the motivation for the present work which
has been reported in greater detail elsewhere
% unknown ';;~friction Period
sparks
+ unknown
46 66 1860-1973
27 44 1949-1973
18 72 1941 -1945
71 1958 -1975
5 30-35 1965 -1980
[9, 10]. The specific objective has been to
answer the following question: under what
circumstances of impact energy, impacting
velocity and type of moving object and anvil,
can single impacts between solid bodies ini-
tiate a dust explosion in a cloud of grain dust
in air?
2. EXPERIMENTAL APPARATUS
2.1. Impact generator
After having considered various principles
for constructing a single-impact generator
suitable for laboratory work, it was decided
to build the apparatus shown in Fig. 1. The
length of the rigid, spring-loaded arm, the
'hammer', from the centre of rotation to the
tip of the test object, can be varied by means
of an adjustment screw. The condition for
obtaining impact is that the arm is slightly
longer than the perpendicular distance from
the arm axis to the anvil. By careful adjust-
ment of this small difference it is possible to
generate impacts of different impact energies
without changing the impact velocity. The
cylindrical tip of the test objects that impact-
ed on the anvil material had a reduced diam-
eter of 4 mm. Details of the end of the arm
are shown in Fig. 2. Springs of different
strengths were used for generating a range of
impacting velocities of approximately 8 - 25
m/s. Fine adjustment of the velocity was pos-
sible by adjusting the spring tension.
When performing an impact test, the arm is
first twisted to the start position, whereby the
helical tension spring is stretched. In this
position the arm is locked by a pneumatically
operated device that can be released electri-
cally. After having been released, the arm
 155

will be further retarded by the spring, and

subsequently arrested by an arm catcher.

2.2. Dust dispersion system

Considerable work was carried o u t to devel-
SPRING op a suitable technique for generating the dust
DRIVEN
iMPACTING clouds to be used in the experiments. A dis-
ARM "----~
INTER-
persion system based on a modified version of
CHANGEABLE
METALOBJECT the dispersion cup of the Hartmann bomb
DUST ~ [11] was adopted. The dispersion cup was
CLOUD
....~.
covered by a cap in order to deflect the dust
cloud being expelled from the cup, from the
vertical to the horizontal direction. F r o m the
dispersion cup, the dust cloud flowed into a
rectangular channel consisting of two parallel
side walls and a b o t t o m plate. The height of
the side walls was 20 cm, the length 50 cm,
30 cm
and the spacing between t h e m 12 cm. Hence,
the volume of the dust cloud was approxi-
mately 12 litres. The front side wall was made
of transparent plastic, which made it possible
to photograph the impacts and the ignitions.
In order to establish the combination of
q u a n t i t y of dust, dispersion pressure and delay
time, that gave the o p t i m u m conditions for
ignition, the minimum electrical ignition ener-
Fig. 1. Sketch of the test rig. gy of the dust clouds generated in the 12-1
channel, was determined by means of cali-
.,,...I.- Atugm'humprofde brated electric sparks. The various parameters
- - 3 0 - 30,, 2
were adjusted until an optimal combination
that gave dust clouds of the lowest minimum
ignition energy had been found.
i
; / ~ljustment screw
M16 ~ 1 2. 3. Dust concentration probes
Steel msert The concentration of the optimal dust
clouds as a function of time and space was
measured using a light attenuation probe con-
structed by The Chr. Michelsen Institute
/~ ~ Lock,rignut (CMI) on the basis of the work by the US
Bureau of Mines [12]. The probe, which is
~l ~ Con~c~tocWmcJnut
shown in Fig. 3, is described in detail else-
Test object where [13]. By repeating the dust-cloud gen-
eration process with the dust concentration
Fig. 2. Detail of the arm of the test rig.

accelerates until it reaches its m a x i m u m an-

gular velocity close to the 'six o'clock' posi-
tion. In this position the spring exerts no force
upon the arm and the velocity remains con-
stant for a short period of time. At this point,
impact occurs between the test object at the
tip of the arm and the plane, horizontal anvil.
The anvil is fixed in a strong vice m o u n t e d
rigidly on the test rig. After impact the arm Fig. 3. Light attenuation dust concentration probe.
156
probe in various locations, the concentration
development of the dust cloud with time, at
any desired location, could be determined.
2.4. Timing system
A microprocessor-based 4-channel timer
was used to control the system. The timer
controlled the triggering of the impacting arm,
dispersion of the dust, as well as high-speed
movie camera and other photographic equip-
ment.
2.5. Equipment for measuring the velocity of
the arm and the impact energy
The kinetic energy of the arm at any instant
was calculated from its mass distribution and
instantaneous velocity. By measuring the velo-
city of the arm immediately before and after
the impact, i.e., immediately before and after
'6 o'clock' position, the kinetic energy lost in
the impact could be calculated. Two sets of
reflective opto-switches were used for this
purpose. The opto-switches were mounted on
the wall of the test rig, with narrow clearance
to the passing arm. A small mirror was fixed
to the arm, and an electrical signal was gener-
ated when the mirror passed the opto-switches.
2.6. Photographic equipment
Different kinds of photographic equipment,
including ordinary single-frame cameras, video
camera, 16 mm movie camera and 16 mm
high-speed movie camera, were used. The high-
speed camera (Hycam 4043020) was primarily
used to make black and white negative film
recordings at 1000 and 2000 frames per sec-
ond. The films were developed and examined
at CMI immediately after each experiment.
2.7. Temperature measurements by a four-
wavelength optical pyrometer
In order to measure the temperature of the
sparks, a four-wavelength optical pyrometer
was used [14] (see 'Acknowledgements').
The method is based on the determination of
the relative magnitude of the radiation inten-
sities from the investigated solid body at four
different wavelengths. The intensity ratios are
compared with the radiation intensity ratios
from black bodies of various temperatures. By
identifying the black body temperature that
gives the best fit, a best estimate of the 'colour
temperature' of the solid body is obtained.
The 'colour temperature' is close to the real
temperature of the body ff the spectral emis-
sivity is approximately constant in the range
of the spectrum of wavelengths used in the
measurement.
2.8. Determination of minimum electric spark
energies for ignition of dust clouds
These energies were determined by means
of the method described by Eckhoff [15].
This method, which is based on using electric
sparks of known energies (integration of
power-versus-time curves), is essentially iden-
tical with the new method proposed recently
in W. Germany [16].
3. DUSTS USED IN THE EXPERIMENTS
Since maize starch is one of the agricultural
materials that have often been involved in
dust explosions in the U.S.A. and is one of
the most sensitive to ignition, it had been de-
cided to use this material in the experiments.
Native maize starch supplied from Amylum,
Aalst, Belgium, was used. On receipt the
starch contained 10 - 11% moisture, which
increased the minimum ignition energy con-
siderably compared with that of dried starch.
In order to ensure maximum sensitivity to
ignition, the starch used in the majority of
the present experiments was first dried for
four hours at about 105 C to remove the
moisture.
A few other dusts of different minimum
electric ignition energies (see Table 2), were
used in order to investigate how the ignition
probability in impact ignition tests ~}aried
with the minimum electric spark ignition
energy.
4. CONCENTRATION DISTRIBUTION AND
ELECTRIC SPARK IGNITION SENSITIVITY OF
THE EXPERIMENTAL DUST CLOUDS
4.1. Dust cloud concentration distribution
The first phase of the experiments concen-
trated on formation of a suitable experimen-
tal dust cloud of sufficient volume and low
minimum ignition energy. This was achieved
by dispersing 8 g of dried maize starch at 2 bar
overpressure in the compressed air line. With
8 g of dust, the theoretical average dust con-
centration of the cloud in the 12-1 channel was
 157

TABLE 2
M i n i m u m i g n i t i o n energies a n d particle sizes of t h e 1oo I i(
dusts used in the e x p e r i m e n t s
g zooI
Dust Min. ign. e n e r g y Particle
(electric sparks) diameter
( m J) (/Jm)

Maize s t a r c h , < 4.5 ~ 5 - 20

dried
Maize s t a r c h , 27 - 36 ~5 -20
'88oi
i , t , t , i , i , ,
10 - 11%
o 0.5 lo 1.5 ;~.0
moisture
TIME (r,.)
Lycopodium < 4.5 ~30 Fig. 5. Dust c o n c e n t r a t i o n d e v e l o p m e n t as a f u n c t i o n
Barley p r o t e i n ]3 o f t i m e at l o c a t i o n s 1 t o 4 (see Fig. 4). S m o o t h e d
average curves based o n t h r e e o r m o r e r e p e a t e d mea-
Barley starch 18 - 22 s u r e m e n t s at each l o c a t i o n .
Barley fibre 47 - 59
W h e a t dust 36 - 47
means of calibrated electric sparks. The elec-
trodes were mounted horizontally a fraction
of a miUimetre above the surface of the plain
anvil, at the same location as a hot spot would
DU5~
DUSI
D|SP::RSIO!~I
Cl~D be expected to occur and metal sparks be gen-
CUe'
erated in an impact test. The electric spark
I 2 3 4
was passed 120 - 300 ms after starting the
+ + + + dust dispersion. Both dried maize starch and
lycopodium ignited with 100% ignition prob-
t
ANVIL ability even at spark energies as low as 4.5 mJ.
Fig. 4. A l t e r n a t i v e l o c a t i o n s o f t h e d u s t c o n c e n t r a t i o n However, any object travelling through a
p r o b e d u r i n g c o n c e n t r a t i o n m e a s u r e m e n t s in t h e ex- dust cloud at a high velocity will create tur-
plosion c h a n n e l .
bulence in its vicinity. This is also the case in
the test rig when the arm passes through the
667 mg/1. However, the dust concentration dust cloud. It would be expected, therefore,
was not uniform. Therefore, the dust concen- to be more difficult to ignite the dust cloud
tration development had to be measured as when the arm is passing than when the cloud
functions o f both time and location, using the is comparatively quiescent.
optical dust concentration probe mounted in The influence of the moving arm on the
different locations of the cloud. The various ignition sensitivity of the dust cloud in the
positions of the dust concentration probe dur- vicinity of the point of impact was studied b y
ing these measurements are shown in Fig. 4. means of electric sparks of known energies.
The results from the measurements are given The spark electrodes were again mounted at
in Fig. 5. the same point as a hot spot would be expect-
The dust concentration measurements ed to occur in case of an impact test. However,
showed that at 120 - 300 ms after opening the in these tests the arm used to generate impact
solenoid valve, the dust concentration was sparks was passed through the dust cloud as it
most uniform. At this m o m e n t the dust con- would be in an impact test, only with a small
centration varied from approximately 400 to clearance to the electric spark electrodes. The
800 g/m 3. For this reason the timing system electric spark was discharged 0.3 - I 0 ms after
was always adjusted so the impact occurred the moving arm had passed the '6 o'clock'
at this point in time. position. Three different peripheral velocities
of the arm were used, viz., 9.5, 15 and 22 m/s.
4.2. Ignition sensitivity for quiescent and Results showing probabilities of ignition of
turbulent dust clouds clouds of lycopodium as functions of electric
The minimum ignition energies o f t h e du.st spark energy at different arm velocities and
clouds in the 12-1 channel were measured b y delay times, are summarized in Fig. 6. When
158
1.0
0.9
I 5m/,~
O.B
0.'/
~ 0.6
E, o.s
>,-
O.l.
g o.~
c)
R~ 0.2
0.1
0.1 I 10 100 1000
SPARK ENERGY , mJ
Fig. 6. Effect o f i m p a c t velocity o n ignition sensi-
tivity o f l y c o p o d i u m . Delay b e t w e e n i m p a c t a n d trig-
gering o f t h e electric spark is 0.3 - 10 ms. E n v e l o p e s
enclose t h e e x p e r i m e n t a l points.
the peripheral velocity of the arm was 22 m/s,
the electric spark energy necessary to ignite
the turbulent dust cloud was two orders of
magnitude higher than the 4.5 mJ necessary
to ignite the quiescent dust cloud. At 9.5 m/s
the electric spark energy required for ignition
was approximately one order of magnitude
higher. This relative ranking of the ignition
sensitivities of the clouds only applies when
ignition occurs close to the point of impact.
For spark~luster ignition away from the
impact point, it would be necessary to perform
separate electric spark ignition tests with the
spark gap at the location where ignition takes
place, and with the appropriate delay between
passing the impacting arm and firing the elec-
tric spark.
5. T E M P E R A T U R E S O F M E T A L S P A R K S F R O M
MECHANICAL IMPACTS AND GRINDING
Temperatures of single-impact sparks from
ordinary mild steel (St37), alloy steel and tita-
nium were measured by means of the four-
wave optical pyrometer. The anvil materials
used were rusty steel and concrete. For
comparison, temperatures of sparks from mild
steel in contact with a rotating grinding wheel
were also measured. In the case of the single
impacts the temperatures were measured on
sparks 40 ram 'downstream' of the impact
point and 10 mm above the anvil surface.
Temperatures of titanium sparks were also
measured 140 mm and 240 mm downstream
of the impact point. The arm velocity just
prior to impact was 24 m/s and the net impact
energy about 20 J in all the experiments. In
the case of grinding-wheel sparks, tempera-
tures were measured 40 mm downstream of
the contact point between the steel rod and
the grinding wheel. The peripheral velocity of
the grinding wheel was 29 m/s.
The results are summarized in Table 3. Ap-
proximately 10 tests were carried out for each
combination of test object and anvil material,
using different combinations of detecting
wavelengths in the pyrometer.
TABLE 3
Results f r o m t h e spark t e m p e r a t u r e m e a s u r e m e n t s b y
the four-wavelength pyrometer
Test o b j e c t Anvil material Spark
temperatures
C~c)
Titanium R u s t y steel 2030 - 2900
Mild steel, St37 R u s t y steel 1730 - 2 3 8 0
Alloy steel R u s t y steel 1530 - 1980
Mild steel, S t 3 7 Concrete 2530 - 2730
Mild steel, St37 Rotating 1600 - 1730
grinding wheel
The results show that the highest tempera-
tures were measured on sparks produced from
impacts of titanium against rusty steel, and
mild steel against concrete. The temperatures
of titanium sparks at 140 - 240 mm down-
stream of the impact point were generally
higher than at 40 mm. There was a clear ten-
dency of increasing temperature with increas-
ing length of travel, i.e., increasing burning
time.
The measured temperatures of sparks from
mild steel against concrete were also high. At
40 mm from the point of impact, some steel
spark temperatures were even higher than
some titanium spark temperatures. However,
the boiling point of titanium is higher than
that of iron, indicating that on average the
titanium sparks should attain higher tempera-
tures than sparks from steel. This was also
observed.
The results from the tests with mild steel
as spark-producing material showed a large

variation in the spark temperatures, depend-
ing u p o n the anvil material. The lowest spark
temperatures were obtained in the tests with
the rotating grinding wheel. The observed
temperatures of 1600 - 1730 C were very
close to those for mild steels measured b y
Hardt [17 ], and also in agreement with Wahl's
[18] t w o ~ o l o u r p y r o m e t e r measurement of
temperatures of steel sparks from grinding
wheels.
The temperature measurements showed
that impact sparks from mild steel attained
higher temperatures when concrete was used
as wall material than when rusty steel was
used. This may in part be due to the heat
conductivity of concrete being much lower
than that of steel, causing more of the fric-
tional heat generated during an impact to be
retained in the particles. This m a y provide
more favourable initial conditions for com-
plete combustion of the particles and hence
contribute indirectly to a significantly higher
spark temperature.
The observation that the temperatures of
single-impact steel sparks were generally
considerably higher than of steel grinding-
wheel sparks, is interesting. One reason for
this effect could be the substantial difference
in the contact pressure between metal and
anvil during spark formation (see next Sec-
tion). A higher contact pressure could lead to
transfer of more heat to the particles during
their formation. Although the maximum pos-
sible difference between the a m o u n t s of heat
transferred to a particle is small compared
with the maximum possible combustion heat,
it could significantly influence the extent to
which the particle will oxidize completely.
This, in turn, could result in significantly
different spark temperatures.
6. ESTIMATION OF THE IMPACT P R E S S U R E
The sparks from single impacts b e t w e e n
carbon steel and concrete or rusty steel, gen-
erated in the present investigation, attained
higher temperatures than the grinding wheel
sparks. A possible explanation could be that
the maximum compression pressure in the
impact apparatus was much higher than the
pressure of the metal against the grinding
wheel. Ritter [8] found that the spark tem-
perature increases with increasing contact
pressure to a certain point where the spark
159
temperature becomes constant and indepen-
dent of the pressure. A similar observation
was made b y Allen and Calcote [7] who
found that low pressure sparks were relatively
small and irregular in shape. The irregular
shape indicated that the sparks had not
reached the melting point. The high pressure
sparks were, however, spherical and coated
with oxide, which means that the particles
had melted and burnt.
In the present investigation, the maximum
contact pressure was estimated b y measuring
the static force exerted on the anvil by the tip
of the impacting arm, when the arm was left in
the vertical position. The measurement was
accomplished b y mounting a mechanical load
cell between the tip of the arm and the anvil.
The force was distributed over a circular area
of diameter 4 ram. The static pressure at the
contact area as a function of excess in arm
length is given in Fig. 7.
Pressure.
[MPa)
200
100
025 050 075 100 T [mml
Excess Am Length
Fig. 7. Static pressure at the c o n t a c t area between the
test object and the w a l l material as a f u n c t i o n o f dif-
ference between arm length and perpendicular dis-
tanee f r o m arm axis to anvil plane.
The dynamic pressures during impacts may
be somewhat different from those measured
under static conditions. It seems reasonable to
assume, however, that the dynamic peak pres-
sures in the present investigation were consid-
erably higher than the contact pressures re-
proted by Ritter [8] (10 - 15 MPa) and Allen
and Calcote [7] {about 1 MPa). It is worth
mentioning that the yield strengths of carbon
and low-alloy steels are approximately 25 MPa.
7. I N F L U E N C E OF IMPACT ENERGY AND
IMPACT VELOCITY ON IGNITION F R E Q U E N C Y
The influence of the net impact energy and
the peripheral velocity of the arm on the fre-
160
quency of ignition of maize starch by tita-
nium sparks, was studied. The peripheral
velocity of the arm was varied using different
springs, whereas the net impact energy was
varied by adjusting the excess arm length.
The results are given in Fig. 8. The frequen-
cy of ignition for each combination of arm
velocity and impact energy is based on obser-
vations from 10 or 20 tests. The scatter (+1 std.
dev.) is indicated. At constant velocity of
approach, the frequency of ignition decreases
systematically with decreasing impact energy,
as would be expected. Low-energy impacts
produce fewer sparks than high-energy im-
pacts, and if the number of sparks is low, the
ignition probability is also low.
10
09
08,
07
c06
o
c,
~ 0~
03
02
0t
I i i i i i i t
2 z. 6 8 10 12 ~z. 16
Impact energy
Fig. 8. F r e q u e n c y o f ignition o f clouds of maize
s t a r c h as ,~ f u n c t i o n o f i m p a c t energy at 16 m / s and
24 m/s, for impaet~ b e t w e e n t i t a n i u m a n d r u s t y steel.
The impacts of similar net energies, but of
different velocities of approach, i.e., different
turbulence levels, gave decreasing frequency
of ignition with increasing turbulence. This is
in agreement with the results from the experi-
ments where the minimum electric spark ener-
gy for ignition was determined when the arm
was moving through the dust cloud at the
m o m e n t of spark discharge (Section 4.2).
8. H I G H - S P E E D F I L M I N G
The 1000 - 2000 frames per second high-
speed films confirmed that impacts with mild
steel against rusty steel or concrete produced
very few sparks compared with titanium im-
pacts. Typical numbers of sparks produced
from mild steel were 5 - 50. Immediately after
impact, no or very few sparks were visible.
But 3 - 5 ms after impact, some sparks became
visible, and after 10 - 20 ms the number of
sparks visible on the f i l l was at maximum.
Impacts with steel never generated a sufficient
number of sparks to ignite clouds of dried
maize starch.
The films of ignition by sparks from tita-
nium against rusty steel showed that the dust
clouds could be ignited in two different ways.
(Reference 9 contains reproductions of some
of the film pictures.) First, ignition frequently
occurred very close to the point of impact,
immediately (i.e., 0 - 5 ms) after the impact
had taken place. At that m o m e n t an extremely
luminous hemispherical 'pocket' close to the
impact point, with a radius of a few centi-
meters and containing 100 - 1000 sparks, was
observed. In addition to the great number of
discrete particles, a 'thermite' flash probably
also contributed to the formation of the lu-
~ m/s minous pocket. Impacts of titanium against
concrete also generated high spark densities
close to the point of impact, but not the addi-
tional continuous, luminous phase as was seen
with impacts against rust. Flame propagation
started somewhere in the region of the 'lumi-
nous pocket', but could neither be referred
specifically to one single metal particle nor to
i i
18 ~0 the hot spot generated at the point of impact.
All the high-speed films of ignition of maize
starch by titanium impacts against rusty steel
showed that ignition and flame propagation
started in this region.
The second mode of ignition was only ob-
served in some of the experiments with lyco-
podium. In this case ignition occurred 1 0 - 50
ms after impact at a distance of 10 - 30 cm
downstream of the point of impact. In this
region the number of sparks per unit volume
of dust cloud was considerably lower than in
the 'luminous pocket' close to the point of
impact. But still it was not possible to trace
the ignition back to any single metal spark.
'Explosion' and fragmentation of single metal
sparks were observed frequently, but never
seemed to cause ignition of the dust cloud.
The high-speed films suggest that a cluster
of a large number of sparks is necessary to
cause ignition of a cloud of grain dust. Im-
pacts with titanium as spark-producing

material generated such conditions, and usual-
ly resulted in ignition.
9. RESULTS FROM IGNITION TRIALS WITH
DIFFERENT DUST CLOUDS AND IMPACTING
MATERIALS
A series o f impact tests with various mate-
rials was carried o u t in clouds o f dried maize
starch in order to identify combinations o f
impacting materials t hat would cause ignition.
The anvil materials used comprised typical
c o n s t r u ctio n steels with either fresh or rusty
surfaces, and different concrete ~ m p l e s . Sup-
p l e men tar y tests were c o n d u c t e d with anvils
o f different minerals and grinding wheels.
T he steel alloys used in the experiments
were obtained f r o m the CMI workshop. Pre-
cise names or compositions o f the alloys were
n o t available, b u t t h e y are likely to be repre-
sentative o f the range of steels in c o m m o n
use. Th e combinations o f steel test objects
and anvil materials investigated, are given in
Table 4.
TABLE 4
Combinations of steel test objects and anvil materials
used in the experiments
Mild steel against
-- mild steel (fresh surface or rusty surface)
-- concrete (incl. two samples supplied by NGFA)
minerals (granite, gneiss, s'andstone, quartz)
- -
- grinding wheel
-
Acid resistant steel, alloyed steel, hardened steel or
stainless steel against
-- mild steel
-- concrete
-- granite
-- grinding wheel
T he peripheral velocity of t he arm in the
test rig just b ef o r e impact was 24 m/s. T he
calculated average impact energy was a b o u t
20 J. After each test, t he anvil was displaced
slightly, so th at every impact t o o k place on
fresh anvil surface.
T he experiments revealed t h a t neither mild
steel against concrete, nor any o t h e r combina-
tions o f steel objects and anvil materials, pro-
duced an y sign o f ignition o f t he dried maize
starch cloud.
F r o m i n t r o d u c t o r y tests with concrete as
anvil material, it was f o u n d t hat the nature
o f the surface was of decisive i m por t ance for
161
the spark f o r m a t i o n process. Impacts by car-
bon steel against a 'fresh' surface containing
a thin layer of cement, produced no visible
sparks at all. However, if t he c e m e n t layer
was removed, so t h a t the impact t o o k place
on t he gravel or crushed stone fraction of t he
concrete, visible sparks were produced. For
this reason experiments using concrete as
anvil material were always carried o u t after
removal of the cem ent layer.
Ignition experiments with impacts between
titanium and several anvil materials were also
carried out. T he combinations o f materials
used, and t he results f r o m the tests, are given
in Table 5. Dust clouds o f dried maize starch
were used in all tests.
TABLE 5
Results from the experiments with impacts between
titanium and different anvil materials
Titanium against Ignition ?
-- rusty, mild steel Yes
-- concrete Yes
-- sandstone Yes
- granite, gneiss, quartz
- No
- grinding wheel
- No
T he experiments showed t h a t visible sparks
were produced f r o m all combinations of im-
pacts between titanium and anvil materials.
T he n u m b e r o f sparks produced d e p e n d e d
u p o n the surface t e x t u r e of t he anvil material.
A rough surface p r o d u c e d m ore sparks t han a
s m o o t h surface, and ignitions were onl y ob-
tained f r o m impacts against anvil materials
having a relatively rough surface.
T he i m port ance of t he surface properties
of t he anvil material was emphasized during
the experiments with titanium against rust y
steel. Surfaces apparently covered by t he
same kind o f rust, would in some cases pro-
duce large numbers o f powerful sparks t h a t
ignited the dust cloud, whereas in o t h e r cases
considerably fewer sparks were produced and
no ignition occurred.
A few experiments with zirconium as spark-
producing material were also carried out.
Rust y steel was used as anvil material. The
impacts produced m ore sparks than titanium
against rusty steel, and ignition of clouds of
dried maize starch were easily obtained.
162
It is worth mentioning that at the very be-
ginning of the present investigation it was
attempted to generate impact sparks by strik-
ing a test object of standard quality alumi-
nium against a rusty steel anvil. However, no
metal sparks or any other luminous thermal
reaction could be observed, only a thin smear
of aluminium was deposited on top of the
rust at the point of impact. This is contrary
to the frequent assumption that impacts of
aluminium against rust will generate energetic
thermite sparks. It should be added, though,
that thermite sparks may be generated if such
a smear o f aluminium on rusty steel is subse-
quently struck by some other object.
In order to produce more steel sparks than
those being produced by single impacts in the
present test rig, a strong rotating cutting disk
was pressed against a piece o f low carbon steel.
A continuous powerful shower of steel sparks
was generated, but dust clouds of lycopodium
and dried maize starch exposed to this heavy
spark shower did not ignite. The reason for
this has not been fully disclosed. It should be
mentioned, though, that the number of sparks
per unit volume of cloud was considerably
lower than in the 'luminous pocket' generated
by titanium impacts.
Finally, the ignition sensitivity of some
further agricultural dusts were compared with
that of dried maize starch and lycopodium
clouds, using sparks from impacts between
titanium and rusty steel. The peripheral velo-
city of the arm was 24 m/s and the net fric-
tional impact energy approximately 20 J. The
minimum electric ignition energy, Emin of the
dusts were measured by calibrated electric
sparks in the Hartmann tube. The results are
given in Table 6.
The results show that the impact sparks
were only able to ignite the dust clouds of
the lowest minimum electric ignition energies.
Apart from dried maize starch and lycopo-
dium, ignition was only obtained with one
dust, with Emin equal to ~ 13 mJ. Maize starch
with 10 - 11% moisture and Emin 27 - 36 mJ
did not ignite. This fairly consistent correla-
tion of minimum electric spark ignition ener-
gies and the frequency of ignition by impact
sparks should probably not be generalized
beyond the group of natural organic sub-
stances to which all the dusts tested belong.
In general, the relationships are likely to be
more complex [8].
TABLE 6
Results from the ignition tests of dust clouds of dif-
ferent 'minimum ignition energy'
Dust Minimum Frequency
electric of ignition
ignition in impact
energy (mJ) tests (%)
Maize starch, dried .~.4.5 100
Lycopodium .: 4.5 100
Barley protein 13 l0
Barley starch 18 - 22 0
Maize starch, 27 - 36 0
10 - 11% moisture
Barley fibre ,17 - 59 0
10. CONCLUSIONS
(1) Metal sparks or hot spots generated in
single impacts between moving steel objects
and anvils of steel, rusty steel, concrete or
minerals failed to ignite clouds of dried maize
starch in air over the entire range of experi-
mental conditions. (Net impact energies of up
to 20 J and velocities of approach from 10 m/s
tp 25 m/s.) Clouds of dried maize starch have
a minimum electric spark ignition energy of
less than 5 mJ, and are probably among the
most ignition-sensitive dusts encountered in
the grain, feed and flour industries.
(2) However, clouds of dried maize starch
could be ignited by single impacts of titanium
on an anvil of either rusty steel, concrete or
sandstone, but if the maize starch contained
about 10% moisture, it could not be ignited
even by titanium impacts.
(3) Single impacts with steel as spark-
producing material produced a very low num-
ber of sparks as compared to the number pro-
duced by titanium under the same impact con-
ditions. The temperatures of individual steel
sparks, however, could reach the same level as
those of titanium sparks (~2500 C). How-
ever, the temperatures of grinding-wheel steel
sparks ate considerably lower, of the order of
1500 - 2000 C.
(4) Impacts of standard quality aluminium
against rusty steel did not generate any sparks
or any other luminous reaction at all, only a
thin smear of aluminium on top of the rust.
Impacts with harder aluminium-containing
alloys were not investigated.

(5) In most cases, ignition by titanium
sparks was observed very close to the point
of impact. However, ignition was occasionally
observed 10 - 30 cm downstream of the im-
pact point. Ignition by one single metal spark
was never observed. A fairly dense cluster of
sparks seemed to be necessary for igniting
the clouds of maize starch.
(6) A n y moving object in the dust cloud
reduces the ignition sensitivity o f the cloud in
the vicinity of the object by inducing turbu-
lence. Experiments with titanium against rusty
steel showed that at a given net impact energy,
the ignition frequency dropped when the im-
pact velocity increased. Thus, at a given net
impact energy, objects generating low turbu-
lence represent a greater ignition hazard than
objects generating high turbulence.
(7) At a given impact velocity, the frequen-
cy of ignition by titanium sparks decreased
systematically with the impact energy, as
would be expected.
(8) The microscopic nature of the anvil sur-
face is decisive for the spark formation pro-
cess. For example, impacts against the gravel
or crushed stone fraction of concrete pro-
duced considerably more sparks than impacts
against a fresh concrete surface covered with
cement.
(9) The overall practical conclusion of the
present investigation is that up to net impact
energies of 20 J, tangential accidental single
impacts between various types of steel, and
between steel and rusty steel or concrete, are
unable to ignite clouds of grain and feed dust,
or flour, even if the dusts are dry. Impacts of
standard quality aluminium against rusty steel
did not even generate any visible sparks. In
the case of titanium or zirconium, the sparks
produced may initiate explosions in clouds of
dried dusts, but not in clouds of dust contain-
ing 10% moisture or more.
ACKNOWLEDGEMENTS
CMI wishes to express its sincere gratitude
to the National Grain and Feed Association
(NGFA), Washington, DC, U.S.A., for the
substantial financial support rendering this
work possible, and for unfailing constructive
advice and support throughout. This paper is
a condensed version o f a comprehensive
report prepared by CMI for NGFA [9].
163
During his stay at CMI in the a u t u m n of
1985, Dr. R. Klemens, Technical University
of Warsaw, gave invaluable help, both theo-
retically and experimentally, with the mea-
surement of metal spark temperatures by a
four-wavelength optical pyrometer borrowed
from TU of Warsaw.
REFERENCES
1 H. de Goijer, Th. M. Groothhuizen and M. E.
Reinders, Literature Investigation into the Dust
Explosion Danger in Industries Storing and Pro-
cessing Cereals and Flour, Report from Tech-
nological Laboratory TNO, October, 1975.
2 D. J. Rasbash, Fires Involving Dusts, Technical
Booklet No. 5, Fire Protection Association,
London, 1949.
3 M. Verkade and P. Chiotti, Literature Survey of
Dust Explosions in Grain Handling Facilities:
Causes and Prevention, Energy and Mineral Re-
sources Institute, Iowa State University, Ames,
IA, 1976.
4 H. Beck and A. Jeske, Documentation Staub-
explosionen, Analyse und Einzelfallderstellung,
Report No. 4/82, Berufsgenossenschaftliches
Institut fiir Arbeitssicherheit, St. Augustin,
F.R.G., 1982.
5 G. Leuschke and J. Zehr, Ziindung yon Staub-
lagerungen und Staub/LuftGemischen dutch me-
chanisch erzeugte Funken, Arbeitsschutz, 6
(1962) 146.
6 T. Zuzuki, S. Takaoka and S. Fujii, The ignition
of coal dust by rubbing, frictional heat and sparks,
Proc. Restricted Int. Conf. of Directors of Safety
in Mines Research, July 1965, Safety in Mines
Research Establishment, Sheffield, England.
7 J. Allen and H. F. Calcote, Grain Dust Ignition
by Friction Sparks, SMS-81-049, National Grain
and Feed Association, Washington, DC 20005,
1981.
8 K. Ritter, Die Ziindwirksamkeit mechanisch
erzeugter Funken gegeniiber Gas/Luft- und
Staub/Luft-Gemischen, Dr.-Ing. Dissertation, Uni-
versit/it Fridericiana, Karlsruhe, 1984.
9 G. H. Pedersen, Initiation of Dust Explosions by
Heat Generated during Single Impact between
Solid Bodies, Fire and Explosion Research Re-
port No. SMS-86-055, National Grain and Feed
Association, Washington, DC, 1986. (Prepared by
Chr. Michelsen Institute, Report No 833310-2,
December 1985.)
10 G. H. Pedersen and R. K. Eckhoff, Initiation of
dust explosions by heat generated during single
impact between solid bodies, Proc. 2nd Int.
Colloquium in Dust Explosions, Jadwisin, Poland,
November, 1986, Report No. 863351-1, CMI,
1986.
11 R. K. Eckhoff, Use of (alP/dr)max from closed-
bomb tests for predicting violence of accidental
dust explosions in industrial plants, Fire Safety
J., 8 (1984/85) 159. (Paper presented at the First
Int. Colloquium on Explosibility of Industrial
164

Dusts in Baranow, Poland, November 8 - 10, 1984.
12 K. L. CashdoUar, I. Liebman and R. S. Conti,
Three Bureau o f Mines Optical Dust Probes, Re-
port o f Investigations No. 8542, U.S. Bureau of
Mines, 1981.
13 R. K. Eckhoff, K. Fuhre and G. tt. Pedersen,
Vented Maize Starch Explosions in a 236 m 3
Experimental Silo, Fire and Explosions Research
Report No. ESV-86-070, National Grain and
Feed Association, Washington, DC, 1986. (Pre-
pared by Chr. Michelsen Institute, Report No.
843307-2, December, 1985.)
14 K. L. Cashdollar and M. Hertzberg, Infrared
Pyrometers for measuring dust explosion temper-
atures, Optical Eng., 21 (1) {1982) 82 - 86.
15 R. K. Eckhoff, Towards absolute minimum igni-
tion energies for dust clouds?, Combust. Flame,
24 (1975) 53 - 64.
16 W. Berthold, Bestimmung der Mindestziindenergie
yon Staub/Luft-Gemischen, Progress Report No.
134, VDI-Verlag, Diisseldorf, 1987.
17 L. Hardt, Temperaturme~ungen an Schleiffun-
ken, Arbeitsschutz, 15 (1954) 430 - 436.
18 H. Wahl, Z. Angew. Phys., 12 {1960) 60 - 62.

Erratum

A simplified characterization of upholstered furniture heat release rates, by V. Babrauskas and

W. D. Walton, published in Fire Safety Journal, 11 (1986) 181 - 192.

In eqn. (4), [style factor] = 1.0 for plain as p r i n t e d o n p. 1 8 9 at t h e t o p o f c o l u m n

primarily rectilinear construction, not 1.5 2.
Journal of Occupational Accidents, 6 (1984) 229-240 229
Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands

DUST EXPLOSION EXPERIMENTS IN A VENTED 500 m3 SILO CELL*

R.K. ECKHOFF and K. FUHRE

The Chr. Michelsen Institute, Fantoftvegen 38, N-5036 Fantoft, Bergen (Norway)

(Received October 16,1983; accepted December 13,1983)

ABSTRACT

Eckhoff, R.K. and Fuhre, K., 1984. Dust explosion experiments in a vented 500 m3 silo
cell. Journal of Occupational Accidents, 6: 229-240.

The general problem of dust explosion venting is discussed. Results of wheat grain
dust, soya meal and maize starch explosions in a vented 500 m3 silo cell are presented
in terms of vent area/P,, correlations. The dramatic increase in P,, caused by strong
turbulence is demonstrated. Comparison of present results and existing venting codes
is made.

NATURE OF THE VENTED DUST EXPLOSION

Basic features

The overpressure P(t) in a vented enclosure, in which a dust cloud deflag-

ration takes place, will always be the net result of two simultaneous, com-
peting processes:

l Burning of the dust, causing the pressure to increase.

l Flow of unburnt and burning dust cloud, and combustion products
through the vent opening, causing the pressure to decrease.

When considering the two competing processes involved, it seems clear that
predicting the rate of heat generation in the enclosure is by far the most
demanding task, and a comprehensive theoretical treatment of this problem
has not yet been undertaken [l] . Appreciation of the decisive influence of
the heat generation rate on Pmax attained in a given vented enclosure, was
the motivation for performing the large-scale silo experiments reported here.

*Presented at the 4th International Symposium on Loss Prevention and Safety Promotion
in the Process Industries organized by the Institution of Chemical Engineers, Harrogate,
England, 12-16 September 1983. Proceedings available from the Book Sales Depart-
ment of IChemE.
230

Factors influencing the heat generation rate

When trying to assess the instantaneous rate of heat production, or the

accumulated released heat at any instant, during a dust explosion in a
vented enclosure, several factors must be considered:

a) chemical composition of the dust, including moisture

b) distributions of particle sizes and shapes in dust, determining the spe-
cific surface of the dust in the fully dispersed state
c) degree of dust dispersion, or of agglomeration of dust particles, deter-
mining the effective specific surface available to the combustion process
in the dust cloud in the actual industrial situation
d) distribution of dust concentration in the actual cloud
e) distribution of initial turbulence in the actual cloud
f) possibility of the generation of explosion-induced turbulence in the still
unburnt part of the cloud (which depends on the location of the
ignition source).

Factors (a) and (b) can be assessed accurately in laboratory tests provided
representative samples are available, whereas factors (c) to (f) are deter-
mined entirely by the actual industrial dust cloud generation process, the
geometry of the confinement and the location of the ignition source. The
latter factors are very difficult to reproduce in laboratory experiments.
However, laboratory tests have indeed demonstrated in general terms the
importance of these factors [ 21.
The conclusion is clear: In future it will be necessary to discard the idea
that the combustion rate in industrial dust clouds can be assumed to depend
on the dust type only, i.e. solely on factors (a) and (b) above (Q-concept).
In addition, the practical circumstances under which the dust cloud is gener-
ated will have to be given careful consideration.

HOW CAN DUST CLOUDS BE GENERATED AND IGNITED IN LARGE SILO CELLS?

This question certainly has more than one answer. In case the main mate-
rial to be stored in the silo is in itself sufficiently fine to give explosible
clouds in air, explosible dust clouds are most likely to arise, at least tran-
siently, somewhere in the silo whenever new material is discharged into it,
whether pneumatically or mechanically. If the main material is coarse, such
as grain, explosible dust clouds may be generated by unburnt dust being
blown into the silo by preceding explosions elsewhere in the plant. Dust
could, for example, be injected through the various openings close to the
silo top. Injection through the hopper exit at the bottom seems a more un-
likely scenario. Another process of dust cloud generation could be that dust
layers, which have accumulated on the inside of the silo wall and roof, are
disturbed and dispersed into a cloud by air blasts or mechanical vibrations
induced, for example, by preceding explosions elsewhere in the plant.
 231

The identification of likely ignition sources is another central problem.

Dust flames from preceding explosions entering the silo through openings
at the top are one possibility. Dispersion of smouldering dust deposits is
another. The possible roles of electrical and mechanical sparks are a topic
of current dispute.

THE EXPERIMENTAL 500 m3 SILO

Eight 500 m3 steel silo cells at Boge, 45 km east of Bergen, were made
available for the present investigation. Most of the experiments were con-
ducted in one of these cells, but after its collapse due to excessive pressure
development in the soya meal test, the instrumentation was transferred to
a second cell , which was then used in the remaining maize starch experi-
ments. A winding staircase extending right to the top of the cell in use per-
mitted easy mounting and inspection of diagnostic instruments at various
desired levels above the silo bottom. In order to enable experiments with
various desired vent areas to be conducted, a strong steel grid was constructed
across the entire top surface of the experimental silo cell, permitting any
desired part of it to be blocked by bolting steel plates to the grid. A cross-
section of the experimental silo cell with point of ignition and diagnostic
instrumentation is illustrated in Fig. 1.
The ignition source used was ca. 50 g of dried nitro-cellulose contained
in a plastic bag and ignited by a pair of electrically fired Ce-Mg (100 J) fuse
heads. The nitro-cellulose flame was fully developed approximately 0.1 s
after firing the fuse heads, and it maintained its full size for approximately
1.0 s. The total energy liberated by complete combustion of the nitro-
cellulose was about 200 kJ.
Four different types of measurements were performed during an explo-
sion :
a) Eight specially constructed dust cloud extraction probes, distributed
throughout the silo volume as indicated in Fig. 1, were used in an attempt at
measuring the dust concentration distribution in the silo just prior to igni-
tion. These probes were used only in the wheat grain dust experiments, and
they were only able to collect the finest particles in the dust cloud.
b) Three piezoelectric Kistler No. 412 pressure transducers were used for
measuring the development with time of the explosion pressure inside the
silo cell. In some tests only the top and bottom transducers were in use.
c) Three narrow-angle optical photo-diode probes were used for detec-
ting the flame front arrival times inside the silo cell.
d) Sixteen mm, 24 frame/s tine recordings of the explosions were taken
from a convenient position on the hillside about 100 m from the silo top.

THE DUSTS USED IN THE 500 m EXPLOSION EXPERIMENTS

Three different dusts were used. The first one was a wheat grain dust
collected in the bag filters of the largest Norwegian grain silo. The second
232

Flame detector 3 Strong steel

grid fitted
,xitil threaded
silo too bolts
surface
"ressure probe 3

Flarie detector 2

iressure nrohe C

Fla:~e detector i

Bolted-steel-vlat~
silo wall

11111tlon sourci' -.

xLorlcrete silo
Sottoll

Fig. 1. Cross-section of the experimental silo cells at Boge, Norway.

type was a soya meal supplied by another Norwegian company, and the third
was a native maize starch quality obtained from Belgium.
X number of samples of the three dusts were taken from a representative
number of bags on site at Boge and transferred to CM1 in sealed containers
 233

for determination of particle size distributions, moisture contents, and

Pmax and in closed-bomb-tests.
(flldth,, Samples were also sent to the
Fire Research Station, UK and Ciba-Geigy AG, Switzerland, for independ-
ent, parallel determinations of Pmax and (dP/dth, in both the Hartmann
bomb and the new 20 1 sphere. The agreement between the results obtained
in the various labs was quite satisfactory. A representative summary of the
results obtained is given in Table 1.

TABLE 1

Properties of dusts used in the 500 m3 silo experiments

Dust type Median Moisture (dP/dt),a, [bar/s1 Pm,, [bar@)1

particle content
size by [ti 51 Hartmann 20-l H.b.120 1 Hartmann 20-l H.b.120 1
weight bomb sphere bomb sphere
&ml

Wheat grain 20 12-15 120 325 0.37 6 7 0.86

Soya meal 50 9 25 175 0.14 4 6.5 0.62
Maize starch 15 10 300 425 0.71 7 7.5 0.93

GENERATION OF DUST CLOUDS IN THE 500 m3 SILO CELL

The method used

It was decided to generate the dust clouds in the silo by simply blowing
into it a known quantity of dust by means of a standard 120 Hp Vac-u-
Vator Model PTA 114144 (Dunbar Kapple Inc., Illinois, USA). As shown in
Fig. 1, the dust jet entered the silo approximately in the silo centre, with
the direction of entry vertically upwards. It is estimated that typical air
velocities during dust injection into the silo were of the order of lo-15 m/s.
To prevent the dust from escaping through the vent opening at the silo top
during the injection period, the vent opening was always sealed with a thin
sheet of plastic. In this way the dust cloud formation process was also kept
independent of the vent area.
In all the experiments, except for the very last one with maize starch, the
dust clouds were practically quiescent at the moment of ignition. This was
achieved as follows: During the dust injection period the blower was oper-
ated at a steady, high speed until a fairly distinct change of its sound indi-
cated that most of the dust had been blown into the silo. At this point the
blower power was reduced to no load, and the automatic system for the
sequential dust concentration sampling, start of Ampex tape recorder, and
ignition source firing was triggered manually by the blower operator. The
duration of this sequence was about 5 s. It should be noted that in the very
last experiment with maize starch, full dust injection was continued right
through the entire explosion event.
234

Dust cloud structures

Although the same dust injection method was used for all three dusts,
the resulting dust clouds in the silo were probably quite dissimilar, because
of differences in the dispersibilities and particle size distributions. In the
case of the wheat grain dust, which is difficult to disperse because of the
high content of fibrous material, considerable quantities of unburnt dust
(up to 100-150 kg) were normally found deposited on the silo bottom
after the explosion. This clearly means that a significant fraction of the dust
being blown into the silo cell was in the form of large agglomerates, able to
settle out of suspension before the dust cloud was ignited. However, this
settling process apparently had a significant homogenizing effect on the
concentration distribution of the remaining cloud of finely dispersed dust,
inasmuch as the automatic dust concentration measurement system revealed
a fairly even distribution of the fine dust fractions throughout the entire
silo volume, both radially and axially.
Contrary to the wheat grain dust, the soya meal was very easy to disperse,
and in this case the dust concentration distribution at the moment of igni-
tion was probably fairly homogeneous throughout. However, no direct
evidence of this is available because of the total collapse of the experimental
silo in the soya meal experiment.
In the case of the maize starch, which is much finer than the soya meal,
but still easy to disperse, there was probably a tendency of the dust con-
centration in the upper part of the silo being somewhat higher than in the
lower part. This is because of the low settling rate in air of the fine maize
starch grains (lo-15 pm diameter). The delayed ignition that was observed
in several of the maize starch experiments is a further indication of the
presence of such a concentration gradient. The dust concentration at the
level of the ignition source was probably too low to propagate the flame at
the moment of firing. However, the plastic bag containing the nitro-cellulose
may have continued to burn for another few seconds, long enough to ignite
the edge of the settling cloud when its concentration close to the silo bottom
had reached a level sufficient for flame propagation.

RESULTS OF EXPERIMENTS WITH INITIALLY QUIESCENT DUST CLOUDS

Comprehensive accounts of the work performed are given elsewhere [ 3,4] .

The present contribution constitutes a brief summary, highlighting some of
the important points.
About 330 kg of wheat grain dust (16 experiments), or 330 kg soya meal
(one experiment only), or 150-200 kg of maize starch (6 experiments) were
blown into the silo in each experiment. Hence, the nominal average dust
concentration was about 660 g/m3 both in the case of the wheat grain dust
and in the case of the soya meal, and from 300-400 g/m3 in the case of the
maize starch. However, in view of what is said about the dust deposits at the
 235

silo bottom in the case of the wheat grain dust, the real average dust concen-
tration in the silo volume was perhaps rather of the order of 400 g/m3.

Explosion pressures

Apart from some characteristic 3-6 Hz oscillations of the order of 5-10

mbar peak-to-peak, probably generated by standing waves in the dust-air
column in the silo, the pressure-time histories were generally one single
distinct bell-shaped peak of about 1 s duration. Two typical examples are
given in Fig. 2.
Figure 3 summarizes all the results obtained with quiescent dust clouds.
Figure 4a gives the pressure-time history of the soya meal explosion leading
to the silo rupture. It seems that the maximum pressure would probably not
have been very much higher than 0.6 bar(g) even if the silo had been able to
withstand the explosion. It is therefore suggested that the conservative
envelope line in Fig. 3, embracing all the CM1 wheat grain dust data, is also

100

0
0 1 2 3 4

Time after ignition [s]

Fig. 2. Pressure-versus-time records obtained during the most powerful of the 500 m
silo explosions with initially quiescent clouds of wheat grain dust and maize starch.
Vent area: 8.6-8.8 m.
236

applicable to both the maize starch and the soya meal containing at least
10% moisture.
Figure 3 also gives data obtained by Pinau et al. [ 5,6] (extrapolated from
100 m3 ) and by Matugek and Stroch [ 71 (500 m3 silo). It is not possible
in the present context to discuss these two other large-scale investigations in
relation to the present work. It should just be indicated that different dust
cloud generation techniques were used, influencing both the dust concentra-
tion distribution, the degree of dust dispersion achieved, and not least the
initial turbulence of the dust cloud at the moment of ignition.

VDI 3673, St1 :

NFPA68(1978), St1
with 'strong' ign.source.
Stat = 0.1 bar(g)

VDI 3673, KSt = 200bar.W

stat = 0.1 bar(g)

:latuiek and ?troch

Polyprooylene. Ipen vent

\
1+1: 103 m3 experiments

dheat grain dust (13Slj a\

0

Soya neal (1981)

Maize starch (1982),

0 quiescent cloud \
Conservative
envelooe line of
Maize starch (13823, Cl41 1981 experiments
turbulent cloud with uheat grain dust

I I I I I I I I I I I I a
0 5 10 l!5

Vent area in tog surface of 500 rn3 silo cm21

Fig. 3. Maximum explosion pressures obtained in 500 m3 silo experiments with three dif-
ferent dust types. Results from two other large-scale test programmes. Vent area require-
ments by two existing venting codes.
 231

For comparison, Fig. 4 also gives the vent area requirements specified by
two codes in current use [8,9]. It should be emphasized that the ignition
source used in the present investigation would indeed be classified as strong
(ca. 200 000 J).

0.6 - Soya meal (330 kg)

Vent area = 2.1 m2
0.5
Pstat of vent cover = 0.15 bar(g)
0.4 - (dP/dt)ma, = 1.6 bar/s

z 0.3

2 0.2

2 0.1
5;
Z 0

ii Time after ignition Is]

2
g 0.6
Maize starch (300 kg) Silo
ruptures 1
g 05* Vent area = 8.8 m2 -7
H
Open vent
z
2 0.4 (dP/dt)max = 14 bar/s
x
w
0.3 Onset of 'abnormal'
pressure rise due
0.2 Onset of to turbulence

0.1 ~~~~sure main pressure rise

0.5 1.0 1.5

Time after ignition Is1

Fig. 4. Pressure-versus-time records obtained during the two explosions that caused
the two silo cells to rupture.

Flame propagation

The vertical flame front speeds in the upper part of the silo were esti-
mated on the basis of the flame-arrival-time data provided by the three
flame detectors. A typical trend for the wheat grain dust explosions would
be a few m/s halfway down in the silo and subsequent acceleration to
238

about 40-50 m/s close to the vent. In the case of the maize starch, the
few data available indicate a slightly higher speed.
It was observed that the duration of the flame signal at a given photo-
diode station was considerably longer than the time required for the flame
to travel from the bottom of the silo to its top. This suggests that the com-
bustion process is to a large extent volumetric, i.e. the flame is very thick.
This opposes idealized models of a thin flame sweeping through the cloud,
which has often been assumed in the past.
Very long burn-out times mean that the maximum rate of heat production
in the silo will occur at the moment when the flame reaches the vent, be-
cause at this moment the quantity of dust that is burning simultaneously is
at the maximum. On this basis one would expect a systematic coincidence
between the moment of flame arrival at the silo top and the occurrence of
the peak pressure in the silo. As Fig. 5 shows, this was in fact observed. On
the other hand it must be added that this coincidence may also be attributed
to other mechanisms, and further clarification is needed before final con-
clusions can be drawn.

6
Time from ignition to flame at photodiode 1.6 m from silo top ISI

Fig. 5. Correlation between data for time of flame arrival at silo top and time of peak of
explosion pressure in silo.
 239

RESULTS OF THE MAIZE STARCH EXPLOSION INVOLVING TURBULENCE

AND CAUSING SILO RUPTURE

The pressure-time history

The pressure-time history, as given in Figure 4b, consists of three phases.

Following an initial small pressure rise of about 10 mbar after about 0.8 s,
the onset of a slightly slower than normal rise of pressure (= 0.17 bar/s)
is observed at 1.3 s. The pressure continues to rise in this normal manner
up to about 55 mbar at about 1.5 s. The overpressure developed at this
point is right in the middle of the range of peak pressures obtained in the
six preceding normal experiments with maize starch (initially quiescent
clouds). However, at this moment the abnormal, very fast pressure rise
begins, resulting in a further increase of the pressure of more than 500
mbar in less than 0.2 s. The steepest part of this phase of the pressure-time
history has a slope of 14 bar/s, i.e. 80 times that in the normal phase.
The maximum overpressure recorded before the silo ruptured was about
580 mbar. From the very high rate of pressure rise at the moment of rup-
ture, it seems reasonable to deduce that had the silo been sufficiently strong
to withstand the explosion, the pressure would probably have risen consider-
ably higher than 580 mbar before reaching its peak.

Flame propagation

The average flame speed between 11.2 m and 6.2 m below the silo top was
ca. 14 m/s. This velocity is of the same order as obtained in the normal
experiments. The flame speed data thus provide a second indication, in
addition to the pressure rise data, of the first phase of explosion develop-
ment in the final experiment being very similar to those in the preceding
six ones where extensive turbulence was absent.

CONCLUSIONS

1. In addition to dust chemistry and fineness, it must be expected

that the properties of the dust cloud in terms of degree of dust dispersion,
and distributions of dust concentration and turbulence in the actual cloud
greatly influence P,, during a vented dust explosion.
2. In the present dust explosion experiments in a 500 m3 silo cell with
initially quiescent dust clouds, significantly smaller vent areas were required
to maintain a given Pmax than those specified in the current VDI and NFPA
codes.
3. However, it was also demonstrated that significant initial turbulence,
in the form of a fairly strong turbulent dust jet, penetrating the upper part
of the silo volume, increased Pmax dramatically, by at least one order of
magnitude, and even beyond the VDI code prediction.
240

4. A new, differentiated philosophy of dust explosion venting is needed,

which must pay appropriate attention to the marked influence on the neces-
sary vent area, of the actual dust cloud generation, ignition and flame
propagation processes. It is urgently needed to discuss, and agree on, which
dust cloud formation and i~ition processes and associated combustion
rates should, under various circumstances in industry, be regarded as the
worst cases that the vents to be designed should be able to accommo-
date.
5. Further realistic large-scale experiments should indeed be encouraged.

ACKNOWLEDGEMENT

The authors are indebted to all the sponsors of both phases of the large-
scale experiments undertaken, and to all persons and agencies providing
other support and assistance. The invaluable assistance of M.J. Henery, S.J.
Parker, G. Pedersen and H.G. Thorsen, all CMI, should be specially acknow-
ledged. Sincere thanks are expressed to the population of Boge for their con-
tinued underst~ding and forbearance. Detailed further acknowledgements
are given in refs. [ 3] and [ 41.

o 1983 The Institution of Chemical Engineers

REFERENCES

Eckhoff, R.K., 1982. Current dust explosion research at the CMI. Proceedings of the
International Conference on Fuel-Air Explosions held at McGill University, Montreal,
Canada, 4-6 November 1981, edited by J.H.S. Lee and C.M. Guirao, University of
Waterloo Press, p. 657.
Eckhoff, R.K., 1983. Relevance of using (dPjdt)n,ax data from laboratory-scale tests
for predicting explosion rates in practical industrial situations. Paper presented at the
international VDI Colloquium: Safe Handling of Flammable Dusts in N~rnber~,
26-28 October.
Eckhoff, R.K., Fuhre, K., Henery, M.J., Parker, S.J. and Thorsen, H.G., 1982.,Dust
Explosion Experiments in a Vented 500 m3 Silo Cell. CMI-report No. 813307-1, June
1982. (Available from Chr. Michelsen Institute)
Eckhoff, R.K., Fuhre, K., Henery, M.J., Pedersen, G., and Thorsen, H.G., 1982. Maize
Starch Explosion Experiments in a Vented 500 m3 Storage Bin. CMI-report No.
823307-1, February 1983. (Available from Chr. Michelsen Institute)
Pineau, J., Giltaire, M., and Dangreaux, J., 1976. Efficacite des events. I.N.R.S. Note
No. 1005-83-76 (3807) CDU 614.839.
Pineau, J., 1982.Private communication to R.K. Eckhoff.
Matuxek, Z. and Stroch, V., 1980. Problematik der Staubexplosionen und Massnahmen
gegen Explosionsgefahren in Grossraumbunker fur Schiittgut. Staub-Reinhalt. Luft
40(12): 503-510.
VDI-Richtlinie 3673, 1979. Druckentlastung von Staubexplosionen.
NFPA 68, 3.978. Guide for Explosion Venting.
Powder Technology. 39 (1984) 263 - 276 263

Study of Mechanisms of Grain Dust Explosion as Affected by Particle Size

and Composition.
Part 2. Characterization of Particle Size and Composition of Grain Dust

F_ S. LA1 and D. W. GARRETT

Agricultural Research Service. US. Department of Agriculture. US. Grain Marketing Research Laboratory. 1515
College Avenue. Manhnffan, KS 66502 (U.S.A.)
and L. T. FAN
Department of Chemical Engineering, Kansas State ti-niuersify. Manhattan. A-S 66506 (USA.)

(Received August 16.1983; in revised form December 19.1983)

SUMMARY In order to study the explosibility of grain

dust in relation to particle size and composi-
Three types of grain dust (corn, wheat, and tion, and the effect of these parameters on
grain sorghum) and cornstarch (used as a minimum explosible concentration, maximum
reference) were each diuided into varying size explosion pressure, maximum rate of pressure
fractions (6 to 11) utilizing air and sieve rise, and average rate of pressure rise, it is
classifications_ The particle size distribution necessary to first characterize the particle size
and the composition (content of moisture, and composition of grain dust. Studying the
ash, protein. and starch and fiber) of each size effect of particle size requires the use of a
fraction were determined_ Dust particles sample with the narrowest possible particle
consisting almost entirely of ash material were size distribution. Studying the effect of
found to concentrate in specific air-classified composition requires that moisture, ash,
size fractions_ The total external surface area, protein, and starch and fiber contents be
the total volume, and the coefficient of known for each sample_ This paper discusses
variability were calculated from the experi- theoretical particle size distribution, and
mental particle size distribution for each size describes the collection of dust samples, the
fraction by utilizing a piecewise log normal separation of each sample into size fractions,
approximation_ These values were compared the determination of the particle size distribu-
with those calculated from the least-squares tion of each fraction, the calculation of the
fitted log normal approximation of the actual average particle diameter of each distribution,
distribution. and the determination of the composition of
each fraction. samples of corn, wheat,
Dust
and grain sorghum collected from cyclone
INTRODUCTION dust control systems in commercial elevators
were tested, and a commercial grade com-
A dust explosion is the rapid combustion starch was used as a reference-
of a solid reactant in the form of fine parti-
cles. Thus, the size and composition of the
dust particles are important parameters in THEORETICAL DETERMIN_4TION OF P_4RTICLE
defining the reaction_ Much has been postu- SIZE DISTRIBUTIOX
lated in the literature about the relationship
of particle size and composition to a dust &lass mean diameter
explosion, and some studies have been per- Two properties of the mass mean diameter
formed on various types of dust_ However, D, render it a convenient choice for describ-
little has actually been done to study either ing the average particle diameter of grain dust.
the explosibility of the different size ranges The fiit is that one-half of the total mass of
of particles or the composition of a specific the sample is contained in particles with
type of grain dust. diameters less than the mass mean diameter_

0032-5910/84/$3.00 0 Elsevier Sequoia/Printed in The Netherlands

264

The second is that the mass mean diameter C21as

approaches the geometric mean diameter
based on weight Da.s and the geometric
D
standard deviation eg_s Cl].
The mass mean diameter is determined
from the cumulative weight distribution.
where Ntot is the total number of particles in
Particle size distributions obtained from air
the sample and n(D) is the particle size
and sieve classifications are typically log
number distribution with diameter D as the
normal with varying degrees of distortion at
distributed variable.
the upper and lower ends of the distribution_
D is equivalent to the diameter of the
This distortion depends on the precision of
pa&ie in a monodispersed particulate system
the cIassification_ Typically there exists an
that has the same total external surface area
interval about the cumulative weight percent
A tot as that of a particulate system with a
of 50 that is log normal_ Therefore, all of the
particle size distribution n(D). In other words,
data in this interval can be utilized in deter-
mining the mass mean diameter_ However, t.he
interval must contain at least two data points A tot = Ntot(xDs2 2, = j- rD2n(D) dD (4)
of which one is above a cumulative weight 0
percent of 50 and the other below.
To determine the interval about the cumu- Herdan et cl. 123 have integrated eqn. (3) by
lative weight percent of 50 that is log normal, assuming n(D) to be a log normal distribution
the following transformation is introduced: and obtained

D a.2 = X3 exp(-2 ln2aaa3) (5)

exp -A,2 dX
( 6)
where w is the cumulative weight percent and
2 is the standard normal deviation.
In the case the data in the interval under
consideration are from a log normal distribu-
tion, we have
(1)
To calculate D,_, for a nonlog normal
distribution, the piecewise log normal ap-
proximation was applied by assuming that
log normality exists between two adjacent
data points (see Fig_ 1) The distribution
function of the fraction of the total mass
contained in particles of diameter D, w(D), is
then given by

(2) zO(D)={Wi(D),i=l,2, . . ..m--l}

where D is the particle diameter, D,., is the where

geometric mean diameter for a log normal 6 1 I , ,,,,,
distribution based on weight (the mass mean Z
diameter D,) and a,., is the standard devia- E- _i;
tion for a log normal distribution based on L-
H -0 m - $
weight. Q - -i=
To determine if the data in the interval Z- Q
>
under consideration are from a log normal S /I E
distribution, we established a criterion based E
w /* -P4
on the 95% confidence interval for the o-u-
population correlation coefficient. The confi- !? 0 I 52
dence interval had to contain 0.99; however, z n, -2 v)
the lower boundary could not be less than !E ;A
0.95. 0 DC.
ZLl 1 1 1 ..lli_
3
Mean diameter based on external surface area DlAMETER
The mean diameter based on the external Fig. 1. Piecewiselog normal approximationof the
surface area D,_2 is defined by Herdan et al. particlesize distribution.
 265

D Mean diameter based on mass

z&(D) =
In U~_s.i(*)
1
exp -
In2 -
( 1
D E.3.i
2 ln*U,_s.i
-- I
D
The mean diameter based on mass Da_3 is

Di< Dd Di+i

and m is the total number of data points. D a_3 can be interpreted as the effective
From eqn. (3) we obtain diameter of the particles in the monodis-
persed system that has the same total mass
D a.2 = lV-tot as the actual particulate system with a
112 particle size distribution of n(D)_ Note that
=p($

1-
m-1 ln2Ug_3.i) I rf~_ )
c
i=l D g.3.i
e x+-I - erf(Yi)l

9 (12)
exp 5 ln%,_,_i
m-1
( > Herdan et ai_ 123 have integrated eqn_ (11)
c
i= D 8.3-i
I erf@i+ 1) - erf(Xj)]
for a log normal distribution and obtained
1
~
(7: D a.3 = 4.3 exp[-l-5 ln2u,_3] (13)
where Again the assumption is made that the non-
log normal type of distribution can be ap-

x, =
In -
( )
D z-3-i
D,
In ug.3.i
n =i,i+l
proximated by a distribution that is log
normal between two adjacent data points
Gin UZ *3 fi (see Fig. 1). Substitution of the relation for
n(D) dD in eqn (3) into eqn_ (11) yields
D
In - D a.3
( D g.3.i1
=
ln ug.3.i
Yn = n = i, i i- 1 m-1 1 ;3
G ln Ug.3.i +Jz 2 lerf(&+l) - erf(Zi)l
i=l
As the entire distribution approaches a log

1
1
normal distribution, the geometric mean Dg.3_i 9
and the geometric standard deviation Ug_s_i of
each log normal section in eqn. (7) reduce,
m_, exp -

2
I 2D,,i3
Ill'U,_3_j

[ erf(_Xi+ i) - erf(Xi)]
i=l . .
respectively, to Dg_3 and u~_~, which are J
common for the entire distribution_ The (14)
resultant expression is
where
D a.2 =%.3 exp(-2 ln20g_,) X

X erfb,)
[ erf(x,)
-
-
erf(yd
erf(x,) 1
12
(8) n=i,i+l

In addition, the maximum diameter D,,, and As the entire distribution approaches a log
the minimum diameter Dmin of the distribu- normal distribution, or as
tion will approach infinity and zero respec-
D g.3.i -D
tively- This causes ym and x, to approach EC.3
positive infinity and yr and x1 to approach
ug.3.i - Us3
negative infinity. This gives, from eqn. (S),
eqn- (14) is transformed into
D a.2 = 13iz.3 exp(-2 ln2u,_,)
[i I~~:~] (9)
D,,, = D,., esp(-l-5 ln%,_,) x

or erf(2,) - erf(Z,) l/3

(IS)
D a.2 = Dg.3 exp(-2 ln20,,,) (19) crf(x,) -_erf(x,) .I
266

Furthermore, as log normality is approached, we obtain

D,-m

D,--0

which cause

.%2,X,--
---cm
2*,X1

Thus, eqn (15) becomes

D a.3 = Dg_3 exp(-1.5 ln*a,_s) (16)
Note that eqn. (16) is identical to eqn. (13).

Coefficient of variability
The coefficient of variability, C-V., is calculated from each distribution. It is a measure of the
variability of particle diameters in the distribution about its mean and is defined as

n(D) dD I!*
s (D -6)2N_
tot
C.V. = O (17)
I3

where

(the arithmetic mean diameter)

Finney [l] has shown that when n(d) is log normally distributed, the coefficient of variability is
C.V. = [exp(ln*o,.,) - l]* (18)
When the distribution is approximated by a piecewise log normal distribution (Fig. I), it can be
shown that the coefficient of variability is
C.V. =

I[ 1
1
ln*o,,s,i exp z In20g.3.i
m-1 m-1 ( >
z IerfWi, d - erWi)l C D lerftYi+ d - erf(ri)l
i=l D g.3.i 3 i=l
il P.3.i
-1

1
r m- 1 exp(2 ln20,_s_r)

t i=l D*g.3.i
lerf0.w 1) - erf(~i)l
I
(19)
MATERIALS AND METHODS

Collection of dust samples

Dust samples from corn, wheat, and grain sorghum were collected from storage bins of dust
removal systems as well as from several other locations in each of three commercial grain eleva-

tars. The wheat dust and corn dust samples
were collected from systems that employed
cyclones to separate dust from air, and the
grain sorghum dust samples were collected
from a system that used a baghouse to make
the separation_ Each sample was 2 to 3 kg in
weight, and a sieve with a 1.0 mm mesh
opening was used to remove very large trash.
One hundred pounds of cornstarch were
obtained in bulk from a mill (General Mills,
Minneapolis, MN).
Separation of dust samples into size fractions
A 250 Tyler mesh sieve was used to initial-
ly divide each dust sample into a coarse
fraction (having particle diameters approxi-
mately greater than 61 pm) and a fine frac-
tion_ This separation was performed because
the series 6000 Microparticle Classifier used
to further separate the fine fraction of each
sample could not effectively classify grain
dust particles with diameters larger than
61 pm..
The series 6000 Microparticle Classifier
(manufactured by A.E. Bahco in Sweden and
distributed by Harry W. Dietert Co_, Detroit,
MI) was employed to separate grain dust
particles with diameter less than 61 ,xm_ The
classifier used the combined effects of cen-
trifugation and elutriation to separate O-02 kg
of dust into a fine fraction and a coarse
fraction. The dust particles were subjected to
a centrifugal force which was opposed by a
current of air. The fine fraction, composed of
dust particles with a terminal velocity less
than the air velocity, was blown into a collec-
tor_ The remaining dust, the coarse fraction,
was thrown by centrifugal force into another
collector.
The fine fraction from the 250 Tyler mesh
sieve of the wheat dust, corn dust, graili
sorghum dust, or cornstarch was further
separated into eight size fractions using the
microparticle classifier_ Size fractions were
obtained by performing a series of separa-
tions, each with a progressively higher air
velocity_ The finest size fraction was fist
separated out of the entire fine fraction by
the lowest air velocity. After increasing the air
velocity, the next fraction was divided again
into a fine fraction and a coarse fraction. This
was repeated until eight size fractions were
obtained; however, the three size fractions
with the finest particies were combined into
26-i
one because each individually did not provide
enough material for an esplosion test.
The coarse fraction from the 250 mesh
sieving of each dust sample was also further
divided into size fractions with a series of
sieves having Tyler mesh numbers of 65,115,
150, li0, and 200 (corresponding openings of
208, 124, 105, 88, and 74 pm). These sis
fractions were accumulated from O-100 kg
portions, which were sieved for 15 min
on a Ro-Tap shaker (W_ S_ Tyler Company,
Cleveland, OH)_ The fraction on top of the
65 mesh sieve was considered trash and
discarded because it coniiained a wide range of
particle sizes. Thus, each type of dust was
separated into 11 size fractions by utilizing
the same air velocities for air classification
and the same series of sieves for sieve classifi-
cation Note that size fractions 1 through 6
were from the air classification and size
fractions 7 through 11 were from the sieve
classification_ During sieve classification,
a large degree of carry-over of particles
with diameters smaller than the sieve aper-
tures occurred unless the sieving was done
carefully_
Determination of particle size distribution
The particle size distribut.ion in each of the
eleven size fractions was determined by the
AACC method SO-10 [ 3 3, namely, the Whitby
sedimentation method which classifies parri-
cle size hydrodynamically_ By centrifugal
sedimentation, dust is allowed to settle in a
capillary tube filled with a liquid termed the
sedimentation liquid_ The diameter obt.ained
corresponds to the diameter of a sphere that
falls with the same velocity as the real parti-
cles_ Though the physical dimension of the
particles obtained by this method might be
different from those of the real particles in
many cases, the distribution is one for spheres
that behave hydrodynamically and is obtained
by measuring the cumulat.ive volume of dust.
Benzene was used as the sedimentation
liquid for the corn, grain sorghum, and wheat
dust; isopropyl alcohol was used for the
cornstarch. The dust was initially dispersed in
a feed solution consisting of the sedimenta-
tion liquid and naptha, which was then placed
on top of the sedimentation liquid in the
tube_ To decrease the settling time, the
particles were centrifuged for increasing
lengths of time at 600,1200, and 1800 rpm.
268
The weight percent of the total dust sample
that had settled out was determined by
measuring the height of the settled dust
column. The diameter of the largest particles
that settled out was determined from the rpm
of the centrifuge and the length of time the
sample was centrifuged.
Determination of composition
The composition of each size fraction was
characterized by determining its content of
moisture, ash, protein, and starch and fiber.
The weight fraction of moisture, ash, and
protein was determined by the AACC meth-
ods 44-40,08-01, and 46-10, respectively [3].
The weight fraction of starch and fiber was
obtained from the difference.
RESULTS AND DISCUSSION
Particle size
Particle size distribution . mation for each distribution for data with
The geometric mean diameter D,q3 and the
natural logarithm of the geometric standard
deviation, In (z&~,of the log normal approxima-
tion of the actual particle size distribution for
data with 121 < 2 were calculated. The
coefficient of determination R2 and the 95%
confidence interval for the population correla-
tion coefficient pr are also calculated (data
not shown). Twenty-one of the 43 particle
size distributions are not complete; they do
not contain data at the two extreme cumula-
tive weight percentages, 0% and 100%. How-
ever, 15 out of these 21 distributions are
essentially complete because each of them
spans more than 95% of the total range of
cumulative weight percents. Five of the
remaining incomplete distributions contain no
data between the cumulative weight percents
of 95% and 100%. Among these five distribu-
tions, four are from size fractions of grain
sorghum dust with maximum cumulative
weight percents of 93.5, 91.0, 76.5, and
4&O%, respectively, and one is from a size
fraction of corn dust with a maximum cumu-
lative weight percent of 84.2%. Only one
contains no data with cumulative weight
percents between 0% and 5%. This is from the
size fraction of cornstarch having a minimum
weight percent of 5.4%.
Data at the upper extreme of the distribu-
tions are missing because there exists in the
Whitby sedimentation method a limit on the
maximum particle diameter that can be
measured accurately. The first cumulative
weight percent measurement is recorded at
the moment when particles with a diameter
equal to the maximum diameter have settled
to the bottom of the capillary tube. Particles
with diameters larger than the maximum
dianreter will have already accumulated in the
tube before the first reading, and this could
result in the first cumulative weight percent
reading of less than 100%.
In this case, it is more critical to have data
in the lower, rather than upper, end of a
weight distribution, because the number of
particles per unit weight of dust is larger at
the lower end of the distribution than at the
upper end. All of the distributions, except for
the size fraction of cornstarch, contain data
with cumulative weight percents less than
14.0%.
The parameter of the log normal approxi-
values of 121 < 2 was recorded. Only data
with 121 > 2 were discarded because the
Whitby sedimentation method does not give
accurate results for data in this region [4].
Note that only 16 of the 43 distributions are
from size fractions that were sieve classified
and the remainder are from size fractions that
were air classified. The four sieve-classified
size fractions are from corn dust. The twelve
aircIassified size fractions consist of two corn
dust fractions, six cornstarch fractions, two
wheat dust fractions, and two grain sorghum
dust fractions. Four of the six fractions of
cornstarch, whose size distributions are
essentially log normal, were from non-freeze-
dried samples (this is 2/3 of the total number
of non-freeze-dried size fractions) and the
remaining two are fractions that were freeze
dried (this is l/3 of the total number of
freeze-dried size fractions). Figure 2 repre-
sents the particle size distribution of size
fraction No. 4 of grain sorghum dust, Fig. 3,
fraction No. 2 of cornstarch, and Fig. 4,
fraction No. 1 of cornstarch by freeze drying.
Notice that without freeze drying, cornstarch
can be separated into fraction 1, since parti-
cles in the small fraction tend to agglomerate
when freeze dried. After freeze drying, starch
granules were broken up as indicated in
Fig. 4. The diameter is plotted on a loga-
rithmic axis, and the cumulative weight
 269

99.9 I 1 1 I 1 1 l3 99.9 I I I I I I 1 1 -3

- N
99.0 - E QS.O-
98.0 - / l -+2 , = 9s.o- 0 --rZ
95.0 - _ u =
5 95.0- :
90.0 - J 5
0 90.0- *
z 60.0 - . - +1
80.0 - + iii

s :
d I-
60.0 - 0
S
_ 0 E
z
10.0 - / -I 20.0 -
s ro.oy a

.O;
10.0
5.0
2.0 / -- -2 I-
z
0
5
1.0 *

I I 8 I 1 I t t !
j-3
4 6 6 II 16 22 31 44 62 68

DIAMETER Ivan1

Fig. 2. Particle size distribution of grain sorghum Fig. 5. Particle size distribution of corn dust.
dust.

percent on a probability axis. The cumulative

weight percent is also presented in terms of
the standard normal deviation.
Even though the particle size distribution
of the original samples of corn dust, wheat
dust, grain sorghum dust, and cornstarch are
approximately log normal (Figs. 2 - 5). the
particle size distributions of the size fractions
from the sieve and air classifications are not
20.0
1 necessarily log normal. Herdan et al. [ 2 3 have
illustrated the shape of the resultant size
L
10.0
5.0 distributions from perfect air and sieve
2.0
1.0
classifications (see Fig. 2). Notice that there
b
exists a range of cumulative weight percents
I r-1 1 I I I t t
-3
W around the cumulative weight percent of
4 6 9 I: I6 22 31 44 62 66
DIAMETER (urn) 5070, where the distribution of particle
Fig. 3. Pnrticle size distribution of cornstarch. diameters is log normal, i.e., the slope of the
line tangent to the distribution is constant;
99.9 however, outside this range, the slope of the
I I I I I I 11.
tangent line increases toward infinity as the
99.0 I- i values of W approach 0% at the diameter of
D min and 100% at the diameter of II,,,,,.
The shapes of the distributions in Figs.
2 - 4 are different from that in Fig. 5. The
distributions in Figs. 2 - 4 can be divided into
two categories. The first category contains
those distributions that are log normal for the
entire range of particle diameters: the slope
of the line tangent to the distributions is
constant. The second category contains those
distributions in which, at the lower end, the
01 I 1 I I 1 1
slope of the tangent line first decreases
-3
6 6 II 16 22 31 44 62 68 toward zero as the particle size increases, and
DIAMETER (urn1 then increases toward a constant value. As the
Fig. 4. Particle size distribution of freeze-dried particle size further increases toward the
cornstarch. upper end, the slope decreases and then
 270

TABLE 1
Amount of weight in particles with diameters less than the lower boundary sieve apertures

Size Sieve Stokes equivalent Weight percent less than ds

fraction aperture da diameter of d,. d,
Grain sorghum dust Wheat dust Corn dust
(Pm) (mn) (wt.%)
(wt.%) (wt.%)

7 61 55 60 92 95
S 74 67 62 ss 9s
9 88 79 so 99 9s
10 105 95 >70 100 >s4
11 124 112 >50 >s4

TABLE 2
Amount of weight in particles with diameters greater than the upper boundary sieve apertures

Size Sieve Stokes equivalent Weight percent less than d,

fraction aperture da diameter of da, d,
Grain sorghum dust Wheat dust Corn dust
(w-4 (Pm)
(wt.%) (wt.%) (wt.%)

7 74 67 1s 3 2
8 ss 79 15 5 0.5
9 105 95 (10 < 0.01 t2
10 124 112 <20 c O-01 ~16
11 20s 188 t16

increases again. The shape of the lower
portion of the distribution is the result of a
range of diameters for a relatively large
number of particles. Correspondingly, the
shape of the upper portion of the distribution
is the result of a range of diameters with
relatively small number of particles followed
by a range of larger diameters for a relatively
large number of particles_
The deviations from the shape predicted by
Herdan et al. 123 in the upper end of the
distribution in Figs. 2 - 4 are more pro-
nounced for the air-classified size fractions
than for the sieveclassified fractions. The
,-deviations in the lower region of the distribu-
tion are more pronounced for the sieve-
classified size fractions than for the air-
classified size fractions except for the fifth air-
classified size fraction. Also, the deviations
that occur in the corn dust size fractions are
less than those in the wheat and the grain
sorghum dust size fractions.
When particles are sized by sieving, the
range of particles with diameters in each size
fraction should fall within the apertures of
the bounding sieves. Therefore, the weight of
particles with diameters less than the lower
boundary sieve aperture and the weight of
particles with diameters larger than the upper
boundary sieve aperture can be estimated.
However, to compare methods of sizing
particles, the shape of the particles should be
considered_ Irani and Callis [ 51 have reported
the value of the shape factor relating the sieve
aperture to a stokes diameter to be approsi-
mately 0.9. Sieve openings, transformed to
equivalent stokes diameters utilizing this
shape factor, are shown in Tables 1 and 2. For
each sieve-classified size fraction, Table 1
gives the lower boundary sieve aperture, da=,
used to obtain the size fraction and the
equivalent stokes diameter, dsL, of the aper-
ture; the value of dsL was obtained with the
use of the shape factor of 0.9 [5]_ In addi-
tion, the fraction of the total weight of dust
in the size fraction that is contained in
particles with diameters less than dsL is given.
In Table 2. the aperture of the upper bound-
ary sieve dau of each size fraction is given
with its equivalent stokes diameter dsu. The
fraction of the total weight of dust in the size
fraction contained in particles with diameters
greater than dsu is also presented. Note that
the fraction of the total weight of the dust
 271

TABLE 3
Results from an analysis of variance of the coefficients of variability for size fractions

Source of variation Degrees of freedom Sum of squares Mean square

Treatment
Dust 2 2154.26 1097.13 60.41*
Size fraction 9 4823.30 535.92 29.51*
Interactions
DxS 18 6508.23 361-57 19_91*
Error 30 544.94 18.16 2
:
Total 59 14070-73

*Significant at the 1% level.

ranges from 80% to 100% in a size fraction
that is contained in particles with diaroeters
less than the stokes equivalent diameter of the
lower sieve aperture_ However, the fraction of
the total weight of the dust ranges from only
0.01% to 20% in the size fraction that is
contained in particles with diameters larger
than the stokes equivalent diameter of the
upper boundary sieve aperture_
Small balls of particles could be seen in the
size fractions of wheat dust and grain sorghum
dust. Martin [ 63 found that wheat dust contains
particles called tricombs which range from 50 to
200 firn in length and 10 to 30 pm in diam-
eter, and which have a large length to diam-
eter ratio in the range of 5 to IO. These
tricombs can trap large quantities of smah
particles during the sieving operation and
prevent them from passing through the sieve.
Martin [6] also found that grain sorghum dust
contains hair-like projections_ During the
sieving operation, these projections can
capture small diameter particles to form a ball
that cannot pass through the sieve.
The coefficients of variability range from
22% for the air-classified size fraction of grain
sorghum dust to 104% for the freezedried
size fraction of cornstarch. When examining
the hypothesis that the size fractions are
monodispersed, the smallest coefficient of
variability, 22%, is relatively large; a value
of 10% is the generally accepted level of
variability in an experiment_
To ascertain if there were any significant
differences between the values of the coeffi-
cient of variability for the different types of
dust or for different size fractions, a two-way
analysis of variance for a 2 X 2 fractional
experiment was used in which the treatments
were the type of dust and the size fraction_
The results in Table 3 show that significant
differences at the 1% level do exist both
among the types of dust and among the size
fractions; however, the interactions are also
significant at the 1% level_ The significant
interactions indicate that the type of dust
has different effects on the coefficients of
variability for various size fractions_ Also. the
effect that the size fraction has on the coeffi-
cient of variability is not the same for every
type of dust. However, a difference can still
exist between air-classified size fractions and
sieveclassified size fractions_ Figure 6 shows
that the values of the coefficient of variabilit-y
for all the air-classified size fractions, except
for wheat dust, are consistently lower than
those for the sieve-classified size fractions_
Figure 7 presents the expected particle size
distributions from the perfect sieve or air
classification of a sample of dust originally
having a log normal particle size distribution,
curve A. Three types of classification are
presented I
Fig_ 6_ Correlation between the coefficient of vari-
ability and the size fraction_
272

determined with the use of a shape factor of

Fig. 7_ Particle size distribution expected
perfect air or sieve classification of a dust with a log
normal particle size distribution A [ 21.
a) particles with diameters less than D,,
are removed, curve B,
b) particles with diameters greater than
D max are removed, curve C, and
c) both a) and b) are performed, curve D.
Mass mean diameter
The mass mean diameter D, of each
distribution, along with the natural logarithm
of the geometric standard deviation, In u8, of
the log normal distribution used in determin-
ing the mass mean diameters were calculated.
The coefficient of determination of the
distribution and the.9590 confidence interval
for the population correlation coefficient pr
were also calculated (data not shown). Thir-
teen of the mass mean diameters are from
distributions which are sufficiently non-log
normal that only two pairs of data points
(each pair consisting of two repetitions) could
be used in determining the mass mean diam-
eter. In Table 4, for each sieve-separated size
fraction, the apertures of the bounding sieves
with their geometric mean are presented_ In
addition, the equivalent stokes diameter was
0.9 [ 53. The mass mean diameters and geo-
metric mean diameter of each size fraction of
the three types of dust samples are compared
in Table 4. For each type of dust, the mass
mean diameters of the air-classified size
fractions increase with the size fraction as
expected; however, the sieve-classified size
fractions do not. Table 4 shows the mass
mean dianreters to be consistently lower than
the geometric average of the bounding sieve
from apertures, corrected for particle shape effects.
This is expected due to the large number of
particles with diameters less than the lower
boundary sieve aperture-
Mean diameter based on external surface
area
To calculate the average diameter based on
external surface area, Da2, from the piecewise
log normal approximation of the actual
distribution, a complete distribution is
necessary. Nineteen of the 43 distributions do
not contain cumulative weight percent data
at the upper extreme of 100%. For these
distributions, the log normal distribution
through the two largest data points was used
to estimate the actual distribution in the
region having no data. Four of the 43 distri-
butions do not have data at the lower extreme
of 0%. For these distributions, a log normal
distribution which contained the data with
the smallest cumulative weight percent and
had a geometric standard deviation o9 equal
to that of the log normal approximation of
the entire distribution was used to estimate
the actual distribution in the region where
there are no data.
The values of Da_2 from two log normal
approximations of the actual particle size

TABLE 4
Comparison of the geometric mean diameter of the boundary sieve apertures to mass mean diameters

Size Sieve aperture Geometric Stokes Weight percent less than ds

fraction mean of sieve equivalent
Lower Upper Grain sorghum Wheat dust Corn dust
aperture da diameter of da
(Wn) (wn) dust (wt_%) (u&_%) wt.%)
(wn) (Mm)

7 61 74 67 60 50 34 24
8 74 88 81 73 63 36 24
9 88 105 96 86 64 31 26
10 105 124 114 103 72 32 34
11 124 208 161 145 89 - 60
 273

distribution and one piecewise log normal Note thzt the right-hand side of the above
approximation of the actual distribution were equation does not contain the quantity
calculated for each size fraction of each type N&W,,,- The values of the quantity on the
of dust. One log normal approximation was left-hand side of eqn. (20) were calculated
determined with only those data having from a log normal approximation of the
121 < 2 and the other data having 121-C 3. actual particle size distribution and from the
The coefficients of variability between each piecewise log normal approximation of each
log normal approximation and the piecewise size fraction_ The coefficient of variability
log normal approximation were calculated- between the value of D,_22(Nt,,/W,,,)(6/~~pd)
The data show differences between values of calculated from the piecewise log normal
D a_2,calculated by approximating the actual distribution, and that from the log normal
distribution with a log normal distribution distribution determined from data with
and with a piecewise log normal distribution. [Z[ < 2 for each size fraction, were also
For 15 size fractions, the coefficient of calculated for each size fraction. Note that
variability is greater than 10% when only none of the coefficients of variability are
data with values of 121 < 2 are considered in greater than iO% and only two are greater
the determination of the log normal approxi- than 5%
mation. When the log normal approximation
is determined using data with 121 < 3, only Nean diameter based on mass
eight size fractions have coefficients of The values of D,_, from tzvo log normal
variability larger than 10% This indicates that approximations of the actual particle size
the extreme lower parts of the distribution distribution and one piecewise log normal
with 2 < -2 can be important in the calcula- distribution were calculated for each size
tion of D,,z; it contains a large fraction of the fraction. These two log normal approxima-
total number of dust particles in the size tions are the same as those used for determin-
fraction_ When the estimate of D,,, from the ing Da_2_ The coefficients of variability
log normal approximation is larger than that between the value of D,_, from each log
from the piecewise log normal approximation, normal approximation and that from the
the log normal approximation underestimates piecewise log normal approsimation were
consistently the weight percents of the fine calculated as are the values of the average
particles. When the estimate of D,,, from the diameter based on the mass. In calculating
log normal approximation is lower than that D a_3 for an incomplete distribution, the same
from the piecewise log normal approximation, methods were used as those employed in the
the log normal approximation overestimates calculation of Da_2_ The differences between
consistently the cumulative weight percents the value of Da_a from each log normal
of the fine particles_ approximation and that from the piecewise
The differences between the value of D,_, log normal approximation are similar to those
calculated from the piecewise log normal previously noted for values of D,_,. The
distribution and that from each of the log coefficients of variability for Da_3 are not so
normal distributions can be attributed to dif- large as those for D,_=; the calculation of
ferences in calculating the quantity N,,J W,,,. D a_3 involves the cubic root of .Nt,JW,,, as
It can be shown that opposed to the square root in the calculation
D a.2 = of D,_,. The cubic root reduces the effect of
the differences in _Vt,,jWtOt more than the
1 l/2
square root does_
m_-l exp 2 ln2ag.3.i
( ) .A comparison between the mass mean
Cerf(Yi+. 1) - erf(Yi)l diameter D, in Table 1 and the values of
c
i=l Dgsi
. .
D a_3 for the same size fraction indicate that
19 of the pairs differ only by 5 pm; however,
the remaining differ as much as 68 pm.
Cerf(xi+1) - erf(xi)l Furthermore, even when two size fractions
g.3.i have the s&e mass mean diameter, they can
have substantially different values of D,_,.
(20) The mass mean diameter indicates that one-
274

h a l f o f t h e w e i g h t o f t h e s a m p l e is i n p a r t i c l e s
w i t h d i a m e t e r s less t h a n D r , ; h o w e v e r , i t
c o n t a i n s n o i n f o r m a t i o n o n h o w t h e w e i g h t is
dist~_buted a m o n g the particles. The mass
mean d i a m e t e r does not characterize the
p a r t i c l e size d i s t r i b u t i o n s u f f i c i e n t l y b e c a u s e
two p a r t i c u l a t e s y s t e m s can have a p p r e c i a b l y
d i f f e r e n t p a r t i c l e size d i s t r i b u t i o n s a n d y e t
have i d e n t i c a l mass m e a n d i a m e t e r s . This
d i f f i c u l t y is a f e a t u r e o f a n y g e o m e t r i c m e a n
d i a m e t e r . T h e g e o m e t r i c d i a m e t e r is o n l y o n e
of two p a r a m e t e r s n e c e s s a r y to c h a r a c t e r i z e a
log n o r m a l d i s t r i b u t i o n . T w o d i f f e r e n t log
n o r m a l d i s t r i b u t i o n s can have the same
g e o m e t r i c m e a n d i a m e t e r s Dm a n d y e t can
have d i f f e r e n t g e o m e t r i c s t a n d a r d d e v i a t i o n s
og.
Composition
T h e w e i g h t p e r c e n t s o f m o i s t u r e , ash,
p r o t e i n , a n d s t a r c h a n d f i b e r for each size
f r a c t i o n a r e p r e s e n t e d in T a b l e 5. T h e s t a r c h

TABLE 5
C o m p o s i t i o n o f each size f r a c t i o n

Identification n u m b e r Moisture ~,Veight percent ( w t . % ) Starch and F i b e r

Ash Protein

CNAC-S01 11.7 2.20 6.7 79.40

CNAC-S02 12.1 1.31 4.8 81_79
CNAC-S03 12.1 1.35 4.6 81.95
CNAC-S04 11.9 6.43 6.9 74.77
CNAC-S05 11.3 14.13 8.5 66.07
CNAC-S06 7.3 46.27 6.1 40.33
CNAC-S07 12.4 3.94 7.7 75.96
CNAC-S08 12.6 3.37 6.9 77.13
CNAC-S09 12.5 3.25 6.8 77.45
CNAC',S10 12.6 3.39 7.5 76.51
CNAC-Sll 12.5 3.84 8.4 75.26
WTAC-S01 10.4 5.19 6.7 77.71
WTAC-S02 11.8 9.06 12.1 67.04
WTAC-S03 11.0 6.97 8.5 73.53
WTAC-S04 9.0 14.96 I 1.1 64.94
WTAC-S05 8.8 24.99 12.7 53.51
~VTAC-S06 6.8 44.83 14.0 34.37
WTAC-S07 10.6 6.79 11.9 70.71
WTAC-S08 10.8 5.95 9.4 73.85
WTAC-S09 10.3 5.24 6.3 78.16
MOAC-S01 10.0 7.14 10.0 72.66
MOAC-S02 10.3 5.14 6.5 78.06
MOAC-S03 11.4 4.49 4.3 79.81
MOAC-S04 10.1 11.74 6.0 72.16
MOAC-S05 10.2 22.78 8.7 58.32
MOAC-S06 7.7 30.09 8.3 44.91
MOAC-S07 11.1 9.73 9.7 69.47
MOAC-S08 11.6 6.25 8.1 74.05
MOAC-S09 12.0 5.52 7.3 75.18
MOAC-S10 12.1 5.82 7.3 74.78
CSAC-S02 11.3 0-00 0.0 88.70
CSAC-S03 10.6 0.00 0.0 89.40
CSAC-S04 9.9 0.00 0.0 90.10
CSAC-S05 8.9 0.00 0.0 91_10
CSAC-S06 9.0 0.00 0.0 91.00
CSAC-F01 4.0 0.00 0.0 96.00
CSAC-F02 4.0 0.00 0.0 96.00
CSAC-F03 14.9 0.00 0.0 85.10
CSAC-F04 12.2 0.00 0.0 87.80
CSAC-F05 12.1 0.00 0.0 87.90
 275

TABLE 6 =
Mean, standard deviation, and coefficient of variability of moisture, ash. protein, or starch and fiber content
among the size fractions within each dust

Composition Sample identification Number of Mean. S Standard Coefficient

component samples (a) deviation of variability
(%I (-1

Moisture Wheat 9 9.9 1.5 15

Grain sorghum 11 10-s 1-3 12
Corn 11 11.7 1.5 13
Cornstarch 9 9.9 3.1 3-I
Ash Wheat 9 13s 13.3 97
Grain sorghum 11 11-4 10.6 93
Corn 11 8.1 13.1 162
Protein Wheat 9 10.3 2-i 26
Grain sorghum 11 7.8 1-s 23
Corn 11 6-8 l-3 19
Starch and fiber Wheat 9 66.0 11-l 21
Grain sorghum 11 70.0 10-o 1-I
Corn 11 73.3 11-t 16
Cornstarch 9 90.1 3.4 4

and fibercontentwas obtained by subtracting
thesum oftheweightpercents ofmoisture,
ash, and protein from lOO%, since the dust
was assumed to contain only those com-
ponents_
Table 6 shows the average values of mois-
ture, ash, protein, and starch and fiber con-
tent for each dust. The standard deviation of
the content of each component among the
size fractions of each kind of dust is also given
as we3 as the coefficient of variability of each
component. The standard deviations of
protein content among the size fractions of
each test dust range from 1_3% to 2_7% and
those for the moisture content from 1.3% to
1.5%. This indicates that the protein content
and the moisture content vary only slightly
among the size fractions of the same type of
dust. The standard deviation of 3_4% for
moisture content among the size fractions of
cornstarch indicates that their variability is
larger than those of other types of dust_ The
larger variability for cornstarch is the result
of freeze drying some of the size fractions to
improve their dispersibility.
The ash content and the starch and fiber
content of each type of dust show more
variability bet-ween size fractions than do
the moisture and the protein content. The
standard deviations of the ash content among
size fractions of each type of dust range from
10.6% to 18.3% and those for the starch and
fiber cont.ent from 10.0% to 14.69o. The
standard deviation of the starch and fiber
content in cornstarch, 3_4%, is less than that
for the other kinds of dust. In contrast, the
standard deviation of 2.8% among the average
values of the ash content for wheat dust,
grain sorghum dust, and corn dust, and that
of 3.7% for t.he starch and fiber content
indicate only a slight variability among the
dust types.
Table 7 contains the average values of all
types of dust, the standard deviation among
these average values, and the coefficient of
variability among these average values For -
exh component_ The standard deviation of
the average value for the moisture content of
all dust samples is 1.8% and that for protein
content is 0.9%. The variability of protein
content and moist.ure content among the
different kinds of dust is approximately the
same as or smaller than the variability among
size fractions within each kind of dust.
In Fig. 8, the moisture content of each t-est
dust is plotted against the size fraction_ The
results correlating ash, protein, and starch and
fiber content with size fraction are presented
in Figs. 9 - 11_
MartW and Lai [7] have shown that air
classification of grain dust results in a large
ash content (approximately 40%) -in the
residue size fraction. Figure 9 indicates a
similar trend in ash content for the size
276

TABLE 7
Mean. standard deviation. and coefficient of variability of moisture, ash, protein, or starch and fiber content
among the average values for each dust

Composition Number-of Mean, Z Standard Coefficient

component samples (%I deviation of variability
(96) (%I

Moisture 4 lo-6 0.86 8

Ash 5 11.1 2.84 26
Protein 3 8-3 1.80 22
Starch and fiber 4 75.0 10.60 14

I 2 3 4 5 F6RACT:ON 8 9 10 11
SIZE

Fig. 8. Comparison of the moisture content of each Fig. 10. Comparison of the protein content of each
size fraction_ size fraction_

O-0 CORN
20 - .---a YIEc..T
=---a SORGU

SIZE FRACTION 0 1 I 1 1 1 3 1 1 1 ,
1 2 3 4 5 6 7 8 9 10 ,1
Fig. 9. Comparison of the ash content of each size SIZE FRACTION
fraction_
Fig. ll_ Comparison of the starch and fiber content
of each size fraction.

fractions in this investigation; the residue size
fractions correspond to the sixth size fraction_
The fifth size fraction of each dust contains
large weight percentages of ash (approxi-
mately 25%). For the wheat dust and the
gram sorghum dust, the fourth size fraction
also contains a relatively large ash content
(approximately 11%) when compared with
the ash content of the remaining size fractions
(approximately 4%).
In Figs. 8 and 11, a similar trend-is noted for
the moisture content and for the starch and
fiber content. For these two quantities. the
fourth, fifth, and sixth size fractions exhibit
lower values than those of the remaining
size fractions. The correlation coefficients
between the ash content and each of those
quantities are significant at the 1% level
(Table 8). Table 8 contains the standard
deviations of compositional content among
the size fractions of each test dust that results
when the data from size fractions 6,5, and 4
are removed one at a time. The highly signi-
cant inverse correlation between the ash and
TABLE 8
Simple correlation coefficients for the correlation between compositional componerits

Correlation Correlation coefficients, r

Grain sorghum dust Wheat dust Corn dust Cornstarch

Ash and moisture -O-83** -0.91** -0.97** -

Ash and starch and fiber -O-98** -0.99** -0_99** -
Moisture and starch and fiber O_i6** O-85** 0.9-I** 1-o**
Protein and ash - O.i2* - -
Protein and starch and fiber - -O-82** - -

*Significant at the 5% 1.~4.

**Significant at the 1% level_

TABLE 9
Effect of removing separation numbers 4.5, and 6 in the standard deviation of the composition components
among the size fractions for each type of grain dust

Composition Type of dust Number of Sample removed

component samples
none, -6 6.6 6,5,-Z

Moisture Wheat 9 1.51 1.0 0.8 0.5

Grain sorghum 11 1.30 0.8 0.8 0-S
Corn 11 1.53 0.4 0.3 0.3
Means 0.80 0.9 0-s 0-S
Ash Wheat 9 13.32 6.9 3-1 1-1
Grain sorghum 11 10.60 5.5 2.3 1.6
Corn 11 13.14 3.-i 1.6 1.1
Means 2.83 2.6 2.4 2.1
Protein Wheat 9 2.70 2.6 2.4 2.5
Grain sorghum 11 1.80 l-9 2.0 2.0
Corn 11 1.27 1.3 1.3 1.3
Means 1.80 l-5 1.1 1.3
Starch and fiber Wheat 9 11-10 S-1 5.0 4.2
Grain sorghum 11 10.00 5-9 3.3 3.4
corn 11 1 l-i3 -I_5 )-
-_I 2.6
Means 3.50 3.4 2-s 2.5

moisture content results from the ash material
bc-ing less hygroscopic than the organic grain
dust. The correlations between the starch
and fiber content and the ash content are
significant.
The large standard deviations of the ash
content and the starch and fiber content
among the size fractions of each test dust
have resulted from the large ash content in
size fractions 4, 5, and 6. The standard
deviations of the compositional contents of
the remaining size fractions are given in
Table 9. For all three types of dust, the stan-
dard deviations of the ash content decrease
to less than 2%, and those of the starch and
fiber content to 4.2%. For the moisture
content, the standard deviations decrease to
less than l%_ The standard deviations of the
protein content among the size fractions
exhibit essentially no change.
The correlat-ion coefficients in Table 8
indicate three additional significant correla-
tions. The correlation between the moisture
content and the starch and fiber content is
significant for all types of dust at the 1%
level. The correlation between the protein
content and either the ash content or the
starch and fiber content is significant for only
the wheat dust at the 5% level_ The correla-
tion between the moisture content and the
starch and fiber content is the result of the
difference between the moisture cont.ent and
the ash content and that between the ash
content and the starch and fiber content.
278
The partial correlation coefficient of -0.70
with the effect of ash removed between the
moisture content and the starch and fiber
content indicates no significant correlation at
the 5% level.
Correlation between particle size and com-
position
The contention that composition is depen-
dent on particle diameter was not verified for
particle diameters ranging from 10 to 90 pm.
The correlation coefficients between each of
the compositional components and the mass
mean diameter indicate that it is significant at
the 5% level for all types of dust except wheat
dust. The correlation coefficient between the
mass mean diameter and the moisture content
is -0-74, and that between the mass mean
diameter and the ash content is 0.673. Both
are significant at the 5% level. Figure 6
indicates that the correlation between mass
mean diameter and ash content is significant
only for wheat dust; by coincidence, those
size fractions (4, 5, and 6) which have the
highest ash content due to air classification
also have the largest mass mean diameters.
The same observation was not made for grain
sorghum dust or corn dust. Therefore, the
correlations for grain sorghum dust and corn
dust are not significant at the 5% level. For
wheat dust, the sieve-separated size fractions
do not have the largest mass mean diameters
because they contain a large number of
relatively small diameter particles. This
results in size fractions 4,5, and 6, by coin-
cidence, having the largest mass mean diam-
eters The high ash content of size fractions 4,
5, and 6 indicates that most of the ash is
contained in separate particles consisting
entirely of ash and not in particles of gram
dust. The ash particles are either distributed
in a narrow size range (approximately 40 to
60 pm) or they are of higher density than the
grain dust particles. Toe ash content of -tin
dust does not depend on the size of the
particle_
The significant correlation between mois-
ture content and mass mean diameter of
wheat dust is due to the significant correla-
tion between moisture content and ash
content, as discussed previously_ The wheat
dust size fractions that contain the lowest
moisture content also contain the highest ash
content and happen to have the largest mass
mean diameters_ Again, grain sorghum dust
and corn dust do not exhibit these trends.
CONCLUSIONS
Wheat dust, corn dust, and grain sorghum
dust were similar in composition and in range
of particle diameters_ The ash content and
starch and fiber content of each type of dust
showed more variability between size frac-
tions than did the moisture content and
protein content_ The correlation coefficients
between each of the compositional compo-
nents and the mass mean diameter indicated
that it was significant at the 5% level for all
types of dust except wheat dust.
For complete derivation details of the
equations in this paper, please contact Dr-
F. S. Lai.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the
technical assistance of C. R. Martin of the
US_ Grain Marketing Research Laboratory,
USDA, and Professor R_ S_ Lee, Department
of Physics, Kansas State University.
REFERENCES
1 D. J_ Firmey, J_ Royal Statistical Society. Vol.
VII, SuppZ. No. 2 (1941).
2 G. Herdan, M. L. Smith and XV_ H. Hardwick,
Small Particle Statistics. Butter-worth, London,
1960.
3 American Association of Cereal Chemists. AP-
proved methods of the AACC. The Associati&,
St. Paul, MI, 12th edn., 1979.
C. R. Martin, personal communication, 1980-
H. Heywood, Proc. Symposium on Particle Size
Analysis, London, 1947.
C_ R_ Martin, Characterization of grain dust
properties. Paper No_ 78-3020, American Society
of Agricultural Engineers, 1978.
C. R. Martin and F. S. Lai, 1979, Physical and
chemical composition of grain dust, Paper No.
79-3090, American Society of Agricultural
Engineers, 1979.
Journal of Hazardous Materials, 8 (1984) 223-238 223
Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands

DUST EXPLOSION PROTECTION - A COMPARATIVE STUDY OF

SELECTED METHODS FOR SIZING EXPLOSION RELIEF VENTS

P. FIELD
Building Research Establishment, Fire Research Station, Melrose Avenue, Borehamwood,
Herts WD6 2BL (United Kingdom)
(Received April 20,1983; accepted in revised form July 15, 1983)

summary
Explosion relief vents are commonly used to discharge safely the combustion pro-
ducts of a dust explosion in items of powder handling plant. A problem associated with
this method of explosion protection is the sizing of the vent area; this must be large
enough to prevent explosion pressures from reaching damaging levels but not so large that
the use of vents becomes impracticable. This report compares three common methods of
estimating vent areas by applying explosibility data determined in the Hartmann bomb
and the 20 litre sphere.

Introduction

It is becoming increasingly acknowledged that many materials handled in

powder processing plant are potentially explosible. Dust explosions have
been the subject of several comprehensive studies [l-3] in recent times. In-
dustries handling materials such as agricultural products, foodstuffs, chemi-
cals, pharmaceuticals, plastics, rubber, wood, confectionery, detergents,
soaps, cosmetics, leather, metals, alloys, and coal products are typical of
those that are faced with the risk of a dust explosion.
For a dust to explode it must be dispersed into an atmosphere containing
sufficient oxygen to support combustion, the dust suspension must be sub-
jected to a sufficiently energetic source of ignition, the dust concentration
must fall within the explosible range and the particle size must be sufficiently
small to promote ignition and propagation of flame through the suspension.

Explosion prevention and protection

Wherever combustible powders are handled, explosion prevention and pro-

tection measures are normally applied simultaneously since only under these
circumstances can the frequency and consequences of an explosion be re-
duced to a tolerable level. However, there are a few situations where total
exclusion of ignition sources can be achieved and under these circumstances
additional explosion protection, particularly if it is not reasonably practical,
224

need not be used. Explosion prevention measures largely involve a common

sense approach to the problem and should include the control of dust sus-
pensions by well designed dust collection systems and general good house-
keeping throughout the plant to prevent the accumulation of dust on ledges,
pipe runs and the like. The latter is extremely important since settled dust
less than 1 mm thick over a wide area can provide sufficient volume of sus-
pension for a very serious explosion when dispersed within a building. A
major contribution to the prevention of a dust explosion is the avoidance of
potential ignition sources such as naked flames, hot surfaces, frictional heat,
welding and cutting operations, electric and electrostatic sparks, spontaneous
combustion and incandescent particles. If these are to be avoided it is essen-
tial that plant is designed so that naked flames or sparks cannot come into
direct contact with dust and that all plant is regularly maintained. Inerting
of the atmosphere into which the dust is dispersed is an effective method of
explosion prevention. However, it cannot be applied in many processes and
in those for which it is appropriate it may be expensive.
Explosion protection methods are somewhat limited but may include the
strengthening of plant, and in extreme cases, particularly for small volumes,
the design of vessels with enough strength to contain an explosion. The most
common choice of effective explosion protection is either explosion relief
venting or explosion suppression. The most convenient and economical ex-
plosion protection technique is explosion relief venting and it should always
be considered as the first option. It cannot be used in cases where toxic dust
is involved or if the vent cannot be sited so that combustion products can be
discharged in a safe and environmentally acceptable manner. Under these
circumstances the second option, explosion suppression, should be consider-
ed. (There are circumstances in which both explosion relief venting and ex-
plosion suppression are used together.)

Sizing of explosion relief vents

A major and sometimes controversial problem arises when prescribing the

size of an explosion relief area for a given dust in a given volume. It is essen-
tial that the vent area is large enough to prevent the explosion within the
vessel from exceeding its design strength (Fig. 1). It is equally important for
practical and financial reasons that the vent is not unnecessarily large. Con-
sideration must also be given to the vent cover needed to keep process mate-
rial within the vessel; it needs to be of low inertia but durable and strong
enough to withstand process pressure fluctuations and has to be designed to
open at a predefined pressure allowing rapid and unhindered passage of the
combustion products. For vents that cannot discharge directly to a safe place
it may be necessary to incorporate ducting to lead the discharged combus-
tion products away to a safe area. In order to avoid unwanted back pressure
effects, which could raise the explosion pressure within the relieved vessel
above desirable levels, the ducting should be designed to withstand pressures
 225

A - Unvented explosion
B - Small vent area
C - Large vent area

Design strength

Time

Fig. 1. Typical pressure/time history of a vented and an unvented explosion.

at least as great as the plant to which it is attached, have a diameter or cross-

section at least as big as the relief area, should contain no bends, and should
be as short as possible (not greater than 3 metres). These factors need to be
considered together with the explosibility of the dust when prescribing the
size of the explosion relief vent. The three methods of specifying the vent
area considered in this paper are detailed below.

The vent-ratio method

For many years in the U.K. and U.S.A. the Hartmann bomb apparatus
(Fig. 2), has been used to determine explosion pressure data, the maximum
rate of pressure rise being related empirically to a relief area for a given vessel
by what has traditionally been referred to as the vent-ratio method [ 1,2],
since the required vent is defined in terms of the ratio of vent area to the
volume of vessel being protected (Table 1).
Important fundamental features of the vent-ratio approach are that the
vent areas are prescribed assuming the use of low inertia relief covers and
that the maximum pressure within the vessel being protected is in the range
0.07-0.14 bar (l-2 lbf/in2) and secondly, that discharge ducts, if incorpo-
rated, are not greater than 3 metres in length.
Since the vent ratio is a dimensional parameter, a situation frequently
arises where for large vessels the prescribed vent area is impracticably large
and in fact cannot be accommodated in some cases. For this reason the vent
ratios given in Table 1 have been traditionally accepted for vessels having
226

- Air
 227

TABLE 1

Vent ratios for dusts having maximum rates of pressure rise determined in the Hartmann
bomb

Maximum rate of Vent ratio

pressure rise
(bar/s) (lbf/in* s) (m/m3) ( ftz /fP)

<345 < 5,000 l/6.1 l/20

345-690 5,000-10,000 l/4.6 l/15
>690 >10,000 l/3.1 l/10

volumes up to 30 m3 while for vessels in the range 30-300 m3 the vent ratio is
reduced linearly from l/6 m-l to l/25 m-l . For vessels having volumes in the
range 300---700 m3, particularly if they have large length to diameter ratios, it
is common to prescribe vent areas equal to half the cross-sectional area of the
vessel for dusts having maximum rates of pressure rise less than 345 bar/s
(5000 lbf/in2 s) and equal to the entire cross-sectional area for more
explosible dusts. The latter is applied to very large vessels in excess of about
700 m3 handling any explosible dust.
The vent-ratio approach to relief venting of dust explosions has been criti-
cised for its tendency to prescribe overlarge, uneconomic and in some cases
impracticable relief areas. This criticism has emanated from industries in
which the powder handling plant is relatively strong, certainly being capable
of withstanding pressures in excess of 0.14 bar (2 lbf/in2), possibly as high as
0.7 bar (10 lbf/in2).
It is not unreasonable to argue that in the cases where a vessel is relatively
stronger the vent areas could be smaller than those prescribed by the vent
ratio, allowing explosion pressures within the vessel to exceed the vent-ratio
limit of 0.14 bar (2 lbf/in2).

The cube root law and nomograph method

Until recently, the argument mentioned above was difficult to substantiate
conclusively from practical experience. Although 1ogicaIly it was generally
agreed that smaller vents could be prescribed, the acceptance of this fact
presented problems in precribing the size of the relief area. This has, how-
ever, largely been overcome by the guidelines given in VDI Richtlinien 3673
[4] and the National Fire Protection Association Venting Guide [5] which
are based on the most comprehensive approach to the sizing of relief vents.
These guidelines, which are now accepted in the U.K. by HSE in appropriate
cases, are based on extensive vented explosion experiments in vessels having

Fig. 2a. The Hartmann bomb apparatus.

Fig. 2b. The Hartmann bomb apparatus.(photograph).

228

Kst bar mls F

Pn,ax(bar)
0.2

1
t
L

2
v rttl

I
i/I ij
1% t
I

10 1 01
i
1
.L
!

10

Vent area - lnz Vessel volume - m3

Fig. 3. Nomograph to determine vent areas for combustible dusts subjected to strong ig-
nition conditions: vent owning Dressure 0.1 bar (using KS, values). pmax = maximum pres-
sure obtained in the veskl during venting [ 41.

P,,,(barJ
02
0.4

Vent area - m2

Fig. 4. Nomograph to determine vent areas for combustible dusts subjected to strong ig-
nition conditions; vent opening pressure 0.1 bar. pmax = maximum pressure obtained in
the vessel during venting.

volumes in the range l-60 m3; the work has been extensively published and
is largely attributed to Rartknecht [ 3,6--11 J and Donat [12-141. A series
of nomographs has been derived from this work. Two typical nomographs
are shown in Figs. 3 and 4 from which the vent area for a given vessel volume
can be determined providing that the I& value or the explosion class of the
 229

TABLE 2

Relationship between Kst and explosion class

Kst (bar m/s) Explosion class

0 St 0
> O-200 St 1
> 200-300 St 2
> 300 St 3

dust (see Table 2), the vent cover opening pressure, and the maximum reduced
pressure in the vessel are known. It should be noted, however, that nomo-
graphs have only been derived for plant having a design strength capable of
withstanding a minimum pressure of 0.2 bar (2.9 lbf/in2). The explosion
class of the dust required for the use of these nomographs is not related to
the traditional classification into Group A and B dusts adopted in the U.K.;
the former is a quantitative classification while the latter is qualitative. The
quantitative classification of dusts is related to the cube root law which has
been found to apply to dusts as well as gases [ 31. It should be noted that the
St nomograph (Fig. 4) is determined from the dust explosion hazard classes
whereas the I& nomograph (Fig. 3) is derived from a mathematical approxi-
mation [4] .
The use of St and Kd nomographs for the same conditions may result in
slight differences of the calculated vent area. This anomaly is most notice-
able for Pm, > 0.6 bar; a condition which should not concern the majority
of those in the U.K. powder handling industry. If any doubt exists concerning
vent sizing by the nomograph method, the St nomograph is recommended
as being the most straightforward to use and interpret.

dp
dtrnL vs
1
= K& (1)

where (dp/dt)max is the maximum rate of pressure rise of a dust (bar/s), V is

the volume of the vessel (m3) in which it was measured, and Kh is a constant
for a given dust (bar m/s). This law has been found to hold for vessels having
length to diameter ratios not greater than 5 to 1, and for volumes not less
than 17 litres, this being the minimum volume for which & values for a given
dust were found to be in agreement with those determined in larger vessels
(1 m3) [3,15].
Since there was likely to be a wide range of & values for the many known
combustible dusts, a classification system, shown in Table 2, was developed
in order that explosion protection measures could be applied more simply,
by referring to the explosion class of the dust.
The relationship given in Table 2 is only applicable if the dust has been
tested in the prescribed manner (see later) in a vessel where length to dia-
meter ratio is less than 5 to 1 and where volume is not less than 17 litres.
230

(a) --lgnttion leads

Fig. 5a. 20 litre spherical

explosion apparatus.

Fig. 5b. 20 litre sphere

(photwaph).
 231

The reason is that the Kst values determined in large vessels were obtained
from explosions in which a specific level of turbulence had been selected,
and unless these conditions are reproduced in experiments with other suitable
vessels, Kd values will not equate to predefined large-scale conditions and
cannot therefore be used uncritically for vent sizing via the nomographs.
The initial work which resulted in the new approach to the sizing of
explosion reliefs was carried out mainly in a 1 m3 vessel, but since the opera-
tion of this vessel on a routine basis requires considerable effort and large
quantities of material, it proved necessary to develop a more convenient
laboratory vessel. A series of experiments involving spherical vessels having
different volumes was carried out and it was established that a vessel volume
of 17 litres was the minimum for which I& values could be correlated to the
1 m3 vessel and hence to the nomographs [ 151. This study resulted in the de-
velopment of the 20 litre sphere (Fig. 5) which, if used in the prescribed
manner, enables I& value and the dust class (St O-3) to be determined and
hence the size of relief vent for a given vessel to be estimated.

Rust theoretical method

A theoretical approach to the sizing of explosion reliefs has recently been
derived by Rust [16] which has subsequently formed the basis of a method
for determining relief vent areas for plant handling soap and detergent
dusts [ 171. This method has a potential advantage over the two methods de-
scribed above since it is capable of being applied to both high- and low-
strength plant. The equation derived by Rust and given below takes account
of the explosibility of the dust, the maximum allowable pressure in the
vessel to be vented together with its volume and shape.
Rust equation :
A = 8.35 X lo- F (PV) 2/3K113
(2)
(P.4 Y2
where A = vent area (ft*); F = shape factor, for a rectangular vessel of sides a,
b, c where a > b and a > c, F = 0.65 (k/a * ) l13; P = maximum pressure in test
vessel (Hartmann) (lbf/ft*); V= volume of vessel to be vented (ft3); K = Rust
constant for dust, where K = [(dp/dt)/1613; (dp/dt) = maximum rate of pres-
sure rise (Hartmann) (lbf /in* s); and PA = maximum allowable pressure (lbf /
ft*) of vessel to be vented.

K factor
A fourth method of vent area calculation involving a term known as K
factor deserves a mention. K factor or K equals the area of cross-section of
a vessel divided by the area of the relief vent and has been shown to be
directly proportional to the maximum explosion pressure for gases. The ex-
tent to which this method can be applied to dust explosions depends on how
closely gas and dust explosions can be equated. Difficulties arise in deter-
mining which of the numerous available equations is most suited to a partic-
232

ular problem. This paper is concerned with commonly used venting methods
and although K factor has been used successfully, its use is not wide enough
to warrant a full discussion here. For those requiring detailed information
reference should be made to the literature, which although extensive, has
been summarized [ 1].

Experimental

Different dusts were tested in the Hartmann bomb and the 20 litre sphere
in the manner described below; the data obtained were then used to cal-
culate vent areas by the methods described earlier.

Hartmann bomb apparatus

This apparatus (Fig. 2) was used to determine the maximum explosion
pressure and the maximum rate of pressure rise of the range of dusts included
in this study. An ignition delay (electric spark ignition only) of 60 ms after
dispersion of the dust was included for tests with a number of different dust
concentrations. The dusts were exploded by dispersing different masses of
dust in the range 0.1-3 g by air pressurised at 8.3 bar (120 lbf/in2) from a
reservoir of volume 50 cm3. The ignition source was either 10 kV inductive
electric spark formed between the points of electrodes set 5 mm apart or a
hot filament of about 1000C. The pressure/time history was measured by
a piezoelectric transducer and digitized by a transient recorder, a micro-com-
puter being used to process the required data.

20 litre sphere
In order that K,t values determined in this apparatus (Fig. 5) could be used
to classify the dusts accurately, prescribed conditions of turbulence and igni-
tion had to be met. This involved evacuating the sphere to 0.4 bar (5.8
lbf/in*), pressurising the dust reservoir to 20 bar (290 lbf/in2), and selecting
an ignition delay (time between initiation of dispersion and ignition) of 60
ms. An automatic test sequence ensures that reproducible ignition of the
dust takes place under the same conditions of turbulence. A range of dust
concentrations was examined by placing in the dispersion reservoir masses of
dust in the range 2-100 g and ignition was affected by two 5 kJ chemical
detonators. The pressure/time history was measured in an identical manner
to that of explosion in the Hartmann bomb.

Results

The results of the experiments in the Hartmann bomb and the 20 litre
sphere are given in Table 3; also included are data from the specimen venting
examples given by Rust [16] and the Soap and Detergent Industry Associa-
tion [17].
The purpose of this paper is to compare the three approaches to the sizing
 233

TABLE 3

Explosibility data

Mediana Hartmann datab K Rust Sphere data,

particle P KSt
dP/dt
size (bar m/s)
(Ib/ft) (lb/ins)
(fim)
LscoDodium 27 18375 14355 7.22 X 10 135
Woociflour 29 18375 8584 1.55 x lo8 104
Aluminium 17 14616 9005 1.78 x 10 155
Polyethylene 14 14407 23867 3.32 x 10 135
Polyester 30 15034 6003 5.28 X 10 85
Acrylic 38 16287 8222 1.36 x lo8 124
Iron 32 8561 421 1.82 X 10 11
Zinc 17 12528 1682 1.16 x lo6 35
Benzoic acid 53 14199 10861 3.13 x loa 199
Epoxy 32 17330 8918 1.73 x loa 125
Soya meal 70 12528 1494 8.14 X 10 26
Calcium stearate 23 17957 16704 1.14 x lo9 122
Zinc stearate 23 16287 22707 1.86 x lo9 74
Magnesium stearate 15 15869 23258 3.07 x lo9 135
Tinuvin 22 13990 23258 3.07 x lo9 244
Irganox 88 12320 10701 2.99 x lo* 138
Filter dust [ 16 ] -. 8640 5080 3.2 x lo7 75c
Toilet soap [ 17 ] 15840 2130 2.36 x lo6 5oc
-

aThe median size, d,,, is the 50% size on a cumulative frequence curve. The sieve frac-
tions were determined by a jet sieve.
bHighest maximum values obtained with either coil or spark ignition.
These are likely maximum values.

TABLE 4

Vessel parameters

Vessel Side a Side b Side c Volume F pA

(ft) (m) (ft) (ml (ft) (m) (W (m3) (lb/ft) (bar)

_
Vl 10 3.0 3.3 1.0 3.3 1.0 100 3 0.3015 403 0.2
v2 25 7.6 6.0 1.8 5.0 1.5 750 21 0.237 403 0.2
V3 [17] 31.6 9.6 10.56 3.2 10.56 3.2 3524 99 0.3123 1044 0.5
V4 [16] 28.0 8.5 18.0 5.5 11.0 3.4 5544 157 0.40 403 0.2
v5 50.0 15.2 15.0 4.6 10.0 3.0 7500 210 0.253 403 0.2

of explosion relief vents which were detailed earlier. In order to do this ef-
fectively, the data given in Table 3 have been applied to five vessels of
volumes in the range 3-200 m3; for simplicity the vessels have been taken as
rectangular. Dimensions of the vessels are given in Table 4 together with the
234

TABLE 5

Vent areas determined by vent-ratio. nomograph and Rust methods

~~
Dust Vent-size Vent area (m*) for vessel
methoda
Vl v2 v3 v4 v5
Lycopodium VRb 0.9 7.1 4.0 9.2 8.5
N 0.5 1.7 3.0 6.3 8.5
R 1.6 4.6 10.9 27.9 23.6
Woodflour VR 0.6 4.3 4.0 6.3 8.5
N 0.4 1.5 2.5 5.3 7.0
R 1.0 2.8 6.5 16.7 14.0
Aluminium VR 0.6 4.3 4.0 6.3 8.5
N 0.5 2.0 3.3 8.0 10.0
R 0.9 2.5 5.8 15.0 12.7
Polyethylene VR 0.9 7.1 4.0 9.2 8.5
N 0.5 1.7 3.0 6.3 8.5
R 2.2 6.5 15.4 39.4 33.4
Polyester VR 0.6 4.3 4.0 6.3 8.5
N 0.3 1.1 2.3 4.2 5.5
R 0.6 1.7 4.0 11.1 7.1
Acrylic VR 0.6 4.3 4.0 6.3 8.5
N 0.4 1.6 2.8 6.2 8.0
R 0.8 2.4 5.3 14.7 12.5
Iron VR 0.5 3.6 4.0 6.3 7.0
N 0.2 0.7 1.2 2.5 3.0
R 0.03 0.08 0.2 0.5 0.4
Zinc VR 0.5 3.6 4.0 6.3 7.0
N 0.2 0.7 1.2 2.5 3.0
R 0.15 0.4 1.0 2.5 2.2
Benzoic acid VR 0.9 7.1 4.0 9.2 8.5
N 0.7 2.8 4.8 10.0 15.0
R 1.0 3.0 6.9 17.3 15.0
EPOXY VR 0.6 4.3 4.0 6.3 8.5
N 0.4 1.6 2.8 6.2 8.0
R 1.0 2.8 6.5 16.7 14.0
Soya meal VR 0.5 3.6 4.0 6.3 7.0
N 0.2 0.7 1.2 2.5 3.0
R 0.13 0.4 0.9 2.3 1.9
Calcium stearate VR 0.9 7.1 4.0 9.2 8.5
N 0.4 1.6 2.7 6.3 8.0
R 1.8 5.3 12.5 32.0 27.0
Zinc stearate VR 0.9 7.1 4.0 9.2 8.5
N 0.3 1.0 1.8 4.0 5.0
R 2.3 6.7 15.9 41.0 34.0
Magnesium stearate VR 0.9 7.1 4.0 9.2 8.5
N 0.5 1.7 3.0 6.5 8.5
R 2.3 6.8 18.0 41.3 34.7
Tinuvin VR 0.9 7.1 4.0 9.2 8.5
N 0.8 3.1 5.5 13.0 17.0
R 2.2 6.2 14.7 38.0 32.0
1rganox VR 0.9 7.1 4.0 9.2 8.5
N 0.5 1.7 3.0 6.5 8.5
R 0.9 2.6 6.2 16.0 13.5
Filter dust [16] VR 0.6 4.3 4.0 6.3 8.5
N 0.3 1.0 1.8 4.0 5.0
R 0.4 1.0 2.3 6.0 5.0
Toilet soap [17] VR 0.5 3.6 4.0 6.3 7.0
N 0.2 0.7 1.2 2.5 3.0
R 0.2 0.6 1.5 3.8 3.2

aVR = Vent ratio, N = Nomograph. Fig. 3, R = Rust.

bFor vessels Vl and V2 the vent ratios l/6 m-. l/5 m-l and l/3 rn+ were used. For vessel V3 the vent
ratio l/25 m was used. For vessel V4 the vent ratio l/25 ml was used for dusts having Hartmann
rates of pressure rise less than 10,000 lbf/in s and half cross-sectional area (cross-section = smallest side)
was used as a vent for dusts having higher explosibility. For vessel V6. the smallest vent was given by
half cross-sectional area and was used for weakly explosible dusts (<5,000 lbf/in* s. in Hartmann bomb);
1 I25 rn.- was used for ail other dusts.
 235

maximum allowable pressure, PA, and the shape factor, F, as defined above.
The vent areas required for a particular dust in a given vessel and determined
by each of the three methods are given in Table 5.

Discussion

Since the nomograph approach is the only method considered in this

paper to be derived from extensive experimental data, it is assumed, in the
absence of other proven work, to prescribe vents that are capable of relieving
pressure without damage to plant, etc. The vent areas determined by this ap-
proach are precise and related to the explosibility of the dust, the vessel
volume (vents can be sized for vessels up to 1,000 m3 providing that the
length to diameter ratio does not exceed about 5), the maximum allowable
pressure within the vessel on venting, which must be capable of withstanding
at least 0.2 bar (2.9 lbf/in) and the pressure required to completely open
the vent cover.
The data required for this approach to be employed can only be obtained
from experiments performed in a prescribed manner in appropriate test
vessels, e.g., 20 litre sphere, 1 m3 vessel [ 1, 31. It should be realised that data
from the Hartmann bomb cannot be used directly in the nomograph method.
The vent-ratio approach is empirical, based on maximum rates of pressure
rise determined in the Hartmann bomb, the vent areas being sized depending
on the volume of the vessel (i.e., it is dimensional). The vents are sized on
the assumption that the vessels being relieved are unable to withstand pres-
sure greater than about 0.14 bar (2 lbf/ir?). As a consequence, excessively
large vent areas may be prescribed for plant which either has a relatively
large volume (> 30 m) or is capable of withstanding pressures in excess of
0.14 bar (2 lbf/in2).
It can be seen from Table 5 that, as would be expected, the nomograph
approach generally gives smaller vent areas than the vent-ratio approach for
vessels capable of withstanding at least 0.2 bar (2.9 lbf/in2); a comparison
cannot be made for relatively weak plant since nomographs have not been
published for such plant.
The Rust approach can be compared to the other two methods since its
equation takes into account the maximum allowable pressure in the vessel
during venting which encompasses both relatively weak and strong plant.
As might be expected, this approach is in better agreement with the nomo-
graph approach than the vent-ratio approach, particularly for dusts having
maximum rates of pressure rise in the Hartmann bomb of less than 345 bar/s
(Table 5). For dust having higher rates of rise there is some agreement in the
middle range (345-690 bar/s) but there is poor agreement at higher rates
of rise and for large volumes (> 100 m). Under these latter conditions, ex-
cessively large vents may be prescribed by the Rust approach with the vent-
ratio method providing smaller, more realistically sized vents.
236

Conclusions

For the dusts and vessels considered in this paper the following is con-
cluded.

General
(1) The vents prescribed by the nomograph are generally smaller than those
determined by either of the other methods but they are for plant that is
relatively strong - being able to withstand at least 0.2 bar (2.9 lbf/in*).
(2) For dusts having rates of pressure rise in the Hartmann bomb less than
about 345 bar/s (5000 lbf/in s) being handled in vessels having volumes
up to at least 200 m3, the Rust approach gives vent areas that are in
fairly close agreement with those determined by the nomograph method.
(3) Since soap and detergent dusts normally give rates of pressure rise in the
Hartmann bomb much less than 345 bar/s (5,000 lbf/in s), vents deter-
mined by the Rust method are likely to give satisfactory results; the
vent-ratio method tends to prescribe oversize vents for weakly ex-
plosible dusts (< 140 bar/s) such as soaps and detergents.
(4) For dusts having rates of pressure rise in the Hartmann bomb in the
range 345-690 bar/s (5,000-10,000 lbf/in s), agreement between the
Rust and nomograph methods is unpredictable; the vent-ratio approach
is often in better agreement with the nomograph method in this range.
(5) For dusts having rates of pressure rise in the Hartmann bomb greater
than 690 bar/s (10,000 lbf/in s), vents prescribed by the Rust method
tend to be excessive, particularly for vessels having volumes greater than
about 100 m3. The vent-ratio method prescribed vents that are closer to
those determined by the nomograph method, although still larger.

Application to plant
(1) Strong plant: For plant that is relatively strong, i.e., capable or with-
standing pressures greater than 0.2 bar (2.9 lbf/in), the nomograph ap-
proach should be used if appropriate data are available (i.e. & or St
values for the dust being handled). If these data are not available, and
cannot be determined, the vent-ratio and Rust approaches can be ap-
plied and the smaller vent area prescribed by the two approaches should
be adopted (since both are likely to be larger than necessary).
(2) Weak plant: For plant that is relatively weak, i.e., capable of withstanding
only about 0.14 bar (2 lbf/in*), the nomograph approach cannot be used.
The vent-ratio and Rust approach can again be employed and the smaller
vent area adopted. The work carried out for this paper indicates that the
Rust method can be used satisfactorily for dust giving rates of pressure
rise in the Hartmann bomb up to about 345 bar/s (5,000 lbf/in* s) (larger
vents are likely to be prescribed by the vent-ratio method). For dusts
giving rates of pressure rise in the range 345-690 bar/s (5,000-10,000
lbf/in* s), a useful approach would be to use the smaller vent prescribed
 237

by the Rust or vent-ratio methods. For dust giving rates in the Hartmann
bomb greater than 690 bar/s (10,000 lbf/in* s), vents determined by the
Rust method are likely to be excessively large, particularly for large
volumes. The vent-ratio method is preferred.
(3) General: If the vent area prescribed by the Rust or vent-ratio method can
be accommodated without difficulty or unreasonable burden, this should
be done; since - although it may be larger than necessary - it will more
than adequately cope with the explosion pressure, i.e., it will err on the
side of safety.

Acknowledgement

The author would like to thank Mr A.R. Abrahamsen of the Fire Research
Station who carried out a significant number of the experiments reported in
this paper.

0 British Crown Copyright 1984.

References

P. Field, Dust Explosions, Handbook of Powder Technology, Vol. 4, Elsevier,

Amsterdam, 1982.
K.N. Palmer, Dust Explosion and Fires, Chapman & Hall, London, 1973.
W. Bartknecht, Explosions, Spinger-Verlag, New York, 1981.
Pressure release of dust explosions, VDI Richtlinien 3673, VDI-Verlag GmbH,
Dusseldorf, 1979.
Guide for Explosion Venting, NFPA Code No. 68, NFPA, Boston, 1978.
W. Bartknecht, Flammable gas and dust explosions, Forschungsbericht F45, Bundes-
institut fur Arbeitsschutz, Koblenz, 1971.
7 W. Bartknecht, Report on the investigations on the problem of pressure relief of ex-
plosions of combustible dusts in vessels, Part 1, Staub-Reinhalt. Luft, 34 (1974) 289.
8 W. Bartknecht, Report on the investigations on the problem of pressure relief of ex-
plosions of combustible dusts in vessels, Part II, Staub-Reinhalt. Luft, 34 (1974) 358.
9 W. Bartknecht, The course of gas and dust explosions and their control, in: C.H.
Buschmann (Ed.), Loss Prevention and Safety Promotion in the Process Industries,
Proceedings of the 1st International Loss Prevention Symposium held in The Hague,
Elsevier, Amsterdam, 1974.
10 W. Bartknecht, Explosion pressure relief, Loss Prevention No. 11, AIChE 83rd Na-
tional Meeting, AIChE, Houston, 1977, p. 93.
11 W. Bartknecht, Gas, vapour and dust explosions, fundamentals, prevention, control,
in: International Symposium on Grain Elevator Explosions, Vol. 1, National Mate-
rials Advisory Board, Washington, D.C., 1978.
12 C. Donat, Selectioning and size of pressure relief devices for dust explosions, Staub-
Reinhalt. Luft, 31 (1971) 154.
13 C. Donat, Explosion pressure relief with bursting discs and explosion hatches, paper
presented at 2nd International Symposium on Prevention of Occupational Risks in
the Chemical Industry, Frankfurt, 1973.
14 C. Donat, Pressure relief as used in explosion protection, Loss Prevention No. 11,
AIChE 83rd National Meeting, AIChE, Houston, 1977, p. 87.
238

15 R.
F.S. LAI
U.S. Grain Marketing Research Labomtoxy.
ture, 15I5 College Avenue, Manhattan, KA
D_ W_ GARRETT and L. T. FAN
Department of Chemical Engineering, Kanscrs State Uniuersity. Mnrrhattan. KA 66506 (U.S.A.)
(Received October 13,198l)
1. INTRODUCTION
Research on dust explosions has been con-
ducted since 1917 but the progress has been
painfully slow El] _ Limited work has been
carried out to systematically explore and
quantify the effects of basic parameters
governing the dust explosions_ A dust explo-
sion consists of a very rapid combustion
reaction that liberates heat and results in the
expansion of gases surrounding the dust.
When this expansion occurs in a vessel or en-
closure, the pressure in it rises rapidly, which
may lead to structural damage to the vessel or
enclosure. Two of the basic param eters zue
the diameter and composition of the gram
dust particles.
The overall objective of this work, cuhni-
nating in a series of articles, was to study the
effects of the particle diameter and composi-
tion on the explosibility of grain dust_ Specify
ically, the explosibihty was characterized by
four parameters from Hartmann bomb tests
[ 5]_ These four parameters were the maximum
explosion pressure, P,,, the maximum rate
of pressure rise, (dP/dt)-, the average rate
of pressure rise, (dP/dt),,,, and the minimum
explosible concentration, C,,_ The first in
this series of articles contains a comprehensive
and critical survey of publications related to
the effects of particle size and composition of
gram dust on its explosion characteristics.
This survey is concerned mainly with dust
from grains, including wheat, corn and grain
s-orghum.
0032-5910/!32/0000-0000/302.75
Science an< Education~Administmtion, U.S. Department of Agrhxd-
66502 (l7S.A.)
..
2. EXPERIMENTAL STUDIJ3S
Many of the limited number of publications
on particle size and composition, as they affect
the four explosion characteristics, present
results of experimental studies graphically
and/or tabularly with minimal analysis. The
majority of such publications are from the
U.S. Bureau of Mines [9,11,153 _ Other pub-
lications include empirical correlations of data
IS, 143 -
2.1. Particle size
Price [221 separated oat and corn dust,
wheat flour, wheat flour dust and potato starch
into four different size fractions by elutria-
tion. Particles in a sample from each fraction
were microscopically sized. The average
particle diameter for each fraction was the
arithmetic average of the maximum particle
diameter in the fraction and that in the next
smaller fraction_ The average particle diameters
ranged from 10 to 100 pm. Price used the
standard Bureau of Mines apparatus for in-
flammability to perform experiments. The
standard procedure was altered by using
0.000075 kg (75 mg) of the dust sample
instead of 0.0001 kg (100 mg), and-by dis-
persing the dust sample with pressurized air
instead of oxygen_ The maximum explosion
pressure measurements from the tests were
plotted against the average particle diameter_
The maximum explosion pressure increased
with decreasing average particle diameters;
however, for wheat flour and wheat flour
0 Elsevier Sequoia/Printed in The Netherlands
194
dust with particle sizes less than approxima-
tely 35pm, the maximum explosion pressure
decreased with decreasing particle size_ It was
concluded that particle size was an extremely
important factor in governing the explosibility
of thedust. . of dust on a single graph. Specifically, they
&rtmann and Nagy [9] used a series of
sieves to separate dust into four size ranges:
20 - 65,65 - 100,100 - 200, and greater than
200 mesh_ They calculated the average particle
diameter for each of the size ranges, Le. the
arithmetic mean of the particle diameter
corresponding to the sieves used to determine
the upper and lower bounds of each range.
The geometric mean, however, should have
been used instead of the arithmetic mean
because particle size distributions are charac-
teristically log-normal (Herdan et al. [lo] ).
According to Herdan et aL, the geometric
mean better represents the particle size than
the arithmetic mean for a log-normal distribu-
tion. For each range, Hartmann and Nagy [9]
plotted the minimum explosible concentra-
tion, the maximum pressure rise, and several
other quantities against the average particle
diameter_ They observed that the minimum
explosible concentration decreased with
decreasing particle size; however, for several
kinds of dust, it changed very little when the
particle sizes were less than approximately
102 pm (O-004 in_)_ For example, for cellulose
acetate molding powder with a dust cloud
concentration of 0.005 kg/m3 (0.005 oz/cu.
ft.), the maximum pressure increased with
decreasing particle size but remained essen-
tially constant for diameters smaller than 102
pm (0.004 in.). The maximum and average
rates increased with a decrease in the particle
size.
By sieving, Jacobson et al. [X4] separated
each of eleven types of dust into two or more
size fractions. They obtained twelve fractions
for alumin um, nine for magnesium, five for
corn starch, and two to four for the rest. With
a series of sieves. they further separated each
of the fractions and calculated the average
particle diameters with a m.ethod described by
Dallavalle [4] _ The arithmetic mean of the
respective size openings of a pair in the series
of sieves was multiplied. by .the fnction of
dust retained on the finer of the.two sieves_ :
The average particle_~diameterw& the summa-
tion of-the productfor the entire series of
sieves. Again, the geometric mean should have
been_ used in place of the arithmetic mean
because of the log-norm&y of the particle-
size distributions [lo] ; The average particle
diameters of these fractions ranged from 18
to 171 pm. Jacobson et uL [14] plotted the
results of the experiments for all eleven types
plotted the ratio of explosion characteristics
against the ratio of average particle diameters.
The ratio of explosion characteristics (e.g. the
maximum explosion pressure) is defined as
the explosion characteristics of a given dust
divided by that of a through No_ --. 200 sieve
sample of the same type of dust. The maxi-
mum explosion pressure, as one of the explo-
sion characteristics, showed a slight increase
as the ratio of average particle diameters
decreased_ The ratio of minimum explosible
concentrations decreased with a decrease in
the ratio of average particle diameters for
coarse dust (, 74 pm); however, for fine
dusts, there was only a slight change with the
ratio of average particle diameters. Jacobson
et al. [14] empirically correlated the ratio of
explosibility indexes, In, with the ratio of
average particle diameter, DR_ The hazard of
a dust is related to its ease of ignition and to
the severity of the ensuing explosion. Among
other parameters the ease of ignition may be
considered as a function of the ignition
temperature, minimum ignition energy, and
minimum explosible concentration_ The
severity of an explosion is related to the
pressure and the rate of pressure rise. They
defined the explosibility index as follows:
(1)
where Tigp = ignition temperature, Ed =
minimum ignition energy, P,, = maximum
explosion pressure, (dP/dt),, = maximum
rate of pressure rise, and C,, = minimum
explosible concentration
The relationship obtained by them is:
IR =DR-3
(2)
To determine the particle size distribution
by sieving and calculating the average particle

diameter for -es@. sample, Jacobson.: et a& .-_
[15] used dust samples of-~various.metals
including zinc, iron .and cobalt, that. .were
obtainedfiom.different processes, e.g. milling,
stamping, and atomizing. The effects of aver-
age particle,diameter on the minimum explo-
sible concentration, maximum pressure rise,
maximum rate of pressure rise, and average
rate of pressure rise for atomized aluminum
dust were similar to those reported by
Jacobson et uL [ 14]_
The explosion of coal dust-air mixtures
was studied by Hertzberg et al. [ll] . Dust
was separated into nine size ranges with aver-
age particle diameters from 2.7 to 65 pm. Six
of the nine size ranges were classified by a
Donaldson centrifugal classifier, two by sieve-
classification, and one was unclassified pulver-
ized coal_ The surface mean diameter was
determined for each size range by a Coulter
size analyzer. The experiments were performed
to determine the minimum explosible con-
centration, the maximum pressure, and the
maximum rate of pressure rise in a modified
Hartmann bomb_ The volume of the explosion
chamber was 7.8 liters instead of the standard
1.2 liters and the inside configuration of the
chamber was altered to achieve a more uniform
dust dispersion than in the original Hartmann
bomb. The standard criterion for determining
the minimum explosible concentration is the
lowest concentration of dust that when ignited
will cause a paper diaphram to rupture [ 5]_
The criterion used by Hertzberg et al. [ll]
was the lowest concentration of dust that
when ignited would create a maximum
pressure that reached a value twice that of the
initial pressure just prior to ignition_ They
altered the standard procedure by partially
evacuating the explosion chamber to a pressure
of 0.2 - O-3 atm so that the pressure reached
approximately 1 atm after the air for dis-
persing the dust was injected into the chamber.
Instead of the standard procedure employing
a continuous electric spark, initiated prior to
the injection of the dust, they used electric
matches ignited after a spatially. uniform dust
cloud was achieved. Hertzberg et al. [ll]
observed the minimum explosible concentra-
tion to be virtually independent of particle
size in the size range 2.7 to 65 pm. At dust
concentrations close to the minimum explo-
sible concentration, the maximum pressure
rise and maximum rate of pressure rise did
195
not depend markedly on the particle size.
However, at higb dust concentrations (O-4 -
O-5 kg/m3), these quantities did increase some-
what with decreasing particle diameter_
Eckhoff [S] determined tbe maximum
explosion pressures and the maximum rates of
pressure rise of several types of dust from
agricultural grams and feedstuffs (e.g. corn,
wheat, rye, fish powder, soya meal and potato
starch). The dust samples were obtained from
elevators in different countries such as the
United States and Norway, and from various
locations in the elevators, such as the bottom
of a b-u&et elevator and the dust filter_ TO
determine the particle size distribution,
Eckhoff [6] used a series of sieves for the
coarse (> 74 pm) end of the distriioution and
a Coulter Counter analyzer for the fine end. -4
specific surface area (the envelope surface
area) was approximated for each sample. The
shape of the dust particle approximated by a
sphere, and the corresponding mass average
particle diameter were used as the radius in
determining the particle size distibution.
B.E.T. measurements were conducted to
obtain another approximate average surface
area The Hartmann apparatus was used to
determine the maximum explosion pressure
and maximum rate of pressure rise for each
sample. A relation was obtained between the
maximum rate of pressure rise and the specific
envelope surface area by means of a least-
squares regression line for starch and protein,
Le.,
=KS
where (dP/dt),, = maximum rate of pressure
rise, K = a constant, and S = specific surface
area
Each regression line was forced through the
origin and two data points were used to
determine the slope. Two different sizes of
fish powder represented protein, and two
different types of starch, potato and maize,
represented starch. Eckhoff obtained a good
correlation with slope values of 450 bar-g/
(s m2) for protein and 1250 bar-g/(s m2) for
starch.
Using two different apparatuses Enomoto
[ 81 generated sufficiently uniform dust clouds
over specific concentration ranges. The range
196

of concentration was from 0 to 0.300 kg/m3 occurred at the ratio-of O/100, ie., the sample
for the apparatus -developed by Ishihama ]12] contained no coarse fraction and 100% fine
and from O-300 to 4-000 kg/m3 for the appa- fraction_ For polyethylene dust, the highest
ratus developed by Ishihama and Enomoto value of the maximum explosion pressure and
[13]. Neither of the apparatuses used com- that of the maximum rate of pressure rise
pressed air to disperse the dust as did the occurred when there was some coarse fraction
standard Hartmann apparatus. The first used a mixed with the fine fraction. Those maxima
mechanical vibrator to shake the dust through occurred approximately at the ratio of 75/25
a sieve into the explosion chamber below_ The for the maximum explosion pressure and 25/
second apparatus generated a dust cloud by 75 for the maximum rate of pressure rise.
rotating the explosion chamber. The faster
2-2. Composition
rotation and the more the dust, the denser the
The effect of volatile matter, ash, and
cloud_ Gun cotton was also used as the ignition
moisture content on the maximum pressure
source in the second apparatus instead of the
rise was examined by Price [22] _ His samples
standard electric spark_ Experiments were
included different types of coal dust and
performed on four types of coal dust:
other types of carbonaceous dust. The appa-
Yamagata lignite, Horonai, Ponbetsu and
ratus and procedure used have been described
Liddel. The Horonai or Ponbetsu had five
earlier. No apparent correlation was obtained
separate particle size fractions, the Yamagata
between the maximum explosion pressure
lignite or Liddel had two_ Small amounts of
developed and the weight percent of volatile
magnesium oxide were added to increase the
matter for the carbonaceous dust. From a
dispersibility of the dust_ Graphs of the explo-
table presented by Bautling [3] on volatile
sion strength plotted against the average par-
matter produced by complete distillation (to
ticle diameter were presented_ The explosion
500 C) of cellulose and starch, Price 1221
strength is the maximum explosion pressure
showed that most of the volatile matter given
multiplied by the highest maximum rate of
off was not inflammable_ He also found that
pressure rise_ The explosion strength increased
the ash content did not have a marked effect
as the particle size decreased_
on the pressure rise by the carbonaceous dust
Bartknecht [2] performed explosion tests
when the weight percent of ash was below
in a 1 m3 spherical explosion chamber He
normal limits; however, at weight percents of
studied four types of dust: flour, methylcellu-
ash above normal limits, the higher the ash
lose, polyethylene and polyvinylchloride. The
content, the lower the explosion pressure rise.
average particle diameters of each dust ranged
His results indicate that an increase in moisture
approximately from 10 to 400 pm_ The maxi-
content decreases the explosion pressure rise
mum explosion pressure and rate of pressure
by both acting as a heat sink and by causing
rise were plotted against the average particle
the effective particle size to increase due to
diameter_ From the methycellulose dust or
agglopleration. The effect is not pronounced
the polyvinylchloride dust, a nonexplosive
enough to make moisture a practical inertant.
coarse fraction and an explosive fine fraction
In his work on .agricuXural dust, Eckhoff
were separated_ From these fractions five
[6] characterized the composition of dust by
samples were prepared, each having a unique
the percent moisture, ash, protein, starch and
ratio, the weight percent of coarse fraction to
fiber. He obtained a reasonable correlation
that of the fine &action_ For example, a ratio
between the maximum rate of pressure rise
of -75/25 indicates that the sample contained
and the percent of starch and fiber. To sepa-
75% coarse &action and 25% fine fraction.
rate the particle size effect from the correla-
The maxim urn explosion pressure and the
tion, he assumed that eqn_.(3) could be used
maximum rate of pressure rise were plotted
to correlate the m aximum rate of pressure rise
against those ratios. The result for each.Qpe
with a specific envelope surface area He
of dust was graphed separately_ For methyl-
assumed the proportionality constant, _K, in
cellulose dust the maximum explosion pressure
eqn_ (3) to be a function of composition_ The
and the maximum rate of pressure rise in-
functionality was
creased when the ratio decreased_ The highest
value of the m aximum explosive pressure and D=g KiXi (4)
that of the maximum rate of p&sure rise i

where n = t&al number of all components, Ki
= the value of K for pure component i, and Xi
= the weight fraction of component i.
The value of Sri-was experimentally deter-
mined for starch using wheat starch and corn
starch and for protein using fine and coarse
fish powder. The value of Ki for starch WAS
considered applicable to the star& and fiber
fractions because of the similarity between
their molecular structures_ The value of Ki for
the fish powder was considered applicable to
the rest of the components: protein, fat, and
nitrogen-free soluble organic compounds
other than starch. The resultant correlation is
(W Starch & Fiber)
K = (SOO- + 450),
100
kp- g/(cm2- s- m2) (5)
Eckhoff [S] also correlated the maximum
rate of pressure rise with the weight percent
moisture_ He selected four types of dusts:
maize starch, cellulose powder, potato starch
and Norwegian oat dust, and prepared several
samples from each with varying weight percents
of moisture_ He obtained an approximate
correlation given by
(6)
where y = wt.% of moisture, and ymin = the
wt.% of moisture above which no explosion
can occur- ymin was assumed to be approxi-
mately 15% for all types of dusts_
To study the effect of moisture on the rate
of pressure rise in the Hartmann bomb tests,
Eckhoff and Mathisen ]7] attempted to elimi-
nate the effects of moisture on the quality of
the dust dispersion, the effective particle size
distribution and the turbulence_ Corn starch
was the only sample used. In a test apparatus
similar to the Hartmann, the effective particle
size distribution was measured. The dispersion
system and the shape of the apparatus are the
same as those in the Hartmann apparatus;
however, there is no ignition system and the
tube is fitted with a piston Double adhesive
tape was attached to the bottom of the piston.
When the dust was dispersed, the tape captured
a sample of the dust, and an effective particle
size distribution was determined by micro-
scopic examination. It was concluded that
moisture does not affect the effective particle
size distribution greatly.
197
The ignition delay (the time interval
between the dispersion and the ignition of the
dust) was taken as an indicator of the degree
of turbulence. For example, if the ignition
delays of two explosions were equal, the
degree of turbulence was considered to be the
same. Therefore, to eliminate the effect of
turbulence on the correlation of the maxi-
mum rate of pressure rise wi& weight percent
moisture, only those data from the explosion
tests with similar ignition delays were com-
pared [6] _ This correlation was significantly
different from the correlation of the maxi-
mum rate of pressure rise with weight percent
moisture where the differences in ignition
delay were not considered_ At short ignition
delay times (approximately 0.005 set) there
were significant maximum rates of pressure
rise for moisture contents considerably higher
than 30%. When the delay time was not con-
sidered, there were few significant maximum
rates of pressure rise for moisture contents
higher than 22%_
The particle size and composition of grain
dust affect the explosion characteristics_ The
exact relation is, however, difficult to quanti-
fy_ Most of the experimental work attempted
to correlate the data and presented the dat,
graphically and/or tabuiarly with minimum
analysis. To improve the experimental studies.
theoretical analyses are needed. Hartmann
and Nagy [9] separated dust into several size
ranges and measured the explosibility cha-
racteristics of the dust in each size range.
They used the arithmetic mean of particle dia-
meter, which has been shown to be an
incorrect method of representing the particle
size distribution [ lOI_ Jacobson et al. [ 143
improved their experimental methods.
However, the characterization of particle size
and composition showed little progress_
Eckhoff [ 63 performed the test of explosibility
of several grain dusts and obtained the correla-
tion with the moisture, ash, protein and fiber
contents. In summary, only limited work has
been conducted on this subject and many
areas of work are still needed, especially in
the characterization of particle size distribu-
tion_
3_ THEORETICAL STUDIES
There are few publications that attempt a
theoretical prediction of the explosion chamc-
 198

teristics: minimum explosible concentration, aT, - a2Tg, a-q;-.

maximum explosion pressure, maximum rate
of pressure rise, and average rate of pressure
---Q-+--~

at t a? r2 ar ).
rise. Jaeckel [ 161 and Tanaka 1231 developed where Q = thermal difffivity of the air, r =
expressions to predict the minimum explosible radial distance Mm-the center of the.ignited
concentration, while Normura and Tanaka particle, and Tgr = temperature of the gas at
[21] and Nagy et al. C-191 developed expres- any radial position r.
sions to predict the maximum explosion The resultant equation is

pressure and rate of pressure rise.
To predict the minimum explosible con-
centration, Jaeckel [16] performed a heat
balance on the dust-air system. He considered
the minimum explosible concentration, Cd,
to be the smallest concentration of dust
capable of producing enough heat. when com-
pletely cornbusted, to elevate the temperature
of the dust cloud and the air to the ignition
temperature; the following expression was
obtained_
c _ = u-i --o)P,G
lnnl
9 --o-l --To) c,,
where q = heat generated per unit mass of
dust, T1 = ignition temperature, To = initial
temperature, pa = density of air, C, = specific
heat capacity of air, and C,, = specific heat
capacity of dust.
This model depicts the system as being a
spatiahy uniform dust cloud throughout the
entire expIosion and does not take into.
account the dependence of C,, on the dust
particle size- Jaeckel [ 161 reported that eqn.
(6) predicted a minimum explosible concentra-
tion of 0.022 kg/m3 for sugar powder; in
contrast, Jacobson et a.L [14] reported an
experimental value of O-045 kg/m3.
Tanaka [23] developed a model to predict
the minimum expiosible concentration. A
particle at the center of. the dust cloud was
ignited by an outside source. The ignited par-
ticle heated its near neighbor, a second par-
ticle, at a distance L_ The minimum concen-
n&ion, C,,, is the smallest concentration
that allov-7 the flame to propagate from the
ignited par,%cle to its neare& neigbbcr. In the
model, the equation for calculating the flame
height, b, was taken from a publication on the
burning of liquid droplets by Miesse [lS] _
Tanaka 1231 has assumed that the gas phase
between the *_o particles is stagnant. He then
estimated the gas temperature at-the gas-solid
interface- of the second particle,. SF,&, by
solving the conduction equation
1+-
Ti(9)
(L-b)
2&zP5
where DP = particle diameter, 11= t/T, T =
time required to completely burn the particle,
&, = proportionality constant between 2 and
Df, Tf = temperature of the flame, and z =
initial temperature of the system.
An energy balance was performed around the
second dust particle to determine the tem-
perature of the particle, TaL, as follows:
where ps = density of the dust particle, C,, =
specific heat capacity of the dust particle,
E, = emissivity of the particle, Ef = emissivity
of the flame, u = Stephan-Boltzmann constant,
F = shape factor, h = heat transfer coefficient,
2K/D$, and k = thermal conductivity of the
air.
The resultant equation is
T,,(o) = 1
e_., Cl$FCcT&n) + D-G3 dn +
+ Tao] (11)
where C = IUr,k/p, -C,,, D = 3KDEPEfFD,/
@SC,,), and Tdo = initial temperature of the
particle.
The value of p, i.e. t/T, at which Td,
reaches the ignition temperature, Ti, can be
calculated from eqns. (9) and (11) by selecting
a distance L. The minimum distance, Lo, is
the distance L which allows the second particle
t.o reach the ignition mmpeE&We, Tan = Ti*
at the moment that the initial particle is com-
pletely consumed. 9 = l_ The minimum explo-
sible concentration is calculated-from-l,, by
the following equation.

Tanaka 123 J has .definedthe ignition tem-
perature, Ti, differently from that experimen-
tally determined by he Godbert-Greenwald
furnace. The ignition temperature defined in
the model is the temperature of the dust par-
ticle that causes it to ignite; however, the
experimentally determined ignition tempera-
ture is the minimum wall temperature that
causes a particle immersed in it to ignite [5] _
Nagy et al. [ 191 have derived an expression
to predict the rate of pressure rise as a function
of time. They have assumed that the pressure
developed is affected by changes in tempera-
ture and increases in the total number of
moles of gas in the explosion chamber. They
have baaed their derivation on an equation
empirically derived from gaseous explosions,
which is
(13)
where A = surface area of the flame front, Vu
= volume of unburned fuel and gas, If, =
burning velocity reference at 1.0013 bar (1
atm) and 298 K (537 X), T, = reference tem-
perature, 298 K (537 R), R, = reference
pressure, 1.013 bar (I atm), andp = aconstant.
They have assumed that the air and combus-
tion gases behave ideally and that any solid
present has negligible heat capacity and
volume. At any time during the explosion, the
combustion chamber is assumed to be divided
into two zones, the burned and unburned
zones, by a very thin combustion zone.
Two cases on the rate of pressure rise have
been considered by Nagy et al. [ZO] . In one
case the system is assumed to be isothermal,
and in the other it is assumed to be adiabatic.
For the isothermal system, the rate of pressure
riseis
(14)
where P_= = maximum explosion. pressure,
PO = initial pressure, TO = initial temperature,
and a = radius of the vessel.
For the adiabatic system, the rate of pres-
sure rise is
d.P SvK,T;eP:g
-= (P& - P$)13 x
dt aT2Pzmu
r
2
p3vs
(15)
In either case, the expression for pressure
as a function of time -SC
obtained by integration
of its equation for the rate of pressure rise,
eqn_ (14) or (15).Note that neither eqn. (14)
nor eqn (15) is an explicit function of particle
size. However, they could have an implicit
size dependence through K,_
The expressions for the rate of pressure rise,
eqns. (14 j and (15), are explicit functions of
neither concentration nor composition. Nagy
et aI_ [19] compared their models with exper-
imental results involving several gases and
corn starch. They obtained good agreement,
except when the explosion was nearly com-
plete and the heat loss to the walls of the
vessel became significant.
Normura and Tanaka [21] have theoreti-
cally derived the experimentally determined
cubic law used by Bartlmecht [ 2 J , which is
Vp3 (16)
_L
= maximum rate of pres-
where (dP/dt),,
sure rise, and V, = enclosuz volume (constant
with time).
Note that Nagy et al. [ZO] also obtained
this result. Equation (15) implies
d.P 1 1
-lx-=-
v--u3
dt a
Normura and Tanaka [21] have assumed
the dust cloud to be stagnant with the par-
ticles uniformly spaced in a fixed geometry.
The geometry is shown in Fig. 1. The flame in
the nth spherical shell of dust particles is
assumed to propagate by transferring heat to
the (n+l)th spherical shell of particles until
the temperature of the particles in the (n+l)th
shell reaches the ignition temperature and the
particles ignite. The flame continues to pro-
2GG
Fig. l_ Model of dust particles (Normura and Tanaka
c=1-
pagate from one shell to the next until the
dust is completely consumed. Normura and
Tanaka 1211 have also assumed the heat is
transported by only conduction and radiation
because of the stagnant cloud assumption. By
performing au energy balance, the time
interz& t,, between the ignition of the par-
ticles in the nth shell and the ignition of the
particles in the (n+l)th shell is calculated for
each pair of shells_ For R between 1 and
approximately 10, the time interval fn
decreases; however, for n greater than 10, it
remains constant. Because tn becomes
constant rapidly, a fixed f is used as the time
interval for the entire flame propagation (n+l).
At any time t, after the ignition of the first
particle, the number of shells that have been
ignited is calculated Tom t_ From the geom-
etry- the total number of particles in each
shell is known, and thus the total number of
particles ignited at any time t can be calculated_
To determine the total mass of dust burned at
any time t, M(t), Normura and Tanaka have
assumed that the burning rate of each particle
is controlled by the diffusion of oxygen
through a layer of burned dust into the un-
burned center_ Ram this, they have found a
relationship between the length of time a par-
ticle had been burning, 19, and the mass of the
particle that w& consumed, m(0 )_ Therefore,
M(f) is the summation over all of the ignited
shells of the product of the total number of
particles in ashell and the mass of each particle
that has been burned, The relationship used
form(f3)i.k~
tihere i = tune necessary to completely burn
the partic.& and m, =.&al amount of mass
initially contained in the particle_
This relationship is, however, for a hetero-
geneous reaction in which the rate of the par-
ticle consumption is controlled by the rateof
a first-order reaction [17] _ The expression,
for which the controlling mechanism k the
diffusion of oxygen through an ash layer, is
(18)
Equation (18) is more complicated than
eqn. (17) and would also complicate the
solution of the model. Therefore, if the frac-
tion of the initial mass that has been con-
sumed, m(e )/me, obtained from eqn. (17) is
within 0.1 of that from eqn_ (18) for all BIT,
then eqn. (17) is the better choice. Figure 2
presents both eqns. (17) and (18) and the
difference between m(O)/m, from eqn. (17)
and that from eqn. (18). The difference is
greater than 0.10 for m(6)/mo between 0.11
and 0.68 with amaximum of 0.21_
By mahing three additionA assumptions,
Normura and Tanaka [21] have derived an
expression for the pressure at any t from the
expression for the amount of mass burned at
any t, M(t). The pressure developed is the
result of a net increase~in the total number of
moles of gas produced by the reaction and the
increase in temperature produced by the heat
released from the reaction, the system of the
dust and air is adiabatic, and the maximum
explosion pressure occurred when ah of the
dust initially present had been burned. The
expression obtained for the pressure, P(t), is
where& = total mass of dust initially present,
v,, = ratio of the heat capacity of the unburned
gas at constant pressure to that at constant
volume,
P = maximum explosion pressure attain-
a=, and PO = initial pressure.
Fig. 2. Comparison of the actual expression for_M(B)/Me in Yagi and Kunii 1243 with
that used by Normura and Tanaka 1211; -eqn. (16), --- eqn. (17), -- --absolute
difference between eqns. (16) and (17).

Hertzberg et al. [ll] showed that dust Therefore,

explosions were oxygen-limited for concentra-
tions greater than 0.4 - 0.5 kg/m3. Thus, all
the initially present dust would not be burned,
and the maximum explosion pressure would
be a function of the total amount of oxygen (21)
present. Therefore, P,_ that Normura and
Tanaka [21] calculated would be too large
for a high dust concentration. They assumed The values of KG calculated from eqn. (21)
their system to be adiabaiic, since the maxi- were significantly larger than those of the
mum explosion pressure and the maximum experimental values; however, they were of
rate of pressure rise occur at the same time. the same order of magnitude.
Therefore, the m aximum rate of pressure rise Until the development of the cubic law by
will also be high. To obtain their derived ex- Bartknecht [2], it was difficult to compare
pression for the cubic law, they differentia- the results of studies by researchers on differ-
ted eqn. (19) with respect to time, and ent types of explosion chambers and different
evaluated the derivative at the time when it methods of characterization of different types
wasm aximum- As previously stated, the rate of dust. The most important finding is that
of pressure rise became a maximum when all for the determination of explosibility data of
of the dust was consumed. After rearrangement, combustible dust a minimum volume of 16
the following is obtained. liters would be required to ensure correlation
with the data from large test vessels (Bartknecht
[2] )_ In general, with everything being equal,
Bartknechts data in a 1 m3 chamber are
much more likely to be right than data from a
Hartmann bomb.

REFERENCES

where pS = the density of dust, Da = diameter 1 D. F. AIdis and F. S_ Lai, USDA Misc_ PubL No.
of the dust particles, and a = O-2-M,/Ve. 1375.1979.
 -.
202

2 Vi_ Bartknecht, ExpZosionen-&Zauf unf Schutz-
massnahmen. Springer Verlag, Berlin, 1978.
3 J_ Bautling. fir Gasbeleuchiung, 57 (1918) 32 -
55.
4 J. M. Dallavalle, The Techhnoiogy of Fi.ze ParticZes.
Pitman, New York, 2nd edn., 1948, Chap_ 5, p_
113.
5 H. G. Dorsett, Jr., M. Jacobson, J. Nagy and R_ P.
Williams, U.S. Bur. Mines, Rep. Invest. RI-5624.
1960_
6 R_ K. Eckhoff, Fire Res.. I (1977/78) 71 - 85.
7 R. K. Eckhoff and K_ P_ Mathisen, Fire Res.. I
(1977/78) 273 - 280_
s H. Enomoto, Proc_ Int. Symp. on Grain Dust Ex-
plosions, Grain Elevator and Processing Society,
Minneapolis, MN, 1977, pp. 143 - 170.
9 I. Hartmann and J_ Nagy, US. Bur. Mines, Rep.
Invest., RI-3751. 1944.
10 G. Herdan, MI_ L_ Smith, W_ H_ Hardwick and P.
Connor, Small PartkZe Statistics, Butterworths,
London, 2nd edn., 1960.
11 M. Hertzberg, K. L. Cashdollar and J. Operman,
U.S. Bur. Mines, Rep. Invest.; RI-8360.
1979.
12 W. Ishihama, II th Int_ Conf. ofDirectors ofSafety
in Mines Research. Warsaw. Poland. I961_
13 W. Ishihama and H_ F.nomoto, Com&st. F&me, . -.
ZI(l973) 177 - 186.
14 M. Jacobson, J. Nagy, A Ri Cooper and F. J. Bi,
US. Bur. Mine-$ Rep. Invest; ~RI-5753. 1961:
15 M. Jacobson, k R. Cooper. and J. Nagy, U.S. Bur.-
Mines, Rep. Invest. RI-6516.1964.
16 G. Jaeckel. 2. Tech; Phys. 3 (1922) 67 - 78.
17 0. Levenspiel, Chemical Reaction Engineering,
Wiley, New York, 1962.
18 C. C. M&se, 6th Symp. on Combustion, 1957. $.
732.
19 J. Nagy, J. W. Conn and H. C. Verakis, U.S. Bur.
Mines, Rep. Invest_, RI-7279 1969.
29 J. Nagy, E. C. Seiter, J. W_ Corm and H. C.
Verakia, U.S. Bur_ Mines. Rep. Invest.. R&7507.
1971.
21 S. Normura and T. Tanaka, Kagcku Kogcku.
(1979) 601- 608_
22 D. J. Price and H. H. Brown, Dust Explosions.
National Fire and Protection Association, Boston,
MA, 1922.
23 T_ Tanaka, Proc In t. Symp. on Grain Dust ExpZo-
sions, Grain Elevator and Processing Society,
Minneapolis, MN, 1977, pp_ 79 - 99.
24 S. Yagi and D. Kunii, 5th Int. Symp. on Combus-
tion, Reinhold, New York, 1955. p_ 231.

-.
Journal of Hazardous Materials, 4 (1980) 177-183 177
Elsevier Scientific PublishingCompany, Amsterdam - Printed in The Netherlands

THE EFFECTS OF FUMIGANTS ON GRAIN DUST EXPLOSIONS

S.R. TAIT, R.G. REPUCCI and J.C. TOU

Analytical Laboratories, Michigan Division, The Dow Chemical Company, Midland, MI
48640 (U.S.A.)
(Received February 21, 1980; accepted March 10,198O)

The effects of three commercial fumigants on grain dust explosions were investigated
under two laboratory conditions. Although the fumigants exhibit some flammable proper-
ties, the results of the test show that they do not increase the severity of grain dust ex-
plosions. In fact, in some cases, the fumigants actually suppressed the explosion.

Introduction

A recent survey [l] reveals that grain elevator dust explosions have oc-
curred at least 216 times from 1958-1977. During this period of time, there
has been a steady increase in governmental regulations designed to prevent
such incidents. The prevention of grain dust explosions is indeed a complex
scientific, social, and economic task.
That grain dust explodes is a well-known fact. According to data from the
United States Bureau of Mines, the index of explosibility of a mixed grain
dust is about 9.2 times that of Pittsburgh coal dust. A comparison of explo-
sion properties from a variety of common dusts, such as starch, soybean meal,
etc., is summarized in the review [l] . Currently, concern has been expressed
[1,6,7,&g] on the effects of fumigants on grain dust explosion properties.
The fumigants are added to the grains for insect control. The concern is essen-
tially based on the following understandable possibilities:
Many of the formulated fumigants contain flammable components such as
ethylene dichloride and carbon disulfide. It is conceivable that the presence
of these flammable components may increase the ease of ignition of the dust
particles and therefore promote the explosion hazard. On the other hand, is
it not unreasonable to speculate that carbon tetrachloride, one of the major
components in the commonly used fumigants and also a known fire retardant,
may present an opposite effect and suppress a dust explosion ? In addition there
are recent reports which suggest the possibility of synergistic behavior of
methane gas in dust explosions when present with certain foreign additives in the
grains. Anthony [ 21 demonstrated that polyvinyl chloride dust, known not to
be explosive alone in a Hartmann tester, exploded in the presence of only 1.5%
178

TABLE 1

The compositions of the commercial fumigants investigated and the flammabilities

of their formulated components

Flammability (% in air)

Lower limit Upper limit Dowfume* Vertifume* Max Kill

High Life*

Ethylene dichloride 6.2 16 70 - -

Carbon tetrachloride none none 30 82.9 82.8
Carbon disulfide 1.3 50 - 16.5 16.4
Inert ingredient - - - 0.6 0.8
.-
*Trademark of The Dow Chemical Company.

methane in air. Singer et al. [3] also reported a similar synergistic effect of
methane in lowering the minimum explosibility concentration of coal dust
in air.
The effects of several commercial fumigants such as Dowfume * 75, Verti-
fume* and Max Kill High Life* on grain dust explosions were studied in a
Hartmann tester. The compositions of the fumigants studied and the flammabili-
ties of their formulated components are shown in Table 1. Their effects on
the maximum pressure and its rate, minimum energy and minimum concen-
tration were characterized. Because of the precision of the Hartmann technique,
the significance of the data obtained was statistically evaluated.

Experimental

The dust sample was obtained from the top of a local grain elevator. The
sample is composed of an unknown mixture of corn, wheat, rye and oats.
The sample was first put through a 50 mesh sieve and then screened
through a 200 mesh sieve of 74 microns. The dust particle size distribution,
as analyzed with a HIAC Model PC-320 particle size analyzer manufactured
by Pacific Scientific Instruments, California, is shown in Fig. 1. A portion
of the grain dust obtained was dried in an oven at 75C for 48 h followed by
an additional 4 h of drying at 105C in a glass cake pan. While drying at 105C,
the dust was stirred every 30 min to insure its uniformity and complete dry-
ness.
The recently proposed ASTM test procedure using the Hartmann Dust Ex-
plosion tester was employed with the exception that a hot coiled wire igni-
tion source was used instead of the continuous spark suggested in the proce-
dure. The coiled wire was 24 turns of 24 gauge Nichrome wire and was heated
with 110 V power. The advantage of this choice has been discussed elsewhere

*Trademark of The Dow Chemical Company.

 179

Cumulative
volume x GRAIN DUST PARTICLE SIZE DISTRIBUTION
loo -

SO-

SO-

70 -

Sn-

SO-

4D-

30 -

Diameter in Microns

Fig.1. Grain dust particle size distribution.

[4]. The performance of the Hartmann tester was checked each day before
and after the sample run against Lycopodium powder, and the screened grain
dust.
The pressure and the pressure rate during an explosion in the Hartmann tester
were monitored with a fast response (1 ps), Piezotronics Model 113A24 pres-
sure transducer connected to a Tectronix Oscilloscope and to a Model4SA
peak meter manufactured by Piezotronics, New York. The maximum pressure
and the maximum pressure rate were obtained from a series of experiments
with different grain dust concentrations.
The determination of the minimum energy, required in a static spark to
ignite the dust cloud at the concentration of the maximum pressure and the
maximum pressure rate, was made from the discharge of selected capacitors of
0.001,0.01,or 0.1 mfd which had been charged to a known potential by a
standard high voltage power supply across l/4 air gap. The discharge was
initiated 0.3 s after the dust was dispersed with 100 p.s.i. air into a clear, heavy,
glass Hartmann tube. The ignition of tine dust was observed by the bursting of
the paper diaphragm at the top of the tube. The experiment was repeated
several times above and below the point of ignition with varied amounts of
energy. The minimum energy was finally determined when 0.01 mfd was used.
The experiments for determining the minimum explosion concentration
limits were carried out also in the Hartmann tester with the pressure trans-
ducer attached to it. The minimum concentrations were determined at the
minimum pressure rise. The coiled wire was used as an ignition source.
The effects of fumigants on the gram dust explosion were determined on
the dried grain dust in order to simulate a worst case situation. The ad-
dition of fumigants to the grain dust was accomplished by the following
180

two ways:
The grain dust was weighed to 1 mg and loaded to the Hartmann tester.
Five drops of a fumigant were added through a hole of about 0.028 in a
coned Whatman# 4 filter paper from a Curtin 5 3/4 transfer pipet. After the
fumigant had been evaporated, the paper was removed, the tube sealed, and
the standard testing procedure then followed.
The second way of preparing the fumigated grain dust samples wasaccom-
plished by mixing the dried gram dust with the recommended amounts of
fumigant. The mixtures were then rolled for 48 h in a sealed container. Care
was taken in the process of sample handling to minimize losses of the fumi-
gants.

Results and discussion

The precision of the maximum pressure and the maximum pressure rate
determined in the Hartmann dust explosion tester is shown in Table 2 for the
undried grain dust and Lycopodium at their maximum pressure rate concen-
trations of 0.24 oz/ft3 and 0.56 oz/ft3 respectively. The relative precisions

TABLE 2

The precision of the maximum pressure and the maximum pressure rate determined in Hartman
dust explosion tester

Experiment Grain dust before drying (0.24 oz/ft3) Lycopodium (0.56 oz/ ft3)

Max. press (p.s.i.) Max press. Max press. (p.s.i.) Max. press.
rate (p.s.i./s) rate (p.s.i./s)

1 76 2580 69 6435
2 73 2714 104 7045
3 65 2174 99 5926
4 73 1962 108 7692
5 63 2391 116 8167
6 66 1500 96 4786
7 77 2354 94 6045
8 67 2580 94 5833
9 68 2016 95 5882
10 74 2260 90 5231
11 62 1683 102 5458
12 60 1434 89 5167
13 75 2386 113 7876
14 69 1608 116 7600
15 70 2100 111 5250
16 70 1700 120 7381
17 64 1550 102 5375
18 75 2019 116 5133
19 63 1618 113 8222
20 62 1775 102 4750

Averagei S.D. 69 + 6 2000 f 360 100 k 12 6300 f 1100

 181

(G/x)100%, where u is the standard deviation and jTthe average value, are
about 18% for both cases. The average value of the maximum pressure rate
for Lycopodium was found to be 6300 p.s.i./s. The data indicate that for a
reliable comparison, it is necessary to evaluate the data statistically. One ex-
periment is definitely not sufficient to define the significance of the observa-
tion. The pressure data reported in this paper are the results from the statis-
tical average of twenty experiments. As described in the experimental section,
the Hartmann bomb was also standardized with the undried grain dust and Lyco-
podium each day before and after the designed experiments were done. No
statistically significant difference was observed before and after the designed
experiments. Part of the data are shown in Table 3. Statistically speaking,
the average values are very close to those shown in Table 2.
In addition to the experimental uncertainty in the measurement of the
maximum pressure and the maximum pressure rate, the concentrations at
these maxima are usually not very well defined. This has been revealed in the
early study [5]. Therefore, we have characterized the effects of the fumigants
on the maximum pressure and maximum pressure rate at the same concen-
tration as that, 0.24 oz/ft3, determined for the grain dust before being dried.
In order to simulate a worst case situation, the basis chosen for compari-
son is the dried grain dust. The data are summarized in Table 4. The statisti-
cal t test was used to evaluate the significance of the effect of the fumi-
gants on the dried grain dust explosions. As shown in Table 4, the statistical dif-
ferences were evaluated at three confidence levels, 90%, 95%, and 99%.
The maximum pressure rate of the dried grain dust is significantly higher
than that before being dried. However, the maximum pressures generated
are about the same for both cases.
Generally speaking, there is no well-defined trend, which can be drawn

TABLE 3

The maximum pressure rates of the grain dust before drying and Lycopodium in a
Hartman dust explosion tester

Day Grain dust before drying (0.24 oz/ft?) Lycopodium (0.56 oz ft)

1 1899 7216
2 1921 5877
3 2090 7050
4 1974 6191
5 1821 6780
6 2050 7014
7 1744 5422
8 1925 6888
9 1800 7062
10 2077 5729
11 2045 6245

Average 1900 f 120 6500 * 520

182

TABLE 4

The grain dust explosion and the effect of fumigants

Sample Max pyp. Max press. rate Cont. at Min. energy Min. cont.
(p.s.i.) (p.s.i./s) max. press. (J) (oz/ft)
rate (oz/ft3)

Grain dust before

drying 70 f 6 2000 f 360 0.24 - -
Dried grain dust 78+ 4 4000 + 570 0.24 0.180 0.050
Dried grain dust +
5 drops of
Dow fume* 7 5 77 f 4 4400 * 630 [l] 0.24 0.125 0.046
Dried grain dust +
5 drops of Max
Kill High Life* 69 f 5 [3] 3300 f; 610 [3] 0.24 0.125 0.066
Drained grain dust
+ 5 drops of
Vertifume* 82 + 4 [2] 4800 i 750 [3] 0.24 0.125 0.056
Dowfume* 75 dried
grain dust (6 gals/
1000 bushel) 56 * 3 [3] 2600 f 280 [3] 0.24 0.125 0.054
Max Kill High Life*/
dried grain dust
(6 gals/1000
barrels) 80t 4 3700 f 590 0.24 0.125 0.049
Vertifume*/dried
grain dust (6 gals/
1000 barrels) 81 + 3 4000 f 400 0.24 0.125 0.052

*Trademark of The Dow Chemical Company.

**Average + 1 standard deviation. Statistical analysis indicates that it is a significant difference
from the control, the dried grain dust, at the (1) 90%; (2) 95%; or (3) 99% confidence level.

from the result shown in Table 4, on the effects of fumigants on the maxi-
mum pressure and the maximum pressure rate. Statistically, the maximum
pressure and the maximum pressure rate generated in the cases of the dried
grain dust with 5 drops of Max Kill High Life and mixed with Dowfume 75
in a proportion of 6 gallons/1000 bushels are definitely lower than those in
the dried grain dust with 5 drops of Vertifume and no significant effect was
observed in the rest of the cases. As shown in Table 1, the compositions of
Max Kill High Life and Vertifume are very similar with minor difference in
the inert ingredients. It would be expected that the effect on the grain dust
explosion should be very similar. The maximum pressure and rates of the
dried grain dust mixed 6 gallons/1000 bushels with the above two fumigants
are almost identical, however, significant differences were observed from the
5 drop addition procedure (Table 4). The reason leading to this observation
is not known.
The minimum energy of ignition of the dried grain dust with the fumigants
 183

added is lower than that for the dried grain dust by itself. The exact differ-
ence was not determined since no values were checked between 0.180 and
0.125 J. It is believed that the reduced ignition energy does not constitute
a significant increase in the hazard. Historically, the known ignition sources
in a grain elevator have very large energies such as welding, open flame, and
electrical failures. Less than 2% may have been caused by static charges [l] .
As indicated in Table 4, four out of the six cases studied show that the
minimum concentrations are higher than that of the dried grain dust.

Conclusion

In conclusion, no statistically unidirectional effects of the fumigants on

the maximum pressure, the maximum pressure rate, or minimum concentra-
tion was observed under the present laboratory controlled conditions. In
fact, in some cases, the fumigants actually suppressed the explosion.
The data indicate that the presence of the fumigants did lower the mini-
mum ignition energy from 0.180 J to 0.125 J. However, it is believed that
the reduced ignition energy does not constitute a significant increase in the
hazard, since ignition sources encountered in a grain elevator typically have
much higher energies than 0.125 J.

Acknowledgement

The authors would like to express their appreciation to Orchid Chiang for
her contributions in the statistical analysis of the data.

References

1 D.D. Bluhm, Grain elevator explosion, A university view, lS-EMRRI-8, Iowa State
University, Ames, Iowa, July 1978.
2 E.J. Anthony, The determination of the minimum explosible concentration of poly-
vinylchloride powder in the presence of methane and air, Fire Research Station, Bore-
hamwood, England, FR Note 1072, August, 1977.
3 J.M. Singer, A.E. Bruszak and J. Grumer, Limits of flame propagation of coal dust-
methane-air mixtures, U.S. Bureau of Mines, RI-7103, 1968.
4 H.G. Dorsett, Jr., H. Jacobson, J. Nagy and R.P. Williams, Laboratory equipment and
test procedure for evaluating explosibility of dusts, U.S. Bureau of Mines, RI-5624,196O.
5 I. Hartmann, and J. Nagy, Inflammability and explosibility of powders used in the plastics
industry, RI-3751, Bureau of Mines, 1944.
6 R. Frye, Elevator explosions, Proc. Internation Symp. Grain Elevator Explosions
(Vol. 2), National Materials Advisory Board 352-2, 1978, p.47.
7 Fumigants and dust a cause of explosions, Chemical Week, February 22,1978, p. 13.
8 Fumigants may cause elevator blast, C&E News, February 20, 1978, p.13.
9 News Release, February 13, 1978, Arthur D. Little Co., by Public Relation Department,
Cambridge, Massachusetts.