Documente Academic
Documente Profesional
Documente Cultură
TABLE ]
Percentage of dust explosions in the grain, feed and food industry assumed t,) be initiated i)y 'friction sparks" or
unknown sources
1 535 20 46 66 1860-1973
1 128 17 27 44 1949-1973
2 91 54 18 72 1941 -1945
3 137 9 62 71 1958 -1975
4 83 28 5 30-35 1965 -1980
The ability of metal sparks or hot spots [9, 10]. The specific objective has been to
from grinding and cutting to ignite dust clouds answer the following question: under what
has been demonstrated by several researchers. circumstances of impact energy, impacting
The experiments by Leuschke and Zehr [5], velocity and type of moving object and anvil,
published in 1962, are probably the first ones can single impacts between solid bodies ini-
in which dust clouds were ignited by grinding tiate a dust explosion in a cloud of grain dust
wheel metal sparks. However, no clouds of in air?
organic dusts ignited. Zuzuki et al. [6] ignited
different coal dusts using both sparks and hot
2. EXPERIMENTAL APPARATUS
spots from a piece of steel in contact with a
grinding wheel rotating at 23 - 47 m/s peri- 2.1. Impact generator
pheral velocity, Later Allen and Calcote [7] After having considered various principles
conducted similar experiments in which metal for constructing a single-impact generator
sparks were generated by pressing a steel rod suitable for laboratory work, it was decided
against a rotating grinding wheel. By retarding to build the apparatus shown in Fig. 1. The
and focusing the spark stream, it was possible length of the rigid, spring-loaded arm, the
to ignite clouds of natural organic dusts such 'hammer', from the centre of rotation to the
as maize starch and wheat grain dusts. An- tip of the test object, can be varied by means
other recent, and very comprehensive study is of an adjustment screw. The condition for
Ritter's [8] investigation. This work confirm- obtaining impact is that the arm is slightly
ed that metal sparks generated by grinding, longer than the perpendicular distance from
rubbing and repeated impacts can ignite the arm axis to the anvil. By careful adjust-
clouds of several dusts, including cellulose and ment of this small difference it is possible to
lycopodium. generate impacts of different impact energies
The second situation, in which metal sparks without changing the impact velocity. The
may be generated in industry, is accidental cylindrical tip of the test objects that impact-
single impacts between solid bodies. In this ed on the anvil material had a reduced diam-
case relevant evidence of the ignition poten- eter of 4 mm. Details of the end of the arm
tial has been lacking. However, as long as the are shown in Fig. 2. Springs of different
necessary conditions for initiation of dust strengths were used for generating a range of
explosions by such impacts remain unidenti- impacting velocities of approximately 8 - 25
fied, one is forced to maintain the hypothesis m/s. Fine adjustment of the velocity was pos-
that such sparks may be hazardous in general. sible by adjusting the spring tension.
This in turn forces industry to take precau- When performing an impact test, the arm is
tions that may be superfluous, and causes fear first twisted to the start position, whereby the
that may be unnecessary. helical tension spring is stretched. In this
This unsatisfactory state of affairs created position the arm is locked by a pneumatically
the motivation for the present work which operated device that can be released electri-
has been reported in greater detail elsewhere cally. After having been released, the arm
155
probe in various locations, the concentration temperature of the body ff the spectral emis-
development of the dust cloud with time, at sivity is approximately constant in the range
any desired location, could be determined. of the spectrum of wavelengths used in the
measurement.
2.4. Timing system
A microprocessor-based 4-channel timer 2.8. Determination of minimum electric spark
was used to control the system. The timer energies for ignition of dust clouds
controlled the triggering of the impacting arm, These energies were determined by means
dispersion of the dust, as well as high-speed of the method described by Eckhoff [15].
movie camera and other photographic equip- This method, which is based on using electric
ment. sparks of known energies (integration of
power-versus-time curves), is essentially iden-
2.5. Equipment for measuring the velocity of tical with the new method proposed recently
the arm and the impact energy in W. Germany [16].
The kinetic energy of the arm at any instant
was calculated from its mass distribution and
instantaneous velocity. By measuring the velo- 3. DUSTS USED IN THE EXPERIMENTS
city of the arm immediately before and after
the impact, i.e., immediately before and after Since maize starch is one of the agricultural
'6 o'clock' position, the kinetic energy lost in materials that have often been involved in
the impact could be calculated. Two sets of dust explosions in the U.S.A. and is one of
reflective opto-switches were used for this the most sensitive to ignition, it had been de-
purpose. The opto-switches were mounted on cided to use this material in the experiments.
the wall of the test rig, with narrow clearance Native maize starch supplied from Amylum,
to the passing arm. A small mirror was fixed Aalst, Belgium, was used. On receipt the
to the arm, and an electrical signal was gener- starch contained 10 - 11% moisture, which
ated when the mirror passed the opto-switches. increased the minimum ignition energy con-
siderably compared with that of dried starch.
2.6. Photographic equipment In order to ensure maximum sensitivity to
Different kinds of photographic equipment, ignition, the starch used in the majority of
including ordinary single-frame cameras, video the present experiments was first dried for
camera, 16 mm movie camera and 16 mm four hours at about 105 C to remove the
high-speed movie camera, were used. The high- moisture.
speed camera (Hycam 4043020) was primarily A few other dusts of different minimum
used to make black and white negative film electric ignition energies (see Table 2), were
recordings at 1000 and 2000 frames per sec- used in order to investigate how the ignition
ond. The films were developed and examined probability in impact ignition tests ~}aried
at CMI immediately after each experiment. with the minimum electric spark ignition
energy.
2.7. Temperature measurements by a four-
wavelength optical pyrometer
In order to measure the temperature of the 4. CONCENTRATION DISTRIBUTION AND
sparks, a four-wavelength optical pyrometer ELECTRIC SPARK IGNITION SENSITIVITY OF
was used [14] (see 'Acknowledgements'). THE EXPERIMENTAL DUST CLOUDS
The method is based on the determination of
the relative magnitude of the radiation inten- 4.1. Dust cloud concentration distribution
sities from the investigated solid body at four The first phase of the experiments concen-
different wavelengths. The intensity ratios are trated on formation of a suitable experimen-
compared with the radiation intensity ratios tal dust cloud of sufficient volume and low
from black bodies of various temperatures. By minimum ignition energy. This was achieved
identifying the black body temperature that by dispersing 8 g of dried maize starch at 2 bar
gives the best fit, a best estimate of the 'colour overpressure in the compressed air line. With
temperature' of the solid body is obtained. 8 g of dust, the theoretical average dust con-
The 'colour temperature' is close to the real 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)
1.0
Temperatures of titanium sparks were also
0.9 measured 140 mm and 240 mm downstream
I 5m/,~ of the impact point. The arm velocity just
O.B
prior to impact was 24 m/s and the net impact
0.'/ energy about 20 J in all the experiments. In
~ 0.6 the case of grinding-wheel sparks, tempera-
tures were measured 40 mm downstream of
E, o.s
>,- the contact point between the steel rod and
O.l. the grinding wheel. The peripheral velocity of
g o.~ the grinding wheel was 29 m/s.
c)
The results are summarized in Table 3. Ap-
R~ 0.2
proximately 10 tests were carried out for each
0.1 combination of test object and anvil material,
using different combinations of detecting
0.1 I 10 100 1000 wavelengths in the pyrometer.
SPARK ENERGY , mJ
Fig. 6. Effect o f i m p a c t velocity o n ignition sensi- TABLE 3
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- 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
gering o f t h e electric spark is 0.3 - 10 ms. E n v e l o p e s the four-wavelength pyrometer
enclose t h e e x p e r i m e n t a l points.
Test o b j e c t Anvil material Spark
the peripheral velocity of the arm was 22 m/s, temperatures
the electric spark energy necessary to ignite C~c)
the turbulent dust cloud was two orders of Titanium R u s t y steel 2030 - 2900
magnitude higher than the 4.5 mJ necessary
Mild steel, St37 R u s t y steel 1730 - 2 3 8 0
to ignite the quiescent dust cloud. At 9.5 m/s
the electric spark energy required for ignition Alloy steel R u s t y steel 1530 - 1980
was approximately one order of magnitude Mild steel, S t 3 7 Concrete 2530 - 2730
higher. This relative ranking of the ignition Mild steel, St37 Rotating 1600 - 1730
sensitivities of the clouds only applies when grinding wheel
ignition occurs close to the point of impact.
For spark~luster ignition away from the
impact point, it would be necessary to perform The results show that the highest tempera-
separate electric spark ignition tests with the tures were measured on sparks produced from
spark gap at the location where ignition takes impacts of titanium against rusty steel, and
place, and with the appropriate delay between mild steel against concrete. The temperatures
passing the impacting arm and firing the elec- of titanium sparks at 140 - 240 mm down-
tric spark. stream of the impact point were generally
higher than at 40 mm. There was a clear ten-
dency of increasing temperature with increas-
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 ing length of travel, i.e., increasing burning
MECHANICAL IMPACTS AND GRINDING time.
The measured temperatures of sparks from
Temperatures of single-impact sparks from mild steel against concrete were also high. At
ordinary mild steel (St37), alloy steel and tita- 40 mm from the point of impact, some steel
nium were measured by means of the four- spark temperatures were even higher than
wave optical pyrometer. The anvil materials some titanium spark temperatures. However,
used were rusty steel and concrete. For the boiling point of titanium is higher than
comparison, temperatures of sparks from mild that of iron, indicating that on average the
steel in contact with a rotating grinding wheel titanium sparks should attain higher tempera-
were also measured. In the case of the single tures than sparks from steel. This was also
impacts the temperatures were measured on observed.
sparks 40 ram 'downstream' of the impact The results from the tests with mild steel
point and 10 mm above the anvil surface. as spark-producing material showed a large
159
variation in the spark temperatures, depend- temperature becomes constant and indepen-
ing u p o n the anvil material. The lowest spark dent of the pressure. A similar observation
temperatures were obtained in the tests with was made b y Allen and Calcote [7] who
the rotating grinding wheel. The observed found that low pressure sparks were relatively
temperatures of 1600 - 1730 C were very small and irregular in shape. The irregular
close to those for mild steels measured b y shape indicated that the sparks had not
Hardt [17 ], and also in agreement with Wahl's reached the melting point. The high pressure
[18] t w o ~ o l o u r p y r o m e t e r measurement of sparks were, however, spherical and coated
temperatures of steel sparks from grinding with oxide, which means that the particles
wheels. had melted and burnt.
The temperature measurements showed In the present investigation, the maximum
that impact sparks from mild steel attained contact pressure was estimated b y measuring
higher temperatures when concrete was used the static force exerted on the anvil by the tip
as wall material than when rusty steel was of the impacting arm, when the arm was left in
used. This may in part be due to the heat the vertical position. The measurement was
conductivity of concrete being much lower accomplished b y mounting a mechanical load
than that of steel, causing more of the fric- cell between the tip of the arm and the anvil.
tional heat generated during an impact to be The force was distributed over a circular area
retained in the particles. This m a y provide of diameter 4 ram. The static pressure at the
more favourable initial conditions for com- contact area as a function of excess in arm
plete combustion of the particles and hence length is given in Fig. 7.
contribute indirectly to a significantly higher
Pressure.
spark temperature. [MPa)
The observation that the temperatures of 200
quency of ignition of maize starch by tita- very few sparks compared with titanium im-
nium sparks, was studied. The peripheral pacts. Typical numbers of sparks produced
velocity of the arm was varied using different from mild steel were 5 - 50. Immediately after
springs, whereas the net impact energy was impact, no or very few sparks were visible.
varied by adjusting the excess arm length. But 3 - 5 ms after impact, some sparks became
The results are given in Fig. 8. The frequen- visible, and after 10 - 20 ms the number of
cy of ignition for each combination of arm sparks visible on the f i l l was at maximum.
velocity and impact energy is based on obser- Impacts with steel never generated a sufficient
vations from 10 or 20 tests. The scatter (+1 std. number of sparks to ignite clouds of dried
dev.) is indicated. At constant velocity of maize starch.
approach, the frequency of ignition decreases The films of ignition by sparks from tita-
systematically with decreasing impact energy, nium against rusty steel showed that the dust
as would be expected. Low-energy impacts clouds could be ignited in two different ways.
produce fewer sparks than high-energy im- (Reference 9 contains reproductions of some
pacts, and if the number of sparks is low, the of the film pictures.) First, ignition frequently
ignition probability is also low. occurred very close to the point of impact,
immediately (i.e., 0 - 5 ms) after the impact
10 had taken place. At that m o m e n t an extremely
luminous hemispherical 'pocket' close to the
09
impact point, with a radius of a few centi-
08, meters and containing 100 - 1000 sparks, was
07
observed. In addition to the great number of
discrete particles, a 'thermite' flash probably
c06 also contributed to the formation of the lu-
o
~ m/s minous pocket. Impacts of titanium against
c,
concrete also generated high spark densities
~ 0~
close to the point of impact, but not the addi-
03 tional continuous, luminous phase as was seen
with impacts against rust. Flame propagation
02
started somewhere in the region of the 'lumi-
0t nous pocket', but could neither be referred
specifically to one single metal particle nor to
I i i i i i i t i i
2 z. 6 8 10 12 ~z. 16 18 ~0 the hot spot generated at the point of impact.
Impact energy All the high-speed films of ignition of maize
Fig. 8. F r e q u e n c y o f ignition o f clouds of maize starch by titanium impacts against rusty steel
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 showed that ignition and flame propagation
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.
started in this region.
The second mode of ignition was only ob-
The impacts of similar net energies, but of served in some of the experiments with lyco-
different velocities of approach, i.e., different podium. In this case ignition occurred 1 0 - 50
turbulence levels, gave decreasing frequency ms after impact at a distance of 10 - 30 cm
of ignition with increasing turbulence. This is downstream of the point of impact. In this
in agreement with the results from the experi- region the number of sparks per unit volume
ments where the minimum electric spark ener- of dust cloud was considerably lower than in
gy for ignition was determined when the arm the 'luminous pocket' close to the point of
was moving through the dust cloud at the impact. But still it was not possible to trace
m o m e n t of spark discharge (Section 4.2). the ignition back to any single metal spark.
'Explosion' and fragmentation of single metal
sparks were observed frequently, but never
8. H I G H - S P E E D F I L M I N G seemed to cause ignition of the dust cloud.
The high-speed films suggest that a cluster
The 1000 - 2000 frames per second high- of a large number of sparks is necessary to
speed films confirmed that impacts with mild cause ignition of a cloud of grain dust. Im-
steel against rusty steel or concrete produced pacts with titanium as spark-producing
161
material generated such conditions, and usual- the spark f o r m a t i o n process. Impacts by car-
ly resulted in ignition. 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
9. RESULTS FROM IGNITION TRIALS WITH was removed, so t h a t the impact t o o k place
DIFFERENT DUST CLOUDS AND IMPACTING on t he gravel or crushed stone fraction of t he
MATERIALS concrete, visible sparks were produced. For
A series o f impact tests with various mate- this reason experiments using concrete as
rials was carried o u t in clouds o f dried maize anvil material were always carried o u t after
starch in order to identify combinations o f removal of the cem ent layer.
impacting materials t hat would cause ignition. Ignition experiments with impacts between
The anvil materials used comprised typical titanium and several anvil materials were also
c o n s t r u ctio n steels with either fresh or rusty carried out. T he combinations o f materials
surfaces, and different concrete ~ m p l e s . Sup- used, and t he results f r o m the tests, are given
p l e men tar y tests were c o n d u c t e d with anvils in Table 5. Dust clouds o f dried maize starch
o f different minerals and grinding wheels. were used in all tests.
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 TABLE 5
n o t available, b u t t h e y are likely to be repre- Results from the experiments with impacts between
sentative o f the range of steels in c o m m o n titanium and different anvil materials
use. Th e combinations o f steel test objects
and anvil materials investigated, are given in Titanium against Ignition ?
Table 4.
-- rusty, mild steel Yes
-- concrete Yes
TABLE 4 -- sandstone Yes
Combinations of steel test objects and anvil materials - granite, gneiss, quartz
- No
used in the experiments - grinding wheel
- No
- grinding wheel
-
T he n u m b e r o f sparks produced d e p e n d e d
Acid resistant steel, alloyed steel, hardened steel or
stainless steel against u p o n the surface t e x t u r e of t he anvil material.
-- mild steel A rough surface p r o d u c e d m ore sparks t han a
-- concrete s m o o t h surface, and ignitions were onl y ob-
-- granite tained f r o m impacts against anvil materials
-- grinding wheel having a relatively rough surface.
T he i m port ance of t he surface properties
T he peripheral velocity of t he arm in the of t he anvil material was emphasized during
test rig just b ef o r e impact was 24 m/s. T he the experiments with titanium against rust y
calculated average impact energy was a b o u t steel. Surfaces apparently covered by t he
20 J. After each test, t he anvil was displaced same kind o f rust, would in some cases pro-
slightly, so th at every impact t o o k place on duce large numbers o f powerful sparks t h a t
fresh anvil surface. ignited the dust cloud, whereas in o t h e r cases
T he experiments revealed t h a t neither mild considerably fewer sparks were produced and
steel against concrete, nor any o t h e r combina- no ignition occurred.
tions o f steel objects and anvil materials, pro- A few experiments with zirconium as spark-
duced an y sign o f ignition o f t he dried maize producing material were also carried out.
starch cloud. Rust y steel was used as anvil material. The
F r o m i n t r o d u c t o r y tests with concrete as impacts produced m ore sparks than titanium
anvil material, it was f o u n d t hat the nature against rusty steel, and ignition of clouds of
o f the surface was of decisive i m por t ance for dried maize starch were easily obtained.
162
(5) In most cases, ignition by titanium During his stay at CMI in the a u t u m n of
sparks was observed very close to the point 1985, Dr. R. Klemens, Technical University
of impact. However, ignition was occasionally of Warsaw, gave invaluable help, both theo-
observed 10 - 30 cm downstream of the im- retically and experimentally, with the mea-
pact point. Ignition by one single metal spark surement of metal spark temperatures by a
was never observed. A fairly dense cluster of four-wavelength optical pyrometer borrowed
sparks seemed to be necessary for igniting from TU of Warsaw.
the clouds of maize starch.
(6) A n y moving object in the dust cloud REFERENCES
reduces the ignition sensitivity o f the cloud in
the vicinity of the object by inducing turbu- 1 H. de Goijer, Th. M. Groothhuizen and M. E.
Reinders, Literature Investigation into the Dust
lence. Experiments with titanium against rusty Explosion Danger in Industries Storing and Pro-
steel showed that at a given net impact energy, cessing Cereals and Flour, Report from Tech-
the ignition frequency dropped when the im- nological Laboratory TNO, October, 1975.
pact velocity increased. Thus, at a given net 2 D. J. Rasbash, Fires Involving Dusts, Technical
impact energy, objects generating low turbu- Booklet No. 5, Fire Protection Association,
London, 1949.
lence represent a greater ignition hazard than 3 M. Verkade and P. Chiotti, Literature Survey of
objects generating high turbulence. Dust Explosions in Grain Handling Facilities:
(7) At a given impact velocity, the frequen- Causes and Prevention, Energy and Mineral Re-
cy of ignition by titanium sparks decreased sources Institute, Iowa State University, Ames,
systematically with the impact energy, as IA, 1976.
4 H. Beck and A. Jeske, Documentation Staub-
would be expected. explosionen, Analyse und Einzelfallderstellung,
(8) The microscopic nature of the anvil sur- Report No. 4/82, Berufsgenossenschaftliches
face is decisive for the spark formation pro- Institut fiir Arbeitssicherheit, St. Augustin,
cess. For example, impacts against the gravel F.R.G., 1982.
or crushed stone fraction of concrete pro- 5 G. Leuschke and J. Zehr, Ziindung yon Staub-
lagerungen und Staub/LuftGemischen dutch me-
duced considerably more sparks than impacts chanisch erzeugte Funken, Arbeitsschutz, 6
against a fresh concrete surface covered with (1962) 146.
cement. 6 T. Zuzuki, S. Takaoka and S. Fujii, The ignition
(9) The overall practical conclusion of the of coal dust by rubbing, frictional heat and sparks,
present investigation is that up to net impact Proc. Restricted Int. Conf. of Directors of Safety
in Mines Research, July 1965, Safety in Mines
energies of 20 J, tangential accidental single Research Establishment, Sheffield, England.
impacts between various types of steel, and 7 J. Allen and H. F. Calcote, Grain Dust Ignition
between steel and rusty steel or concrete, are by Friction Sparks, SMS-81-049, National Grain
unable to ignite clouds of grain and feed dust, and Feed Association, Washington, DC 20005,
or flour, even if the dusts are dry. Impacts of 1981.
8 K. Ritter, Die Ziindwirksamkeit mechanisch
standard quality aluminium against rusty steel erzeugter Funken gegeniiber Gas/Luft- und
did not even generate any visible sparks. In Staub/Luft-Gemischen, Dr.-Ing. Dissertation, Uni-
the case of titanium or zirconium, the sparks versit/it Fridericiana, Karlsruhe, 1984.
produced may initiate explosions in clouds of 9 G. H. Pedersen, Initiation of Dust Explosions by
dried dusts, but not in clouds of dust contain- Heat Generated during Single Impact between
Solid Bodies, Fire and Explosion Research Re-
ing 10% moisture or more.
port No. SMS-86-055, National Grain and Feed
Association, Washington, DC, 1986. (Prepared by
Chr. Michelsen Institute, Report No 833310-2,
ACKNOWLEDGEMENTS December 1985.)
10 G. H. Pedersen and R. K. Eckhoff, Initiation of
dust explosions by heat generated during single
CMI wishes to express its sincere gratitude impact between solid bodies, Proc. 2nd Int.
to the National Grain and Feed Association Colloquium in Dust Explosions, Jadwisin, Poland,
(NGFA), Washington, DC, U.S.A., for the November, 1986, Report No. 863351-1, CMI,
substantial financial support rendering this 1986.
work possible, and for unfailing constructive 11 R. K. Eckhoff, Use of (alP/dr)max from closed-
bomb tests for predicting violence of accidental
advice and support throughout. This paper is dust explosions in industrial plants, Fire Safety
a condensed version o f a comprehensive J., 8 (1984/85) 159. (Paper presented at the First
report prepared by CMI for NGFA [9]. Int. Colloquium on Explosibility of Industrial
164
Dusts in Baranow, Poland, November 8 - 10, 1984. 14 K. L. Cashdollar and M. Hertzberg, Infrared
12 K. L. CashdoUar, I. Liebman and R. S. Conti, Pyrometers for measuring dust explosion temper-
Three Bureau o f Mines Optical Dust Probes, Re- atures, Optical Eng., 21 (1) {1982) 82 - 86.
port o f Investigations No. 8542, U.S. Bureau of 15 R. K. Eckhoff, Towards absolute minimum igni-
Mines, 1981. tion energies for dust clouds?, Combust. Flame,
13 R. K. Eckhoff, K. Fuhre and G. tt. Pedersen, 24 (1975) 53 - 64.
Vented Maize Starch Explosions in a 236 m 3 16 W. Berthold, Bestimmung der Mindestziindenergie
Experimental Silo, Fire and Explosions Research yon Staub/Luft-Gemischen, Progress Report No.
Report No. ESV-86-070, National Grain and 134, VDI-Verlag, Diisseldorf, 1987.
Feed Association, Washington, DC, 1986. (Pre- 17 L. Hardt, Temperaturme~ungen an Schleiffun-
pared by Chr. Michelsen Institute, Report No. ken, Arbeitsschutz, 15 (1954) 430 - 436.
843307-2, December, 1985.) 18 H. Wahl, Z. Angew. Phys., 12 {1960) 60 - 62.
Erratum
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
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, loading pipe inside the silo. The ionizer is equipped with a
we developed an eliminator equipped with passive-type needle electrode grounded through a resistor of 50 MV
ionizers without a power supply [19]. This paper describes within the insulated metal nozzle. The resistor was used to
improvements of this type of eliminator demonstrated by prevent the occurrence of spark discharges. Each nozzle was
real-size experiments using polypropylene pellets. 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
2. Passive-type eliminator on the pellets during the loading, probably due to the electric
discharges from the charged pellets. Consequently, the metal
The newly developed passive-type eliminator consists of nozzle was electrified with the same polarity due to the
six or eight passive-type ionizers mounted in a row on the electrostatic induction and discharge from the charged PVC
outside of a short PVC pipe that is installed at the end of the 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
Fig. 3. Schema of the silo in which a nozzle-type static eliminator was Photo 2. Bulk surface discharges taken with an image-intensifier camera;
installed. exposure time = 10 s.
212 T. Kodama et al. / Powder Technology 135 136 (2003) 209215
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 213
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.
214 T. Kodama et al. / Powder Technology 135 136 (2003) 209215
6. Conclusions
and Safety Promotion in the Process Industries, Antwerp, Elsevier, [17] T. Suzuki, T. Mogami, T. Kodama, A. Ohsawa, S. Watano, Online
Amsterdam, 1995. measuring and controlling method for electrostatic charged powder in
[12] M. Glor, K. Schwenzfeuer, Occurrence of cone discharges in produc- pneumatic handling process, Proceedings of Annual Meeting on
tion silos, J. Electrost. 40&41 (1997) 511 516. Safety Engineering, November, Japan Society for Safety Engineering,
[13] J. Taillet, P. Poidras, M. Glor, B. Maurer, Testing a new device for 2001, pp. 15 18, in Japanese.
suppressing electrostatic charges on polymer granules, J. Loss Prev. [18] T. Suzuki, T. Mogami, T. Kodama, S. Watano, Charge elimination
Process Ind. 5 (1992) 235 237. performance of nozzle-type eliminator for powdereffect of wave-
[14] J. Taillet, Elimination of static charges in the processing of bulk form and frequency of applied AC voltage, Proceedings of Annual
material, J. Electrost. 30 (1993) 181 190. Meeting of the Institute of Electrostatics Japan, August, 2002, pp.
[15] T. Kodama, T. Suzuki, K. Nishimura, S. Yagi, S. Watano, Prevention 231 232, in Japanese.
of incendiary discharge in a silo using a newly developed electrostatic [19] T. Kodama, T. Suzuki, K. Nishimura, T. Mogami, S. Watano, A new
eliminator, Conf. Record (CD) of 2000 IEEE/IAS Annual Meeting, nozzle-type static charge eliminator used in a silo during pneumatic
October 2000, Rome, 2000, pp. 652 656. powder transport, Conf. Record (CD) of World Congress on Particle
[16] T. Kodama, T. Suzuki, K. Nishimura, S. Yagi, S. Watano, Static Technology 4, July, Sydney, 2002, p. 326.
charge elimination on pellets in a silo using a new nozzle-type elim-
inator, J. Electrost. 55 (2002) 289 297.
Powder Technology 135 136 (2003) 223 233
www.elsevier.com/locate/powtec
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.
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
tion. However, it must be kept in mind that not only high- 5. Discharges associated with installations and
resistivity materials such as polyethylene, polypropylene, equipment
etc. lead to resistances to earth far above this order of
magnitude. Most pigments, dyestuffs, additives, drugs, 5.1. Spark discharges
foodstuffs, etc. show resistivities above 1010 V m.
As accumulation of charges grows higher, the resulting Spark discharges occur between two conductors at
electric field in the air may reach its limit value. This limit different potential as soon as the electrical field in the
value is also called the dielectric strength of air. Under gap reaches the breakdown value of about 3 MV/m under
normal conditions, it amounts to about 3 MV/m. At this normal atmospheric conditions. In practice, one of these
limit a discharge may occur. All or only part of the energy conductors (including personnel) reaches a high potential
stored in the charge accumulation may be released in a hot since it is not connected to earth; the other conductor is
discharge channel that may ignite a given flammable normally at earth potential. Nearly the total energy stored
atmosphere. in such systems (capacitors) is released in a single spark,
The energy released in the discharge and the sensitivity which generates a single discharge channel of rather high
of a given flammable atmosphere measured in terms of its current density (Fig. 4). Therefore, these discharges are
minimum ignition energy determine whether ignition will rather incendive and their energy release can simply be
occur or not. The assessment of the occurrence and calculated by the energy W stored in the capacitor using
incendivity of discharges in different situations in practice the potential U (in Volts) and the capacitance C (in
represents the most important, but also the most difficult, Farads):
step in the analysis of electrostatic hazards. The basic
principles and typical examples are given in the following 1 2
paragraphs. W CU 1
2
I q dm=dt 2
U IR 3
Q UC 4
Fig. 5. Electrical circuit simulating the charge buildup on a conductor.
This is the case if all the current flows to ground via the
In principle, gases, vapours and dust clouds can be leakage resistor R. Based on Eqs. (1) (4), the maximum
ignited by spark discharges. Earthing and bonding of all energy of spark discharges generated by a small plant item
conductive parts, which may become charged, is therefore (conductor) not safely connected to ground can be calcu-
mandatory in areas where flammable atmospheres formed lated. Fig. 6 shows the results for such a plant item
by these substances have to be expected. It is commonly having a capacitance of 20 pF. Since the potential is
agreed [8,9] that a resistance to earth of less than 106 V is limited by the geometrical arrangement and the dielectric
generally sufficient in the case of equipment and 108 V in strength of air, the limit value of the spark energy for
the case of personnel to avoid spark discharges. Even different potentials is indicated.
higher resistances to ground may not necessarily lead to Since the energy increases with the square of the charg-
hazardous potentials under moderate charging conditions. ing current and with the square of the leakage resistance,
If, however, only minor changes in the operating proce- minor changes of these quantities have a huge effect on the
dures (slightly different resistance to ground and/or charg- ignition hazard. These considerations explain why many
ing current) occur, this may have a drastic effect on the operations may run without any problems for a long time
potential acquired. The potential buildup on a container and all of a sudden an explosion occurs.
filled with charged product or on a piece of pipe in a
pneumatic transport system or on any capacitor, which 5.2. Brush discharges
becomes charged by static electricity, can be calculated by
using the electrical circuit shown in Fig. 5. For example, Charges of one polarity accumulated on the surface of
in the case of filling a container with powder, the charging nonconducting installations and equipment may produce
current I is calculated from the charge to mass ratio q of very high electric fields in the surroundings. As soon as
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.
228 M. Glor / Powder Technology 135 136 (2003) 223233
an earthed conductive electrode is approached to the sur- Projection of internal fittings into a highly charged dust
face, the electric field is locally distorted and the dielectric cloud or a highly charged droplet cloud.
strength of air may be reached (3 MV/m at atmospheric Charging of nonconductive powder into packing drums,
conditions) at the surface of the electrode. As a result, the containers or silos, approach of the highly charged dust
so-called brush discharges will occur. An exact mathemat- surface to the internal fittings or lowering of a conductive
ical treatment for a special arrangement of the field gener- earthed sample baker for sampling or a level probe to
ating charges and of the electrode (cylindrical symmetry) is determine the level.
given in Ref. [5]. Fig. 7 shows a sequence of several brush Projection of flagpoles, ships masts, antennae or ice axes
discharges generated at the surface of an earthed spherical into powerful atmospheric electric fields (St. Elmos fire
electrode. during thunderstorms).
It depends on many parameters such as on the radius of
curvature of the electrode, on the speed at which the The characteristics and incendivity of brush discharges
electrode is introduced into the electric field and on the have been investigated by different authors (see Refs. [10,11]
polarity, whether a corona or a brush type discharge will for details). The values reported for the equivalent energy
actually occur in such situations. In practice, it can generally have been determined with gas/air-mixtures and lie in the
be assumed that brush discharges occur, if the radius of range of 1 to 3.6 mJ. The parameters affecting the incendivity
curvature of the electrode is larger than 5 mm. of brush discharges are the radius of curvature of the
In addition to the situation with a charged insulating electrode, the polarity of the electric field andif the electric
surface, brush discharges may also occur, if an earthed field is generated by a highly charged nonconducting sur-
conductive electrode is introduced into a high electric field facethe surface charge density and the area of the surface.
of any origin. The high electric field may, for example, Brush discharges from an earthed, metallic electrode opposite
result from a highly charged insulating liquid or suspen- to a negatively charged surface are by far more incendive than
sion, from a highly charged mist, from highly charged from an electrode opposite to a positively charged surface.
insulating powder in bulk or from a highly charged dust On the basis of the values of the experimentally deter-
cloud. In what follows, examples of plant operations that mined equivalent energy of brush discharges, it is expected
are the most frequent sources of brush discharges are that most flammable gas or solvent vapour atmospheres and
summarized: hybrid mixtures can be ignited by brush discharges.
The question, whether brush discharges from highly
Approach of an earthed conductive electrode such as a charged insulating surfaces of equipment or packaging are
tool or a finger tip to highly charged insulating surfaces able to ignite powders in the absence of flammable gases or
(e.g., plastic pipe for the conveyance of liquids or dusts, solvent vapours is still the subject of an ongoing debate. The
plastic bags, common flexible intermediate bulk contain- reason being contradictory results obtained from gases and
ers, plastic packing drums or filter cloths as well as film vapours on the one hand and from powders on the other
webs, nonconductive conveyor belts or V-belts). hand. Although the minimum ignition energy of certain
Discharging of solids from plastic bags, or shaking out of powders lie in the range of 1 10 mJ, no ignition of a dust
plastic bags in the vicinity of metal fittings (e.g. above an cloud clearly caused by a brush discharge has yet been
access port of a reaction vessel). 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
Fig. 7. Brush discharge [9]. surface resistance, measurement of charge decay times is
M. Glor / Powder Technology 135 136 (2003) 223233 229
approach of an earthed electrode to a highly charged The energy released in propagating brush discharges is
insulating surface on top of a metal plate. high enough to ignite most flammable gases and vapours as
For a long time, it has been assumed that these discharges well as most combustible powders. Persons may suffer a
can only occur if one surface of the insulating layer is in serious shock when the human body acts as the initiating
intimate contact with an earthed metallic plate. In this case, electrode for such a discharge. Propagating brush discharges
the second charge layer of opposite polarity is accomplished must be excluded in areas where flammable atmospheres
by the induced charges on the metallic plate. It can, however, formed by flammable gases or vapours or by combustible
be demonstrated, by spraying charges of opposite polarity to powders may be present.
the two surfaces of a dielectric sheet, that the metallic plate in Propagating brush discharges can be excluded by the use
intimate contact with one surface of the insulating sheet is of conductive materials or insulating materials of low di-
not essential for the buildup of the double layer. If a dielectric electric strength at all locations where the buildup of high
sheet in free space (e.g., the insulating wall of a container) surface charge densities may occur. If the breakdown voltage
becomes highly charged on one surface and the electric field across a nonconducting layer or sheet is less than 4 kV,
is mainly directed through the dielectric sheet towards earth, propagating brush discharges will not occur.
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 6. Discharges associated with powder
filling an insulating container such as a FIBC with highly
charged polymeric powder. 6.1. Brush and corona discharges
From practical experience and experimental evidence, it
is generally agreed that the high surface charge densities In addition to the situation with a charged insulating
necessary for propagating brush discharges cannot be surface described in Section 5.2, brush discharges may
achieved by separation processes in manual operations such always occur if an earthed conductive electrode is intro-
as wiping of insulating surfaces or discharging a powder duced into a high electric field resulting from a highly
from a plastic bag. Charging mechanisms associated with charged insulating powder in bulk or from a highly
high separation velocity are necessary to build up the high charged dust cloud. With respect to energy and incendivity,
surface charge densities necessary for propagating brush the same arguments and figures apply as mentioned in
discharges. Such high-charge buildup may be observed Section 5.2.
during the following industrial operations:
6.2. Lightning-like discharges
High-velocity pneumatic transfer of powder through an
insulating pipe or a conductive pipe with an insulating Before the phenomenon of cone discharges described in
internal coating. the next section was discovered, the charged dust cloud
Use of inspection windows made from glass or Plexiglas was considered to be the most dangerous source of elec-
in pneumatic transfer pipes. trostatic ignition in a silo or container. This hypothesis was
High-velocity transportation of a highly insulating originally based on the lightning flashes observed in the
suspension through an insulating pipe or a conductive clouds of dust and ashes generated by erupting volcanoes.
pipe with an insulating internal coating. Since then, systematic tests have been carried out in a 60-
Continuous impact of powder particles onto an insulating m3 container [15]. The results obtained from these tests
surface (e.g., a coated dust deflector plate in the cyclone indicate that in silos with a volume of 60 m3, or in those
of a dust separator). with any other volume but a maximum diameter of 3 m,
Fast rotation of conveyor or transmission belts made discharges of the lightning type are not to be expected. For
from insulating material or from conductive material, larger silos, no predictions based on firm experimental
which is coated with an insulating layer of high dielectric evidence can be made. Nevertheless, the prevailing opinion
strength. is that lightning-type discharges in industrial-size silos are
Filling of large containers or silos made from insulating highly improbable. The investigations now described have
material (e.g., flexible intermediate bulk containers) or of shown that much more importance must be paid to cone
metallic containers or silos internally coated with an discharges.
insulating layer of high dielectric strength.
6.3. Cone discharges
According to practical experience and experiments in the
laboratory, it is not expected that layers of insulating pow- Cone or bulking discharges were first observed in 1979
ders that are, for example, formed in metallic pipes during by Maurer [16] while silos were being filled with polyeth-
pneumatic transfer, or paints, will give rise to propagating ylene granules. Results published in the same year [17] on
brush discharges unless the powder is subjected to a sintering the electrostatic discharges occurring during the filling of
or melting process. 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
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 ing potentially explosive atmospheres arising (for example
range: by excluding flammable gases, vapours and fine dust or by
reducing the oxygen content), or designing the plant with
0:5 m < D < 3:0 m explosion protection elements.
and
This does not mean that the empirical equation is a priori Based on experience and industrial research, the ignition
no longer valid outside the stated range. It is, however, no hazards associated with static electricity can nowadays be
longer backed up by measured results. Extrapolations can assessed rather well. The present knowledge in this field is
therefore only be recommended after consultation with an sufficient to prescribe safety measures for most common
expert. industrial processes. It must, however, be kept in mind that
The occurrence of cone discharges depends on whether electrostatic phenomena are encountered so widely that it is
such a strong electric field can be formed in the silo from not possible to cover all cases and that research in the field
the product being filled into it such that the dielectric continues providing new information. In addition, new
strength is reached. Experience indicates that this is espe- processes are developed, common processes are run under
cially the case when silos are filled from pneumatic con- more hazardous conditions or conflicts of interests do not
veying systems and if the charge introduced into the silo allow a standard solution. Thus, there are and there will be
from the system cannot be dissipated back to ground open questions concerning the assessment of electrostatic
quickly enough. ignition hazards.
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- References
edge, this is probably a rather conservative value. Initial
model calculations have been carried out for orientation [1] H. Beck, VDI Ber. Nr. 975 (1992) 47.
purposes, taking into account the silo geometry, the product [2] W. Bartknecht, Explosionsschutz, Grundlagen und Anwendung,
specific charge and the filling rate. These have shown that, Springer Verlag, Berlin, 1993.
[3] R.K. Eckhoff, Dust Explosions in the Process Industries, Butterworth-
depending on the situation, values up to two orders of 10
Heinemann, Oxford, 1991.
greater than this limit may possibly be regarded as still not [4] R. Siwek, C. Cesana, Process Saf. Prog. 14 (1995) 107.
being critical. More research is needed on this point. [5] M. Glor, Electrostatic Hazards in Powder Handling, Research Studies
In general, it is hardly possible to increase the conduc- Press, Letchworth, Hertfordshire, 1988.
tivity of bulk materials in industrial practice. In situations [6] European Standard EN 1127-1, 1997.
[7] A. Gitzi, K. Schwenzfeuer, M. Glor, Proceedings of the 10th Interna-
where cone discharges cannot be excluded, and the antici-
tional Symposium on Loss Prevention and Safety Promotion in the
pated equivalent energy is the same or greater than the Process Industries, 19 21 June 2001, Stockholm, Elsevier, Amster-
minimum ignition energy of the product to be processed, dam, 2001, p. 909.
other measures therefore need to be taken, such as prevent- [8] CENELEC Report R044-001 Safety of machineryGuidance and
M. Glor / Powder Technology 135 136 (2003) 223233 233
recommendations for the avoidance of hazards due to static electric- [16] B. Maurer, Chem. Eng. Technol. 51 (1979) 98.
ity 1999. The document is available from BSI-British Standards [17] A.R. Blythe, W. Reddish, Inst. Phys. Conf. Ser. No. 48, The Institute
Institution as published document PD R044-001, 1999. of Physics, Bristol and London, 1979, p. 107, Oxford.
[9] ISSA Prevention Series No. 2017 (E), Static Electricity-Ignition Haz- [18] B. Maurer, VDI Rep. 494 (1984) 119.
ards and protection measures, Heidelberg, 1996. [19] B. Maurer, M. Glor, G. Luttgens, L. Post, J. Electrost. 23 (1989) 25.
[10] N. Gibson, F.C. Lloyd, Br. J. Appl. Phys. 16 (1965) 1619. [20] M. Glor, B. Maurer, G. Luttgens, L. Post, J. Electrost. 23 (1989) 35.
[11] M. Glor, J. Electrost. 10 (1981) 327. [21] M. Glor, B. Maurer, J. Electrost. 30 (1993) 123.
[12] K. Schwenzfeuer, M. Glor, J. Electrost. 30 (1993) 115. [22] M. Glor, B. Maurer, VDI Rep. 1272 (1996) 87.
[13] K. Schwenzfeuer, M. Glor, J. Electrost. 40 and 41 (1997) 383. [23] M. Glor, J. Electrost. 15 (1984) 223.
[14] O. Larsen, H.J. Hagen, K. van Wingerden, J. Loss Prev. 14 (2001) [24] A. Kuttler, M. Glor, J. Electrost. 30 (1993) 285.
111. [25] V. Ebadat, Thesis, University of Southampton, 1989.
[15] P. Boschung, W. Hilgner, G. Luttgens, B. Maurer, A. Wimer, J. Elec- [26] M. Glor, K. Schwenzfeuer, J. Electrost. 40 and 41 (1997) 511.
trost. 3 (1977) 303.
Journal of Loss Prevention in the Process Industries 16 (2003) 333340
www.elsevier.com/locate/jlp
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.
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
2. Experimental
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
mm 220 mm and was 1000 mm high and was bolted under pressure from an external injection system and
directly to the top of the test vessel (Fig. 4). was ignited either at the centre or rear of the vessel using
The principle of using small gaps to prevent ignition a 50 kJ ignition source. The charge chute was attached
breakthrough, for example in rotary valves, is well to the vessel at a central position on the top of the cylin-
recognised. Typically, in a 250 mm rotary valve the gap drical body of the vessel. The propagation prevention
is likely to be 1 mm. However, whilst small gaps are device was bolted directly to the top of the charge chute.
desired to prevent ignition breakthrough, the gaps in the
propagation prevention device need to be significantly 2.2.3. Test procedure
greater because of the requirement for the free flow of The prototype device was tested under a range of
material. In vented vessels where the explosion pressures explosion conditions. Initial tests were carried out to
are expected to be low (500 mbar) a larger gap is likely establish the test conditions that gave the correct reduced
to be suitable. In the design used for this work a gap of explosion pressure for the KSt of the dust in accordance
40 mm at the tips of the baffle plates was employed. with vent sizing equation described in VDI 3673 Part 1
The free flow of material over the inclined plates also for homogeneous dust distribution. These conditions
needed to be considered and the angle of repose of bulk were then used in the tests with and without the propa-
granular materials was taken as evidence of its likely gation device attached to the chute. The effectiveness of
flow behaviour. Woodcock (Woodcock & Mason, 1987) the device with respect to its ability to prevent flame
considered 2530 degrees to be very free flowing and propagation was assessed by comparing the length of the
3038 degrees to be free flowing. In this device, the flame from the chute without the device with the length
short plate was inclined at 30 degrees and the long plate of the flame (if any) when the device was installed in
inclined at 20 degrees. the chute.
Fig. 4. Test vessel with charge chute. Fig. 5. Propagation prevention device for pipes.
336 P. Holbrow, A. Tyldesley / Journal of Loss Prevention in the Process Industries 16 (2003) 333340
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)
3.1.3. Effect of device on flame lengthvent area 0.26 3.3.2. Effect of device on flame lengthvent area
m2 0.126 m2
With a larger relief vent of 0.26 m2, the jet of flame Without the device installed in the chute, and with an
from the chute was slightly smaller at 1.5 m. When the explosion relief vent area of 0.13 m2, a small tongue of
test was repeated with the device installed, the jet of flame approximately 0.25 m in height was released from
flame from the chute was completely eliminated. 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
4.1.1. Effect of device on Pred 4.2.3. Effect of flame lengthvent area 0.66 m2
The device acted as an obstruction in the pipe and With a vent area of 0.66 m2, a flame of approximately
tended to increase the reduced explosion pressures when 4 m long was ejected from the 1.2 m long pipe. The
compared with an open pipe. For example, with a vent two-cone and three-cone devices were tested, with a
area of 0.96 m2, the Pred was 150 mbar without the reduction in the flame length to 1 m using each of the
device and increased to 161 mbar with the device. 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
mbar with the one-cone device fitted. The Pred did not in the chute, the venting capability of the chute was
exceed 100 mbar with the two-cone device fitted. reduced and this increased the explosion pressure.
4.3.2. Effect of flame lengthvent area 0.27 m2 5.2. Propagation prevention device for use in
With a vent area of 0.27 m2, a jet of flame approxi- pipelines
mately 4 m long was ejected from the 0.6 m long pipe.
When this test was repeated with a one-cone device fit- The device was successful in either eliminating or
ted, the flame length was reduced to 0.25 m and when achieving a significant reduction in the flame from the
the test was repeated with a two-cone device fitted, the vessel over a range of vent areas. Without the device
flame was completely eliminated. 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
5. Conclusions vessel. The pipe stub was approximately the same length
as the propagation prevention device (0.30.9 m). Flame
5.1. Propagation prevention device for use in charge lengths measured from the pipe stub, without the device
chutes installed, were approximately 4 m using milk powder,
13 m using coal dust and 45 m using anthraquinone.
The tests have demonstrated that dust explosions in a When the propagation device, comprising two or three
2 m3 vessel, with low explosion pressures, will propagate cones, was installed the flame from the pipe stub was
significant flames through an attached open charge chute. generally eliminated or greatly attenuated.
Coal and anthraquinone flames from the open charge Flame from the coal and milk powder was completely
chute were up to 2 m long and had the potential to cause eliminated using a two-cone device. A three cone device
serious injury to any operator close to the chute. Milk was successful in the complete elimination of the anthra-
powder, the least reactive of the three test dusts, was the quinone flame when a relatively large vent was installed
exception and had a flame of only 0.25 m long. (0.96 m2), and 75% reduction in flame length was achi-
The attachment of the explosion propagation preven- eved when a smaller vent configuration was used (0.66
tion device to the charge chute resulted in the total elim- m2). Even though it was not possible to completely elim-
ination or significant reduction in the flame from the inate the anthraquinone flame, the 75% reduction in
charge chute. The results demonstrated the effectiveness flame length would reduce the potential for ignition of
of the device with explosion pressures within the range dust in the pipe. However, the attenuated anthraquinone
111 mbar to 303 mbar. The device was particularly flame was still capable of igniting the dust in a 5 m long
effective with coal dust, the results showing a complete pipe attached to the device and causing propagation of
elimination of the propagating jet of flame from the top flame down the length of the pipe.
of the charge chute. With the more reactive anthraqui- The device, in its current form is therefore not suitable
none dust, the device was again very effective, but a for reactive dusts similar to anthraquinone (KSt 308 bar
small flame of 0.5 m was observed. The milk powder m/s). However, with further development it may be poss-
flame was completely eliminated with the device. ible to achieve a more effective device for use with the
In the event of a dust explosion in a 2 m3 process more reactive dusts.
vessel (or similar size vessel) where low explosion press- In general, the presence of the device in the pipe acted
ures are anticipated, this type of device would provide as an obstruction and resulted in an increase in the
a high level of protection to personnel. The dusts tested reduced explosion pressure within the vessel.
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. References
Although the device described in this paper has been
very effective in these tests, it may require further Holbrow, P., Hawksworth, S. J., & Tyldesley, A. (2000). Thermal radi-
ation from vented dust explosions. Journal of Loss Prevention in
refinement and development to suit other situations. the Process Industries, 13, 467476.
Without the device, the open chute was effectively an Woodcock, C. R., & Mason, J. S. (1987). Bulk solids handling: An
explosion vent. However, when the device was installed 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
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.
Studying silo vulnerability involves conducting Evidently, for an explosion to actually occur, the
research into all possible scenarios. That is why analyz- explosible atmosphere must encounter a source of
ing accidents occurring in silos (Fig. 1) for instance, inflammation such as a hot point, an electrostatic dis-
demonstrates that in addition to a diagnostic approach charge, etc. Confinement is important because it causes
to the explosion hazard, it is imperative that a diagnostic pressure effects to increase significantly.
approach to the structures behavior be conducted. Dust explosibility is a difficult matter to study because
of the high number of properties the dust has, unlike
1.2. Contingencies: dust explosion gases. These characteristics are (Eckhoff, 1991): dust
granulometry and humidity rate, concentration, oxygen
1.2.1. Explosible atmosphere rate, temperature, pressure, turbulence, confinement, vol-
The explosion hazard diagnosis approach should be ume and dusty atmosphere flammability minimum
organized by following the process of an explosion. In energy.
order for an explosible atmosphere to occur, the follow-
ing four conditions should be present simultaneously: 1.2.2. Enclosure wall pressure
An enclosure rupture mode is contingent upon the
the dust must be flammable; pressure signal received by it. This signal is, at some
the atmosphere must contain combustible material; point, the evolution of pressure over time P(x,y,z,t). Two
situations should be considered:
Corresponding author. Tel.: +33-2-48-48-40-46;fax: +33-2-48-48- the combustion mechanism occurs at low speed in
40-40.
E-mail addresses: guy.mavrot@ensi-bourges.fr (G. Mavrot); isab- relation to the celerity of sound waves through the gases
elle.sochet@ensi-bourges.fr (I. Sochet); patrice.bailly@ensi-bourges.fr contained in the enclosure; in this case one has to con-
(P. Bailly). 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 Let us consider a reinforced concrete wall receiving
which, in a number of instances, is a relatively slow a dynamic load to one side.
deflagration, The phenomena observed (Fig. 2) are as follows:
the burning rate may increase for different reasons,
i.e. because of turbulence or overpressure effects, and compaction of concrete near loaded side,
the explosion may move toward a supersonic deton- scaling of non-loaded side, and
ation rate. shear failure or bending of the structure.
It is usually accepted that a pressure increase rate not The mechanisms are herein described separately, but
exceeding 1 kPa/s leads to a loading that may be mod- they are actually related and simultaneous.
eled almost statically, while a rate of the order of 100
kPa/s induces a dynamic that should not be neglected 1.3.2. Compaction of loaded side
(Baker, Cox, Kulescz, & Strehlow, 1983). A few micro-seconds after the explosion and in the
The rupture occurring as a result of the explosion cre- case of high-intensity strains, the loaded side is subjected
ates breaks and throws fragments to the exterior. Frag- to high pressure; the concrete may get superficially pul-
mentation is accompanied by a blast wave through air verized. A decrease in volume occurs as a result of the
which propagates through the environment. Secondary collapse of the porous structure of the material. Pressure
explosions may occur as a result of the wave propagation levels in silo explosions are generally too low for this
(Crowhurst, Colwells, & Hoare, 1995). The internal to happen.
overpressure then evolves according to the reactants
being discharged through the breaks resulting from frag- 1.3.3. Wave reflection scaling to non-loaded side
mentation and through natural openings. Scaling of a reinforced concrete element results from
It should be underscored that all of these loading a break caused by a traction occurring perpendicularly
modes have been taken into account in the EFFEX to its free surface. This very rapid phenomenon is linked
software (Proust, 1997; Proust, Roux, & Chhuon, 2000).
Table 1
Fragment typology
Types of Load elements related to Elements affecting fragment size Elements affecting fragment projection speed Ve
fragments this rupture
pose of this is to provide information on concrete quality other general arrangements. Tower silos are sets of
and on the nature and geometrical layout of the frames. contiguous cells whose fundamental aspect is that they
In principle, it should be pointed out that sampling are much higher than they are wide, regardless of the
locations are to be determined in a timely manner, so that geometric design of each tower. The following are com-
relevant information may be obtained from a minimum monplace:
number of samplings.
It is assumed that the determination of sampling cylindrical cells with a square or rectangular section,
locations depends upon: (i) the activity and functions circular cylindrical cells which, when laid flat, may
related thereto, within the structure; (ii) the nature of the either be fitted into one another in a staggered pattern,
structure and of the various components. There will be or arranged mutually tangentially four by four, thereby
a focus on structure resistance under the combined action enclosing a space which may or may not be in the shape
of the contract loads and of the loads resulting from acci- of a wedge whose cross-section is that of an ace of
dental overpressure that may occur within the structure. diamonds and is usually referred to as such. There usu-
When accidental overpressure occurs within the struc- ally are two or three rows enclosing one or two rows of
ture it should be underscored, first and foremost, that aces of diamonds. Using an ace of diamonds for stor-
horizontal structural elements (floors and associated age purposes poses specific strength problems. Again, it
components) are usually and exclusively calculated to should be underscored that in some very rare cases, these
carry loads, hence to withstand descending actions. cells may be divided up into two, three or four segments
by the use of vertical webs whose cross-sections are
2.2.2. Tower silo diametral, Y-shaped or cross-shaped. These webs may
2.2.2.1. Design characteristics The structure of a con- be built at the time of silo construction or added at a
crete storage tower silo (Fig. 4) has some variations and later stage. Below are statements concerning this latter
type of situation which poses specific strength problems.
at various heights on the cell wall itself. A distinction projectiles, as well as where these projectiles may be
should be drawn here between circular and rectangular thrown.
cross-section cells. Therefore, in practice, three sam- The purpose of this article is to provide diagnostic
plings should be performed: one in the higher section, elements only statically in response to the first
one in the lower section and at least another one at mid- issues just discussed. In other words, the matter of prime
height. It is absolutely imperative that a sampling be per- concern is the Ultimate Limit State (ULS), defined to
formed slightly below the cell roof. Under normal oper- mean that the works or an element thereof has been dam-
ational conditions, this area is not awash in grain. The aged as a result of loss of static balance, rupture, buck-
analysis of some concrete examples demonstrated that ling, etc. (Perchat & Roux, 1999). Evidently, this ulti-
those sections not subject to ensiled matter load under mate limit state should in no way be confused with the
normal operational conditions have a very low level of Serviceability Limit State (SLS), defined as the limit
reinforcement and are incapable of resisting any signifi- state for normal operational and durability conditions.
cant internal overpressure; This approach is obviously analytical and takes the
in the drawbell, designating the funnel-shaped device various structural elements into consideration in suc-
designed to facilitate grain flow by gravity. It is cession. A conclusion can be drawn with regard to the
important to know how the drawbell is attached to the entire structure under study and by highlighting the
post. Samplings therefore concern the securing arrange- weakest parts and the actions likely to cause damage. It
ment as well as the drawbell wall. An additional sam- is only after such diagnosis and having regard to the
pling should be performed to search for radial reinforce- various explosion modeling patterns irrelevant for
ment. discussion here that the analyst can, in practice, define
the possible measures to be taken, keeping in mind the
The following samplings concern the cell roof slab, protection of a storage unit against the effects of an
insofar as the slab includes all or part of the reinforced explosion which could originate therein.
concrete. Also, the samplings to be performed should
allow the characterization of the roof slab design with 3.2. Data to be considered
regard to its capability to withstand ascending loads. It
should be clearly understood that those ascending efforts The data to be considered primarily comprise the
result from accidental internal overpressure. characteristics of the ensiled matter: its volume weight
The types of wall designs and top gallery roofing g kg/m3 and angle of repose f in degrees (Fig. 5). Also
should then be examined. 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
3. Vulnerability study (m) of this section, the dimension of its maximum
medium filling plane. It is therefore inferred that when
3.1. Procedure the ensiled matter is flowing from an opening in the bot-
tom of the cell roof, it follows the shape of a cone typi-
The aim of the approach outlined hereafter is to evalu- fied by the internal friction angle. This maximum
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:
medium filling plane is by definition the horizontal plane Comparing both curves allows us to determine the
including the center of gravity of the said cone. Through- area of the cylindrical cell post where the difference
out the following text, the dimensions of this plane are between the capable total pressure and the pressure
regarded as the origin of the depth vertical axis. resulting from the ensiled matter load is the weakest.
Regarding this, the authors found out that in some
3.3. Stability analysis of various types of cells cases the said area of concern was located just below
the roof bottom, especially as a result of very weak
For each type of cell, the approach is about evaluating reinforcement in the area which is not awash in grain.
the side thrusts resulting from the ensiled matter and, A number of methods are available to evaluate side
subsequently, comparing them with the total pressure thrusts on the walls, while filling or draining. These cal-
compatible with the reinforcement. culations are of interest only when verifying that the
From experience the DIN 1055 approach should be drawbell is securely attached. From experience, in most
accepted with some degree of caution among all the cases due consideration is given to this aspect at the time
approaches put forward by Reimbert & Reimbert, 1959, of construction.
Albiges & Lumbroso, 1964, Laredo, 1964, Guerrin,
1969, Lumbroso, 1970, 1977. In most cases, its use pro- 3.3.2. Aces of diamonds: a specific issue
vides the most significant thrust values and as a result The horizontal and vertical components resulting from
leaves the narrowest possible margin to statically evalu- ensiled matter thrust are analogous to the foregoing (see:
ate the accidental overpressure the structure can with- circular cylindrical cell).
stand without sustaining important damage. The walls are composed of four quarter-circle shaped
webs. As a result, the loads are strictly evaluated accord-
3.3.1. Circular cylindrical cell ing to the type of links between the webs and the adjac-
The DIN 1055 standard allows us to define the two ent circular cells.
corresponding curves: It should be underscored that under the load of the
ensiled matter, each of these four webs applies a certain
for the first curve, to the extreme shell of the side degree of tangential thrust (p.r) to the cells with the pur-
thrusts to which the wall is subjected while filling. The pose of separating them.
points on this curve correspond to a completely full cell;
A number of authors, including Lumbroso, 1970,
for the second curve, to the extreme shell of the same 1977, Albiges & Lumbroso, 1964, Laredo, 1964 et al.
thrusts, only while draining. have attempted to theoretically identify the value of that
thrust taking into account the unquestionable aspect of
After evaluating the side thrusts, the resistance offered
redundancy. However, they have based themselves upon
by the cylindrical cell must be calculated on the basis
assumptions that have never been applied in the field.
of its reinforcement level.
In practice, the area of contact between two circular
Insofar as such a cylindrical cell is only under traction
cells even if its outline resembles a beam does
load, a simple calculation, depending upon ring
not feature the specific framing allowing it to be verti-
reinforcement, allows us to plot on the same diagram
cally flexible.
as previously and according to the various steel sections
Consequently, the structure is balanced through the
highlighted at the various sampled dimensions the
use of adjacent circular cells that should reasonably be
total side thrusts (ensiled matter+overpressure) the struc-
considered empty in order to deal with the most unfavor-
ture is capable of withstanding (Fig. 6).
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.
3.3.6. Fragment projection show that the examination of the cell framework
A method to assess the projection velocity of large resistance brings us around to considering admissible
fragments is proposed in the regulation TM 5-1300. static overpressure values for accidental explosions, in
Fragment velocity depends upon the excessive impulse ultimate limit state, by a maximum of a few hundreds
defined as the difference between blast impulse and the of millibars,
section ripped away as a result of the rupture. This is look at the attachment and strength of the drawbell
the area limited by the time resistance curve (Fig. 7). and how the abutment slab is constructed. This step is
Although different fragments can travel at different part of a general arrangement which includes the top
speeds, the average debris projection speed may be gallery, the handling tower, the undercell crawlspace as
approximately assessed from excessive impulse ie and well as the ace of diamonds, if any,
wall unit weight m: 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 explosion characteristics of vented dust explosions. IchemE Symp.
between the adjacent circular cells, providing an appar- Series no. 139.
Guerrin, A. (1969). Traite de beton arme. France: Dunod.
ent beam outline vertically, shall be reinforced. Eckhoff, R. K. (1991). Dust explosions in the process industries.
inform risk analysts of the opening possibilities in Oxford: ButterworthHeinemann.
the cell covering floor, undercell gallery, handling tower, Laredo M. (1964). Etude tridimensionnelle des cellules as de carreau
show that to provide protection against explosions, des grands silos cylindriques. Memoire enregistre au Secretariat
all newly erected concrete storage units shall technically des Annales des Ponts et Chaussees, France.
Lumbroso A. (1970). Determination numerique des sollicitations exer-
include: (i) frames whose section provides enough mar- cees par la matiere ensilee dans les silos. Annales de lInstitut Tech-
gin for overpressure, especially in the area not awash in nique du batiment et des travaux publics no. 267268, France.
grain, and under vertical bending; (ii) large openings on Lumbroso A. (1977). Bases scientifiques du calcul des enceintes
top of each cell and in the top gallery, handling tower renfermant des massifs pulverulents et du calcul des silos. Annales
and undercell crawlspace, equipped with explosion vents de lInstitut Techniques du batiment et des travaux publics no.
346, France.
and restraints; (iii) efficient separations between each Perchat, J., & Roux, J. (1999). Pratique du B.A.E.L. 91. France: Edi-
space and adjoining spaces to avoid spreading and, poss- tions Eyrolles.
ibly, building up of the explosion to neighboring attach- Pontiroli C. (1995). Comportement au souffle des structures en beton
ments. 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-
References sentee a la journee DRIRE organisee par le ministere de lenviron-
nement, Paris.
Albiges M., & Lumbroso A. (1964). Silos a cellules principales circul- Proust Ch., Roux P., & Chhuon B. (2000). Prevoir les effets des
aires et intermediaires en as de carreau. Annales de lInstitut Tech- explosions de poussieres sur lenvironnement, EFFEX, un outil de
nique du batiment et des travaux publics, no. 204. simulation. Rapport INERIS, France.
Baker, W. E., Cox, P. A., Kulescz, J. J., & Strehlow, R. A. (1983). Reimbert, A., & Reimbert, M. (1959). Silos, theorie et pratique.
Explosion hazards and evaluation. New York, USA: Elsevier. France: Editions Eyrolles.
Brara, A. (1999). Etude experimentale de la traction dynamique du TM 5-1300. Structures to resist the effects of accidental explosions.
beton par ecaillage. These, Universite de Metz, France. Department of the Army, the Navy and the Air Force, Technical
Crowhurst D., Colwells S.A., & Hoare D.P. (1995). The external manual NAFVAC 397/AFM 88.
Journal of Loss Prevention in the Process Industries 16 (2003) 2732
www.elsevier.com/locate/jlp
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.
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 distance as a unit. Furthermore, combustion behaviour
(Zockoll, 1989; Alfert et al., 1988). in plant with flowing air will be different to when the
The risk of an explosion due to hot nests encountering air is stationary, as Zockoll (1989) has shown.
a dust cloud has been studied in this project. Burning At present, however, it is unclear which properties of
nests with different characteristics have been introduced a dust nest make it an effective ignition source. Despite
into clouds of dusts covering a range of Minimum all the reports of ignition incidents in industrial plant,
Ignition Temperatures (MITs). Dusts with different experimental studies have in the main indicated that
Train Firing properties have been used as the nest ignition of dust clouds by hot nests is not easy. Pink-
material and the dust clouds have MIT values of wasser (1985) showed that smouldering material
approximately 400, 500 and 600 C. entering a pneumatic conveying line was soon extingu-
ishedthe distance to extinguishment depending on the
dust concentration.
2. Literature review Pinkwasser used an 80 m length of 100/110 mm i.d.
pipe, with six 90 elbows, which ended in a cyclone.
Harper, Plain and Gibson (1997) have discussed the Nests of smouldering material were introduced through
burning behaviour of powder accumulations on hot sur- an air-lock at the end of the pipe remote from the cyc-
faces. The stages of ignition and the form of the combus- lone. The powder conveying rate was measured from the
tion zone can be complex. Some powders burn directly weight of powder introduced into a known volume of
in the solid phase either with a flame or by smouldering, conveying air over a given time, and the mean powder
others melt and burn as a liquid, whilst some burn with concentration calculated. The temperature of the smould-
a large amount of flame. Some dusts can evolve large ering material was measured by thermocouples, and the
amounts of flammable gas when subjected to heat. A distance of travel of the smouldering material by spark
change from solid to liquid or agglomeration/expansion detectors and flame detectors. The powders examined
of dust particles to give an extended mass of material were three grades of flour, with Kst-values below 100
can block burning if diffusion of oxygen to the seat of bar m s1, and CC ratings of 5.
burning is prevented. Only one of the powders was capable of producing
A measure of the ignitability of a dust layer and inten- smouldering nests. The powder properties were: moist-
sity of burning of a dust layer is the Combustion Class ure content 8.9%, bulk density 290 kg/m3, median par-
(CC) (ISSA, 1998). This classification is based on the ticle size 120 microns and minimum ignition energy
behaviour of a defined dust heap when subjected to a approximately 100 mJ. A much coarser-grained flour
gas flame or hot platinum wire: with a bulk density of 510 kg/m3 failed to produce nests,
as did a finer grained flour with the properties: moisture
1. CC1: No ignition; no self-sustained combustion content 13.1%, bulk density 440 kg/m3, median particle
2. CC2: Short ignition and quick extinguishing; local size 55 microns and minimum ignition energy approxi-
combustion of short duration mately 500 mJ.
3. CC3: Local burning or glowing without spreading; Glowing clumps up to 15 mm diameter, with tempera-
local sustained combustion but no propagation tures of 500550 C, were fed into the line. In dust-
4. CC4: Spreading of a glowing fire; propagation free air, glowing particles were transported, in conveying
smouldering combustion velocities of 10 and 20 m/s, as far as 68 m. But as the
5. CC5: Spreading of an open fire; propagating open powder loading in the airstream increased the distance
flame to extinction of nests of approximately 10 g decreased
6. CC6: Explosible burning; explosive combustion. substantially. Extinction was promoted by breaking up
of a smouldering nest into individual glowing particles.
The train firing test assesses flammability with refer- No dust explosions were detected when powder-loading
ence to different ignition sources, and the Combustion was within the explosive range.
Class is allotted based on the result. Not all dusts have Pinkwasser (1986) concluded that it was impossible,
the ability to form a coherent burning nest. with the dusts tested, for smouldering nests of approxi-
Once a smouldering or burning deposit has developed, mately 10 g weight to be conveyed in a powder loading
nests, ranging in size from several millimetres to several of greater than 1 kg/m3 of air, but that in conditions
centimetres, may break off and, carried along by an air where powder loading is lowexhaust systems and under
stream until they reach an extensive dust cloud, then act startupsmouldering material could be conveyed over
as an ignition source. relatively long distances.
However, the train firing test does not, at first sight, Alfert et al. (1988) used a pneumatic transport system
give an indication of the likelihood that a burning nest ending in a 5.8 m3 filter unit. At an air speed of 35 m/s,
will form. The powder or dust needs to coagulate and only very strong nests could be transported. Charcoal
any accumulations that detach need to travel for some 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 29
agglomerations ( 1 cm3) even after a relatively short had internal temperatures of 700800 C. Nests at 1200
distance (11 m), and with maize starch as the explosive C did ignite the dust clouds but only after impact on
atmosphere in the filter even large nests (0.5 litre) pro- the floor of the explosion vessel. When dust was dropped
duced no ignitions in the system. When fine wood dust over smouldering nests on the vessel floor, cloud
was used to give the explosive atmosphere, ignition explosions could occur at temperatures of about 860 C.
occurred in the filter, but not in the pipeline. A flaming nest could, however, be practically extingu-
Pinkwasser (1986) showed those smouldering nests ished by the dispersal of milk powder around it in the
with a temperature of 700 C, free-falling into dust explosion vessel. Tests on the development of smoulder-
clouds, did not produce an ignition in explosive atmos- ing in nests under the influence of a 0.5m s1 air stream
pheres of wheat flour or wheat starch. Only when nests showed that smouldering developed differently
of at least 25 mm diameter and weight of at least 15 g depending on the type of milk powder. At higher air
landed on the bottom of the 1 m tall test column did speeds open fires occurred in relatively large quantities
some ignitions occur. Jaeger (1989) found that smould- of skimmed milk powder. At air speeds of about 10 m/s
ering nests could be produced only with dusts having a compact smouldering nests reached temperatures of
Combustion Class greater than 3. A minimum nest area 1200 C in the hottest spots. The transition from smould-
of 75 cm2 and surface temperature of 900 C were ering into open fire occurs around 800850 C,
required for igniting clouds of dust with Minimum Ignit- depending on the type of milk.
ing Temperatures less than 600 C. Alfert et al. (1988) Work by Bailey and Walker at Syngenta (Bailey &
noted that nests of low mechanical strength disintegrated
Walker, 2000) has shown that clouds can be ignited by
during a fall and generated a large fireball that acted as
various burning or smouldering ignition sources. Three
an ignition source. Mechanically stable nests were cap-
ignition sourcesparaformaldehyde, which burns with
able of igniting the cloud only when they reached the
a flame, Fe3 (H2), which smoulders, and incandescent
silo floor, but could get covered with dust before an
particles of sawdustwere used. Sulphur and lyco-
explosion had time to start. In these tests, nests of known
size were dropped through dust free air in a silo with a podium dust clouds of various concentrations were
height of 22 m. The dusts were charcoal, cork dust and blown over the first two of the ignition sources, and both
wood dust. Charcoal has a strong nesting structure; no dusts ignited. The incandescent particles were introduced
burning of the charcoal particles occurred, and when the into the dust clouds soon after the clouds had been pro-
nest reached the floor of the silo, a shower of glowing duced. The sulphur clouds ignited, but the lycopodium
particles was produced. Cork dust forms stable nests; did not.
flaring up of the nests was noted at approximately half These tests were repeated with dusts of various MIT
the height of the silo, and on the silo floor. Wood dust values, from 270 C to above 1000 C, as measured in
produces unstable nests; these could break up in the very the Godbert Greenwald furnace. With the burning layer,
top part of the silo creating a fireball, remain intact and dusts with MITs above 600800 C did not ignite; with
flare up in the upper half of the silo or reach the floor the smouldering layer, dusts with MITs above 340 C,
and flare up on impact or not at all. approximately, did not ignite; with the incandescent par-
When nests were dropped into an explosive atmos- ticles, dusts with MITs above about 330 C did not
phere of maize starch, nests of wood dust with a size of ignite. Some tests using layer ignition sources of various
0.51 litre produced no ignition in 40% of the tests. Nest areas and temperatures showed that as the area
with volumes of 11.5 litres ignited equally in the upper decreased, for a given temperature, the dust MIT above
part of the silo or on the floor. Nests of cork dust pro- which a dust cloud did not ignite increased.
duced no ignitions, but could start fires in settled maize In summary, the likelihood of ignition of a dust cloud
starch powder. Charcoal nests could produce ignitions if by a hot nest depends on the temperature of the nest, its
they were mechanically broken up, and also a short time residence time in the dust cloud and the availability of
after reaching the floor. oxygen to the burning area. Griesche and Brandt (1976)
Zokoll (1989) has reported some tests using milk pow- have shown that the Minimum Ignition Temperature of
ders as both the nest material and the explosive dust a dust-cloud decreases substantially when the residence-
cloud. Initial tests in which both dust and smouldering time of a dust in a Godbert-Greenwald furnace increases.
nest were dropped into the test chamber simultaneously The longer a local part of a dust cloud remains in contact
showed that fist-sized nests at temperatures approxi- with a smouldering nest, the more likely it is that an
mately 100 C or more above bulk powder Minimum ignition will occur. It appears, also, that dust clouds can
Ignition Temperature as measured in the BAM furnace extinguish burning nests and an ignition be prevented,
test did not ignite ground corn and milk powder clouds. if conditions are right.
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
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. 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 ering nests did not ignite dust clouds with MIT values
listed in Table 2. above 400 C, but flaming nests were able to ignite
A burning nest of dust was positioned inside the verti- clouds of all the dusts used up to the maximum MIT
cal tube close to the bottom and its burning mode noted. used, 600675 C. Smouldering nests with a temperature
A dust cloud was then created within the tube using the above approximately 700800 C ignited sulphur clouds.
screw feed arrangement. A review of the wider literature also shows that
In some of the tests the burning nest was dispersed smouldering nests are poor sources of ignition. The like-
by an air blast from a conical nozzle buried in the dust, lihood of ignition of a dust cloud by a hot nest is low
to see if the action of breaking-up the smouldering if the nest burns only by smouldering. Ignition depends
deposit would result in ignition of the dust cloud. 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.
4. Results and discussion When it is considered that in order for a nest to ignite
a dust cloud the following are needed: a dust capable of
An analysis of the results is shown in Table 3, where holding together as a nest; a means of heating this; a
the temperatures of the nests are divided into three means of transporting it through the system, a means
bandsapproximately 700, 800900 C; and approxi- of breaking the nest open and producing flames or high
mately 1000 C and above. The type of burning is listed, temperatures, and a dust cloud with an explosible con-
along with the Minimum Ignition Temperature of the centration at the right time, then the risks from glowing
dust cloud and whether or not ignition occurred. In the nests are probably overestimated and many events
majority of tests where smouldering was the mode of ascribed to them may have been caused some other way.
combustion, ignitions did not take place even when high A more critical look at actual incidents, including in
smouldering temperatures were evident and the differ- particular the nest forming properties of dusts where
ence between this temperature and the cloud MIT was nests have been blamed for the incident, might well lead
high. Only when the nest was dispersed and the tempera- to a revision of the previously assigned causes of
ture difference was high did ignition occur. Sulphur dust ignition.
clouds were the only ones which would ignite on In practical situations, a test similar to the Train Firing
smouldering nests, but even then the temperature differ- Test is a useful method for determining whether a dust
ence between the nest and the cloud MIT was greater deposit will either propagate smouldering or produce
than 500 C. By contrast, if flaming combustion took flame, although both the temperature and airflow inside
place ignition of a dust cloud was practically guaranteed, dust-handling plant can have an effect on the burning
even when the flames were small. behaviour. If the air above the dust deposit is at a tem-
The results from the current project are in agreement perature higher than normal room temperature, the possi-
with those of Bailey and Walker. In their tests, smould- bility for ignition of a dust cloud may rise.
Table 2
Burning temperature of dust deposits
Table 3
Ignition conditions
Temperature of Nest C Mode of Combustion MIT of Dust Cloud (C) Ignition Y/N
Various tests have been developed to measure the Eckhoff, R. K. (1997). Dust explosions in the process industries (2nd
ignition behaviour of dust deposits in a stream of hot ed.). Butterworth Heinemann.
Griesche, G., & Brandt, D. (1976). Einflussfaktoren auf die Zundtem-
air. Similar tests could be used to tell whether smoulder- peratur von Staub-Luft-Gemischen. Die Technik, 31, 504.
ing dusts were in danger of flaming. The I.ChemE Guide, Harper, D.J., Plain, & Gibson (1997). Use of intrinsically safe circuits
Prevention of fires and explosions in dryers described and enclosures to control ignition risk from equipment in powder
tests developed to simulate various conditions and obtain handling plant. I.Chem.E. Symposium Series No. 141 p. 463. Haz-
measurements of the temperature at which exothermic ards XIII. Process Safetythe Future.
ISSA (1998). Determination of the combustion and explosion charac-
reaction begins (Abbot, 1990). If the dust deposits and teristics of dusts. International Social Security Agency,
surrounding conditions properly simulate practical situ- Mannheim (1998).
ations, the temperature at which deposit burning pro- Jaeger, N. (1989). Zundwirksamkeit von Glimmnestern in Staub/Luft-
gresses to flaming combustion could be used as a basis Gemischem. VDI-Berichte 701, p. 263. Dusseldorf: VDI-Verlag,
for safe procedures with an adequate safety margin GmbH (in German).
Jeske, A., & Beck, H. (1989). Evaluation of dust explosions in the
incorporated. 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.
References Pinkwasser, T. (1986). On the Ignition capacity of Free-falling
Smouldering Fires. Euromech Colloquium 208, Explosions in
Abbot, J. (1990). Prevention of fires and explosions in dryersA user Industry.
guide (2nd ed.). Rugby, UK: The Institution of Chemical Engin- Porter, B. (1989). Industrial incidents. Paper presented at Dust
eers. Explosions: Assessment, Prevention and Protection, 24th Nov-
Abbott, J.A. (1985). BMHB Survey of dust fire and explosions in the ember, London.
UK 197984. British Materials Handling Board, ISBN 0 85624 Testing methods for electrical apparatus installed in a dusty environ-
4554. ment with a potential risk of explosion (2001). EU Contract No.
Alfert, F., et al. (1988). The ignition capability of nests of smouldering SMT-CT98-2273. Project EADE. Synthesis report available on
material and hot objects in industrial plants. VDI-Berichte No 701. www.safetynet.de.
pp. 303319. Zockoll, C. (1989). Ignition effect of smouldering pockets in dust-air
Bailey, M., & Walker, N. (2000). Private communication. Syngenta, mixtures. VDI-Berichte. 701, p295. Dusseldorf: VDI-Verlag
Manchester, UK. GmbH.
Journal of Loss Prevention in the Process Industries 16 (2003) 3340
www.elsevier.com/locate/jlp
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.
European Directive 94/9/ECthe ATEX Directive environments. The European Union funded this project
comes fully into force in June 2003 (DTI, 1994). The (EU, 2001).
Directive applies to equipment and protective systems, The partners in the project were the Health and Safety
among other things, that are intended for use in poten- Laboratory (HSL), UK, Laboratorio Oficial J.M. Madari-
tially explosive atmospheres. For the first time in regu- aga (LOM), Spain and Deutsche Montan Technologie
lations, explosion hazards arising from atmospheres con- (DMT), Germany. The objectives of the project were:
taining flammable dusts are specifically included.
Consequently, where appropriate, harmonised European To review the current state of the art in testing electri-
standards are being prepared to specify methods for cal apparatus in dusty environments through an
either preventing dust explosions or protecting against enquiry directed at test houses, certification organis-
their effects. ations and laboratories in Europe and the USA.
The hazards arising from dusty environments are: To review current scientific and technical knowledge
on ignitions of dust clouds and deposits.
ignition and explosion of a dust cloud by electrical or To develop a test apparatus and test method for
mechanical means, assessing the ignition hazard of electrical apparatus
ignition and burning of a dust layer or accumulation, intended for use in potentially explosive dust atmos-
ignition of an explosive dust atmosphere by a burning pheres.
deposit or burning nest. To develop a test apparatus and test method for
assessing the ignition hazard of electrical apparatus
intended for use in dusty environments where layers
To aid the development of standards CENELEC pro- and accumulations of flammable dust may be present.
posed a research project on electrical ignitions of dusty To propose test procedures for standardisation.
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 are transient inductive sparks, and the theoretical energy
test apparatus was constructed and a draft test method is given by 0.5Li2, where L is the circuit inductance and
developed. A duplicate test apparatus was built and sent i is the current in the circuit. Discharges from compo-
to the partners for a series of round-robin tests. These nents with significant inductance or resistance generally
tests showed that reproducible results could be obtained produce discharges with a longer duration than purely
from the test apparatus when the test procedure was fol- capacitative discharges, with the result that less total
lowed. 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
1. Literature review substantially when a series resistance was included in
the circuit. Because the series resistance absorbs a large
At the start of the project a number of testing and fraction of the capacitor energy during discharge, the
certifying organisations and laboratories were contacted energy going into the spark gap is only a small part of
throughout the European Union and the USA. Enquiries the theoretical stored energy, generally about 510%
were made of these organisations as to the current (Eckhoff, 1997). Thus, in Boyle and Llewellyns experi-
methods used for testing for ignition of flammable dusts ments, at a series resistance of 104105 ohms, for some
by electrical apparatus, both as regards ignition of clouds dusts the ignition energy was reduced to approximately
by sparking and ignition of dust layers and accumu- 1% of the energy necessary without the resistance. Simi-
lations by hot surfaces. lar changes were measured by Line, Rhodes, and Gilmer
None of the eleven responses received described spe- (1959). Smielkow and Rutkowski (1971) increased the
cific tests for electrical apparatus used in dust clouds. spark duration either by adding a large inductance or by
DMT suggested that gasair mixtures should be used as adding a large resistance to the circuit, and again the
substitutes for dust clouds, with the ignition energy of theoretical ignition energy from the capacitor decreased
the gasair mixture matching the ignition energy of the by a factor of ten, approximately.
dusts. This type of substitute test is used in the testing The optimum spark discharge duration measured by
of hand-held electrostatic spray guns according to EN Matsuda and Naito (1983), and estimated from the data
50 050: 1986 (EN, 1986). This standard specifies gas in Boyle and Llewellyn (1950) and Line et al. (1959)
air mixtures with ignition energies of 0.24 mJ (for paint were 0.11.0 ms, decreasing with the net spark ignition
spray guns) and 5 mJ (for powder spray guns) for use energy (Eckhoff, 1997).
in ignition tests. Measurements by Parker (1985) showed that for some
Underwriters Laboratories Inc. of the United States dusts there was a fairly distinct region where the dis-
responded that their test method for intrinsically safe charge duration, produced the lowest ignition energies.
apparatus in dusty atmospheres was given in the Stan- With other dusts, however, no such region occurred.
dard for Safety (Underwriters Laboratories, 1997). the Parker used four dusts, and those with the higher ignition
Factory Mutual Researchs Approval Standard Class No. energies showed the optimum spark duration effect.
3610: 1999(Factory Mutual Research, 1999) demands The physical length of the spark gap also has an effect
the same requirements as UL 913: on the ignition energy. Tests by Ballal (1980), using
metal dusts and carbon, demonstrated that there was an
Intrinsically safe apparatus not enclosed in a dust- optimum spark gap length for each dust, and that it
tight enclosure shall comply with the spark ignition increased as the ignition energy at the optimum gap
requirements of some gases, e.g ethylene for metal length increased, although the result for carbon did not
dusts and propane or methane for coal and grain dusts fit in with the results for the metal dusts. The optimum
and fibres and flyings. spark gaps fell in the range 27 mm. Measurements by
Norberg, Zu, and Zhang (1988) showed the optimum
Electric sparks and arcs have long been known to be spark gap length to be in the range 68 mm, for short
capable of igniting dust clouds. The minimum energy duration capacitative sparks.
required for ignition varies with the properties of the There is little published literature on research into the
dust, characteristics of the spark such as the spatial and ignition of dusts by the type of circuits used in intrinsi-
temporal distribution of energy, and the geometry of the cally safe apparatus, but two programs of work have
electrode gap, the electrode size and shape and the dis- been performed using the make and break sparking
charge triggering mechanism. apparatus.
Electrostatic spark discharges are capacitative, and the The make and break sparking apparatus is known as
theoretical spark energy, neglecting external circuit the break-flash apparatus and is described in BS-
losses, is given by 0.5CV2, where C is the capacitance EN50020 (BSI, 1995). Harper, Plain, Wilston, and Gib-
and V is the voltage difference. son (1997) used the standard disc and wires and fixed
Sparks or arcs generated when live currents are broken 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 35
resembled the Hartmann tube that is used for dust Both the standard test for assessing dust explosibility
explosibility testing. The sparking apparatus was pos- and for measuring the Minimum Ignition Energy (MIE)
itioned about 5 cm above the Hartmann dispersion cup. utilise very similar designs of apparatus. The Hartmann
By repeated dispersion of the dust a continuous cloud tubeor vertical tubeconsists of a 1.2 l tube, fitted
was produced, with a concentration passing through the with an ignition source that is either a high voltage con-
explosion range. Following calibration of the apparatus tinuous spark or a heating coil, with a means for introd-
with methane, ten dusts were used with low ignition ucing a blast of air to disperse a dust sample. The con-
energies, as measured with capacitative sparks. centration of the dispersed dust is varied by changing
Of the dusts tested, only sulphur had a minimum ignit- the amount of dust placed in the tube prior to the test.
ing current less than that of methane. In a resistive cur- By this means the potential explosion range of the dust
rent, at low voltage, the minimum igniting current is covered. If an explosion occurs the dust is considered
measurements for sulphur followed the gas Group IIB to be flammable at ambient conditions.
data, and then passed through the gas Group IIA to gas The MIE, defined as the lowest capacititively stored
Group I as the voltage increased. In an inductive circuit, electrical energy that just ignites the most ignitable
the results for sulphur fell between gas Groups IIB and dust/air following discharge across a spark gap, is meas-
IIA. These measurements show that circuits certified as ured in an apparatus similar to the vertical tube. A
intrinsically safe for gases and vapours would not capacitor of known capacitance is charged to a specified
present an ignition risk for all but the most sensitive high voltage and then discharged across a gap through
dusts. a dust cloud dispersed by an air blast. The energy of the
A similar apparatus to that used by Harper et al. spark discharge is lowered in steps, and at every step
(1997) was devised by Tolson (Lunn, Rowland, & Tol- tests are performed across the explosible range. This
son, 1999) to measure the ignitability of coal dust clouds continues until no ignition occurs at any dust concen-
when subjected to discharges produced by the IEC spark tration.
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 2. Development of the proto-type test apparatus
and broken using a modified version of the IEC spark
test apparatus described in CENELEC Standard EN The literature survey and discussions with users and
50020. The main modifications were: manufacturers indicated that an acceptable way forward
was to combine the vertical tube apparatus familiar from
the spark test apparatus was mounted horizontally; dust explosibility testing and measurement of MIEs with
the disc electrode was copper instead of cadmium; the spark test apparatus and its associated test method.
the rotating electrodes were inside a Hartmann verti- The vertical tube is an accepted way of producing an
cal tube apparatus. explosive dust cloud for testing purposes, for example
the MIKE 3 apparatus for MIE measurements (Adolf
The current in the circuit was set by adjusting the Kuhner, 1997). The spark test apparatus (STA) is an
power supply voltage and the coal dustair mixture pro- accepted means of testing the igniting capabilities of
duced by introducing a blast of air through a measured electrical circuits for gases. This combination has the
amount of coal dust placed in the dispersion cup at the advantages that there is expertise throughout the world
base of the Hartmann apparatus. The concentration of in using both types of test and the apparatus and test
coal dust in air could be changed by altering either the procedure would fit well into an established series of
amount of the dust placed in the dispersion cup, or the tests for dust ignitability and explosibility. A research
air introduced by the blast. The current was progress- apparatus on similar lines has been used successfully by
ively increased to give spark energies of from 50 to 300 two groups of workers in the United Kingdom and this
mJ but there were no ignitions. work had demonstrated that the proposed test was tech-
The test equipment was then validated using lyco- nically feasible.
podiumair mixtures. These are known to have ignition In addition, a new version of the spark test apparatus
energies of about 8 mJ and were readily ignited with for circuits up to 20 A has been developed under an EU
spark energies of 11 mJ which indicated that the appar- funded contract SMT4-CT98-2217 (EU, 1999).
atus was functioning correctly. A prototype apparatus was constructed using the
Coal dustair mixtures were eventually ignited with a Spark Test Apparatus (STA) as described in EN50 020
spark energy of 400 mJ; this was the maximum possible (BSI, 1995) as modified in project SMT4-CT98-2217.
with the circuit used. Ignitions at this energy level were The dust dispersion chamber is based on the principles,
repeatable. This energy level is approximately 1000 and where practicable, the dimensions of the MIKE 3
times the level necessary for ignitions of methaneair minimum ignition apparatus. Photographs of the appar-
mixtures. 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. Fig. 3. Dust dispersion system.
The apparatus thus consists of the STA disc and trai- mechanism involving the potential for many variables to
ling electrode system mounted horizontally inside a ver- effect the outcome. In view of the need to produce a test
tical Hartmann-type tube with a mushroom shaped dis- system that is reproducible and can be used universally
persion system in the base. Successive blasts of air at the procedure has to be kept as simple as possible, and
timed intervals disperse dust placed in the base to pro- some of the variables involved need to be confined to
duce a cloud around the STA disc and the location of standardised settings. Otherwise the operation becomes
the sparks. Provision is made to pass a gas/air mix too complicated and unwieldy and not workable in prac-
through the chamber for calibration of the STA. The tice. Following the development research work a test
number of turns of the STA motor are counted and dis- procedure has been produced that gives reproducible
played with a provision to automatically end the test results.
when a preset count is reached. An ignition detector
mechanism based on detection of the negative pressure
following ignition has been developed. 3. Test procedure
Tests have been carried out with three dusts, lyco-
podium, calcium stearate and sulphur. The research has A maximum particle size and drying procedure have
been aimed at devising a workable test method that takes been defined (63 m and 105 oC for 1 h are values com-
into account the variables that may affect the ignition monly used in the past), but care has to be taken as
result. Ignition of dusts by electrical means is a complex 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 (methane) curve and indicate that the apparatus is giv- 5. Round robin tests
ing reliable results.
Three dusts have been tested using a resistive circuit 5.1. Tests at LOM
calcium stearate, lycopodium and sulphur. The ignition
results based on the proposed draft test method are In the tests performed by LOM, around 25 V were
shown in Fig. 5 in a comparison with the published cur- used for the motor supply, about 28 air pulses were
rent voltage ignition curves from EN 50 020 (BSI, 1995). obtained for 400 revolutions and various resistor combi-
Both lycopodium and calcium stearate show ignition nations were used to obtain the desired spark intensities.
conditions higher than the methane curves. Sulphur, A calibration check with methane at 8.3% was perfor-
however, is much easier to ignite, and falls close to the med, with ignition obtained at 0.120 A. Three dust
ignition curve for IIB gasestypically ethylene. This samples delivered by HSL were tested: lycopodium, cal-
result supports previous work that shows ignitions of sul- cium stearate and sulphur. Dust weights ranged from 2.5
phur falling on or close to the IIB ignition curve. to 20 g.
6. Discussion
Fig. 5. Ignition results for lycopodium, calcium stearate and sulphur
dusts compared to published ignition curves. 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
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
this apparatus and method indicates that the procedure potentially explosive atmospheres. Available from DTI. The docu-
is satisfactory in this regard. ment contains the text of the Directive as an Appendix.
Eckhoff, R. K. (1997). Dust explosions in the process industries.
Based on the dusts tested and the results of the round Oxford: Butterworth Heinemann.
robin tests the equipment can be used as a viable test EN 50 050:1986 (1986). Electrostatic painting and finishing equipment
for individual pieces of electrical equipment. Based on using flammable materials. Part 1. Specification for hand-held
a comparison of the results obtained with the curves for spray guns and associated apparatus.
gases and vapors in the European Standard EN 50 020 EU (2001). Testing methods for electrical apparatus installed in a
dusty environment with a potential risk of explosion (EADE). EU
(BSI, 1995), then there seems to be a good possibility funded project contract no. SMT-CT98-2273. Synthesis report
of using the equipment to produce similar curves for available on the SAFETYNET web site: www.safetynet.de.
various dust groups. The results with sulphur dust fall EU (1999). EU funded project SMT4-CT98-2217gas ignition test for
very close to the curve for ethylene and show that some heavy current low voltage circuits. Final report.
dusts are very sensitive and equipment would have to Factory Mutual Research (1999). Approved Standard. Intrinsically safe
apparatus and associated apparatus for use in Class 1, 11, and
be certified accordingly. The results obtained using a 111, Division 1 hazardous (classified) locations. Class number
similar experimental apparatus by Harper et al. (1997), 3610.
further verify that the HSL test produces reproducible Harper, D. J., Plain, K. E., Wilton, J. S., & Gibson, N. (1997). Use
results. of intrinsically safety circuits and enclosures to control ignition risk
On the other hand the lycopodium and calcium stear- from equipment in powder handling plant. I.Chem.E. Symposium
Series No. 141. p463. Hazards XIII. Process safety. The future.
ate results indicate that a less stringent classification for Line, L. E., Rhodes, H. A., & Gilmer, T. E. (1959). The spark ignition
these and probably a great many more dusts would be of dust clouds. Journal of Physical Chemistry, 63, 290.
appropriate. The fact that their MIKE 3 MIE values are Lunn, G. A., Rowland, D. B., & Tolson, P. (1999). Electrical ignitions
fairly different when the STA apparatus produces almost and use of flameproof enclosures in coal dust and methane atmos-
identical results suggests that dusts may lend themselves pheres. Transactions of the Institution of Mining and Metallurgy
A, 108, 71.
to groupings. 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.
References 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,
Adolf Kuhner, A. G. (1997). MIKE 3 technical and operations manual. Uppsala, Sweden.
Switzerland: Adolf Kuhner AG. Parker, S. J. (1985). Electric Spark ignition of gases and dust. Ph.D.
Ballal, D. R. (1980). Ignition and flame quenching of quiescent dust thesis, (August) City University, London. Dust part also published
clouds of solid fuels. Proc. Roy. Soc. Lond. A, July, 479500. as report no. 8 53351-4 (December) Chr. Michelsen Institute,
Boyle, A. R., & Llewellyn, F. J. (1950). The electrostatic ignitability Bergen, Norway.
of dust clouds and powders. J. Soc. Chem. Ind. Trans, 69, 73181. Smielkow, G. I., & Rutkowski, J. D. (1971). Badania Zjawiska
BS EN 50 020:1995 (1995) Electrical apparatus for potentially explos- Zaplonu Mieszanin Pylowopowietrznych wywolanego wyladonani-
ive atmospheres. Instrinsic safety (i). British Standards Insti- ami iskrowymi. Chemia Stosowana, XV3, 283.
tution. London. Underwriters Laboratory Inc. (1997). Standard for Safety. Intrinsically
DTI (1994). Guidelines on the application of Directive 94/9/EC of 23 safe apparatus and associated apparatus for use in Class 1, 11 and
March 1994 on the approximation of the laws of the Member States 111, Division 1, hazardous (classified) locations. UL 913. 5th edi-
concerning equipment and protective systems intended for use in tion.
Journal of Loss Prevention in the Process Industries 15 (2002) 373383
www.elsevier.com/locate/jlp
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.
Bucket elevators are widely used in the handling of spacing of 6 m between vents is used. The guidance also
large quantities of bulk powders and are normally the requires that the top casing and the boot must be
preferred means of conveying where long vertical lifts explosion relieved (Lunn, 1992).
are required for free flowing powders/granular products. There appears to be no guidance from other European
As such they are to be found in nearly all animal feed organisations such as the German VDI, but the USA
mills, bulk grain stores and many of the larger instal- National Fire Protection Association document NFPA-
lations handling powders in the food industry. Powder 91: Fires and dust explosions in agricultural and food
or granular products inevitably spill from the buckets products facilities gives guidance in somewhat more
during operation and fall down the upleg of the elevator. detail than Lunn (1992) but still recommends a vent
The finer material is likely to remain in suspension while spacing of approximately 6 m.
the coarser material falls back to the boot. At the top of There is, however, no evidence that any current guid-
the elevator most of the powder will discharge down the ance spells out the optimum venting requirements of
off-take chute, but some will normally be carried over elevators, and there is little published work on elevator
into the downleg of the elevator. Thus both legs are explosion tests. Gillis and Fishlock (1982) carried out
likely to contain a dust cloud of unknown concentration, venting and suppression experiments on a twin leg elev-
constantly agitated by the moving buckets, all the time ator and some guidance was given.
the elevator is in operation. Various sources of ignition This paper describes an experimental programme, a
are foreseeable in such units and explosion incidents collaborative effort by the Health and Safety Executive
have been reported.
and manufacturers and users of bucket elevators under
Explosion venting is one method for explosion protec-
the British Materials Handling Board, in which
tion of bucket elevators. The current Institution of
explosion venting of bucket elevators was studied using
Chemical Engineers guidance requires that the vents
a single leg elevator and a twin leg elevator with a num-
equal in cross sectional area to the limbare positioned
according to the guidance for ducting. Alternatively, a 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.
Corresponding author. Tel.: +44-114-2892005; fax: +44-114-
2892010.
E-mail address: geoff.lunn@hsl.gov.uk (G.A. Lunn).
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
Fig. 1. Single leg bucket elevator. 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
2.4.1. Cornflour B
Fig. 3 shows all relevant test results using Cornflour
B, for various explosion-venting arrangements.
Table 1
Explosion pressures in single-leg elevator tests
Level 1 (mbar) Level 3 (mbar) Level 5 (mbar) Level 7 (mbar) Level 9 (mbar)
Table 3
Pressures in some twin-leg elevator tests
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
Table 5
Peak Pressures in the twin leg elevator with different dust cloud forma-
tions
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- that vents should be fitted in the head and as close as is
ing information. A comparison of these data with the practicable to the boot. This generally means within 6
information in Fig. 8 shows that the lines for guidance m of the boot or within the recommended vent spacing,
drawn in Fig. 8 envelop the results from this twin leg whichever is the lesser. The static burst pressure of the
elevator. vent closure should not exceed 0.1 bar.
The spacing of additional vents depends on the KSt
value of the dust.
4. Guidance
(a) Although explosions are possible with dusts of low
4.1. Single leg elevators KSt, generally the pressures developed by dusts with
KSt values below 100 bar m s1 are not significant,
Vent openings should have an area equal to the cross- and no additional vents are required.
sectional area of the elevator leg and the minimum (b) Dusts with a KSt value of 150 bar m s1 are able to
requirement is that vents should be fitted in the head and develop significant pressures, although the likeli-
as close as is practicable to the boot. This generally hood of explosion propagation through the elevator
means a vent within 6 m of the boot or within the rec- is low. Vents additional to those at the head and boot
ommended spacing, whichever is the lesser. The spacing may be required on long elevators if the casing is
between vents along the elevator is listed as a function comparatively weak. The graphs in Figs. 7, 8 and
of the dust KSt value, the vent burst pressure and the 11 should be used to estimate the reduced explosion
reduced explosion pressure in Table 6. pressure for a given KSt value and vent spacing.
For dusts with KSt values of 150 bar m s1 or less, a (c) Dusts with KSt values above 150 bar m s1 will
vent spacing of 6 m will limit the reduced explosion propagate explosions, and vents additional to those
pressure to 300 mbar, when the vent static burst pressure in the head and boot are required on elevators taller
is 0.1 bar. than 6 m. The graphs in Figs. 7, 8 and 11 should be
For dusts with KSt values of 100 bar m s1 or less, used to estimate the reduced explosion pressure for
vents installed in the head and boot of the elevator, with a given KSt value and vent spacing. The strength of
none intervening, will limit the reduced explosion press- the elevator should then be designed appropriately.
ure to 0.5 bar. For dusts with KSt values of 80 bar m s1 (d) No data are available for dusts with KSt values
or less, a vent spacing of 14 m will limit the reduced greater than 210 bar m s1.
explosion pressure to the vent bursting pressure if this
is no greater than 0.1 bar. The guidance given in this section differs to that given
For dusts with a KSt value of 80 bar m s1, a vent in NFPA-91 in some important respects. In this guidance
spacing of 20 m will limit the reduced explosion press- the area of a vent is assumed to be equal in cross-sec-
ure to 250 mbar. 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
4.2. Twin leg elevators cross-sectional area of the leg. NFPA-91 suggests the
vent area at a given position be divided into two equal
Vent openings should have area equal to the cross- vents installed in opposite sides of the leg; the present
section of the elevator leg and the least requirement is 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
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
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 3. Experimental assessment of MIE of dust clouds
contact between the energised bare rod and any
grounded metal part of the silo. This possibility has to The strong influence of the spark discharge time on
be considered in Zones 20 and 21, at least when taking MIE for dust clouds was first indicated by the results
into account possible faults. of Boyle and Llewellyn (1950) and Line et al. (1959).
The intrinsic safety requirements currently adopted in Experiments by Eckhoff and Enstad (1976) gave further
such cases by many producers and users of electrical support to the validity of the observation by Line et al.
apparatus are those of gas group IIB. This has, in fact, that the shock wave produced by sparks of very short
also been proposed in the current draft ExDi standard duration can, in fact, prohibit ignition by blowing the
for dusts, produced by IEC (2001). One argument put dust away from the spark before the ignition process gets
forward in favour of this approach is that the same types underway. Section 5.3 of Eckhoff (1997) provides a
of electrical apparatuses can be used both for practically more recent summary of the state-of-the-art on electric
all kinds of combustible gases and vapours, and for all spark ignition of dust clouds.
dusts, even the most spark-ignition-sensitive ones, such One important conclusion is that conservative experi-
as very fine sulphur. This obviously simplifies life for mental determination of MIE of dust clouds requires the
the producers of electrical apparatuses, who can stan- use of electric sparks of sufficiently long discharge times
dardise on only one type. to prevent significant disturbance of the dust cloud by
However, the purpose of the kind of standards in ques- the shock wave generated by the spark discharge. This
tion should be to draw objective borderlines between is achieved by introducing a 12 mH inductance in the
safe and unsafe conditions, with reasonable safety mar- discharge circuit, and this feature has been included as
gins. Therefore the approach currently adopted by IEC a compulsory option in the current international (IEC,
is unreasonable if the electrical spark ignition sensitivity 1994) and new European (CEN, 2001) standard methods
of dusts vary substantially. In that case, an intrinsic for MIE determination for dust clouds. (This inductance
safety standard for dusts should exhibit a corresponding is to be removed when the purpose of the test is to assess
marked differentiation of the maximum permissible cur- the sensitivity to ignition by electrostatic discharges.)
rents and voltages for preventing incendiary sparks in A major limitation of the test apparatuses commonly
electrical apparatus. The main objective of the present used so far is that they cannot generate sparks of ener-
paper is to suggest that this is necessary, and can be gies significantly below 13 mJ. Efforts should be made,
achieved by basing the specification of maximum per- therefore, to develop generators for synchronised sparks
missible circuit parameters on the experimentally of energies down to 0.1 mJ, or even down to 0.01 mJ.
determined MIE value of the dust in question. Research to meet this need is currently being initiated
in Norway.
2. Range of MIEs for dust clouds
In the past it was thought the absolute lower limit for 4. MIE and ignition curves for gases
MIEs of dust clouds in air was of the order of 10 mJ,
i.e., about two orders of magnitude above the values of In the current Exi standards for gases CENELEC
IIA gases. For example, Dorsett and Nagy (1968) pub- (2000) and IEC (1999), the ignition curves play a central
lished a figure of 15 mJ for a very fine sulphur powder role. These are empirical borderlines separating the com-
of average particle size 4 um. However, Eckhoff (1975) binations of circuit parameters that cannot produce
showed that the true MIE for very sensitive dusts was incendiary electric spark discharges from those that can.
considerably lower, by at least one order of magnitude, The curves are determined experimentally by means of
than the values indicated by USBM in the 1960s. More a special ignition apparatus specified in the same stan-
recently, Bartknecht (1993) indicated an extreme lower dards. In this apparatus the explosive gas atmosphere is
limit, by reporting a MIE value of very fine sulphur pow- exposed to electrical sparks generated by breaking or
der of only 0.01 mJ, i.e., practically identical with the closing live electrical circuits. Three different types of
MIE of carbon disulphide. ignition curves are determined, viz. capacitive, inductive
As documented by Bartknecht (1993) extensively and and resistive. The dependence of the ignition curves on
by Beck et al. (1997), MIEs of clouds of various dusts the ignition sensitivity of the gas is taken care of by
in air span from perhaps as low as 0.01 mJ at the lower having complete sets of such curves for the four different
end, to beyond the 1KJ range at the upper. Hence, there standard gases: methane, propane, ethylene and hydro-
is an obvious need to establish a differentiated system gen. All other gases and vapours are then grouped in
for specifying the maximum permissible spark energy four corresponding categories according to their standard
requirements to be satisfied by ExiD equipment for minimum ignition current, determined in a closely speci-
dusts. fied way.
R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 15 (2002) 305310 307
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 relation between current and voltage, using MIE as a
spark-generating circuit, is an obvious, conservative sensitivity parameter, may be a solution. The empiri-
basic criterion for such differentiation. cal relation in Eq. (1), where I is in A, MIE in J and
3. For capacitive circuits, conservative safe combi- U in V, was shown to apply to the four standard gases
nations of circuit parameters are obtained by methane, propane, ethylene and hydrogen, as well as
satisfying the criterion 0.5 CU2 MIE. to clouds in air of lycopodium powder. This powder
4. For inductive circuits, conservative safe combinations is the only one for which both the resistive ignition
of circuit parameters are obtained by satisfying the curve and the MIE have been reported. Whether the
criterion 0.5 LI2 MIE as long as L 1 mH. For proposed relation is also valid for other dust, in parti-
L 1 mH, the circuit is essentially resistive. cular those of significantly higher MIE values than
5. For resistive circuits, the use of a simple empirical that of lycopodium, remains to be investigated.
310 R.K. Eckhoff / Journal of Loss Prevention in the Process Industries 15 (2002) 305310
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
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.
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
W1/2C(V21V22), (1) known that the greater the particle size, the more energy
that is needed to ignite the dust. If the size distribution
where C (F) is the capacitance of the capacitor and
changed, it could affect the flammability of the anthra-
V1 (V) and V2 (V) are the voltage of the power source
quinone dust.
before and after sparking, respectively.
As shown in Fig. 3(a), all lycopodium particles had
2. When a resistor was connected, the calculation was
almost the same size and round shape. Polyacrylonitrile
done with the PC using the formula below.
(b) particles come in various sizes, but their shapes were
t all round. Neither lycopodium nor polyacrylonitrile par-
W n(t)i(t)dt (2) ticles had any other sub-components. Anthraquinone
particles (c), however, consisted of numerous small slen-
0
der crystals of irregular shape. It was also found that the
where n(t) (V) and i(t) (A) are the voltage and the median diameter of particles passing through the vibrat-
current of the spark, respectively, and t (s) is the dur- ing mesh, as measured by the laser diffraction method,
ation of the spark. was 178 m, as shown in Table 2. Since the anthraqui-
none powder was sieved with a 100-mesh (150 m in
In order to evaluate the MIEs obtained by the new bore) prior to the test, this value of 178 m is much
method, the data were also taken using the same samples larger than expected. There are several reasons for the
with the conventional Hartmann tube apparatus enlargement of the particles. First, there are attractive
(MIKE-3). 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
3. Results and discussion different materials. The charge amounts of the samples
measured by the Faraday cage method are shown in
3.1. Evaluation of MIE data obtained by the new Table 3.
method It is obvious that anthraquinone obtained a conspicu-
ously large amount of charges in comparison with the
The MIEs obtained by both the new and conventional others. Taking the electrical attractive forces exerted
apparatuses together with the target values stated in the among particles into consideration, aggregation through
IEC standard are given in Table 1. In this case, we this process would be quite possible. Second, the slen-
believe that the data taken with MIKE-3 should be der, filamentary structure of the crystal could also be
regarded as the benchmarks rather than those found in responsible for the aggregation. Filamentary substances
the IEC standard (rightmost column) because our naturally tend to become intertwined. A visual evidence
samples might have not been exactly the same as those of the aggregation of anthraquinone particles is given in
expected by the standard. The MIEs taken with the Fig. 4. The picture depicts how anthraquinone particles
vibrating mesh and the Hartmann tube were similar in stick together on the mesh.
the case of lycopodium and polyacrylonitrile. The MIEs These facts suggest that a method to enhance the dis-
for polyacrylonitrile were slightly different, yet this persion of particles should be devised for making the
problem can be solved by tuning-up carefully the new new apparatus more practical.
apparatus. In the case of anthraquinone, however, the
difference seems to be too large to be made up by mere 3.2. Estimation of dust concentration
adjustment, and the reason of the difference should be
clarified. One of the weak points of our apparatus is that the
We finally noticed that the shape and size distribution dust concentration is not easily estimable. Generally
of the anthraquinone dust would be attributable. It is well speaking, as the vibration strength increases, the concen-
tration becomes higher. Therefore, it will be possible to
Table 1
obtain the MIE for a certain dust without knowing the
MIEs measured by the new (vibrating mesh) and the conventional exact concentration. In a real situation, however, it is
(Hartmann Tube) method (L=0.94 mH) often required to determine the ignition energy at a cer-
tain concentration level. Consequently, the relationship
Substance MIE (mJ) between the vibration level and the dust concentration
Vibrating mesh Hartmann tube Target value
(MIKE-3) recommended
needs to be clarified.
by IEC As a trial, we employed a rather simplified model in
which each particle flowed at a constant velocity,
Lycopodium 2329 1030 510 obtaining an even concentration at any point in the dust
Anthraquinone 2934 13 26 cloud. Thus, we obtained the following formula:
Polyacrylonitrile 3441 1030 26
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.
gave similar MIE values. This can be attributable to the sparking circuit and the spacing of the sparking elec-
relatively long discharge duration and the impartial dissi- trodes was set at 4 mm, the lowest level of MIE, 4.6
pation of electric power during that time. mJ, was attained for lycopodium powder.
The circuit resistance of 100 k gave the lowest 2. Electrodes with sharp tips gave smaller MIEs than
energy for ignition, 4.6 mJ, where the electrode spacing those with round tips in an LC sparking circuit, while
was 4 mm and both the electrodes were round. These shape made no difference in an RC circuit.
results suggest that the ignition of dust depends on the
electric power and the discharge duration.
Acknowledgements
4. Conclusion
This work was supported financially by the Korea
The new apparatuses for testing dust MIEs were put
Research Foundation in 1999.
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 References
anthraquinone was much larger than its reference. It is
Matsuda, T., & Yamaguma, M. (2000). Tantalum dust deflagration in
concluded that the high-level tribo-charge and the fila- a filter bag collecting device. Journal of Hazardous Materials A,
mentary crystal sub-component of the particle enhance 77, 3342.
the aggregation of the particles, leading to insensi- Cheung, W. L., Kodama, T., & Yamaguma, M. (1997). Vibrating mini-
tiveness to electric sparks. If this shortcoming is over- mum ignition energy measurement system. The lecture paper
come, this method will be another reliable testing presented at the Institute of Static Electricity of Japan (pp. 181
186).
method. Cheung, W. L., Kodama, T., & Yamaguma, M. (1998). Development
Other factors essential to the device are as follows: of vibrating minimum ignition energy for powder. In ESAIJ Joint
Symposium on Electrostatics Proceedings (pp. 143150).
1. When a 100 k resistance was connected to the IEC, International Standard (1994). 1241-2-3.
Journal of Loss Prevention in the Process Industries 14 (2001) 495502
www.elsevier.com/locate/jlp
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.
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
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
( 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.
d
up
mp f dmp
G (9)
dt
dx
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
|
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
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
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 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
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 (Luttgens & Glor, 1989). A dust cloud in between the
non-conducting plate and the electrode may hamper the
An increased use of non-conductive materials in the charge transfer by reducing the electrical field. Some
process industry may lead to a higher probability of theories focus on the differences in the spark discharge
charge accumulation. Common earthing techniques are characteristics between a brush discharge and a capaci-
often not suitable and the frequency of electrostatic dis- tive spark between two electrodes. The character of these
charges may increase. Brush discharges are among the two types of discharges is rather different in time and
most usual forms of electrostatic discharges occurring space.
when a charged isolator comes into the neighbourhood Various critical parameters for ignition of gas by
of a blunt earthed conducting object. Work has shown brush discharges have been investigated. These are: the
that common hydrocarbons are easily ignited by brush non-conducting plate surface humidity and charge trans-
discharges (Gibson & Lloyd, 1965; Heidelberg, 1967). fer in the brush discharge (Gibson & Lloyd, 1965), the
An equivalent energy of brush discharges in gasair charge density on the non-conducting plate and the non-
atmospheres was found to be in the range 14 mJ (Glor, conducting plate surface polarity (Lovstrand, 1981), the
1981). This energy should be sufficient to ignite dust size of the electrode (Heidelberg, 1967) and the non-
with low minimum ignition energy (MIE). However conducting plate surface potential and the time charac-
such ignitions have never been reported in spite of com- teristics of the brush discharge (Johnson, 1977/78;
prehensive laboratory tests. Kramer & Asano, 1979).
Potential reasons have been discussed. Electrical It is well known from literature (Eckhoff, 1991), that
forces may attract dust towards the electrode and the adding oxygen to an atmosphere where a dust cloud is
non-conducting plate. A dust deposit on the electrode generated will make the dust more sensitive to ignition
divides the discharge into many weaker discharges 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
* Corresponding author. Tel.: +47-55-57-4040; fax: +47-55-57- concentrations just giving ignition of the dust can be
4041. 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 plate until a discharge occurred. For reasons of safety,
reasonable to believe that brush discharges are not able movement of the various parts was done pneumatically
to ignite dust clouds in pure air. yielding remote control.
Gasair mixtures introduced into the chamber are pre-
pared using a gas-mixing panel where the desired gas
2. Experimental apparatus and procedure concentration is easily mixed. To reduce any pressure
build up in case of an explosion, the backside and the
2.1. The brush discharge generator top of the rig work as explosion vents. A thin plastic
sheet covers these openings during tests where explosive
A brush discharge generator, based on the apparatus mixtures were attempted to ignite. From a dispersion
used by Heidelberg (1970) was developed. The appar- nozzle through a hole in the bottom of the rig nitrogen
atus is shown in Fig. 1 and consists of a gas tight cubic can be introduced in case of fire. The brush discharge
chamber of 0.550.550.55 m. The left side wall and generator is placed in a fireproof tent. Two extraction
the bottom as seen in the figure are made of steel and ducts, one above and one below the generator, carry the
PVC. The front wall is made of plexigass, allowing for excess gas in the tent outdoors via a ventilation system.
visual inspection during the experiments. All conductive
parts are shielded to avoid charge dissipating to ground. 2.2. Dust dispersion and control system
The right wall consists of a 6 mm thick PVC plate. On
to this, non-conducting plates can be mounted on the A circular dust dispersion cup with a mushroom
inside by plastic screws. A metal plate with more than nozzle is mounted underneath the bottom plate of the
200 sharp-pointed corona needles is used to charge the generator near the non-conducting plate. Through a cir-
non-conducting plate. The metal plate was supplied with cular hole dust can then be dispersed up into the area
high voltage from an adjustable high voltage generator where the brush discharges are generated. The air blast
(030 kV). When this plate was placed near the non- used to disperse the dust is delivered from a 725 ml
conductive plate, charges were transferred to the non- pressurised container via a magnetic valve. The dust dis-
conducting surface. Discharges between ground and persion and the electrode movement are controlled from
negatively charged non-conducting surfaces are tra- a PC and a control unit. Various delay times between
ditionally named positive surface discharges. These are the dust dispersion and start of the electrode can be
found to be far more incendive than discharges between selected in order to generate brush discharges within the
ground and positive charged surfaces (Lovstrand, 1981). dust cloud and not before or after the dust cloud is
The charging voltage was therefore 30 kV. present. Various perforated plastic discs can be mounted
In order to obtain high charge density on the non- on top of the dispersion opening in order to avoid
conducting plate and avoid charge dissipation to the sur- agglomeration of the dust sample.
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 2.3. Charge measurement circuit
through a circular opening in the centre of the PVC wall.
Brush discharges were generated by moving a spherical An RC circuit in series with the electrode measures
electrode pneumatically towards the non-conducting 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
Fig. 1. Brush discharge generator. be determined from the video recordings.
. Larsen et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 111122 113
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 results in Fig. 7 show a marked increase in the
the electrode speed. In almost all tests where several charge transfer when the speed is increased from 0.15
sparks were initiated, the first brush discharge was the m/s to 0.6 m/s for all three diameters. For the 40 mm
discharge with the highest charge transfer. An oscillo- which shows the most distinct tendency, this can be
scope signal showing three successive brush discharges explained by the division of the charge transfer into more
in one test is seen in Fig. 6. brush discharges and the fact that the average charge is
In Fig. 7 the average charge transfer in these tests is calculated only from the largest brush discharge. The
plotted. A dashed horizontal line divides the plot area. same explanation can also be given for the fact that the
Above this line all the charge was released in one single 20 mm electrode curve is well below the 40 mm and the
brush discharge. Below the dashed line the charge trans- 60 mm electrode curves. If the electrode speed is further
fer from the non-conducting plate was divided into sev- increased the curves flatten for the 20 mm and the 40
eral discharges. The average value of the charge transfer mm electrode around 0.8 m/s. For the largest 60 mm
is calculated from the largest of the discharges. The electrode, further increase in the speed gave no increase
charge average is therefore rather low for the 20 mm but rather a slight decrease in the charge releases.
electrode for all velocities and the 40 mm electrode for If the total charge transfer (the sum of all charge trans-
the lowest velocity. 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 115
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).
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.
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 is preferable that all the available energy is discharged
thin the electrical forces may hold on to more of the in one single breakdown. It should therefore be favour-
electrons impeding them for being released to the elec- able to use relatively large electrodes and a fast move-
trode during the brush discharge. However, if the insu- ment of the electrode towards the non-conducting plate.
lator is too thick, the electrical forces set up by the When dust is introduced into the chamber high electrode
grounded metal plate is lower due to higher volume res- velocity will also make the timing between the start of
istivity. This results in a lower charge density on the electrode and the dispersion of the dust easier. This is
inner surface of the non-conducting plate and a lower because of the limited period of time the dust is present
charge transfer to the brush discharges. A certain optimal in the area where the brush discharge occurs. The disad-
non-conducting plate thickness therefore exists. vantage may be generation of turbulence in the dust
Varying the speed and the size of the electrode shows cloud by the electrode, which may hamper ignition. The
that a certain electrode size and speed is necessary to minimum ignition energy of dust may increase by sev-
collect all the charge transfer into one single brush dis- eral orders of magnitude if initial turbulence is present.
charge. When the electrode diameter is small and the From Fig. 9 it can be seen that for the 60 mm elec-
electrode speed low, several weaker discharges may trode the distance from the non-conducting plate where
occur instead of one powerful brush discharge, see Fig. the brush discharge most often occurred was 56 cm.
7. To increase the incendivity of the brush discharge it 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- mixture was constant at 4.2 vol% while the nitrogen con-
tance will make the timing of dust dispersion easier. tent in the air was increased in steps of 0.2 vol% from
Due to the results in Fig. 10 the humidity in the 24 vol% up to 32 vol%. The results presented in Fig. 11
chamber should be kept as low as possible. High air show the frequency of ignition for various mixtures of
humidity leads to condensation at the surface, which propane, air and nitrogen compared with results obtained
reduces the surface resistance. by Glor (1981) and Schwenzfeuer and Glor (1993). The
In addition to the parameters presented above an opti- tendency is clear. For both of the tested electrode speeds,
mal charging period was found to be approximately 30 0.4 m/s and 1.5 m/s, the frequency of ignition drops sys-
seconds. It was also seen that it was important to gener- tematically as the content of nitrogen content in the
ate the brush discharge within a relative short time inter- atmosphere increases above 22 vol% and 24 vol%
val after charging was completed. The conclusions of respectively. The highest nitrogen content giving
the optimisation of the procedures and apparatus are ignition is 30 vol% for both electrode velocities.
summarised in Table 1. The result of the present study are in between those
of Glor (1981) and Schwenzfeuer and Glor (1993) indi-
3.2. Determination of the equivalent energy in brush cating a maximum equivalent energy of 3.6 mJ in the
discharges brush discharges (see Fig. 12). A slightly higher prob-
ability of ignition can be seen for the highest electrode
Brush discharges were generated in various mixtures velocity. Flow velocity and turbulence intensity is
of propane, air and nitrogen. The propane content of the 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
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 119
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).
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 Average spark length (cm) Average charge transfer % of max charge transfer
(nC) without dust
No dust 5 449
Dust layer on the electrode 44.2 g/m2 14.1 304 67.7
Dust layer on the non-conducting 4.0 g/m2 5 330 73.5
plate
Dust cloud g/l 6.2 308 68.6
Table 3
Condition during tests with suIphur resulting in ignition by brush discharges
Test no. Vol% oxygen in air Dust weight (g) Dispersion pressure Spark length (cm) Charge transfer (nC)
(barg)
4. Conclusion
affects the brush discharge duration and appearances will meters on minimum ignition energy and quenching distance. In:
be investigated. 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.
Acknowledgements Eckhoff, R. K. (1991). Dust explosions in the process industry.
England: Butterworth-Heinemann.
Gibson, N., & Lloyd, F. C. (1965). Incendivity of discharges from
The authors acknowledge the support given by the fol- electrostatically charged plastics. British Journal of Applied Phys-
lowing companies: Alpharma A/S, Borregaard Industries ics, 16, 1619.
Limited, Brandforsk, DSM Research, Elkem ASA, FK Glor, M. (1981). Ignition of gas/air mixtures by discharges between
Group, Schweizerisches Institut zur Forderung der Sich- electrostatically charged plastic surfaces and metallic electrodes.
Journal of Electrostatics, 10, 327.
erheit, Norgesmllene DA, Norsk Hydro ASA, Forestia Heidelberg, H. (1967). Generation of igniting brush discharges by lay-
AS, Tine Norwegian Dairies BA, Pharmacia & Upjohn, ers on earthed conductors. In: Static Electrification Conference,
SCA Hygiene Products AB, RhonePoulence Industrial- 1967, p. 147.
isation, Forsikring & Pension. They also want to thank Heidelberg, H. (1970). Entladnungen an electrostatisch aufgeladenen
Dr. Martin Glor and Dr. Klaus Schwenzfeuer nichtleitfahigen Metallbeschichtungen. PTB-Mitteilungen, 6, 440.
Johnson, J. K. (1977/78) The ignition of vapour and droplets by liquid-
(Schweizerisches Institut zur Forderung der Sicherheit) to-metal sparks. Journal of Electrostatics, 4, 53.
and Prof. Rolf K. Eckhoff for several fruitful dis- Kramer, H., & Asano, K. (1979). Incendivity of sparks from surfaces
cussions. of electrostatically charged liquids. Journal of Electrostatics, 6,
361.
Luttgens, G., & Glor, M. (1989). Understanding and controlling static
electricity. Germany: Expert Verlag.
References Lovstrand, K. L. (1981). The ignition power of brush discharges
experimental work on the critical charge density. Journal of Elec-
Al-Arainy, A. A. (1997). Influence of dust particles on DC discharges trostatics, 10, 161.
in symmetrical air gaps. European Transactions on Electrical Schwenzfeuer, K., & Glor, M. (1993). Test to determine the ignition
Power, 7, 275. frequency of dust by brush discharges. Journal of Electrostatics,
Ballal, D. R., & Lefebvre, A. H. (1975). The influence of flow para- 30, 115.
Journal of Loss Prevention in the Process Industries 14 (2001) 139151
www.elsevier.com/locate/jlp
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
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
chemicals
physical processes
plants
energy
personnel
external influences
3.2. The steps in risk analysis 4. Dust explosion hazards in the blending unit
Deviations from safe working conditions can lead to An explosion hazard can exist when dusts are pro-
hazards. A systematic search for hazards is therefore duced, stored or processed in a plant and these materials
only possible if safe working conditions and their limits are present as a mixture in air. An explosible mixture
are defined and the consequences of deviations from is present, when combustible dusts are present in such
them are known. In the forefront, therefore, are the quantities in air that an explosion occurs after an
chemicals involved in the process, the intended reactions ignition.
and the criteria, which must be maintained for safe pro- An explosion requires three conditions to exist simul-
cessing. taneously (see Fig. 4):
avoiding the development of explosible mixtures tially the extent, and hence the cost of the protective
(combustible dusts, flammable gases), or measures, to be used. This is especially true for the use
replacing the atmospheric oxygen with an inert gas, of the protective measure prevention of ignition sour-
working in a vacuum or using inert dust, or ces and also for the understanding of the ignition
preventing the occurrence of effective ignition sourc- phenomena regarding static electricity, e.g. brush dis-
es. charges, bulk surface discharges, spark discharges, pro-
pagating brush discharges. Knowledge of a powders
All three measures are summarized under the heading minimum ignition energy, particle size distribution and
preventive explosion protection. It should be ensured specific resistivity allows one to define the necessary
that at least one of the three conditions is eliminated or protective measures for different operations based on the
so strongly reduced, so that an explosion is no longer vessel size and the conductivity of the material of con-
possible or at least very rare. Their appropriate elimin- struction Luettgens & Glor, 1989.
ation prevents an explosion from starting.
To determine the necessary explosion protection 4.1.1. Burning class
safety measures, the knowledge of the safety data of the The test is designed to determine the ignitibility and
materials being used in the production unit and the effec- the combustibility of a dried product according to the
tiveness of the identified ignition sources must be following criteria:
known.
Can the product be ignited at all?
4.1. Basic safety data information Is the ignition propagated?
Is there a fire or a flameless reaction?
In order to sufficiently survey the danger potential of
a dust, one is normally forced to carry out a multitude The tests are carried out in a ventilated laboratory hood.
of laboratory tests that are based on standardized The air velocity at the test location is approx. 0.2 m/s.
methods. The test results are described in the form of The superimposed light air stream is necessary to vent
parameters known as safety characteristics. Thorough the inert gas which are generated upon ignition of the
knowledge of the ignition behavior of dustair mixtures product and which may inhibit the combustion behavior
towards mechanically generated sparks or electrostatic of the dust (see Fig. 6).
discharges (Minimum Ignition Energy, MIE) and hot If the combustibility of a product is of interest at elev-
surfaces (Minimum Autoignition Temperature, MAIT) is ated temperatures, the combustibility test can be done
extremely important for assessing the hazards in dust- at elevated temperatures (e.g. at the anticipated drying
carrying plants. The ignition behavior determines essen- temperature, see Fig. 7).
Table 1
Definition burning classes
5. Protective measures
outside the explosive range is rarely possible due to sedi- transport duct, there is however, no need to consider
mentation or whirling up of the material being handled. ignition hazards due to electrostatic discharges if the fol-
Thus, as a matter of principle, an explosive atmosphere lowing conditions are observed:
can only be avoided with certainty by reducing the oxy-
gen concentration, e.g. inerting. In practice, however, the The diameter of the duct is less than 1 m;
possibilities to apply inerting are also limited. For such The MIE of the bulk goods is 1 mJ,
situations, avoidance of effective ignition sources and/or The duct is electrically conducting and is grounded;
explosion-proof design are the only measures available. The duct is not lined with an insulating material of
In the following, protective measures are outlined for 2 mm or with a breakdown voltage of greater than
typical powder handling operations in the blending 4 kV. In our experience, deposits of insulating pro-
facility. As previously stated, the following product and ducts in the duct are not critical if they are not burnt
plant properties are important for an accurate hazard or melted on.
assessment: Normally, pneumatic transport is installed for fill-
ing silos or large vessels; consequently, the same
Minimum ignition energy, MIE, of the bulk material requirements and considerations must be observed as
(measured without inductance in the discharge for normal filling and emptying (see Section 5.2).
circuit),
Minimum Autoignition Temperature, MAIT, of the
bulk material, 5.2. Filling and emptying operations
Volume resistivity rR of the powder,
Particle size distribution of the bulk material and In filling and emptying operations, electrostatic
mean value, M, ignition hazards are of prime importance due to the elec-
Volume and shape of the silo or container (volume trostatic charging occurring during separation process.
and shape of the product heap and of the dust cloud). The hazard comprises possible charge accumulation not
only on plant units, incl. drum and container, but also
Unless otherwise stated, the following sections are based in the case of insulating bulk materialon the bulk
on the assumption that the bulk materials are handled material itself which is shown in Fig. 17.
without flammable gases or vapors being present. Assuming the insulating bulk material carries a
charge, filling represents the more hazardous of the two
5.1. Pneumatic conveying system operations for the following reasons. In a filling oper-
ation, the bulk material undergoes dispersion (e.g. grav-
With pneumatic transport, potential ignition sources ity feed, pneumatic transport, etc.) and can therefore be
include: glowing particles from the product, mechanical charged in the separation processes occurring in trans-
sparks from connecting equipment and, in the case of port. The bulk material and hence its associated charge
highly insulating materials, electrostatic discharges (such is then packed in a small space in which the charge is
as Propagating Brush Discharges, Fig. 16). Inside the 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. 16. Examples of propagating brush discharges ISSA, 1996. Fig. 17. Filling (charging) and emptying (discharging) operations.
148 N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151
Table 5
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- 5.4. Dust separation
viding guidance for the required safety measures
Jaeger & Siwek, 1999. In the case of dust separators, especially in filters, the
dust explosion hazard must not be underestimated. The
5.3. Mixing probability of occurrence of a fine dust atmosphere
sensitive to ignition is large. In addition, the entrainment
Even when handling highly sensitive dusts, the occur- of ignition sources (e.g. formation of smoldering lumps)
rence of effective ignition sources in mixing is unlikely, and the danger of ignition through electrostatic charging
provided the following conditions are met: are of prime importance (see Fig. 19).
Electrostatic charging must be prevented by the fol-
In the filling and emptying of the mixer, the measures lowing measures:
applied are the same as those in the filling and empty-
ing of containers. Grounding of all conductive apparatus parts. Parti-
In the filling and emptying of the mixer, the mixing cular attention must be paid to the grounding of all
elements must be off or run at a circumferential speed conductive parts which could possibly be insulated
(tip speed) which does not exceed 1 m/s. This restric- from ground if a filter cloth made of insulating
tion must be assured by technical safeguards. material is used (e.g. filter supports, clamps). This
If the mixer is closed and is filled to fill level of 70 must be specially checked after repair and mainte-
vol.% or more, the circumferential speed of the mix- nance work.
ing elements is no longer restricted (see Fig. 14). With a MIE3 mJ or in the presence of flammable
Any insulating coating must have a breakdown volt- gases or vapors in the air being cleaned, electrically
age of less than 4 kV. Product build-up must be conducting filter materials must be used. An excep-
checked if a homogeneous layer can be formed. tion to this rule can be applied when the protective
Circumferential speeds up to 10 m/s can be tolerated measure inerting is employed. Continuity of the
in a mixer with a product fill height of less than 70 conductivity and safe grounding must be checked.
vol.%, provided the combination of material values Multiple washing can have an adverse effect on the
as listed in Table 5 are present. continuity of the filter material conductivity and thus
Nauta mixers with bottom support of the helical screw require repeat checking.
can heat up during operation and care must be exer- All inner walls on which dust can impact at high
cised with substances capable of spontaneous speed must not have any insulating inner coatings
decomposition. (Fig. 18) with a high electrical breakdown strength (breakdown
voltage must be less than 4 kV; periodic checks are
required).
Fig. 18. Mixing operation. Fig. 19. Electrostatic ignition sources (dust collector).
N. Jaeger / Journal of Loss Prevention in the Process Industries 14 (2001) 139151 149
Flexible intermediate bulk containers are used on ever To meet the specification as a type C bag, the following
increasing scale in the powder handling industry. requirements are recommended:
Depending the hazard situation at the location where
they are used, they must meet different requirements (see The FIBC must be clearly labeled indicating that it
Table 6) in order to avoid ignition hazards caused by is conductive and that grounding is required during
electrostatic charging (Fig. 20). charging and discharging (see Fig. 21).
The FIBC must have a clearly marked area for the
A: No special requirements. attachment of the grounding clamps.
B: Breakdown voltage of the FIBC wall material must The lifting naps must also be made of conductive
not exceed 4 kV in order to prevent propagating material and have a leakage resistance of less than
brush discharges. 108 to the FIBC body.
C: The bag material including the slings must be elec-
trostatic conductive. The resistance to the ground It is extremely important to keep in mind that the dis-
measured at any bag location (inside and outside) charge from an ungrounded bag can occur at single
must be fess than 100 M (108 ). The flexible bulk point. Such a discharge is strong enough to ignite dust
bag must have a grounding strap. The conductivity
and the necessity for grounding must be clearly mark-
ed.
All electrical design within the confines of the facility extinguishing barriers (see Fig. 23) for each convey-
are Class II Division 2 Group G with the exception ing line entering the silo. Silos and the conveying
of a three foot radius around the packaging machines lines are rated to withstand a reduced explosion press-
which will be Class II Division 1 Group G. The ure of 15 psig. Explosion venting was ruled out as a
classification was done in accordance to NFPA 499 protective measure in this case due to the indoor
NFPA, 1997a. location of these silos.
There are no moving parts at tip speeds of greater The three 3500 ft3 silos for the storage of the basic
than 1 m/s. materials is protected by an explosion venting system
All components are made out of conductive material. and extinguishing barriers for each conveying line
All conductive parts are reliably bonded and entering the silo. These three silos are rated to with-
grounded. stand a reduced explosion pressure of 50 psig.
Tramp metal is rejected from the system through Installed vent discharge ducts will ensure a safe
metal detectors. release from the enclosure to the outdoors NFPA,
All units with a volume of greater than 1 m3 are pro- 1997b (see Fig. 24).
tected by the use of constructional measures against Silos that do not receive product through pneumatic
explosions. This was necessary because the formation conveyors are inerted with nitrogen.
of a conical pile discharge cannot be excluded with All dust collectors are protected with explosion sup-
certainty. pression systems.
Silos with a volume of 500 ft3 or less are equipped All pneumatic conyeyings are done by nitrogen.
with explosion suppression systems (see Fig. 22) and
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 151
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-
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.
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.
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 185
diameter (23 m) and the mass median diameter (42 m) Hartmann tube (Nagy & Verakis, 1983; Dorsett et al.,
can be determined from the 50% points on the cumulat- 1960) is often used for preliminary screening tests and
ive curves in Fig. 1A. The cumulative curve also shows for minimum ignition energy (MIE) measurements.
that the dust has 82% by mass minus 200 mesh (75 However, it may yield false negatives for dusts that are
m). The differential curves in Fig. 1B are often more difficult to ignite with a spark but that are ignitable by
useful in visualizing the size distribution. Other ways of stronger ignition sources. It is also not recommended
identifying a representative particle size for the dust in (ASTM, 1999a) for measuring rates of pressure rise. The
Fig. 1 are the surface mean diameter (DS=30 m) and 20-L chambers are used for explosibility measurements
the mass mean diameter (DW=50 m), calculated from such as maximum explosion pressures, maximum rates
the data in Fig. 1B. Because the combustion of the dust of pressure rise, minimum explosible concentrations, and
cloud is greatly dependent on the surface area of the inerting effects. An example of a 20-L laboratory
dust, a mean particle diameter based on surface area is chamber (Cashdollar & Hertzberg, 1985) is shown in
perhaps more appropriate than one based on mass. Vari- Fig. 2. This is the standard laboratory test chamber
ous books on particle size analysis (e.g. Allen, 1975; designed and used at the PRL for studying the explos-
Irani & Callis, 1963), may be useful in better under- ibility and inerting of combustible dusts. There is another
standing this aspect of dusts. It should be noted that dif- style of 20-L chamber designed by R. Siwek (Bartknecht
ferent particle size analysis instrumentation may give 1981, 1989; Siwek 1977, 1985, 1988) that is in wide use
somewhat different results for the same dust because of in Europe and elsewhere. There are also 1-m3 (1000-L)
the different particle sizing methods used. chambers (Bartknecht 1981, 1989; Cashdollar & Chatra-
The combustion properties of a dust depend on its thi, 1993). The 1-m3 chambers may give more realistic
chemical and physical characteristics, especially its par- measurements of minimum explosible concentrations,
ticle size distribution. Published dust explosibility data maximum explosion pressures, and maximum rates of
can give an indication of the hazards associated with a pressure rise, but the testing is more time consuming and
particular type of dust. However, it is preferable to deter- requires much larger dust samples than the 20-L cham-
mine the explosibility characteristics of an industrial dust bers.
by test, because published data are for a particular size The PRL 20-L chamber is made of stainless steel, and
distribution that may be different from the dust in ques- has a pressure rating of 21 bar. Two optical dust probes
tion. Particle shape and porosity are also important con- (Cashdollar, Liebman & Conti, 1981; Conti, Cashdol-
siderations in the explosibility of a dust. In general, lar & Liebman, 1982) are used to measure the uniformity
shapes with greater surface area will propagate flame of the dust dispersion at the positions shown in Fig. 2.
more readily and therefore be more hazardous. The optical probes measure the transmission through the
It should be noted that there is no US standardized dust cloud, with path lengths of 38 or 95 mm. The strain
test for whether or not a dust is explosible. There are gauge pressure transducer measures the explosion press-
tests to determine whether a dust can be ignited by an ure and rate of pressure rise (dP/dt). The data from the
electric spark (Dorsett, Jacobson, Nagy & Williams, various instruments are collected by a high speed per-
1960) or what the maximum explosion pressures sonal computer (PC) based data acquisition system. The
(ASTM, 1999a) or minimum explosible concentrations experimental dust concentration reported for the 20-L
(ASTM, 1999b) are using stronger chemical ignitors. chamber is the mass of dust divided by the chamber vol-
One reason for this lack of an explosibility test is that ume. After the dust and ignitor have been placed in the
the question of whether or not a dust can be ignited and chamber, the chamber is partially evacuated to an absol-
propagate a flame depends greatly on the ignition source. ute pressure of 0.14 bar,a. Then a short blast of dry air
However, different industries have different views about (from a reservoir at 9 bar) disperses the dust and raises
what would be an appropriate or likely ignition source. the chamber pressure to about 1 bar,a. There is a total
For some industries, it could be an electrostatic spark; ignition delay of 0.4 s from the start of dispersion until
for others, it could be a flame; and for the mining indus- ignition for the standard test procedure in the PRL
try, it could a very large flame from blown-out explos- 20-L chamber. The standard procedure for the Siwek 20-
ives. Therefore, each industry has to decide what are the L chamber has an ignition delay of 0.06 s and a reser-
likely or possible ignition sources and which dusts could voir pressure of 20 bar, resulting in a higher level of
be ignited by them. turbulence. The usual ignition sources used for the 20-
L tests are electrically activated, pyrotechnic ignitors
manufactured by Fr. Sobbe2 of Germany. These ignitors
3. Laboratory equipment for dust explosibility are available in various energies from 250 to 10,000 J.
evaluation
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 The absolute pressure (Figs. 3A and 4A) and rate of
book of 20 pocket matches, all ignited at once. The pressure rise (Figs. 3B and 4B) are plotted versus time.
Sobbe ignitors are much stronger than the electric sparks Fig. 3 shows the data for a 20-L chamber explosion test
used in the 1.2-L Hartmann tests. 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
4. Explosion characteristics 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
4.1. Pressures and rates of pressure rise on the pressuretime trace. The ignitor is activated at
0.5 s at a chamber pressure of 1.0 bar,a. The maximum
Examples of the pressure data for a weak and a mod- explosion pressure is about 3 bar,a or a pressure rise of
erate coal dust explosion are shown in Figs. 3 and 4. about 2 bar. In Fig. 3B, the rate of pressure rise,
(dP/dt)ignitor, for the ignitor is larger than (dP/dt)ex for of reactivity is the magnesium (Mg) dust, followed by
the dust explosion itself. It is important to determine the two carbonaceous dusts (polyethylene and coal). The
(dP/dt)ex rather than the ignitor effects. Fig. 4 shows data polyethylene and the high volatile bituminous coal (hvb)
for a larger explosion of the low volatile coal dust at a have similar maximum pressures, but the polyethylene
higher concentration of 200 g/m3. The maximum has a faster rate of pressure rise. The titanium (Ti) dust
explosion pressure is about 5.5 bar,a or a pressure rise has a lower explosion pressure than the carbonaceous
of 4.5 bar. For this explosion, the (dP/dt)ex for the dust dusts, and the iron (Fe) and zinc (Zn) dusts are even
explosion is greater than (dP/dt)ignitor. lower. The dust with the lowest reactivity in Fig. 5 is the
Examples of absolute pressure versus time traces for tantalum (Ta) dust, which barely reaches its maximum
typical dust explosions at a concentration of 600 g/m3 pressure by 250 ms. The relative reactivities of these
in the constant volume 20-L chamber are shown in Fig. dusts are dependent not only on the intrinsic reactivities
5. The traces are for two carbonaceous dusts and six of the materials but also on the specific particle sizes of
metal dusts. The relative reactivity of the dusts can be the dusts.
estimated from either the peak explosion pressure or the The pressure evolution of an explosion in a constant
maximum rate of pressure rise. The aluminum (Al) has volume system is predicted by classical combustion
the highest reactivity, in part because it is much finer in theory (Lewis & von Elbe, 1961, pp. 367381). For the
size than any of the other dusts in Fig. 5. Next in order ideal case, the absolute pressure as a function of time,
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
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
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 at two magnifications on the left side of the figure. They
for polyethylene dust at very high concentrations. This are compared to the burned post-explosion particles in
shows that these dusts explode even at concentrations the four frames on the right side of the figure. The
beyond 4000 g/m3. There is, of course, an increased burned particles are mainly char residues that are often
uncertainty in the dust dispersion effectiveness at these larger than the original particles. In the flame, the bitum-
very high concentrations. The decrease in pressure at inous coal particles become molten as shown by the
higher concentrations may be due to the increased heat rounded particles on the right. Some particles form
sink of the very large dust concentrations. The decrease cenospheres. The particles also devolatilize in the flame,
in dP/dt at higher concentrations may be due to the and the volatiles are emitted through the blow holes
increased heat sink effect and/or to the possible decrease seen in the char residues. Additional SEM photomicro-
in turbulence due to the large mass of dust. Deguigand & graphs for various post-explosion dust residues are in
Galant (1981) had previously observed an apparent Ng, Cashdollar, Hertzberg & Lazzara (1983).
upper limit at 4 kg/m3 for coal dust, but this may have Metal dusts show similar explosibility data to carbon-
been only an ignitability limit because they used an elec- aceous dusts, as shown by the data for two sizes of iron
tric spark ignition source that was much weaker than the dust in Fig. 10. The Fe-1 dust was finer in size and had
2500-J Sobbe ignitor used here. Mintz (1993) observed Dmed4 m; the Fe-2 dust had Dmed45 m. The
some upper limits under conditions of reduced oxygen explosion pressures, rates of pressure rise, and measured
and at large coal particle sizes. In principle, there are explosion temperatures are shown as a function of dust
rich limits for dusts. Eventually, the large mass of excess concentration. Fig. 10C shows the measured explosion
fuel will become too much of a heat sink and the flame pressure (absolute) for each test, corrected for the press-
temperature will be reduced below its limit value. How- ure rise due to the ignitor. Fig. 10B shows the size nor-
ever, for most practical purposes, dusts can be con- malized maximum rate of pressure rise, (dP/dt)V1/3, for
sidered to have no rich limit of explosibility. This obser- each explosion test. As for the coal dust data in Fig. 6,
vation has also been made by Wolanski (1992). the iron data in Fig. 10 show that explosions are not
Examples of scanning electron microscope (SEM) observed below a certain dust concentration. The MECs
photomicrographs of coal before and after explosions are for the Fe-1 and Fe-2 dusts are about 220 and 500 g/m3,
shown in Fig. 9. The dust was a narrow size distribution respectively, based on the procedures of ASTM E1515
of Pittsburgh coal with a mass median diameter, (ASTM, 1999b). However, there is considerable uncer-
Dmed23 m. The original unburned particles are shown 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 191
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- with the data crosses for the broad size distributions of
trations, Pmax and (dP/dt)maxV1/3 level off as all of the coal dusts. The MEC-values in the bottom section of the
oxygen in the chamber is consumed. Similar to the car- figure are relatively independent of particle size for the
bonaceous dusts in Fig. 8, the iron metal dusts show no finer sizes. At the larger sizes, above 100 m, the MEC-
evidence of a normal rich limit. values increase with particle size until a size is reached
The explosion temperatures shown in Fig. 10A were that can not be ignited. The top two sections of Fig. 11
measured with a six-wavelength infrared pyrometer show that the maximum pressures and rates of pressure
(Cashdollar & Hertzberg, 1982). The pyrometer rise are found at the finest sizes tested, as expected. The
observed the continuum radiation from the particles, and pressures decline slowly and the pressure rise rates
temperatures were calculated from the best Planck curve decrease faster with increasing particle size. At some
fit to the infrared radiance data. The maximum measured size between 200 and 300 m, the narrow sizes of Pitts-
particle temperatures for the Fe-1 dust were 1800 K, burgh coal dust can no longer be ignited. These data are
well below the maximum calculated adiabatic tempera- typical for narrow size distributions of carbonaceous fuel
ture, Tad,max=2250 K, for ideal combustion at constant dusts. A broad size distribution is just a combination of
pressure (Cashdollar, 1994). The maximum measured narrow distributions, and these data show that it is the
particle temperatures for the Fe-2 dust were even lower. finer particles in a broad distribution that contribute the
These experimental temperatures are only those of the most to its hazard. The MEC data crosses for the broad
particles in the explosion; the gas temperatures may be size distributions show little difference from the narrow
different. For all three explosion characteristics shown size distribution data below Dmed100 m. However,
in Fig. 10, the Fe-1 dust has higher values than the Fe- the broad size distributions ignite and propagate at larger
2 dust, showing that it is more reactive, due to its finer Dmed sizes than the narrow size distributions. The press-
particle size. ure and dP/dt data for the broad size distributions are
somewhat higher than those for the narrow size distri-
4.3. Particle size effects butions, even in the Dmed range of 20100 m. These
effects are probably due to the tail of fine particles in
Most of the previous explosibility data were measured the broad size distributions. These fine particles were
using rather broad size distributions of the dusts. Fig. removed from the narrow size distributions. The main
11 shows explosibility data from the 20-L chamber for conclusion of Fig. 11 is that particle size has an
Pittsburgh bituminous coal dust as a function of mass important effect on the explosibility of coal dusts and
median particle diameter. The data for the narrow size other carbonaceous dusts.
distributions are shown as the solid circles and solid Data showing the effect of particle size for iron dust
curve. These data for narrow distributions are compared 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
broader and the Dmed values less certain, the data are the dusts were dispersed with various oxygennitrogen
shown as bars rather than points. The explosion data are mixtures instead of normal air at 20.95% O2. Fig. 13 is
similar to those for the coal dust. The maximum values an example of the reduced oxygen data for coal dust.
for pressure and rate of pressure rise are found at the The explosions are denoted by the solid circles and the
finest particle size. The MEC values are relatively size- nonexplosions by the open circles. The data for coal dust
independent at the finer sizes and increase above 30 m in air are shown at the top of the figure. In air, the dust
until a size is reached that can not be ignited. Additional ignites and burns at all coal concentrations above the
data for size effects of aluminum dusts are in Cashdol- MEC of 80 g/m3. At the bottom of the figure,
lar (1994). explosions still occur at 14% down to 11.5% O2. At 11%
O2, the coal dust ignited only in one out of eight tests.
4.4. Effects of oxygen concentration At even lower oxygen concentrations, the dust could not
be ignited. The boundary between oxygen concentrations
One of the ways to prevent a dust explosion is to inert that support combustion and those that do not support
the atmosphere so that there is insufficient oxygen for a combustion is the limiting oxygen concentration, LOC.
flame to propagate. This removes one side of the fire As a safe margin, NFPA 69 (NFPA, 1997) recommends
triangle or explosion pentagon, thereby preventing com- keeping the system oxygen concentration at least 2%
bustion. One of the most common inerting gases is nitro- lower than the measured LOC. Gases other than nitrogen
gen, which is the main constituent of air. To determine can also be used to reduce the oxygen concentration.
the limiting oxygen concentration for coal dust Carbon dioxide is usually more efficient than nitrogen
explosions in the 20-L chamber (with 2500-J ignitors), for inerting carbonaceous dusts, but it is often less effec-
K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199 193
Fig. 11. Effect of particle size on the explosibility of coal dusts: , Fig. 12. Effect of particle size on the explosibility of iron dusts.
for narrow distributions; , for broad distributions.
tive than nitrogen in inerting metal dusts (Nagy, Dor- decreasing oxygen concentration, until the mixture
sett & Jacobson, 1964). would no longer be explosible.
The oxidant for a dust explosion is usually the oxygen Because of the large effect of varying oxygen concen-
in air, although other gases can also be oxidizers. Oxy- tration on the explosion characteristics of dusts, it is
gen concentrations greater than 21% tend to increase the important to test the dust at the appropriate O2 concen-
burning velocity, and concentrations less than 21% tration. When determining the explosion characteristics
reduce the burning velocity. An example of the effect of for a dust in air, it is important to measure the O2 content
varying oxygen concentration on the explosion pressure of the air cylinders used for the tests. Gas cylinders
and rate of pressure rise for a carbonaceous dust is that are filled with air that has been compressed and
shown in Fig. 14. The data are from DiPalma (1998). dried have the normal 20.95% O2. However, many air
The solid data symbols are at a dust concentration of cylinders are filled with synthetic or reconstituted air
500 g/m3 and the open circle data symbols are at a dust that has been mixed from liquified oxygen and nitrogen.
concentration of 375 g/m3. Fig. 14A shows that the rate The O2 content of these cylinders has been observed to
of pressure rise varies almost exponentially with oxygen vary considerablyfrom 19% to 26% O2.
concentration on this semi-logarithmic plot. The
explosion pressure in Fig. 14B varies roughly linearly 4.5. Effect of temperature
with oxygen concentration, although there is some scat-
ter in the data. If the dust concentration was varied at The thermal ignitability of coal dust is shown in Fig.
each oxygen concentration to obtain the highest Pmax 15, as measured in the PRL 6.8-L furnace (Conti, Cash-
value, the explosion pressure would be expected to dollar & Thomas, 1993). The tests resulting in ignitions
increase linearly with oxygen concentration, based on (solid circles) and non-ignitions (open circles) are plot-
the ideal gas law (Eq. (1)). However, this linear relation- ted on a graph of initial furnace temperature versus dust
ship would change as the LOC is approached. Near the cloud concentration. The solid curve is the temperature
LOC, the pressure would decrease very rapidly with boundary between the upper region of the graph where
194 K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199
dust would autoignite. For these tests, the 20-L chamber where CT is the limit in terms of mass concentration at
was wrapped with electrical heater tape and insulated to temperature T, CT0 is the limit at T0, and the temperatures
reach the elevated temperature. A thermocouple meas- are in C. The dust concentrations to the right of the
ured the set temperature of the chamber before the test. solid curve are flammable (explosible) and the region to
The solid circle data points in Fig. 16 show the MEC- the left of the curve is nonflammable. For comparison,
data (Cashdollar, 1996) for the coal dust at near ambient the measured lean flammable limit data for methane gas
(60C) and at an elevated temperature of 180C. The as a function of temperature (dashed curve, from Cow-
experimental data points are extrapolated to even higher ard & Jones, 1952, p. 43) are also shown. The decrease
temperatures (solid curve) using the modified Burgess in the LFL or MEC with increase in temperature is simi-
Wheeler law (Zabetakis, 1965; Conti, Cashdollar, Hertz- lar in form for the dust and gas.
berg & Liebman, 1983) for hydrocarbons: At higher dust concentrations, the maximum
K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199 195
Fig. 18. Lower flammable limits for hybrid mixtures of low and high
volatile coal dusts with methane gas.
Fig. 17. Effect of pressure on the lower flammable limits for coal
and polyethylene dusts, compared with methane gas.
the symbol in the figure. The data for hybrid mixtures
of the low volatile Pocahontas coal and CH4 in Fig. 18A
4.7. Hybrid mixtures of dusts and gases show some curvature. This is probably due to the even
greater difference in ignitability between the low volatile
Another important factor in the explosibility hazard coal and the CH4. That is, the dust becomes more easily
of a dust is the possible co-presence of a flammable gas. ignited as more CH4 is added. Therefore, the curvature
Hybrid mixtures of a combustible dust (coal) and a is more likely an effect of ignitability rather than an
flammable gas (CH4) were also studied in the 20-L effect of flammability. Ideally, the true mixing relation-
chamber using 2500-J ignitors (Cashdollar, 1996). Data ship would be determined in a much larger chamber,
for a low volatile bituminous (lvb) coal are shown in such as a 1-m3 chamber, where a very strong ignition
Fig. 18A, and for a high volatile bituminous (hvb) coal source could be used for the dusts without overdriving
in Fig. 18B. The flammable limits for mixtures of coal
the CH4 gas. For most practical situations for mixtures
and CH4 are shown by the data points and solid curves.
of hydrocarbon dusts and gases, the linear mixing law
The areas above and to the right of the curves are explos-
of Le Chatelier would be sufficient. This approximately
ible (flammable) and the areas below and to the left of
linear relationship for the lean limits of coal dust and
the curves are nonexplosible (nonflammable). The data
CH4 gas mixtures was also observed by Amyotte and
for mixtures of Pittsburgh coal and CH4 in Fig. 18B
colleagues (Amyotte, Mintz, Pegg, Sun & Wilkie, 1991;
show a linear or near-linear mixing relationship similar
to Le Chateliers Law for hydrocarbon gases (Zabetakis, Amyotte, Mintz, Pegg & Sun, 1993) using 5000-J igni-
1965; Kuchta, 1985, pp. 4850). All of the solid circle tors in a 26-L chamber. This linear mixing relationship
data symbols are for 2500-J ignitors. The measured LFL is also applicable to mixtures of two carbonaceous dusts
for the pure CH4 with this 2500-J ignitor is 4.4%, but (Hertzberg & Cashdollar, 1987). However, it is not
this is an overdriven system as shown by tests in a larger applicable to mixtures where the two components have
120-L chamber (Hertzberg et al., 1988a). The more greatly different limit flame temperatures, such as a car-
appropriate LFL for CH4 is the 4.9% value measured bonaceous dust and hydrogen gas (Hertzberg & Cashdol-
with a 1000-J ignitor in the 20-L chamber and shown as lar, 1987).
K.L. Cashdollar / Journal of Loss Prevention in the Process Industries 13 (2000) 183199 197
4.8. Effect of added inert dust the LFL or MEC of the coal dust shows almost no effect
with added rock dust until there is over 50% rock dust
The addition of an inert powder to a combustible dust in the mixture. At higher rock dust percentages, the LFL
and air mixture can reduce the explosibility through the increases, until the mixture can not be ignited at 75%
absorption of heat. In the mining industry, coal dust rock dust. The two lower parts of the figure show the
explosions are prevented by the addition of limestone maximum pressure and rate of pressure rise as a function
rock dust to the deposited coal dust (Nagy, 1981). Since of rock dust percentage in the mixture. At each rock dust
the limestone is incombustible, it acts as a heat sink to percentage, the coal dust concentration was varied over
reduce the flame temperature of the dust mixture below a series of tests to determine the maximum pressure and
its limit value. The inerting of coal dust by the addition dP/dt. The explosion pressures show only a slight
of limestone rock dust has been studied in the PRL decrease with added rock dust content up about 70%.
20-L laboratory chamber, and the results were compared Between 70% and 80% rock dust, the pressures drop
to those from full-scale experimental mine tests rapidly as the mixture becomes totally inerted and flame
(Cashdollar et al., 1992; Cashdollar, 1996; Cashdollar & no longer can propagate. The rates of pressure rise
Hertzberg, 1989; Greninger et al., 1991). The laboratory decline almost linearly with increased rock dust content
data are shown in Fig. 19. In the top part of the figure, 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
Acknowledgements
explosions and helping him to understand combustion dust probe for monitoring dust explosions. Review of Scientific
theory. 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.
References Conti, R.S., Cashdollar, K.L., & Thomas, R.A. (1993). Improved 6.8-L
furnace for measuring the autoignition temperatures of dust clouds.
Allen, T. (1975). Particle size measurement. London: Chapman and US Bureau of Mines RI 9467, 26 pp.
Hall. Coward, H.F., & Jones, G.W. (1952). Limits of flammability of gases
Amyotte, P. R., Mintz, K. J., Pegg, M. J., Sun, Y. H., & Wilkie, K. and vapors. US Bureau of Mines Bulletin 503, 155 pp.
I. (1991). Laboratory investigation of the dust explosibility charac- Deguigand, B., & Galant, S. (1981). Upper flammability limits of coal
teristics of three Nova Scotia coals. Journal of Loss Prevention in dustair mixtures. In 18th Symposium (International) on Combus-
the Process Industries, 4, 102109. tion (pp. 705715). Pittsburgh, PA: The Combustion Institute.
Amyotte, P. R., Mintz, K. J., Pegg, M. J., & Sun, Y. H. (1993). The DiPalma, J. (1998). Cytec Industries, Inc., Stamford, CN, personal
ignitability of coal dustair and methanecoal dustair mixtures. communication.
Fuel, 72, 671679. Dorsett, H.G., Jr., Jacobson, M., Nagy, J., & Williams, R.P. (1960).
ASTM (1999a). Standard test method for pressure and rate of pressure Laboratory equipment and test procedures for evaluating explos-
rise for combustible dusts E1226-94. In ASTM annual book of stan- ibility of dusts. US Bureau of Mines RI 5624, 21 pp.
dards, vol. 14.02. West Conshohocken, PA: American Society for Eckhoff, R. K. (1991). Dust explosions in the process industries.
Testing and Materials. Oxford: Butterworth Heinemann.
ASTM (1999b). Standard test method for minimum explosible concen- Field, P. (1982). Dust explosions. Handbook of powder technology,
tration of combustible dusts E1515-98. In ASTM annual book of vol. 4. Amsterdam: Elsevier.
standards, vol. 14.02. West Conshohocken, PA: American Society Greninger, N. B., Cashdollar, K. L., Weiss, E. S., & Sapko, M. J.
for Testing and Materials. (1991). Suppression of dust explosions involving fuels of inter-
ASTM (1999c). Standard test method for minimum autoignition tem- mediate and high volatile content. In Proceedings of the 4th Inter-
perature of dust clouds E1491-97. In ASTM annual book of stan- national Colloquium on Dust Explosions (November 49, 1990,
dards, vol. 14.02. West Conshohocken, PA: American Society for Porabka-Kozubnik, Poland) (pp. 208228).
Testing and Materials. Hertzberg, M., & Cashdollar, K. L. (1987). Introduction to dust
Bartknecht, W. (1981). Explosions: course, prevention, protection. explosions. In Industrial dust explosions, STP 958 (pp. 532). West
Berlin: Springer. Conshohocken, PA: American Society for Testing and Materials.
Bartknecht, W. (1989). Dust explosions: course, prevention, protec- Hertzberg, M., Cashdollar, K. L., & Zlochower, I. A. (1988a). Flamm-
tion. Berlin: Springer. ability limit measurements for dusts and gases: ignition energy
Bartknecht, W. (1993). Explosions-Schutz (explosion protection). requirements and pressure dependences. In 21st Symposium
Berlin: Springer. (International) on Combustion (pp. 303313). Pittsburgh, PA: The
Cashdollar, K. L. (1994). Flammability of metals and other elemental Combustion Institute.
dusts. Process Safety Progress, 13, 139145. Hertzberg, M., Zlochower, I. A., & Cashdollar, K. L. (1988b). Vola-
Cashdollar, K. L. (1996). Coal dust explosibility. Journal of Loss Pre- tility model for coal dust flame propagation and extinguishment.
vention in the Process Industries, 9, 6576. In 21st Symposium (International) on Combustion (pp. 325333).
Cashdollar, K. L., & Chatrathi, K. (1993). Minimum explosible dust Pittsburgh, PA: The Combustion Institute.
concentrations measured in 20-L and 1-m3 chambers. Combustion Hertzberg, M., Zlochower, I. A., & Cashdollar, K. L. (1992). Metal
and Science Technology, 87, 157171. dust combustion. In 24th Symposium (International) on Combus-
Cashdollar, K. L., & Hertzberg, M. (1982). Infrared pyrometers for tion (pp. 18271835). Pittsburgh, PA: The Combustion Institute.
measuring dust explosion temperatures. Optical Engineering, 21, Irani, R. R., & Callis, C. F. (1963). Particle size: measurement,
8286. interpretation, and application. New York: John Wiley and Sons.
Cashdollar, K. L., & Hertzberg, M. (1985). 20-L explosibility test ISO (1985). Explosion protection systemsPart 1: determination of
chamber for dusts and gases. Review of Scientific Instruments, 56, explosion indices of combustible dusts in air, ISO 6184/1. Switzer-
592602. land: International Organization for Standardization.
Cashdollar, K. L., & Hertzberg, M. (1987). In: Cashdollar, K. L., & Kuchta, J. M. (1985). Investigation of fire and explosion accidents in
Hertzberg, M. (Eds), Industrial Dust Explosions (Proceedings of the chemical, mining, and fuel-related industriesa manual. US
the Symposium on Industrial Dust Explosions, Pittsburgh, PA, June Bureau of Mines Bulletin 680. pp. 4850.
1013, 1986), STP 958 (363 pp). West Conshohocken, PA: Amer- Lewis, B., & von Elbe, G. (1961). Combustion, flames, and explosions
ican Society for Testing and Materials. of gases. New York: Academic Press.
Cashdollar, K. L., & Hertzberg, M. (1989). Laboratory study of rock Mintz, K. J. (1993). Upper explosive limit of dusts: experimental evi-
dust inerting requirements: effects of coal volatility, particle size, dence for its existence under certain circumstances. Combustion
and methane addition. In Proceedings of the 23rd International and Flame, 94, 125130.
Conference of Safety in Mines Research Institutes (September 11 Nagy, J. (1981). The explosion hazard in mining. US Mine Safety and
15, 1989, Washington, DC) (pp. 965977). Health Administration IR 1119, 69 pp.
Cashdollar, K.L., Liebman, I., & Conti, R.S. (1981). Three bureau of Nagy, J., & Verakis, H. C. (1983). Development and control of dust
mines optical dust probes. US Bureau of Mines RI 8542, 26 pp. explosions. New York: Marcel Dekker.
Cashdollar, K. L., Hertzberg, M., & Zlochower, I. A. (1989). Effect Nagy, J., Dorsett, H.G., Jr., & Jacobson, M. (1964). Preventing ignition
of volatility on dust flammability limits for coals, gilsonite, and of dust dispersions by inerting. US Bureau of Mines RI 6543, 29
polyethylene. In 22nd Symposium (International) on Combustion pp.
(pp. 17571765). Pittsburgh, PA: The Combustion Institute. NFPA (1997). Standard on explosion prevention systems, NFPA 69.
Cashdollar, K. L., Weiss, E. S., Greninger, N. B., & Chatrathi, K. Quincy, MA: National Fire Protection Association.
(1992). Laboratory and large-scale dust explosion research. NFPA (1998). Guide for venting of deflagrations, NFPA 68. Quincy,
Plant/Operations Progress, 11, 247255. MA: National Fire Protection Association.
Conti, R. S., Cashdollar, K. L., & Liebman, I. (1982). Improved optical 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 199
Electron microscopy studies of explosion and fire residues. US Weiss, E. S., Greninger, N. B., & Sapko, M. J. (1989). Recent results
Bureau of Mines IC 8936, 63 pp. of dust explosion studies at the Lake Lynn experimental mine. In
Siwek, R. (1977). 20-L-Laborapparatur fur die Bestimmung der Proceedings of the 23rd International Conference of Safety in
Explosionskenngrossen brennbarer Staube (20-L laboratory appar- Mines Research Institutes, (September 1115, 1989, Washington,
atus for the determination of the explosion characteristics of flam- DC) (pp. 843856).
mable dusts). Thesis at Winterthur Engineering College, Win- Wiemann, W. (1987). Influence of temperature and pressure on the
terthur, Switzerland, 109 pp., in German. explosion characteristics of dust/air and dust/air/inert gas mixtures.
Siwek, R. (1985). Development of a 20 ltr laboratory apparatus and In Industrial Dust Explosions, STP 958 (pp. 3344). West Consho-
its application for the investigation of combustible dusts. Basel, hocken, PA: American Society for Testing and Materials.
Switzerland: Ciba Geigy AG. Wolanski, P. (1992). Dust explosion research in Poland. Powder Tech-
Siwek, R. (1988). Reliable determination of the safety characteristics nology, 71, 197206.
in 20-L apparatus. In Flammable Dust Explosion (November 24, Zabetakis, M.G. (1965). Flammability characteristics of combustible
1988, St. Louis, MO) (pp. 529573). Holliston, MA: IBC USA. gases and vapors. US Bureau of Mines Bulletin 627, 121 pp.
Stephan, C.R. (1990). Coal dust as a fuel for fires and explosions. Zhen, G., & Leuckel, W. (1997). Effects of ignitors and turbulence on
Report No. 01-066-90, US Mine Safety and Health Administration, dust explosions. Journal of Loss Prevention in the Process Indus-
Pittsburgh, PA, 10 pp. tries, 10, 317324.
VDI (1983). Pressure release of dust explosions, VDI 3673. Dussel-
dorf, Germany: Verein Deutscher Ingeniere.
Journal of Loss Prevention in the Process Industries 13 (2000) 209219
www.elsevier.com/locate/jlp
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.
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
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 211
puter (PC) based data acquisition system. The dust to be of the ball valve disperses the dust and air into the 1-
tested can be placed either in the dust reservoir or on m3 chamber through the rebound nozzle and raises the
top of the dispersion nozzle at the bottom of the chamber chamber pressure to about 1 bara. The ignitor is fired
(Fig. 1). After the dust and igniter have been placed in 0.6 s after activation of the ball valve. The reported
the chamber, the top is bolted on and the chamber is experimental dust concentration for the 1-m3 chamber is
partially evacuated to an absolute pressure of 0.14 bara. the mass of dust divided by the vessel volume.
Then a short blast of dry air (0.3 s duration from a 16- The dispersion time and measured Kst values (and pre-
L reservoir at 9 barg) disperses the dust and raises the sumably the turbulence level) in the Fike 1-m3 chamber
chamber pressure to about 1 bara. The ignitor is acti- are comparable to those in European 1-m3 chambers
vated after an additional delay of 0.1 s. This results in (Bartknecht, 1989). This is the turbulence level in VDI
a total ignition delay of 0.4 s from the start of dispersion Standard 3673, ISO Standard 6184/1, and ASTM Stan-
until ignition. The experimental dust concentration dard E1226 used to determine the maximum rate of
reported in this paper is the mass of dust divided by the pressure rise of a dust explosion. The Kst and turbulence
chamber volume, i.e. the nominal dust loading. levels in the PRL 20-L chamber are lower, but this
The Fike Corporation 1-m3 chamber (Figs. 3 and 4) should not significantly affect measurements of the MEC
was also used to measure dust explosibilities. The 1-m3 or LOC (Cashdollar & Chatrathi, 1992). The main effect
chamber is spherical with an internal diameter of 1.22 of increased turbulence at low dust concentrations is to
m and a wall thickness of 9.5 mm. It has a pressure make the dust cloud more difficult to ignite (Amyotte,
rating of 21 barg. The two halves of the sphere are con- Chippett & Pegg, 1989). However, with the strong igni-
nected by 12 bolts of 51 mm diameter. Two variable tors used for the tests, the somewhat higher turbulence
reluctance pressure transducers were used to measure the level in the 1-m3 chamber should have little effect on
explosion pressure. Data from the instruments were col- the measurements.
lected by a high speed PC based data acquisition system. The ignition sources used for the tests were electri-
The dust injection system for the 1-m3 chamber con- cally activated chemical ignitors manufactured by Fr.
sists of a 5-L dispersion reservoir, a 19-mm pneumati- Sobbe of Germany. The ignitors are composed of 40%
cally activated ball valve, and a rebound nozzle (Fig. 3). zirconium, 30% barium nitrate, and 30% barium per-
In previous work (Cashdollar & Chatrathi, 1992), a ring oxide. They are activated electrically with an internal
nozzle was used. To create a dust cloud, a weighed sam- fuse wire and deliver their energy in about 10 ms. The
ple of dust is placed in the dispersion reservoir. The res- Sobbe ignitors are available in energies of 0.25, 0.50,
ervoir is pressurized with dry air to 20 bara and the 1.0, 2.5, 5.0 and 10.0 kJ. These are nominal calorimetric
chamber is partially evacuated to 0.88 bara. Activation energies based on the mass of pyrotechnic powder in
each ignitor. The 5000-J ignitor by itself produces a 3. MEC data and discussion
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 In a previous report (Cashdollar & Chatrathi, 1992)
chemical properties of the dusts are shown in Table 1. in this series, the MECs of three dusts were evaluated
The gilsonite is an asphaltic material mined in Utah. in the PRL 20-L chamber and in the Fike 1-m3 chamber
The bituminous coal is from the Pittsburgh seam; this following the test procedures in ASTM E1515. The
dust has been used for decades as a standard test dust MEC values for gilsonite dust and bituminous coal dust
(Rice & Greenwald, 1929; Cashdollar, Sapko, Weiss & were measured in each chamber at several ignition ener-
Hertzberg, 1987). RoRo93 was distributed worldwide by gies. Chemical ignitors with energies from 0.5 to 10 kJ
A. Kuhner AG, of Switzerland, in 1993 as a round robin were used in the tests. These results, given as the top
test material for Pmax and KSt testing. RoRo93 is a two dusts in Table 2, indicated that the 20-L chamber
2,2,6,6-tetramethylpiperidine derivative (light stabilizer). may be overdriven with high energy ignitors. The MEC
Lycopodium clavatum (reticulate form) is a plant spore values measured in the 20-L chamber with 2.5-kJ igni-
obtained from the Meer Corporation. The iron is a tors were comparable to those measured in the 1-m3
minus-325-mesh powder. The aluminum was Alcoa ato- chamber with 10-kJ ignitors. At higher ignition energies
mized aluminum powder, grade 123. The size distri- in the 20-L chamber, there was evidence of overdriving.
butions were determined from a combination of sonic The explosibility of anthracite coal was also studied in
sieving data, Coulter counter data, and laser diffraction the two chambers, but the data are not listed in the table.
particle size data. The anthracite did not ignite at 2.5 kJ, but appeared to
Table 1
Physical and chemical properties of fuels
Table 2
Summary of MEC testing results (g/m3)
1 kJ 2.5 kJ 5 kJ 10 kJ 2.5 kJ 5 kJ 10 kJ
Fig. 9. MEC data for iron powder from the 20-L chamber. Fig. 10. MEC data for iron powder from the 1-m3 chamber.
1-m3 results for the carbonaceous dusts. This is a result overdriving was occurring, at least for some of the dusts.
of overdriving the smaller vessel with too strong an Another difference between Siweks data and the data
ignition source. For the carbonaceous dusts, the closest presented in this paper is that he reports the MEC as the
agreement to the 1-m3 data was found with a 2.5-kJ igni- highest concentration that does not produce an explosion
tor in the 20-L vessel. rather than as the lowest concentration that just produces
The iron results appear to be somewhat different in an explosion.
that the 20-L MEC at 2.5-kJ ignition energy was higher In conclusion, the MEC data presented in this paper
than the 1-m3 MEC. It must be noted that these MEC show that overdriving can occur in the 20-L vessel using
values are three to five times higher than those found 5-kJ or 10-kJ ignitors. For the dusts tested, the MEC
with the carbon-based fuels. The data in Fig. 11D show data using 2.5-kJ or 5-kJ ignitors in the 20-L vessel
that for the more difficult-to-ignite dusts (such as this agreed best with the MEC data from the 1-m3 vessel
iron powder) with higher MEC values, the use of a 5- using a 10-kJ ignitor. The extent of overdriving in the
kJ ignitor in the 20-L chamber may be more appropriate 20-L vessel is dependent on the type of dust. There was
for agreement with the 1-m3 data at 10 kJ. less of an effect (Fig. 11) for the RoRo93, gilsonite, and
Siwek (1988) compared the MEC values measured lycopodium, all of which have a high volatility. The
with a 10-kJ ignitor in 20-L and 1-m3 vessels. He con- overdriving in the 20-L vessel was greatest for the coal
cluded that there was no significant effect of vessel vol- dust, which had a much lower volatility.
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- 4. LOC data and discussion
tration increment was 10 g/m3. A close examination of
the data in Fig. 8 of his paper, however, suggests that All of the LOC testing in the 1-m3 vessel was conduc-
the MECs for six of the 16 dusts studied were between ted with the 10-kJ ignition source and the 20-L tests were
10 and 30% lower in the 20-L vessel than in the 1-m3 conducted mainly with 2.5- or 5-kJ ignition sources. The
vessel. Two of the dusts were less than 10% higher and criteria for ignition were unchanged from the MEC test-
the rest were about the same. This suggests that some 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
centration on the measurement. In effect, the fuel con- RoRo93 also ignited at 11% but not at 10% oxygen, so
centration was varied in order to determine the lowest the LOC value is also 10.5%. In this case the LOC value
possible LOC at any concentration, i.e. the worst case. determined in the 20-L chamber using a 2.5-J ignitor
The oxygen concentration was reduced by adding nitro- was in agreement with the data from the 1-m3 chamber
gen to the air. This was done in both test vessels as well using 10-kJ ignitor. All of the LOC data are summarized
as in the dust injection gas reservoirs for the 20-L and in Table 3.
1-m3 chambers. This approach eliminated any question The 20-L data for Pittsburgh coal in Fig. 13 clearly
about the actual O2 concentration in the initial stages of show that the effect of the ignition source is significant
dispersion and ignition. in this volume. As ignition energy increases, lower quan-
The LOC data for the RoRo93 dust are shown in Fig. tities of fuel and oxygen are required to create 1 bar
12. Tests were made over a range of dust concentrations overpressures. A comparison of the 14% oxygen and 150
in order to determine the LOC at the worst case. The g/m3 coal coordinates on the three 20-L graphs provides
ignitions/explosions are shown as the solid data symbols a good example of the effect of ignition source. With a
and the nonignitions/nonexplosions are shown as the 1-kJ ignition source, the 14% oxygen and 150 g/m3 point
open data symbols. The 1-m3 chamber data using 10-kJ is outside the flammability envelope. With a 2.5-kJ
ignitors are shown in the top part of the figure and the ignition source, this point is just inside the flammable
20-L data using 2.5-kJ ignitors are shown in the bottom envelope and with a 5-kJ ignition source, this point is
of the figure. In the 1-m3 chamber, the RoRo93 dust well within the flammability envelope. Increasing the
ignited and burned at 11% oxygen but not at 10% oxy- ignition energy increases the size of the flammability
gen. The reported LOC value is then taken as the average zone. As listed in Table 3, the LOC in the 20-L chamber
of these values or 10.5%. In the 20-L chamber, the 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
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. Fig. 15. LOC data for aluminum.
source may be the most appropriate ignition source to safety margin. NFPA 69 recommends keeping the oxy-
use for LOC measurement in the 20-L vessel to match gen concentration at least 2% below the measured LOC
the LOC data from the 1-m3 vessel using a 10-kJ ignitor. value when protecting equipment. It is important to
However, there may be some dusts for which overly con- remember that the LOC values listed in Table 3 are only
servative LOC values will be obtained in the 20-L for the specific dusts tested and may not be applicable
chamber with the 2.5-kJ ignitor. to other particle sizes of the same materials. Often, finer
Siwek (1988) recommended that a 5-kJ ignition sizes of dusts have lower LOC values. It is also
source should be used for dusts with LOC values below important to recognize that these LOC data are for nitro-
10%. However, neither the current study nor Siweks gen inerting of air. Inerting with other gases such as car-
work investigated dusts with LOC values significantly bon dioxide may give different results.
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 5. Conclusion
the ignition source is to either overdrive or underdrive
these systems, then it might be expected that a 5-kJ The data from this study and the previous work
ignition source would more likely overdrive these sys- (Cashdollar & Chatrathi, 1992) demonstrate that over-
tems. Additional data need to be gathered for dusts with driving can occur when using strong chemical ignitors
LOC values below 10% to establish a clear recommen- in the 20-L chamber. The result is that apparent MEC
dation. values are found which are lower than the true values.
To obtain unambiguous and practically usable LOC For most dusts tested, the best agreement is found
values, the 1-m3 chamber is preferred, especially when between 20-L chamber data with 2.5-kJ ignitors and 1-
LOC values are below 10%. The 20-L LOC values may m3 data with 10-kJ ignitors. Overdriving is not a concern
be somewhat more uncertain, but the 2.5-kJ data from when testing with 10 kJ in the 1-m3 vessel. The advan-
the 20-L vessel appear to be conservative, at least for tage of the 20-L chamber is that the explosion tests can
LOC values above 10%. be conducted more quickly and with much smaller dust
In applying the LOC data, one should use a reasonable 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 219
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.
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- different for each kind of material (A=0.63; B=17.2 for
ing vessel. Two parameters can be defined upon the propane. A=1; B=68.9 for hydrogen).
curve, the maximum explosion pressure (MEP) and the For the pressure gradient Palmer (1973) establishes
maximum rate of pressure rise (MRPR). the following relation for coal:
When the initial pressure is elevated, the explosion
dP
severity rises with the initial pressure value, following AP0.5
i B (4)
a multiplicative law. dt max
MEPMEPa 10.75Pi
Pa (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-
where MEP, MEPa, Pi and Pa have the same meaning ning of the dispersion (normally fixed in 600 ms) two
as in Eq. (1). chemical ignitors, with a total energy of 10,000 J,
Also Bartknecht (1981) and Wiemann (1987) carried located at the centre of the sphere, are activated to
out experiments at high initial pressures (0.22 bar initiate the explosion, and then the explosion course is
absolute with starch for Bartknecht and 14 bar absolute recorded and processed. Fig. 1 shows a photograph of
with coal dust for Wiemmann), finding a proportinal the testing facility.
relation between the MEP and the initial pressure. Later, These values of maximum explosion pressure and
Borisov and Gelfand (Eckhoff, 1991) found a similar lin- pressure gradient were obtained by varying the concen-
ear correlation up to very high initial pressures,
approaching 100 bar. Table 1
With regard to the pressure rate, Bodurtha (1980) Bituminous coal (36% limestone) severity parameters at different
states that for gases it rises linearly as initial pressure is pressures
increased, according to the following expression:
Initial pressure (bar) MEP (bar) MRPR (bar/s)
dP
dt max
APiB (3) 1
5
9.7
43.7
155
354
10 84.0 434
where: dP/dt is the calculated value of the maximum rate 15 117.7 521
of pressure rise at Piand A and B are constants that are
E. Conde Lazaro, J. Garca Torrent / Journal of Loss Prevention in the Process Industries 13 (2000) 221228 223
tration in air of the coallimestone mixture in wide An experimental study was developed over several
increasing intervals with the initial pressure rise. Fig. 2 subbituminous Spanish coals used in a demonstration
shows the variation of MEP versus concentration for dif- PFBC power plant belonging to the electrical utility
ferent initial pressures. As it can be observed, the ENDESA, situated in Escatron (Zaragoza, Spain), that
maximum value corresponds to higher concentrations has an operating pressure of 12 bar (Torrent & Menen-
and the explosible interval includes higher concen- dez, 1993).
trations as initial pressure increases. The aim of the work was to investigate the influence
Fig. 3 illustrates the tendency of MEP and MRPR with of inertization, both adding inert matter to the combust-
initial pressure showed by the same material. It can be ible (in this case the limestone used for the sulphur reten-
noticed in the MRPR trend a slight levelling out at high tion is also an inhibitor of the explosion) and reducing
initial pressures, that may be explained by the high the oxygen content of the operating air, and to determine
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
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)
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 portional to the initial pressure. In the case of the
rate greatly rises. explosion pressure gradient, it has been observed that
Thus, it is remarkable that, in Table 8, there is a great there is a tendency to level out at high initial pressures.
difference between the Kst value at 15 bar at low turbu- Furthermore, this can be aggravated by the appearance
lence (524 bar/m/s) and at high turbulence (1369 of high turbulence situations. It has been found that
bar/m/s). The high turbulence value is 2.5 times the low increased turbulence has much more effect on the rate
turbulence one. This phenomenon is not shown accu- of pressure rise than on the explosion pressure, and this
rately, because in the high pressure turbulence test at 15 should be considered for safety measures design. An
bar, the sphere rupture disks burst out because of the increase in turbulence produces a higher mobility of par-
great severity of the explosion, releasing the high rise of ticles, the access of oxygen to the surface of the dust
pressure. The Kst is expected to reach these high values particles is promoted and then the burning rate greatly
at elevated pressures, which are normal operating press- rises. A series of experiments carried out at initial press-
ures at pressurized processes. 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.
5. Conclusions A study carried out at an initial pressure of 12 bar,
with seven coals with different levels of limestone con-
Every process involving the handling of combustible tent, percentage of oxygen in air and humidity content
dusts is exposed to explosion risks. When the process pointed to oxygen as the most influential factor on the
operating pressure rises over atmospheric conditions, the occurrence or not of an explosion. Limestone and
severity of the explosion increases following a multipli- humidity were the main influential factors on the
cative law. The maximum explosion pressure is pro- 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
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
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
opening and the under sieve is used as the test powder. of dustcup, air reflector, opening between dustcup and
It is recommended to dry the sample which contains a air reflector, compressed air supply pipes, and timing
lot of water content or which is absorptive to moisture, of ignition.
avoiding chemical change of the powder itself. The test Attempts were made to produce dust clouds as hom-
powder is recommended to have a water content of 1 ogenous as possible, because the dust cloud does not
2% or less. The dried test powder is to be contained in usually have sufficient uniformity. Many explosion tests
a dessicater. An additive can be mixed in the sample were carried out to compare the tapping sieve equipment
with poor dispersibility to improve the dust cloud hom- with the Hartmann-type explosion tube, and between
ogeneity. Fullers earth, light magnesium oxide or sili- experimenters. It was found that the value for the mini-
con dioxide are samples of the additive. mum explosive concentration is smaller and that the
The dust cloud concentration is measured by weighing reproducibility of the data is better than the case of Hart-
the dust dispersed in the explosion chamber, divided by mann-type explosion tube. The major reason for these
the chamber volume. The ignition is confirmed by the differences is due to the difference of homogeneity of
flame propagation for 10 cm above the ignition point. dust cloud between the two. Therefore it is rec-
The ignition source is provided by the power source of ommended that the Hartmann-type tube is used for
15 kV and 20 mA output. In the ignition test, it is neces- roughly investigating the degree of sensitivity, not for
sary to pay attention to avoid oxygen suffocation, obtaining the minimum explosive concentration because
because repetitive ignition tests consume oxygen. of experimental simplicity. However, it is indicated that
The strictly specified Hartmann-type explosion tube is similar results could be obtained if careful attention is
shown in Figs. 37. Other details are also described in paid to the Hartmann-type tube. For example, the mini-
the standard. Special attention is paid to the dimensions mum explosive dust concentration of standard lyco-
M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251 247
podium for calibration obtained with the tapping sieve Fig. 3. Stricly specified Hartmann-type explosion tube.
equipment is 40 g/m3, whereas that obtained with the
specified Hartmann-type explosion tube is 45 g/m3. stances. The inert materials used and their median diam-
In other words, this indicates the possible application eter are shown in Table 3. The ignition tests were carried
of the Hartmann-type explosion tube even for somewhat out with both testing equipments.
precise evaluation of the explosibility. The published Fig. 8 shows the relation between the mixing ratio of
data so far from the Hartmann tube is not consistent and inert material and the minimum explosive concentration
it may be possible to mention that the specified Hart- obtained with the tapping sieve equipment. Fig. 9 also
mann-type explosion tube will produce more reliable shows the same relation, using the strictly specified Hart-
data. mann-type explosion tube.
To calibrate the testing equipments, it is recommended It can be understood that the data fluctuation is small
to use both the lycopodium and the 50/50 mixture of the in the case of lycopodium-talc mixture. Also, the lyco-
lycopodium and talc. It is noted that the equipment has podium-talc mixture has a larger explosion limit concen-
the normal performance if the minimum explosive con- tration than other cases and the minimum explosive con-
centration of lycopodium obtained is between 3545 centration obtained with both testing equipments are
g/m3, and 8595 g/m3 in the case of the mixture of lyco- roughly the same value.
podium and talc. The standard lycopodium and talc are The 50 wt% mixing ratio of inert material was chosen
supplied by APPIE. because the minimum explosive concentration is appro-
The lycopodium was selected as the standard powder, priately high, that is ca. 100 g/m3 and the mixing ratio
because its dispersibility is excellent, and its chemical is rather simple (the weight ratio is 1:1). It should be
composition and size distribution are known. Also, also noted that even if the mixing ratio is not very accur-
another important reason is that the minimum explosive ate due to the inadequate mixing operation for example,
concentration is adequately low. the minimum explosive concentration will still stand at
It is also necessary to calibrate the testing equipments around 100 g/m3.
at higher dust cloud concentrations, because some pow- The results of the explosion test are classified into 3
ders are difficult to ignite at such a low concentration. classes: high explosibility is for 35 g/m3 or less, moder-
Since it was hard to find the appropriate sample, an ate explosibility between 35 and 100 g/m3, and low
attempt was made to mix lycopodium with inert sub- explosibility 100 g/m3 or more.
248 M. Nifuku et al. / Journal of Loss Prevention in the Process Industries 13 (2000) 243251
Fig. 5. Dustcup.
The reason for this classification is somewhat unclear, 25% of volatile content can be ignited only with a rela-
because these classes are specified basically by experi- tively large ignition source such as flames of a match and
ences. However, there are some supports for this classi- torch. And the minimum explosive dust concentration of
fication. For example, coal dust with volatile content of coal dust with 25% volatile content is around 100 g/m3.
25% shows a clear difference in the development of Also powders with smaller ignition energy, such as poly-
explosion. This implies that the coal dust with less than 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.
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. 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
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.
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
Superscript
Subscripts
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
l r
16 s 3
3 Cr
T , (7)
1
Pt1exp t ,
t (15)
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
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
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.
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)
Table 1
Size distribution of the maize starch particles
a
Average diameter used in the calculation.
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 3.3. The pressure development after ignition
distance to the wall and is very low near the wall. In
positions below 3.75 m, dust concentration decreases The explosion pressuretime profiles of their different
from 800 mm to 600 mm and increases from 400 mm points along the axis (top, center and bottom) are shown
to the wall owing to the gas flow circulation near the in Fig. 8, and the time begins from ignition. After that
bottom of the silo. time, there are some differences because blast waves
make a different fluctuation of pressure in different pos-
The initial particle size distribution is listed in Table 1. itions. After 250 ms the pressure is about 0.85 MPa, and
the maximum rate of pressure rise is 10.6 MPa m1 s1.
3.2. The gas temperature and turbulence fields after
ignition 3.4. The distribution of temperature along the axis
The initial turbulence energy (calculated from RMS
velocity), flow velocity and dust concentration fields The temperature distribution along the axis at different
after reprocessing are shown in Fig. 3. times is shown in Fig. 9. The flame front locates where
Dust ignition occurs at time zero and then the silo is the maximum gradient of temperature is found. After
closed (no more feeding), and only the cylindrical part maize starch is ignited at a position 1 m from the bottom,
of the silo is considered. The ignition is modeled by an the flame propagates towards the bottom and upwards
energy discharge with a specified distribution with space with increasing temperature. Behind the flame front,
and time at 1 m from the bottom. there is a slight decrease of temperature.
The temperature and turbulent kinetic energy fields at
9.9 ms, 40.0 ms, 80.3 ms and 120.2 ms after ignition 3.5. The flame trajectories along the axis
are shown in Figs. 47.
Comparing Fig. 4(b) and Fig. 3(a), it is evident that A flame trajectory along the axis is obtained by find-
turbulence generated by the flame front is greater than ing the maximum gas temperature gradient of a tempera-
the existing turbulence before ignition. From Fig. 5, it ture profile along the axis. The flame trajectories at dif-
can be observed that the flame front near the axis of the ferent radius positions along the axis in the 12 m3 silo
cylinder drapes because of the cool effect of a high dust are shown in Fig. 10.
concentration. Figs. 5, 6 and 7 show that the maize dust In general, the flame first accelerates at about 60 ms
flame propagates along the axis in the 12 m3 silo with after ignition and then the it propagates more slowly
a relatively slower flame propagation velocity near the before about 90 ms because the top of the silo is closed.
axis. The flame front at the axis seems to have a jump at about
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).
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 verify the model, CFD code and a series relevant
to compare the predicted result with further explosion parameters by means of laboratory scale experiments;
experiments. to obtain the initial conditions in industrial equipment
Owing to the complexity of explosions it is difficult by measurement or pre-calculation;
to use empirical measures to predict the explosion devel- to model the explosion development in the industrial
opment in industrial equipment well. However, it will scale equipment and to verify it with a large scale
be possible by using the numerial modeling method: experiment.
S. Zhong, X. Deng / Journal of Loss Prevention in the Process Industries 13 (2000) 299309 309
Acknowledgements
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
Fig. 9. The temperature profile along the axis in the 12 m3 in different 26, 1996.
10 (about 10 40 80, 120 ms). 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
Fig. 10. The flame trajectory at different radius positions along the (pp. 843849).
axis in the 12 m3 silo. (PosiR0 0 m from the axis; PosiR1 0.28 Smirnov, N.N., Nikitin, V.F. (1996). Dustair mixtures evolution and
m from the axis; PosiR2 0.56 m from the axis; PosiRs wall). 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-
In this simulation it is also found that turbulence gen- way, June 2326, 1996.
erated by the flame front is much larger than the initial Smoot, L. D., Horton, M. D., & Williams, G. A. (1977). Propagation
turbulence. The mechanism of the relationship between of laminar pulverized coal-air flames. Proceedings of the 16th sym-
turbulent RMS velocity, dust dispersion effect and Kmax posium (international) on combustion. The Combustion Institute
(pp. 375387).
should be investigated in detail. Zhou, L. (1994). The theory research and numerical modeling of tur-
In the multi-phase turbulent combustion model, dis- bulent gas-particle flows and combustion. Beijing: Science Press
persion efficiency must be considered. The proposed of China.
Journal of Loss Prevention in the Process Industries 12 (1999) 227234
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.
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.
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
Fig. 5. Maximum reduced explosion pressure against Av/V2/3. Fig. 7. Maximum reduced explosion pressure against Av/V2/3.
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.
3.2.1. Notes for guidance Figs. 416 show relationships between the maximum
1. The larger of the vessels should not exceed 20 m3. reduced explosion pressure and the dimensionless vent
2. The length of the interconnecting pipe must not area for the conditions listed on each graph. The figures
exceed 15 m, and must not be less than 5 m. can be used in two ways: to estimate the maximum
3. The pipe area to volume of the smaller vessel in the reduced explosion pressure in an interconnected system
system must not exceed 0.1. when the vent area is known; and to estimate the neces-
4. The Pmax of the dust should not exceed 10 bar a. sary vent area to limit the maximum reduced explosion
5. The vent area is always divided between the vessels pressure to a given value.
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 References
of the connecting pipeline is ignored.
3.2.2. Estimating vent areas for interconnected vented Holbrow, P., Andrews, S., & Lunn, G.A. (1996). Dust explosions in
interconnected vented vessels. J. Loss. Prev. Proc. Ind., 9(1), 91.
vessels Lunn, G.A., Holbrow, P., Andrews, S., & Gummer, J. (1996). Dust
This guidance is applicable to compact vessels (L/D explosions in totally enclosed interconnected vessel systems. J.
2), Pstat 0.1 bar g and for vessel volumes 20 m3. 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
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
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
G.A. Lunn
Journal of
ELECTROSTATICS
ELSEVIER Journal of Electrostatics 40&41 (1997) 511-516
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.
2. MEASURING TECHNIQUES
? 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
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
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.
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
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]
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
Abstract
Kew'ords: Ignition conditions; Powder ignition; Minimum ignition energy: Dust explosion:
Electrostatic particle suspension
1. Introduction
* 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
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
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
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.
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
~ 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
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
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].
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].
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
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).
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.
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.
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].
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
= 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.
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].
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.
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
(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.
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].
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).
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
:
Fig. 12. Nondimensional spark ignition energy versus ND 2 parameter for propane-air mixtures;
q~ = fuel/air ratio (mole/mole).
where
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
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- workplaces; primary dust explosions normally occur in
mining the minimum explosible concentration (MEC), process equipment. The simple representation in Figure 1
also termed the lower or lean flammability limit, of also illustrates the potential hazard of secondary dust
combustible dusts. The MEC is the concentration explosions occurring by layered dust being raised into
boundary above which a dust-oxidant mixture will suspension and ignited from a primary explosion.
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. S T A N D A R D I Z A T I O N OF MEC
The MEC has been measured in different types (sizes MEASUREMENTS
and shapes) of laboratory chambers for many years. The Measurement of the MEC of a heterogeneous fuel-
most common laboratory chamber used has a volume of oxidant mixture requires consideration of factors that
20L. Among the test methodologies available for may not prevail in similar measurements for homo-
determining the MEC are those proposed by the geneous mixtures. For example, with gaseous mixtures
American Society for Testing and Materials (ASTM) the fuel and oxidant are separated by only molecular
and the International Electrotechnical Commission distances; further, the attainment of a quiescent state
(IEC). In this work, tests were conducted to allow a prior to ignition is usually not difficult experimentally.
comparison between the two methods. Solid fuel-gaseous oxidant mixtures present different
The minimum explosible concentration and the upper challenges. The heterogeneous nature of the combustion
or rich flammability limit form the boundaries of the process poses the particular problem of ensuring that a
flammability envelope of a combustible dust. Figure 1 flammability limit, not an ignitability limit, is being
shows the flammability envelope (explosive range) at measured.
ambient temperature and atmospheric pressure for maize Assuming that the conditions of suitable test apparatus
starch, which is representative of a typical natural geometry and size are met, the prerequisites for accurate
organic dust. The MEC and upper limit are shown in determination of limit data for combustible dusts are:
relation to typical dust concentrations permitted in the
context of industrial hygiene and to a typical density of (1) careful control of the particle size distribution of the
deposits of natural organic dusts. As noted by Eckhoff 1, dust;
MECs of dusts correspond to concentrations of (2) physical and chemical homogeneity of the dust
airborne particles that are unlikely to occur regularly in particles;
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.
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
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.
REFERENCES
1 Eckhoff, R. K. 'Dust Explosions in the Process Industries', But-
4 terworth-Heinemann, Oxford, 1991
0 2 American Society for Testing and Materials. ASTM Standard E
1 2 3 4 5 6 7 8 9 tO 1515-93, 'Standard Test Method for Minimum Explosible Con-
centration of Combustible Dusts', Annual Book of ASTM Stan-
Ignition Energy [kJ]
dards, Vol. 14.02, pp. 774-781
3 Cashdollar, K. L. and Chatrathi, K. Combust. Sci. TechnoL
Figure6 Effect of volatility on minimum explosible concentration 1992, 87, 157
(Pocahontas coal and oil shale; MEC from ASTM criterion) 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.
The MECs of oil shale at various ignition energies are Hertzberg), STP 958, American Society for Testing and Materi-
higher than for Pocahontas coal even though the volatile als, Philadelphia, 1987, pp. 107-123
matter of oil shale is higher than that of Pocahontas coal 6 Greninger, N. B., Cashdollar, K. L., Weiss, E. S. and
Sapko, M. J. In Proceedings of Fourth International Collo-
(see Figure 6). This behaviour of the oil shale is due to its quium on Dust Explosions, Warsaw, 1991, pp. 208-228
high ash content (the ash content of Pocahontas coal is 7 Amyotte, P. R., Mintz, K. J. and Pegg, M. J. In Proceedings of
5 wt% and that of the oil shale is 70 wt%). Ash, being an 27th Annual Loss Prevention Symposium, AIChE, Houston,
incombustible material, arrests flame propagation by 1993, Paper No. llc
absorption of thermal energy from the combustion 8 Amyotte, P. R., Mintz, K. J., Pegg, M. J. and Sun, Y. H. Fuel
1993, 72, 671
reaction. This can be understood from the work of 9 Conti, R. S., Zlochower, I. A. and Sapko, M. J. Combust. Sci.
Amyotte et al. 12 on inerting requirements for coal dust. TechnoL 1991, 75, 195
They demonstrated that the percentage of dolomite l0 Hertzberg, M., Cashdollar, K. L. and Lazzara, C. P. In 'Twenty-
required to inert a coal dust explosion decreased in direct first Symposium (International) on Combustion', The Combus-
tion Institute, Pittsburgh, 1981, pp. 717-729
proportion to decreasing volatile matter, from 80% for a I1 Hertzberg, M. and Zlochower, A. I. In 'Twenty-third Sympo-
sample of Prince coal (36wt% volatiles) to 74% for sium (International) on Combustion', The Combustion Insti-
a sample of Lingan coal (32 wt% volatiles) and 64% for a tute, Pittsburgh, 1990, pp. 1247-1255
sample of Phalen coal (26 wt% volatiles), all three coals 12 Amyotte, P. R., Mintz, K. J., Pegg, M. J., Sun, Y. H. and
having similar particle size distributions. However, when Wilkie, K. I. J. Haz. Mater. 1991, 27, 187
In the introduction to his paper in this issue of the jour- tive tool for assessing the pressure development of dust
nal, Eckhoff comments that: The total amount ofexisting explosions and suggests a modification of the cubic
knowledge on industrial dust explosions, their origin, relationship based on a function of flame thickness and
propagation, prevention and mitigation, is vast. And yet, vessel radius; there are thus both fundamental and practi-
further information is continually being generated cal implications of their study.
through on-going research in a large number of coun- The work of Lunn et al., as described in the first
tries. The aim of this special issue is to gather in one of two papers from the UK Health and Safety Labora-
reference some of the results of this on-going research, tory, deals with the particular problem of dust explosions
consistent with the theme of Huzard Evaluation, Preven- in systems of linked, enclosed vessels. Hazard evaluation
tion and Mitigation of Dust Explosions. It would, of in this case is aimed at gaining a better understanding
course, be impossible to Capture in one instant al1 of the of the phenomenon known as pressure-piling; such
excellent research on dust explosions being conducted knowledge is critical in applying the mitigation tech-
worldwide. The very definition of research prohibits this nique of containment, whereby explosion products and
freezing of al1 information. Nevertheless, it is hoped overpressure are contained entirely within the process
that the current document wil1 provide the reader with unit. The authors begin with a review of recent research
a sample of recent work on dust explosions and wil1 per- in this area and then describe the results and analysis of
haps indicate future directions for concerted effort. their experimentation with vessels having volumes of 2,
The first paper, by Eckhoff, gives a thorough review 4 and 20 m.
of the authors own work and that of others worldwide, While most of the papers in this issue deal with
primarily dealing with the period from 1990 onwards. dust/air suspensions, the work of Nakajima and Tanaka
Al1 elements of the special issues theme are addressed, offers an interesting look at the potential hazards of dust
with particular emphasis on existing and potential link- layers. In this fundamental study, the critical state for
ages between fundamental research and industrial prac- spontaneous ignition is classified as being of either the
tice. In many respects, Eckhoff sets the tone for the Frank-Kamenetskii type or the oxygen-deficient type.
remainder of the papers and offers a foundation for The authors present guidance for type discrimination
future work in the field. based largely on the bed material and porosity. Bearing
Siwek continues with the theme of hazard evalu- in mind that the primary use of hazard evaluation data
ation by drawing on his extensive work at Ciba-Geigy. is the design of preventive or protective measures, Naka-
He gives practica1 information on laboratory-scale tests jima and Tanaka illustrate the practica1 significante of
commonly used in Europe and elsewhere to determine their work through the example of aeration cooling -
the flammabilitylexplosibility parameters of dust layers an effective technique for Frank-Kamenetskii-type lay-
and dust suspensions. An important feature of his paper ers, but one that can actually increase the hazard of oxy-
is the reference to testing standards developed by organi- gen-deficient-type layers.
zations such as the International Electrotechnical Com- Cashdollars paper on coal dust explosibility pro-
mission (IEC) and the International Standardization vides a thorough review of research conducted at the US
Organization (ISO). Bureau of Mines. Although the work described is spec-
Centra1 to the matter of hazard evaluation (and sub- ific to a particular dust, the lessons learned are applicable
sequently the mitigation technique of venting) is the to many other materials with respect to how hazard
cubic relationship by which the K,, parameter is evaluation is affected by changes in dust composition,
obtained. The fact that K,, is not an invariant property dust particle size and the co-presence of a flammable
of a material is amply demonstrated in the work of gas. Of particular note in this paper is the validation of
Dahoe et al. They focus on the role of flame thickness laboratory-scale data through comparison with full-scale
in pressure development during a dust/air mixture test results. The special issue theme of explosion preven-
explosion and present a three-zone model which tion is addressed in Cashdollars work by his examin-
accounts for this influence. Their work presents a predic- ation of the requirement in coal mines for admixture of
inert rock dust with combustible coal dust.
1
2 Editorial
Explosion prevention can be achieved by use of an volumes of 2, 6.3 and 20 m3. Much practica1 guidance
inert gas (e.g. nitrogen) to ensure that process operation is given on the variables affecting vent area requirements
occurs at oxygen concentrations below the maximum for interconnected vessels (e.g. diameter and length of
permissible level. This approach is aimed at removing interconnecting pipe); particularly significant is the
the oxidant criterion for explosion. Alternatively, authors attention to the relationship between these vari-
removal of the heat necessary for combustion can be ables and vent areas determined by Ks, nomograph
achieved by use of an inert dust as described in Cash- guidelines.
dollars paper. This concept, however, has not been Further advances in venting technology are likely
applied extensively in industries other than coal mining. to arise through a combination of several factors: on-
The paper by Mintz et al. describes one such application going programmes of industrial-scale testing, develop-
where the inerting dust is already present in the final ment and use of first-principle mathematica1 models,
product; product degradation is not, therefore, a prohibi- and continued efforts at computational fluid dynamics
tive factor in the use of explosion-inhibiting additives. (CFD) modelling. The paper by Tamanini and Valiulis
The authors provide data which show how the safe prep- describes how the first two of these approaches have
aration of powdered metals (Al/Mg) for refractory been incorporated in a comprehensive and long-standing
materials can be enhanced by early introduction into the programme at Factory Mutual Research Corporation.
manufacturing process of inert Mg0 powder. The authors draw on their own experimental results and
The theme of explosion mitigation (or protection) the test data of other researchers, coupled with the use
is the subject of the remaining four papers in the special of theoretical models, to present a new approach to the
issue. The first three deal with pressure relief venting sizing of explosion relief vents. It is significant to note
while the final paper covers automatie suppression of that in their conclusions they cal1 for contributions to the
explosions. The common feature of these papers is that advancement of venting technology from organizations
they al1 address the situation where, despite having worldwide, with consensus on the conceptual framework
evaluated the hazard and taken appropriate preventive for such work. This fits wel1 with Eckhoffs encourage-
measures, an explosion has occurred and the primary ment (this issue) of current efforts to establish inter-
concern is to mitigate the damage to personnel and plant. national cooperation in joint research programmes.
Siweks paper on explosion venting technology Automatie explosion suppression is the subject of
contains a wealth of practica1 information obtained from the final paper, by Moore. The author draws on his
industrial-scale testing. Such testing, and the empirical extensive industrial experience to provide a practica1
correlations derived from the data, form the backbone of overview of suppressant choices and the basis for mak-
the venting guidelines issued by the Verein Deutscher ing selections. He emphasizes the importante of choos-
Ingenieure (VDI) and the Institute of Chemical Engi- ing the best agent for suppression, best often being a
neers (IChemE) in Europe, and the National Fire Protec- compromise between effectiveness and practica1 accept-
tion Association (NFPA) in North America. It is interes- ante. The issue of acceptance is particularly important
ting to note the emergence of regression equations in the with the phase-out of chlorine- and bromine-containing
new VDI guideline, with a decreased emphasis on the compounds.
use of nomographs. As a final note, 1 wish to express my gratitude to
The absente of adequate venting theories, and the both the manuscript authors and the reviewers for the
quality of their work and the timeliness of their efforts.
subsequent reliance on experimentation, is illustrated
1 would also like to acknowledge with appreciation the
further in the paper by Holbrow et al. They describe an
dedication to this special issue of the editorial and pro-
experimental study which extends the work of Lunn et
duction staff at Elsevier Science Ltd and the office staff
aE. (this issue) on interconnected enclosed vessel systems
of the Department of Chemical Engineering at the Tech-
to interconnected vented vessels. As in their previous
nical University of Nova Scotia.
paper, the authors first give a thorough review of recent
research in the field and then describe the results and Paul R. Amyotte
analysis of their experimentation with vessels having 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
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.
The total amount of existing knowledge on industrial aspects. In the second part of the paper, knowledge and
dust explosions, their origin, propagation, prevention and technology related to preventing and mitigating dust
mitigation, is vast. And yet, further information is con- explosions in industrial practice wil1 be addressed
tinually being generated through on-going research in a specifically. A separate third part is devoted to the
large number of countries. A few years ago, 1 had the important area of test methods for assessing the ignit-
opportunity to review some of the knowledge accumu- ability and explosibility of dusts. One main aim of the
lated up to about 1990. About 900 references were paper is to indicate areas where a more intimate inte-
covered, but they almost exclusively originate from gration of fundamental research and practica1 develop-
English/American and German literature. There is little ment in industry is timely.
doubt that a lot more interesting and useful information
can be retrieved by screening the large amount of Status and outstanding problems in
research reported in languages other than English and fundamental research related to dwt
German. Perhaps this could be accomplished by organiz- explosions
ing a joint international translation/edition effort.
The present paper mainly summarizes works pub- Dust cloud formation processes. What is turbulente
lished from 1990 and onwards, and only more recent of a dwt cloud?
works than those quoted in the above mentioned text- This is an important and sometimes overlooked aspect
book are cited explicitly. Earlier versions of the present of dust explosions. It is wel1 established experimentally
survey were presented in Poland in April 1993 (and sub- that the initial state of a cloud of a given dust in a given
sequently published) and in the Peoples Republic of gas (dust concentration, degree of dispersion into indi-
Chinaj. vidual particles, dynamic state) has a strong influence on
There is a growing acceptance of the fact that both the ease with which the dust cloud ignites and the
adequate dust explosion prevention and mitigation in rate at which it burns.
practice must be based on a proper understanding of the However, comparatively little fundamental quanti-
basic physical and chemical phenomena involved. tative knowledge exists about the processes by which
Therefore, when reviewing existing knowledge on the dust clouds may be generated. Some work has been done
prevention and mitigation of industrial dust explosions, on entrainment of dust particles by turbulent gas flows
it is appropriate to start by addressing the fundamental passing over dust layers and ridges in wind tunnels.
3
4 Prevention and mitigation of dust explosions: R.K. Eckhoff
However, quantitative relationships between essential theoretical studies of the influence of blast waves on dust
parameters of the dust cloud generation processes and clouds and dust layers/deposits need to be conducted.
the states of the resulting dust clouds are scarce, and Work on this problem has also been performed by UraP,
more work is needed in this area. When designing exper- Gelfand and Tsyganov7, and others. Gelfand and Tsy-
iments for fundamental studies, one should select con- ganov showed that the presence of dust layers on solid
figurations that can be related to industrial practice. surfaces exposed to blast waves changed the blast wave
Examples are bulk dust dropped into an air flow, and characteristics as compared with the characteristics in
entrainment of dust layers by air flows in channels and the case of dust-free surfaces. Kauffman et aLx and
galleries with and without flow-obstructing obstacles. Austin et aL9 have summarized their extensive research
Hauert et ~1.~ measured RMS turbulent velocities, on entrainment of dust layers in long tubes by the blast
global velocities and dust concentrations in experimental wave heading a dust explosion propagating along the
dust clouds in the standard 1 m3 ISO vessel, and in a tube. Tamanini and Ural have summarized their work
12 m3 silo. In the latter case, the dust was either blown on characterizing the dispersibility of dust layers when
into the silo via a pneumatic transport pipe, or dropped exposed to air blasts.
into it from a screw conveyor at the silo top. This kind Increased emphasis should be put on investigating the
of work is essential, e.g. for testing (validating) the connection between the parameters of dust cloud gener-
performance of numerical codes for simulating dust ation processes and the structures of the resulting dust
cloud formation and development in real industrial situ- clouds. The structures of the clouds produced must be
ations. Such codes, in turn, wil1 be an essential element defined in terms of distribution of dust concentration,
in future comprehensive codes for simulating dust quality of dust dispersion (de-agglomeration), turbulente
explosion propagation in industrial situations. leve1 and global velocities. For example, fundamental
Hauert et ~1.~applied laser Doppler anemometry for theoretical studies are being performed on the generation
measuring the turbulente in their experimental clouds of dust clouds behind shock waves sweeping across a
of maize starch. As tracer particles, they simply used the dust layer at 90 to the layer surface. Frolov et al.
starch grains of the experimental cloud itself. It has been developed a mathematica1 diffusion model describing
argued against this procedure that such comparatively such a process. A similar model was developed by Lu
large particles are unable to follow the rapid turbulent et al. *. This kind of work is relevant in the context of
movement of the gas phase and, therefore, the measure- propagation of secondary dust explosions. Nikolova
ments performed do not represent the real turbulente of developed a source code for numerical models for simul-
the cloud. ation of dust cloud flow. The code can be used for cold
However, this argument raises a basic question: flows as wel1 as flows with combustion. Medvedev et
What is the turbulente of a dust cloud in the context al. l4 conducted experimental studies of the entrainment
of flame propagation and dust explosion? Could it be of dust from dust layers by the short-duration flow gen-
that the most important turbulente mechanism is in erated by the rarefaction wave in a shock tube. Dushin
some cases the movement of burning particles in relation et al. l5 developed a mathematica1 model of the evolution
to unburnt ones, rather than the movement of the gas of a cloud of entrained dust in the atmosphere, following
phase? Could it be, therefore, that the movement of the a huge explosion on the ground. Nikitin et al. l6 perfor-
particles is in fact in some cases a more appropriate mea- med a theoretical numerical study of the evolution of a
sure of turbulente than the movement of the gas phase? dust cloud in a turbulent atmosphere. The cloud could
The answer depends on the nature of the particles. Very be generated instantaneously as a result of a sudden
rapid devolatilization, with the combustion occurring explosion blast, or continuously from multiple moving
mainly in the bulk of the gas phase, is an extreme case sources. Geng et al. l7 studied the influence of clouds of
where the gas movement is indeed important. If, on the maize starch in oxygen on the peak pressures of incident
other hand, the combustion mainly occurs close to the shock waves of Mach numbers 5.4 and 6.0.
particle surface, the particle movement is of primary In order to guide fundamental research in this area
importante. in the direction of maximum practica1 relevante, infor-
The state of a dust cloud that influences the ignition mation about dust cloud structures that are typical in
sensitivity of the cloud wil1 clearly be the state at the industrial operation is required. This not only means the
moment of ignition (initial state). When it comes to cloud structures in normal plant operation, but perhaps
flame propagation, however, the picture is complicated even more the structures existing during abnormal transi-
further by the strong coupling between the flame propa- ent phases, including plant start-up and close-down, dur-
gation process and the state of the unburnt cloud ahead ing which the occurrence of dust explosions may some-
of the flame. Therefore, the initial state of the dust cloud times seem more likely than under normal steady-state
at the moment of ignition may only be of secondary sig- conditions.
nificance for the flame propagation once the explosion
has developed beyond the initial stages. Dust layer and dust cloud ignition processes
As soon as a significant blast wave has been gener- Ignition is a broad field of research. The concept of ther-
ated by the primary dust flame, this blast may generate mul runaway is a common basis for understanding and
secondary explosible clouds ahead of the flame from describing ignition processes. Theories have been
dust deposits and layers there. Lebecki et ~1.~ investi- developed even for complex cases with reactant con-
gated such processes in a 100 m long gallery of cross- sumption during the process leading to ignition. How-
section 3 m2. In order to establish an improved under- ever, it does not seem realistic for the time being to fore-
standing of these processes, further experimental and see the development of one single unified theory, usable
Prevention and mitigation of dust explosions: R.K. Eckhoff 5
in practice, which covers al1 types of ignition sources. light pulse (100 hs or 10 ns) from an Nd:YAG laser
It is instead expected that separate theories, in terms of (1064 nm wavelength). This kind of experiment can pro-
dynamic computer models, wil1 be developed for various vide basic information about dust cloud ignition pro-
categories of ignition sources, such as hot surfaces and cesses, and also about flame propagation processes in
electric sparks. Work conducted on ignition of pre-mixed dust clouds. Van der Wel et aLX described how a sim-
gases should be studied to see whether elements of gas ple, modified Semenov theory (no temperature gradients
ignition theory can be applied to ignition of dust clouds. inside the heated volume) for auto-ignition was used for
However, development of such theories needs to be transforming the experimental laser-light-pulse ignition
backed by careful experimentation. Many variables must results to predicted minimum ignition temperatures and
be considered, which are partly related to the nature of energies. The predicted values were in approximate
the potential ignition source itself and partly to the com- agreement with measured minimum temperatures and
bustible dust layer or cloud that is exposed to the source. energies for ignition. Gieras and Klemens studied the
Definition of good parameters and design of correspond- critical conditions for ignition of single coal particles in
ing experimental test methods remain a centra1 chal- air, and in air containing methane. They also investigated
lenge. the critical ignition conditions for clouds of coal dust in
Hensel and John8,9 provided further insight into air and in methane/air.
the important relationship between the conditions In the past, the minimum hot-surface temperature
required for initiating smouldering combustion of a dust for ignition qf a dust cloud has often been regarded as
layer on a hut plate, and the layer thickness. Hensel et if it were a universal constant for a given cloud. Conse-
aZ.* presented a mathematica1 model for numerical quently, results from small-scale laboratory tests were
simulation of temperature development in powder often applied directly in industrial plant design. How-
deposits enclosing a heat source of constant power. The ever, it has been known for some time that minimum
model was tested against experiments in which a metal ignition temperatures of dust clouds vary significantly
sphere, heated by a constant power source, was embed- with scale, and this has recently been confirmed by Wol-
ded in the powder sample. Further promising work along anskP2. Further experimental and theoretical work is
this line, focusing on the critical heat flux for ignition needed in this area.
rather than on the critical surface temperature, was Ignition of dust clouds by smal1 burning metal par-
reported by Krause and Hensel*. It was pointed out that ticles (impact sparks, metal sparks) generated by mech-
a critical heat flux relates more to real practice than a anical impact is a very complex problem. A comprehen-
critical constant surface temperature. sive, practically useful theory does not seem to be within
Kauffman et al. summarized their research on sight. Such a theory must comprise several steps. The
smouldering combustion of dust heaps and dust layers. first is the generation and initial heating of the metal
Chernenko et al.* investigated flame propagation along particle by the impact, which is in itself a very complex
the surface of layers of metal powders, and mixtures of problem. The second is the ignition of the flying hot
metal powders and metal oxides. The influence of the particle and the subsequent burning process. The third
chemical composition of the powder layer, and to some is the heat transfer to the dust cloud, which ultimately
extent also of the particle size, on the burning velocity determines whether ignition occurs or not.
along the surface of the layer was determined experi- In dust explosion statistics, the frequency data for
mentally. Matyukhina and Babushok developed a occurrence of various categories of ignition sources are
mathematica1 model for self-heating in coal deposits, sometimes confusing, because the frequently used term
taking into account the effect of diffusion of air into the friction sparks is ambiguous. Sometimes hot surfaces,
deposit. Sobolev* developed a discrete, two-temperature generated either by repeated impacts on the same spot
mathematica1 model for heat conduction in a dust or by sliding friction, are included in the friction spark
deposit. An informative overview of research on self- or mechanica1 spark category. Furthermore, one does
heating/self-ignition in dust deposits was given by not always distinguish between burning metal particles
Crowhurst25. Zhang and Deng2h studied the combustion from one single accidental impact and from repeated
rate of coal dust layers on constant-temperature hot sur- impacts in rotationary machinery such as grinding and
faces. Influences of hot surface temperature, oxygen con- cutting equipment. The categorization of ignition by
tent in the atmosphere and flow rate of air across the burning metal particles from grinding and cutting oper-
layer on the oxidation rate were investigated. Itagaki and ations is not always clear either. This is because ignition
Matsuda used both differential scanning calorimetry may have been caused by the hot surface being produced
and accelerating rate calorimetry for measuring the exo- at the object being ground, rather than by the burning
thermic reactivity of activated carbon dusts. Adsorbed metal particles generated in the grinding process.
nitrogen oxides or fluorine on the coal surface increased Ignition of dust clouds by electric/electrostatic dis-
the exothermic reactivity markedly, with the onset tem- charges is another very complex topic. Theories have
perature of exothermic reactions being as low as 30- been developed for ignition of dust clouds by electric
40C. sparks between two metal electrodes, which is the sim-
Glinka et al.* conducted an experimental and plest case, but even such theories are only rough
theoretical study of ignition of dust layers by thermal approximations. The variables of the electric spark
radiation. The ignition process was resolved in detail by ignition problem include voltage and current character-
means of high-speed Mach-Zehnder interferometry. istics across the spark gap, spark gap geometry and elec-
Van der Wel29 conducted a series of experiments trode material, and dust cloud variables such as particle
in which laminar dust clouds were ignited by a short material and particle size/shape distributions, dust moist-
6 Prevention and mitigation of dust explosions: R.K. Eckhoff
ure content, dust concentration, and the dynamic state of Flume propugution processes in dust clouds
the dust cloud with respect to the spark gap. Some centra1 topics are the same as for flame propa-
Dahn et ~1.~~reviewed some published work on gation in premixed gases, viz.:
electric spark ignition of combustible dust clouds. Their
own experiments with clouds of lycopodium confirmed 0 laminar flames
the dramatic influence of the combination of capacitance ?? flame acceleration mechanisms
Krishenik and Shkadinskiiso developed a mathe- carbon/active coke) of considerably lower volatile con-
matical model for flame propagation in dust clouds of tent than 7-8% could give fully developed dust
mixtures of two monosized particle fractions. Both con-. explosions in the 1 m3 ISO vessel.
ductive and radiative heat transfer were incorporated in Turbulent combustion in dust clouds has been stud-
the model. Deng et ~1.~proposed that the burning velo- ied experimentally and theoretically by several investi-
city of a dust Zayer under specific laminar conditions gators. Eckhoff6 summarized some work on the influ-
be regarded as the fundamental combustion property of ence of initial and explosion-induced turbulente on dust
a dust. They did not, however, elaborate the idea to the explosions in closed and vented vessels. Wel et a1.6S62
extent needed to resolve the implications of the proposal re-emphasized the important role played by turbulente
with respect to dust cloud combustion. in dust explosion propagation in closed vessels. Kauff-
Lee et aLS2 showed that theoretical equilibrium man et uZ.* and Austin et aL9 summarized their research
properties of dust cloud combustion (constant-pressure on turbulent combustion of premixed dust clouds, and
adiabatic flame temperatures, and maximum constant- Tamanini and Ural summarized their work on the
volume explosion pressures) calculated by standard effect of initial turbulente on flame propagation in dust
computer codes are in good agreement with experimen- clouds. Scheuermann63 also investigated the influence of
tal data obtained by various workers. the initial dust cloud turbulente on the development of
Wlodarczyk et ~1.~conducted experiments in a 5 dust explosions in vented enclosures. Rzal-Rebire and
litre spherical explosion bomb to determine the influence Veyssire 64 addressed some centra1 basic aspects of tur-
of the dust concentration in exploding aluminium/air bulent dust flames. Veyssire65 summarized the funda-
clouds on the content of aluminium oxide in the reaction mental studies on flame propagation in dust clouds con-
products. Fan et cds4 developed a dynamic numerical ducted at LED in Poitiers, comprising laminar flame
simulation model for the propagation of spherical Al propagation in dust clouds, the role of turbulente in
dust explosions in closed vessels. Good agreement was flame acceleration, and the conditions for propagation of
found between experiments and theoretical predictions detonation-like, but non-ideal, combustion waves. In
for the influence of particle size on the pressure develop- their basic studies Rzal-Rebire and Veyssire% investi-
ment in the vessel. gated the interaction of a laminar maize starch/air flame
Limiting cloud compositionsfor jame propagation with an obstacle, viz. a sphere, a disk or a vortex ring.
is an important fundamental research topic for at least
With the ring, flame-quenching phenomena were
three different practica1 applications, viz. explosible/non-
observed, which were attributed to centrifugal separation
explosible assessment, assessment of minimum explos-
of dust particles and air in the turbulent eddies. This is
ible dust concentration, and assessment of maximum
a very important observation, indicating that the burning
permissible oxygen concentration for inerting. Mintz5j
rate of a dust cloud may not respond to turbulente in
found evidente for the existente of a maximum explos-
the same way as the burning rate of a premixed gas.
ible dust concentration for dust clouds under certain cir-
Further work towards improved understanding of
cumstances. For a narrow size fraction (106-125 Pm) of
the relationship between the dynamic state of a dust
maize starch, a reasonably wel1 defined limit of 800-
cloud and its combustion rate should be encouraged.
1000 g me3 was found. The results were interpreted in
Basic parameters need to be identified. Various strategies
terms of a simple oxygen depletion model. The influ-
ence of the particle size distribution on the minimum have been indicated. For example, an induction time for
explosible dust concentration was investigated by Pole- ignition may be taken as a global characteristic of the
taev and Korolchenko56, using data from experiments combustion chemistry (shock tube or stirred reactor).
with polysized polyethylene dusts. Promising agreement Geng et aL6measured the ignition delay behind an inci-
between theory and experiments was obtained. dent shock wave of Mach number 4-6 passing through
Hertzberg et al.* measured minimum explosible a cloud of maize starch in oxygen. UraP emphasized
concentrations, maximum explosion pressures and the fact that different induction times are observed with
maximum rates of pressure rise at constant volume, and incident and reflected shock waves, due to different
maximum flame temperatures, for clouds in air of dusts ignition mechanisms. Boiko et al.69 measured ignition
of 14 different metals. They found that for some metals delays for coal dust clouds exposed to reflected and inci-
flame propagation appears to occur in a mixture of metal dent shock waves.
vapour and air, similarly to the gas phase flame propa- An alternative approach is to consider the laminar
gation mechanism in clouds of organic dusts. burning velocity as the fundamental parameter, as sug-
Hertzberg et ~1.determined the same parameters gested by Bradley et al.. Empirical relationships
for nine different dusts of solid explosives when dis- between turbulent burning velocity and turbulente inten-
persed as clouds in air in a closed bomb. In the low- sity are then established, using the laminar burning velo-
concentration range (6 400 g m-), the dusts behaved as city as a normalizing parameter. As pointed out by van
dusts of normal carbonaceous and plastic materials. At Wingerden49, numerical flame libraries can then be for-
higher concentrations, they became more hazardous. mulated and used for closing the positive-feedback loop
It is wel1 known that pulverized coal and coal dust combustion-expansion-jow-turbulente-combustion in
in mines do not represent a dust explosion hazard unless numerical dust explosion simulation codes. The on-
the content of volatiles exceeds 7-8%. However, this going research and debate on numerical modelling of
does not apply to carbon dusts of specific surface areas premixed gas combustion should be watched carefully
exceeding the order of 100 m* g- (N2 adsorption). Wie- to ensure that any elements that may contribute to solv-
mann59 showed that dusts of such materials (active ing the dust explosion modelling problem are explored.
8 Prevention and mitigation of dust explosions: R.K. Eckhoff
Understanding flame acceleration, due to flame dis- excellent review of the state-of-the-art and outstanding
tortion and turbulente produced by the propagating problems in dust cloud detonation research has been
explosion itself, is centra1 for understanding both dust given by Kauffman et al. *. The current knowledge on
and gas explosions. Extensive experimental research dust cloud detonations was also summarized by Alex-
programmes have been conducted to study these ander et al.79. Kauffman et uZ.* and Austin et ~1.~ sum-
phenomena for gas explosions in obstructed geometries. marized their extensive work on how detonations can
If the experimental facilities used in these experiments develop from accelerating turbulent combustion of dust
are stil1 available, it would seem relatively straightfor- clouds in long tubes. Sichel and Kauffmannso studied
ward to repeat the experiments in the various vessels, the transition from deflagration to detonation during dust
using dust clouds instead of premixed gas. Systematic explosions in long ducts. The dust was initially deposited
comparison of results with previous data from gas as a layer along the duet floor, and the dust cloud was
explosions would yield a valuable overview of simi- generated by the entrainment of the dust layer by the
larities and discrepancies, which would help to focus blast wave propagating ahead of the flame. Khomik et
basic research efforts on important areas where dust uE.~~,~* determined experimentally the minimum critical
cloud combustion may differ significantly from combus- tube diameter for detonation propagation in suspensions
tion of premixed gases. Proust7 has given an informative of a fine aluminium flake dust in air. The critical value
review of the state-of-the-art on propagation of dust found was in the range 0.040-0.055 m. Korobeinikov83
explosions in pipelines in relation to gas explosion conducted a theoretical study of the propagation of det-
propagation in pipelines. Pu et aL7* studied experimen- onation waves in dust clouds. The problem of estab-
tally the acceleration of an Al dust/air flame in a one- lishing adequate staling rules was given particular atten-
end-open horizontal tube of diameter 140 mm and length tion. Markovg4 presented a new method for numerical
5 m. With ignition at the closed end, maximum flame simulation of non-steady detonations in dust clouds.
speeds at the open exit end amounted to 1200 m s-. Hu Two-dimensional computations yield a multi-wave
and Sun73 studied the mechanisms of fast combustion of structure of the detonation process. Ding and Huang8
Al powder suspended in atmospheres of different oxygen analysed the mathematica1 theory for the reaction zone
contents, using an explosion shock tube technique. in a detonation wave passing through a dust cloud, and
Much work has been, and is currently being done, proposed a new numerical criterion describing the
on turbulent combustion of sprays and mists74, which is Chapman-Jouguet condition. Tulis et ~1.~~ conducted
in part also relevant in the context of dust explosions. detailed experimental studies of the structure of deton-
Huang et UZ.75 observed that burning clouds of alu- ation processes in clouds of aluminium in air. The influ-
minium dust in air are electrically conductive. They attri- ence of particle size and shape was studied, and various
buted this effect to generation of metal vapour by evap- detonation wave structures were identified. Paplinski and
oration of the particles prior to combustion. Besides WlodarczykE7 analysed the critical conditions for direct
being of fundamental interest, the observed effect also initiation of detonations in dust clouds of infinite size.
has implications with respect to industrial safety. Klemens et al.87 performed experiments in which deton-
The hope is that adequate mathematica1 codes for ation waves in hybrid mixtures of methane, air and oats
computer simulation of dust JEame propagation pro- dust were studied.
cesses wil1 be available in a not too distant future. Kjld-
man76 has summarized his work on applying compu- Blast waves generated by burning dust clouds
tational fluid dynamics to turbulent dust explosion One case of practica1 interest is blast waves from
propagation. Application of the numerical model to peat explosions in partly confined geometries, e.g. deliber-
dust explosions in a closed 20 litre vessel showed prom- ately vented, or bursting process equipment, and work
ising agreement with experiments. Krause7 presented a rooms. The strength and shape of blast waves from dust
comparatively simple two-dimensional model for explosions depend on the way in which the dust clouds
numerical simulation of explosions in vented enclosures. burn. For example, Wirkner-Bott et aL8* conducted a
The turbulente sub-model was empirical, and the fairly detailed study of the nature of the secondary
explosible medium was regarded as a homogeneous, pre- explosion, i.e. the explosion of unburnt dust cloud out-
mixed gas. Comparison with more complex models gave side the vent opening. This phenomenon was discussed
reasonable agreement. Comparison with dust explosion further by Schumann and Wirkner-Botts9. Centra1 vari-
experimental data was not reported. In a subsequent con- ables influencing blast wave generation, in addition to
tribution, Krause7* used this simulation model to predict type of dust cloud and geometry of system, include the
the maximum explosion pressure in a vented explosion dynamic state of the dust cloud at the moment of
as a function of the vent area and the turbulente intensity ignition, ignition point in relation to vent, the vent size,
in the dust cloud just prior to ignition. He was able to and the vent cover opening pressure. Van Wingerden
predict the earlier experimental finding of Tamanini et gave an informative overview of pressure and flame
al. (see Reference l), that the maximum explosion press- effects in the direct surroundings of installations pro-
ure in the vented vessel increases with increasing initial tected by dust explosion venting. Some basic studies of
turbulente intensity in the exploding cloud. shock wave emission from burning dust clouds were per-
Finally, the singular phenomenon of dust cloud det- formed by Gelfand et ~1.~.
onation should be mentioned. It is now generally Medvedev et a1.92 studied, experimentally as wel1
accepted that this can occur, but further work is needed as theoretically, the blast wave generated by sudden
in order to establish an adequate understanding of the expansion of a dust-filled enclosure, such as a hopper or
deflagration-to-detonation-transition process (DDT). An a pipe. The same authors 93 also studied the interaction
Prevention and mitigation of dust explosions: R.K. Eckhoff 9
between blast waves and dust deposits, using a specially Deng and He pointed out the need to use thermo-
developed shock tube technique. The experiments dynamics, chemical reaction kinetics, and fluid dynamics
revealed a strong dependence of the pressure amplitude for proper description of ignition and flame propagation
transmitted through the dust on the duration of the com- phenomena in dust clouds and layers. These topics con-
pression phase of the primary air shock wave. Smirnov stitute centra1 elements of classica1 chemical reaction
et aLy4 presented a new mathematica1 model of shock engineering, and Deng and He proposed a corresponding
wave propagation in dust clouds comprising poly-dis- concept Dust explosion reaction engineering (DERE)
perse systems, i.e. a range of different particle sizes for the dust explosion domain.
within the same cloud. Inter-particle collision was not Siwekio3 presented a concentrated overview of
considered. An experimental and numerical study of the current methods for dust explosion prevention and miti-
supersonic flow behind a shock wave passing through a gation in the process industries, based mainly on experi-
dust cloud was performed by Boiko et al.. Gelfand et mental research and development performed within the
aLy6 investigated experimentally the attenuation of shock Swiss/German domain. Siweks paper reflects the
waves propagating through dust clouds in a 50 mm important fact that industry needs practicable solutions
diameter shock tube. Reasonable agreement between for today, and cannot wait for more ideal solutions that
experimental data and analytical and numerical predic- may become available in the future. However, industrial
tions was found for incident shock waves of Mach num- pragmatism must not, on the other hand, black the con-
ber < 3. stant strive for better solutions based on improved basic
A useful condensed introduction to the complex understanding of the phenomena involved. It seems as
field of properties and effects of blast waves from if the mutual understanding and respect between the two
explosions has been given by Harmanny. parties, industry and the researchers, is growing.
The effect of a given blast wave on humans, build- EckhoffOJ reviewed the state-of-the-art on pre-
ings and process equipment is an important area where venting and mitigating dust explosions in the ferro-alloys
more research is needed. Valuable reviews are given by industry. In general, both the electric spark ignition sen-
MercxcJx and LAbb. Britan er al. 0studied the inter- sitivity and the explosion violente (K,,) of metal dusts
action of shock waves with layers of water-based foam increase with decreasing particle size, right down to the
used in fire fighting. Induction times and time constants 1 Pm region. In the past, this has not always been taken
for foam layer destruction were determined. They also into account. Often particle size was merely specified in
analysed the features of the transmitted shock wave and terms of less than 74 Pm or less than 63 Pm, which
the waves reflected from the air/foam boundary and from is by no means satisfactory. More research is needed, in
the walls of the experimental channel. particular on alloys, where the most hazardous compo-
nents may sometimes accumulate in the fine tail of the
Status and outstanding problems in particle size distribution.
preventing and mitigatingkontrolling dwt
explosions in industrial practice
Generation and states of industriul dust clouds
The role of fundamental knorvledge in developing Comparatively little quantitative knowledge of practica1
practica1 solutions use has been generated. It is known, however, that the
The status on the fundamental side, outlined above, is dynamic state of a dust cloud dramatically influences
necessarily also the status as regards the genuine under- both its ignition sensitivity and its combustion rate.
standing of the various corresponding phenomena in Therefore, it is clear that systematic quantitative assess-
industrial practice. indicated in Table 1. As already ment of this state is of great practica1 significante.
emphasized, the strengthening of the links between fun- Experimental investigation of typical processes of gener-
damental research and industrial research and develop- ation of dust clouds and the resulting states of the clouds,
ment should be encouraged. This wil1 promote the use in various types of process equipment and modes of
of available basic knowledge in the design of practica1 operation, should be encouraged. This is a typical area
measures, as wel1 as direct the basic research into areas where close interaction of fundamental and applied
useful for practice. research can be highly beneficial. The work of Hauert
Table 1 Means for preventing and mitigatingicontrolling dwt explosions in the process industries. Topics for applied research
and development
Prevention
et ak4, discussed above, constitutes an important step in Preventing explosible dust clouds
the right direction. 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
Preventing ignition sources a leve1 at which the dust cloud can no longer propagate
A considerable amount of fundamental knowledge is a self-sustained flame. A fair amount of data exists for
available, and the time is ripe for accepting that the con- maximum permissible oxygen content in the atmosphere
cept of ignition comprises a range of very complex pro- for inerting. However, there is room for improving the
cesses. Simple parameters such as a minimum ignition test methods by which such data are obtained. Further-
energy or temperature are not true constants for a given more, most data are for atmospheric pressure and normal
dust, but vary significantly with the geometry and other temperature. Data for other conditions, in particular for
properties of the ignition source, as wel1 as with the state elevated temperatures and pressures, are sometimes
of the dust cloud. Further development of correlations required, and adequate test methods should be
between results from standardized tests and actual criti- developed. Glor determined maximum permissible O2
cal ignition conditions in practice should be encouraged. contents for inerting clouds of coal dusts at elevated tem-
Gibson gave a valuable summary of methods for peratures and pressures. Wolinski and Hayashi18
preventing ignition of powders and dusts in drying oper- determined standard explosibility parameters of dusts
ations. ZockollOh described a new system for early and various rare metal alloys in air containing halon
detection of se(f-heating/self-ignition in deposits of 130 1 and added nitrogen. Theories for flame propagation
organic powders in spray driers, based on detection of limits (see discussion above with reference to the third
low concentrations of CO. column of Table 1) would be useful even in this context,
and their development should be encouraged. Krause et
Xu et al. found that very weak ekctric spark dis-
al. 19 proposed a simplified, approximate method for
charges, in the range 0.1-1.0 mJ, are probably able to
first-order estimation of the minimum oxygen concen-
initiate smouldering combustion in linen flax. Electro-
tration for flame propagation in dust clouds. Maddison*O
static hazards in connection with industrial use of flex-
summarized some important aspects of inerting powder
ible big bags were discussed by Rogers and by Dahn
handling plants using nitrogen, carbon dioxide and other
et al. 09. Pratt presented three case histories in which
inert gases.
electrostatic spark discharges were generated during Gao er al. 12described a new system for injection
pneumatic transport of powders. Sun Keping con- of pulverized bituminous coal into a blast furnace for
structed a new type of electric charge eliminator for steel production, for which nitrogen inerting constitutes
charged powders being pneumatically conveyed in pipes. a centra1 safety measure. Full or partial inerting by inert
The method was tested out successfully in industrial gas also constitutes a primary element in the strategy for
practice during a period of one year. Wang and Lou~ reducing the coal dust explosion hazard in melted slag
discussed the electrostatic hazards in powder handling reduction furnaces, as outlined by Wang Junyi et al.22.
and storage in general, and proposed methods for Whilst reducing the oxygen content in the atmosphere
preventing/reducing this hazard. prevents dust explosions, it can introduce a suffocation
An example of commercially available equipment hazard. However, it has been shown that adding a few
for preventing ignition in industrial plants was given by vol% CO2 to the gas mixture considerably reduces the
Kleinschmidt3, who reported that the GreCon system critical oxygen threshold for suffocation. An inert gas
for detection and extinction of sparks in terms of Jlying mixture (INERGEN) utilizing this effect is now being
burning particles is being developed further with respect marketed by Dansk Fire Eater A/S23. Further work to
to optimizing system performance. The same applies to identify gas mixtures that keep the dust cloud inert with-
the Firefly system, which, according to Janssen, offers out presenting a suffocation hazard should be welcomed.
an adjustable lower particle temperature limit of detec- Inerting by adding non-combustible dust is not gen-
tion, down to 150C. A multi-zone checkpoint system erally applicable, because the inert dust wil1 in most
prevents false alarms, and indicates the size of the hot cases cause unacceptable contamination. However, there
object (single particle, several particles or extensive are cases where the dust/powder processed is already a
flame). Depending on the detection temperature and the mixture of combustible and non-combustible dusts, and
nature of the industrial process, detection of a hot object where control of the composition ensures that dust
may give rise to either activation of an extinction system, clouds are non-flammable. It is then essential to avoid
close-down of plant, or simply adjustment of plant run- segregation of combustible and non-combustible compo-
ning conditions to prevent further hot object generation. nents throughout the process. Zhang Zhancheng et al. 24
A similar spark detection and extinguishing system has studied the possibility of preventing grain dust
also been developed by T&B Electronic . explosions in grain storage facilities by adding talc pow-
Lloyd and Laar presented the state-of-the-art der to the grain stream. The method works, as far as
on international standardization of classication ofareas explosion prevention is concerned, but the industrial
hygiene aspects require careful consideration.
containing combustible dust, and guidelines for selection
Keeping the dust concentration below the minimum
of electrical equipment for such areas. Centra1 organiza-
explosible concentration is a third means of maintaining
tions involved in this work are the International Electro-
dust clouds non-explosible. More work is needed to
technical Commission (IEC), and CEN and CENELEC
establish procedures for the determination of dust con-
of the European Community.
centrations in industrial situations. Zockoll25 described
Prevention and mitigation of dust explosions: R.K. Eckhoff 11
an apparatus, based on infra-red light attenuation, for German concept of pressure-shock-resistant design
determining rhe dusr concenrrarion of dust clouds gener- Should be further developed. Crowhurst* gave a USefUl
ated in various situations in industry. Typical measure- overview of the state-of-the-art on desig? of equipment
ment results were reported, demonstrating the feasibility to withstand a given overpressure caused by a fully con-
of the method. Xu Bowen et al. 26proposed a systematic fined or vented explosion. Bartenev et al. 33 developed
method for estimating the dust explosion hazards in a mathematica1 model for the failure of a process vessel
industrial plants. The method implies an empirical subjected to an internal exothermic, comparatively slow
relationship between local dust concentration and the process creating an overpressure, allowing the maximum
local rate of dust deposition from the cloud [mass (time initial velocities of fragments or vent covers to be esti-
area)-]. Properties of the dust emission source must be mated. Bartenev et al. 34extended this work to the case
known. Also, research is needed to establish the mini- where the pressurized vessel is filled with dust. Exper-
mum mass of dust deposit per unit of surface area that iments revealed that the presence of dust can have a sig-
is required for maintaining a self-sustained explosive nificant effect on the pressure development inside the
combustion along the surface. The assumption that bursting vessel, and on the kinematic parameters of the
entrained dust wil1 become distributed evenly throughout ejected fragments.
the available space may not hold. For example, a thin The objective of explosion isolation is to prevent
dust layer on a floor may be dispersed into just a shallow, dust explosions from spreading from the primary
dense dust cloud close to the floor, through which the explosion location to other process units, workrooms,
flame can sweep. The conditions required for producing etc. Van Wingerden et al. reported on dust explosion
this kind of self-sustained shallow sweeping Aames need experiments in integrated systems of various process
to be investigated further in the context of the frst col- equipment connected with ducts. Due to pressure piling,
umn of Table 1. jet-initiated high initial turbulente and turbulent jet
Wolanski27 suggested that the concept of mini- ignition, very high pressure peaks can be generated even
mum hazardous mass of dust be introduced in the evalu- in generously vented vessels. This investigation is a
ation of dust explosion hazards in practice. This para- further demonstration of the need for effective means of
meter is defined as the minimum mass of the actual dust explosion isolation in coupled systems. Various passive
that can generate a dust explosion of destructive and active techniques for explosion isolation have been
strength. The parameter is not a constant for a given developed and are being used, but there is room for
dust, but depends also on the characteristics of the actual further improvement. If adequate performance can be
enclosure in which the explosion takes place. The basic achieved, passive techniques are clearly more attractive
idea is as follows: a quantity of dust, unable to generate than active ones.
a cloud of concentration above the minimum explosible A basic understanding of flame propagation and
value when being dispersed evenly throughout the entire pressure build-up in coupled geometries (interconnected
enclosure volume, can present a significant explosion vessels) is important for the prediction of the perform-
hazard when being dispersed in the enclosure as a ance of various active and passive isolation equipment.
smaller and correspondingly denser cloud. Valuable large-scale experimental work in this area was
Xu Bowen et al. 2Xfound that the minimum explos- reported by Lunn and by van Wingerden and Alfert.
ible concentration of linen flax dust in air was inde- The understanding of flame propagation in long ducts
pendent of particle size up to about 100 Pm. For larger and pipes is also very important. Vog138 presented
particles, a systematic increase of MEC with increasing results from a comprehensive experimental investigation
particle size was found. Mittal12 discussed various of the propagation of dust explosions in pipes for pneu-
mathematica1 models for calculating minimum explos- matic powder transportation. Pipe lengths up to 48 m and
ible concentrations of dust clouds. Shao and Wang130 pipe diameters up to 200 mm were used. The influences
reviewed some non-intrusive methods for measuring the of a range of experimental parameters on the flame speed
dust concentration in dust clouds, primarily in pipes and explosion pressure were studied. The parameters
and ducts. were: initial air velocity in the pipe, pipe diameter,
location of the ignition source, dust concentration in the
Mitigating and controlling measures pipe, and the Ks, value of the dust. Zellweger
The use of explosion-pressure-resistant process equip- described further work to improve passive and active
ment for full explosion conjinement is limited because isolation valves of the VENTEX type. Closing times
of high equipment costs. Current experimental methods (from sensing of the explosion to valve fully closed)
allow sufficiently accurate prediction of maximum down to 12 ms were obtained for active valves. A sim-
explosion pressures in simple vessels with point source plified VENTEX valve, operating in one direction only,
ignition. This also applies to explosions at elevated has also been developed. Passive explosion interrupters
initial pressures. Torrent and Menndez found that the based on venting at a bend have been in use for some
proportionality between initial and final pressure found time. However, there is room for further exploration of
previously in smal1 laboratory-scale vessels also holds the potential of this attractive, simple principle of
in vessels of 1 m3 volume, up to 12 bar initial pressure. explosion isolation. A new low-pressure-drop design
However, if complex dynamic pressure development, was described by Alfert and Fuhre140 and by van Wing-
e.g. with pressure piling, is to be expected, closed-bomb erden and Alfert13. Glor reported work on the per-
test data are of limited value. There is also room for formance of explosion barriers in ducting connected to
further improvement in the design of the process equip- vessels with venting or automatie explosion suppression.
ment itself, with respect to minimizing its heaviness. The Klincewicz and Kordylewski described a new, prom-
12 Prevention and mtgaton of dust explosons: R.K. Eckhoff
ising design of an explosion diverter for interrupting simulation code DYNAMICS. Fan Xisheng et al. Is0
explosions in pipe lines. The new design avoids the studied, theoretically as wel1 as experimentally, the
pressure drop created in normal operation by passive influences of details of the mounting flange arrangement
diverters, but requires active triggering. Cybulski et on the static bursting pressure of bursting
al. 42, addressing the problem of coal dust explosions in panels/membranes. They observed two distinctly differ-
coal mines, used solar panels for automatie detection of ent bursting patterns, viz. bursting of the membrane in
the coal dust flame and simultaneous actuation of water its centra1 region, and bursting along the edge.
barriers. The water was contained in plastic bags, which The paper by Crowhurst2 on the design ofenclos-
were opened by a detonating cord triggered by the flame- ures to withstand a given maximum explosion pressure
generated power from the solar panel. also applies to vented enclosures. Harmanny pre-
Partial inerting by inert gas is a promising means sented a new formula for predicting the duration of
for mitigating dust explosions, which deserves further vented dust explosions in enclosures of volumes from
attention. The idea is that as the oxygen content in the 10 to 60 m3. This is useful for evaluating whether static
atmosphere is decreased, there is a gradual decrease of pressure considerations or impulse considerations apply
both ignition sensitivity and combustion rate of the dust when predicting the response of the enclosure structure
cloud. In some cases, the explosion hazard may be to the explosion load.
reduced substantially by only a moderate reduction of A further dimension of complexity is added to the
the oxygen content in the gas. However, more research venting problem if the initial pressure (and/or
seems necessary in this area to establish a correlation temperature) deviates from atmospheric. Results from
between the oxygen content in the gas and various ignit- venting of dust explosions in air of elevated initial press-
ability and explosibility parameters. Interesting new ure were reported by Siwek et al. 15*. The effect of press-
information is currently being produced in a joint Euro- ure piling and turbulent flame jet ignition on vent area
pean research programme (CREDIT), focusing on the requirements in systems of interconnected vessels were
influence of oxygen content of the atmosphere on the studied by Lunn6.
minimum energy and temperature for ignition of dust
In dust explosion venting, maintaining the integrity
clouds.
of the enclosure is not the only concern. Venting implies
Dust explosion venting remains a complex and con-
that both pressure waves and flames are emitted into the
troversial subject. Adequate understanding of flame
surroundings, and this may present a hazard, depending
propagation processes in dust clouds is essential for the
on the size of the emitted flame and the magnitude of
design of optimal venting arrangements for industry.
the blast wave. Crowhurst et al. reported a series of
Useful reviews of various aspects of dust explosion vent-
large-scale venting experiments (40 m3 and 20 m)
ing in practice were given by Scholl4 and Lunn44*45.
where external blast waves and flame lengths were
The basic understanding of flame propagation processes
determined and compared with existing empirical corre-
inside and outside vented enclosures is stil1 unsatisfac-
lations (see also the section above on Preventing explos-
tory. This implies that neither the processes by which
ible dust clouds). Peng et al. Is4 developed a labyrinth
dust clouds of given structures are generated, nor the
type flame arrestor for mitigating flame and pressure
way in which clouds of given initial structures burn, are
effects from dust explosion venting. Experiments using
wel1 understood. Consequently, adequate venting the-
a 2.7 m3 enclosure and a 370 mm diameter vent opening
ories do not exist, and one must rely on experiments.
suggested that this arrestor concept works satisfactorily
During the last few years, the need for differentiating
for Stl dusts. The maximum explosion pressures in the
vent area requirements in view of the different turbu-
vented enclosure, with the arrestor mounted in the vent,
lente levels, degrees of dust dispersion and concen-
were only a few per cent higher than the pressures with-
tration distributions of dust clouds which occur in prac-
tice has become widely accepted. Eckhoff46 discussed out the arrestor, whereas both flame and pressure effects
this problem with particular reference to venting of large were substantially reduced. Li Gang et al.55 presented
silo cells. Deng et al. 14conducted vented maize starch a new version of the quenching tube first developed in
explosions in a 95 m3 vertical experimental silo of Norway (see Reference l), in which the conventional
UD = 3. The 3.4 m2 vent was located in the silo roof. bursting diaphragm had been replaced by a re-usable
The maximum vented explosion pressures were com- vent cover. This solution is of interest in situations where
paratively low. Hchst et al. 4X conducted systematic the expected frequency of explosions is comparatively
maize starch/air explosion experiments in a top-vented high.
silo of volume 50 m3 and UD = 4. Both dust concen- Venting of industrial buildings requires special con-
tration and initial turbulente (prior to ignition) were siderations. A useful overview was given by
monitored. The dust injection process as wel1 as the Crowhurst.
ignition point location were varied. Flame propagation The injluence of vent ducts on the maximum
and pressure build-up during the vented explosion were explosion pressure in the vented vessel has been studied
measured. Comparative experiments with methane/air in experimentally by several workers. Recently Ural pre-
the silo were also performed. sented a theoretical model for vented gas explosions by
Molkov et al.14 studied experimentally the influ- which he has been able to calculate pressure-versus-time
ence of the inertia of the vent cover on the gas dynamics characteristics in the vented vessel that agree wel1 with
of the venting process. Good agreement was obtained corresponding experimental data. It remains to be inves-
between pressure-versus-time traces from experiments tigated, however, whether this theory can reproduce
and corresponding traces obtained using the numerical existing experimental data.
Prevention and mitigation of dust explosions: R.K. Eckhoff 13
Dust explosion venting remains an area in which explosions in short, narrow tunnels in grain elevators can
considerably more work is required. The new German be extinguished by properly designed, actively triggered
draft VDI-3673 venting guideline, issued in 1992, has water barriers. Cybulski et al. M presented results of
met some objections in other countries. This in particular experiments with propagation of weak coal dust
applies to the concept of heterogeneous dust clouds. explosions in a network of full-scale mine galleries. A
Nevertheless, computer-based software for easier use main conclusion was that, under the conditions prevail-
of this draft guideline was presented by Alfertls8. ing, the possibility of flame penetration into blind gallery
Hattwig and Hensel 59 discussed deficiencies in the new branches was small. This kind of work may also be of
VDI draft guideline on the basis of dust explosion exper- relevante to the analysis of flame propagation in large
iments in 4.5 m3 (UD = 2) and 85 m3 (UD = 4) vented industrial systems, e.g. in grain storage and handling
silos of square cross-sections. plants.
Automatie explosion suppression is an active, com- Gieras et al. 67 studied the development of com-
paratively sophisticated method of dust explosion bustible gases (H2, CO and CH,) during combustion of
mitigation/control, which is used when simpler and less fuel-rich clouds (up to 5 kg mmR)of grain dust in air.
expensive methods cannot be applied. Although this This is an important aspect in the context of industrial
method has been in use for many years, there is stil1 a safety, because mixtures of combustible gases and air
need for research and development. Moore60 reported can give rise to severe secondary gas explosions, follow-
that the number of suppressant bottles of a given size ing comparatively slow primary fires in fuel-rich dust
required for suppressing explosions of a given dust in a clouds.
given vessel was reduced by a factor of 0.2-0.3 when Quantitative risk analysis (QRA) is receiving
the dust clouds were generated by industrial pneumatic increasing attention as a potential means of controlling
injection rather than by the VDI method used in previous explosion risks in the process industries. This also
experiments. G10r~ and Moore6 reported current work applies to the dust explosion hazard. The subject is com-
on the possibility of applying this method even in the plex, and in part controversial, not least because of a
case of highly explosible organic dusts of K,, > 300 lack of relevant quantitative failure rate data. However,
bar m s-. In the case of aluminium powders, satisfactory attempts are nevertheless being made to apply QRA to
suppression has not yet been achieved for powders of the dust explosion problem, as demonstrated by
K,, > 200 bar m ss, which means that only dust Wagner6x. Mittal69 discussed the risk of dust
explosions in clouds of relatively coarse aluminium explosions in pneumatic powder transportation systems.
powders can be suppressed. The influence of the Potential sources of ignition were identified, with the
dynamic state of the dust cloud at the moment of sup- main focus on electrostatic discharges. A systematic risk
pressant injection, the influence of the suppressant injec- evaluation and risk reduction approach was outlined. Li
tion on this state, and development of improved sup- and Wangi7 proposed a quantitative, computer-based
pressants are some of the areas where further work decision support system for assessing the explosion
would seem useful. Recent experiments have indicated hazard and taking appropriate actions to prevent and
that water can be an effective suppressant, if injected at mitigate coal dust explosions in metal production plants.
a temperature > 180C. According to Tyldesleyh2, opti- Moore and Frehill7 outlined a philosophy for system-
mum suppression is found to require about 0.5 litres of atic analysis of the risks of dust explosions in industrial
water per m of vessel volume: 16-18% of the super- plants. Specific technical solutions for preventing and
heated water is flashed to steam, the remainder forms mitigating explosions for some typical cases were pro-
very smal1 droplets. Reduced explosion pressures of 0.3- posed.
0.4 bar(g) were obtained in experiments in a 28 m3
experimental vessel, using a dust of K,, = 150 bar m ss. Status and outstanding problems in testing
Gieras et af. conducted a series of experiments in order of dust ignitability and explosibility
to optimize the shape and mass of the explosive charge Historica1 background and standardization of methods
used for automatie release of suppressant (powder) for When some of the older test methods were designed, the
suppressing dust explosions. The overall aim of the ambition was in fact quite modest. The original intention
research was to minimize the powder ejection time. was just to establish some relative measures of properties
Siwek described experiments where a combi- of practica1 relevante to preventing and control-
nation of explosion venting and automatie suppression lingimitigating dust explosions. Later some of these
was adopted for mitigating/controlling dust explosions methods were adopted as official standards, and test data
in various enclosures. Sliz et a1.1h4investigated the per- were sometimes treated as basic physical constants for
formance of such a combined system for mitigating grain a given dust to an extent far beyond the original purpose
dust explosions in an 8 m experimental vessel. Different of the test. As more knowledge from systematic research
types of vent covers and suppressors were tested. became available, the lack of justification for this use of
Prevention of secondary explosions outside process these test data was pointed out, and the arbitrary, relative
equipment remains an important issue. Proper house- nature of the various test methods was brought to light
keeping is an esstntial means of achieving this aim. again.
However, further work is required for assessment of the The situation today is complex. It is realized that
maximum acceptable mass of deposited dust per unit only a few of the dust parameters that are currently being
area of surface for preventing secondary dust flame used for characterizing ignition sensitivity and explos-
propagation under various conditions. Cybulski et al. M ibility of dusts can be regarded even as approximate
showed that comparatively weak secondary dust physical constants for a given dust. In most cases, a
14 Prevention and mitigation of dust explosions: R.K. Eckhoff
great number of variables are involved and a differen- In the case of minimum ignition energy, the trans-
tiated view is required. Typical examples are the mini- formation would imply that the basic test value be con-
mum ignition energy and the explosion violente of verted to minimum discharge energies for ignition of the
dust clouds. same dust cloud by various kinds of discharges, such
A number of standard test methods have been as direct electrostatic two-electrode sparks, break flashes
developed through the years, e.g. by IEC (International (e.g. live wire rupture) and one-electrode discharges.
Electrotechnical Commission) and ISO (International However, theories are scarce and the time may not yet
Standardization Organization). In the USA, ASTM be ripe for this approach. Siwek and Cesana173 presented
(American Society for Testing and Materials) has issued a new test apparatus for the determination of minimum
a number of standards in this area. Recently, the stan- electric spark ignition energies of dust clouds down to
dardization organization of the European Union, CEN, below 1 mJ spark energy. The measurements in the low-
has launched a multi-year programme to produce a series energy range are demanding and require carefully
of standard methods for the testing of explosion propa- designed equipment and skilled personnel. Zhou
gation limits, ignition sensitivity and explosion violente Benmou et aZ.74 discussed the features of an apparatus
of combustible dust clouds. Reference 1 gives a compre- for determining the minimum electric spark energy for
hensive presentation of the wide range of test methods igniting very sensitive explosives such as lead azide.
in use in Europe and the USA. Spark energies as low as 1 PJ were attainable, using a
travelling-electrode system.
Two alternative approaches for achieving Limits of JEame propagation - a special problem of
differentiation scale
A first approach for handling this situation is to have Determining limits of flame propagation constitutes an
several test methods for any given parameter, allowing important test objective (explosible/non-explosible
test conditions to be selected according to the practica1 assessment, and assessment of minimum explosible dust
use of the test result. In the case of minimum ignition concentration and maximum permissible oxygen con-
energy, such an approach has in fact been incorporated centration for inerting). However, special care must be
in the new IEC standard issued in 1994. Whereas an exercised in designing flame-propagation-limit tests.
appreciable inductance is to be included in the capacitive One basic problem is that near the limits self-sustained
discharge circuit in the case of standard testing, to obtain flame propagation cannot be established unless a con-
the most incendiary sparks, it is argued that this is not siderable amount of energy is supplied for initiating
relevant if the test result is to be used for assessing the flame propagation. Hence, if the volume of the experi-
electrostatic spark discharge hazard. mental dust cloud is too small, it is difficult to assess
A similar situation should arise when testing for whether observed flame propagation is truly independent
explosion violente. It is now appreciated that the stan- of the ignition source. Some recent results by Cashdollar
dard ISO Ks, value of a given dust reflects a rather et aZ.175 and Cashdollar and Chatrathi76 are of funda-
extreme combustion rate in the conservative direction, mental significante in this context. They found that
because the turbulente leve1 and the degree of dust dis- clouds in air, at normal ambient conditions, of an anthra-
persion in the test are rather extreme. If, for example, cite coal dust of 8% volatile matter did not show self-
K,, values are stil1 to be used for sizing of explosion sustained flame propagation in a 1 m3 test chamber, even
vents, differentiation by varying the intensity of the dust when being exposed to a 30 kJ chemical igniter. How-
cloud formation process used in the test to fit the practi- ever, in a 20 litre chamber, fully developed explosions
cal situation of interest could be worthwhile considering. were generated even with a 5 kJ chemical igniter. The
The work by Liu et al. 17is relevant in this context. They reason for this could be that in the smal1 chamber, due
described introductory studies of the turbulente structure to the initial combustion and expansion of the dust cloud
in experimental dust clouds in the 1.2 litre Hartmann directly affected by the ignition source, the pressure and
bomb under various conditions of dust injection. This temperature in the unbumt cloud ahead of the flame
apparatus was traditionally used for characterizing the increase significantly before flame propagation no longer
explosion violente to be expected from various dusts. receives support from the ignition source. Consequently
Because of its smal1 size, the Hartmann bomb has in the self-sustained flame propagation, if any, occurs in an
recent years been replaced by larger test vessels. adiabatically pre-compressed dust cloud, rather than in
However, there is a second, alternative approach for a cloud of normal ambient temperature and pressure.
handling the need for differentiation. In this case, the These results suggest that great care must be exercised
measured dust ignitability or explosibility parameter whenever comparatively smal1 chambers, in particular
should be of a more basic nature, and differentiation to closed ones, are used for any explosion limit determi-
meet various practica1 situations in industry should be nation.
via adequate theory, using the basic parameter as input. Zhou Congzhang et a1.77 proposed a new, interest-
In the case of explosion violente assessment, the exist- ing procedure for determining minimum explosible dust
ing K,, concept may be replaced by the laminar burning concentration values in closed-bomb explosion exper-
velocity of the dust cloud, for example. It is clear, how- iments. They presented experimental evidente indicating
ever, that theories for transforming such basic para- that at the minimum explosible concentration, the time
meters to flame propagation rates in practice must interval from ignition to the pressure peak has its highest
involve complex fluid dynamic computer simulation value. They proposed that this criterion be used instead
codes. of some arbitrary pressure rise criterion of explosion.
Prevention and mitigation of dust explosions: R.K. Eckhoff 15
Matsuda and Itagaki, comparing dust explosions in a physics and chemistry of the phenomena treated, expert
30 litre explosion bomb, and in a 1 m3 vessel, found that systems should indeed be welcomed. However, there
the ranges of explosible concentrations in the 30 litre may be a possibility of the future market place being
vessel were considerably wider than those in the 1 m3 offered software that is not up to acceptable standards
vessel for the same dust. A marked increase of the with respect to the physics and chemistry. As long as the
explosible range was found in the 30 litre bomb when interior of the system is not fully exposed, deficiencies in
increasing the ignition energy from 1 to 10 kJ. This the basics may not be obvious to the user.
effect was practically absent in the 1 m vessel in the The time may very soon be ripe for making
ignition energy range 4-20 kJ. Tian Renqu er al. , adequate quality assurance of this kind of system com-
using a 20 litre explosion bomb, found that the minimum pulsory. A need may emerge for establishing some inter-
explosible concentrations of coal dusts decreased by a nationally recognized body of experts that can ensure
factor of two or more when the ignition energy was that expert systems offered in the area of explosion pre-
increased from 2.5 tol0 kJ. When using a 2.5 kJ igniter, vention and mitigation are up to acceptable standards.
and adding 2 vol% methane to the air, the minimum
explosible dust concentration dropped by at least a factor The human hazard factors
of two, compared with the values for dust in air. This
The present survey deals with the chemistry, physics and
hybrid effect has been studied previously by several
technology of dust explosion prevention and mitigation.
other workers. Xu Tianrui et al. also arrived at the
However, a brief mention should also be made of the
conclusion that the apparent minimum explosible dust
importante of the human factors in this effort. This
concentration determined in a 20 litre bomb depends
aspect has been discussed by Fernandoxh.
markedly on the ignition energy. A value for ignition
energy of 10 kJ was found to be too high to yield
Joint research efforts in Europe, and
realistic results.
research and development in the Peoples
Pu et al. concluded that the turbulente structure
Republic of China
of experimental dust clouds in a standard 20 litre spheri-
cal dust explosion test bomb had little resemblance to Over recent years, a steadily growing potential for
turbulente structures in dust clouds in accidental dust organizing joint European research efforts has emerged
explosions in industry. within the EU/EFTA/EUREKA system. This also
applies to dust explosion research. The British Materials
Miscellaneous Handling Board (BMHB) in the UK has played a centra1
Wang and ZhanglX determined the minimum ignition role in this processx7. A number of parallel research pro-
energy, the minimum explosible concentration, and the grammes have been started within the European Unions
maximum explosion pressure for clouds of TNT dusts Credit Project. Gibsonxx summarized the areas requir-
in air. The values are similar to those of natura1 organic ing further work under the headlines:
materials. The results confirm that dilute clouds of dusts ?? combustion processes in dust clouds (experiments,
of explosives do not exhibit explosive properties, but theoretical models)
behave as clouds of ordinary combustible dusts. ?? identification and control of ignition sources
?? design of methods to prevent/protect against dust
Expert systems - friends or enemies? explosions
During recent years, there has been an increasing interest These headlines in fact cover most of the research needs
in developing sophisticated computer-based expert sys- identified in the present review. In addition, mechanisms
tems for evaluation of dust explosion hazards and assess- of generation of dust clouds, which are important
ment of optimal safety design features. Haefen and because they set the stage for subsequent ignition and
Scheckerix presented such a system for assessment of combustion, require further exploration.
dust explosion hazards in industry and selection of Wang DongyanXy, characterizing the Peoples
appropriate means of prevention and mitigation, which Republic of China as a developing country, emphasized
is in al1 essentials based on the German protection philo- the need for increasing the efforts to prevent dust
sophy. Wachix4 presented another expert system explosion accidents in Chinas rapidly growing industry.
designed for the same purpose, but the technical and Of the number of dust explosions recorded in this coun-
philosophical basis was not explicitly stated. A compre- try during the decade 1980-1989, 65% were in the grain
hensive expert system developed in the UK was pre- industry, 17% in the textile industry, 12% in the coal
sented by Tyldesleys, and the need for quality assur- industry and 6% in the metallurgical industry. With the
ante of such systems was emphasized. rapid development of the chemical and metallurgical
The development of this kind of expert system is a industries, the annual number of explosions may easily
natura1 consequente of two main factors. The first is the rise if adequate precautions are not taken. There is a
almost explosive development of the performance of strong need for education and training on al1 levels, and
personal computers. The second is the steadily increas- for adequate safety technology. The 6th International
ing knowledge about ignition and explosion phenomena, Colloquium on Dust Explosions in Shenyang, Peoples
which demands a steadily more differentiated and com- Republic of China, in August/September 1994, demon-
plex approach for solving the practica1 design problems. strated that research and development on dust explosion
As long as this development is conducted by people prevention and protection in this enormous country is
who are not only experts on computers, but also on the growing at great pace.
Prevention and mitigation of dust explosions: R.K. Eckhoff
trol, Cambridge, UK, organized by IBC Technical Services Ltd 59 Wiemann, W. Rauchgasreinigung mit brennbaren Stuben, VDI
in association with BMHB and IELG Betichte No. 975, 1992, pp 607-625
36 Glor, M. and Maurer, B. Zndversuche mit Schttkegelentladun- 60 Eckhoff, R. K. Powder Technol. 1992, 71, 181-187
gen, Paper presented at VDI Colloquium Sichere Handhabung 61 van der Wel, P. G. J., van Veen, J. P. W., Lemkowitz, S. M.,
brennbarer Stube, 4-6 November 1992, Nrnberg, Germany Scarlett, B. and van Wingerden, K. Powder Technol. 1992, 71,
37 Glor, M. Private communication, Ciba-Geigy AG, Basle, Switzer- 207-215
land, 1993 62 van der Wel, P. G. J., Lemkowitz, S. M., Timmers, P. and Scar-
38 Wolanski, P. Deflagration and detonation combustion of dust lett, B. The role of turbulente on the propagation mechanism and
mixtures, American Institute of Aeronautics and Astronautics, behaviour of dust exolosions. Proc. 5th Int. Coll. Dust
1990, pp 3-3 1 Exnlosions, Pultusk near Warsaw, Poland, 19-22 April 1993,
39 Boiko, V. M. and Papyrin, A. N. Laser diagnostics processes of pp 199-209
mixture formation and combustion of dusts in shock waves, Proc. 63 Scheuermann, K. Studies about the influence of turbulente on the
6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.- course of exolosions. Proc. 28th AIChE Ann. Loss Prev. Svmp.,
2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 302-305 Section No. 13 on Dust Explosions, Atlanta, USA, 17-21 April
40 Geng, J. H., Liao, S. P. and Tang, M. J. Dynamics effects on 1994
ignition of a dust suspension, Proc. 6th Int. Coll. Dust 64 Rzal-Rebire, F. and Veyssire, B. Interaction of a vortex ring
xplosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds with gaseous and dust flames, Paper submitted for presentation
Dene Xufan and Wolanski. P.. DD 366-370 at the 24th Symposium (International) on Combustion, Sydney,
41 Ge& J. H., van de Ven, A., Zhang, F. and Gronig, H. A new Australia, 5-10 July 1992 (International Combustion Institute)
setup to measure ignition delay of a dust suspension behind an 65 Veyssire, B. Powder Technol. 1992, 71, 171-180
incident shock wave, Proc. 6th Int. Coll. Dust Explosions, Shen- 66 Rzal-Rebire, F. and Veyssire, B. Propagation mechanisms of
yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and starch particles-air flames, Proc. 6th Int. Coll. Dust Explosions,
Wolanski, P., pp 309-314 Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan
42 Lu, S. X. and Fan, B. C. Dust combustion in shock-induced and Wolanski, P., pp 186200
boundary layer, Proc. 6th Int. Coll. Dust Explosions, Shenyang, 67 Geng, J. H., Tang, M. J. and Grnig, H. Shock-induced ignition
P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, delay of cornstarch dusts, Proc. 5th Int. Coll. Dust Explosions,
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 68 Ural, E. A. Private communication, Factory Mutual Research,
on highSpeed riaction behaviiur of wheaten flour, Proc. 6th Int. PO Box 9102, Norwood, MA 02062, USA, 1992
Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 69 Boiko, V. M., Papyrin, A. N. and Poplavski, S. V. On peculiari-
1994, eds Deng Xufan and Wolanski, P., pp 355-359 ties of coal dust ignition in incident shock waves, Proc. 5th Int.
44 Mazurkiewicz, J. and Jarosinski, J. Gas composition near a com- Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
starch dust-air flame front, Proc. 5th Int. Coll. Dust Explosions, 1993, pp 329-334
Pultusk near Warsaw, Poland, 19-22 April 1993, pp 153-160 70 Bradley, D., Chen, Z. and Swithenbank, J. R. Buming rates in
45 Proust, Ch. Experimental determination of the maximum flame turbulent fine dust/air explosions, Proc. 22nd Symp. (Int.) on
temperatures and of the laminar buming velocities for some com- Combustion, The Combustion Institute, Pittsburgh, USA,
bustible dust-air mixtures, Proc. 5th Int. Coll. Dust Explosions, pp 1767-1775
Pultusk near Warsaw, Poland, 19-22 April 1993, pp 161-184 71 Proust, Ch. Dust explosions in pipelines, Paper given at the
European Summer School on Dust Explosion Hazards: Their
46 Seshardi, K., Berlad, A. L. and Tangirala, V. ComDust. Flame
Assessment and Control, Cambridge, UK, organized by IBC
1992,89, 333-342
Technical Services Ltd in association with BMHB and IELG,
47 Bradley, D., Chen, Z., El-Sherif, S., El-Din Habik, S. and John,
1993
G. Combust. Flame, 1994, 96, 80-96
72 Pu, Y. K., Hu, S., Yang, Q. Z., Bao, G. G. and Wang, Z. M.
48 Gieras, M., Klemens, R., Wolanski, P. er al. Suppression of dust
Combustion tube studies of dust flame acceleration, Proc.
explosion triggered by explosive charge, Proc. 6th Int. Coll. Dust
(suppl.) 5th Int. Coll. Dust Explosions, Pultusk near Warsaw,
Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
Poland, 19-22 April 1993, pp 33-46
Deng Xufan and Wolanski, P., pp 484490
73 Hu Dong and Sun Zhumei Studies of the behaviour of aluminium
49 van Wingerden, K. Private communication, Christian Michelsen
powder reaction in the gas phase reaction environment, Proc. 6th
Research AS, Fantoftvegen 38, N-5036 Fantoft, Bergen, Nor-
Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.-
way, 1993
2 Sent. 1994. eds Dena Xufan and Wolanski. P.. DD 349-354
50 Krishenik, P. M. and Shkadinskii, K. G. Modeling of combustion
74 Eckhoff, R. K. Generition. ignition, combustion and explosion
modes in Bi-size and Bi-material dust-air mixture, Proc. 5th Int.
of sprays and mists of flammable liquids in air. A literature sur-
Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22
vey, Report No. CMI-91-A25014 from Christian Michelsen Insti-
April, 1993, pp 191-198 tute. N-5036 Fantoft. Norwav. 1991
51 Deng Xufan, Xu Renxian, Xie Lin, Dang Junxian, Zang Tingan, 75 Huang Wanli, Pu Yikang and Ding Hua The electrical conduc-
Gao Jun and Tan Feng Gui Explosibility and ignitability of 16 tion phenomenon in the process of combustion of aluminium
types of dust and some opinions on fundamental research for dust
powder air mixtures, Proc. 6th Int. Coll. Dust Explosions, Shen-
explosions, Proc. 5th Int. Coll. Dust Explosions, Pultusk near vang, P.R. China. 29 Atm-2 Seot. 1994, eds Dena Xufan and
Warsaw, Poland, 19-22 April 1993, pp 217-224 Woanski, P., pp 201-205- 1
52 Lee, J. H. S., Zhang, F. and Knystautas. R. Powder Technol. 1992, 76 Kjldman, L. Powder Technol. 1992, 71, 163-169
71, 153-162 77 Krause, U. A two-dimensional model for the numerical simul-
53 Wlodarczyk, E., Maranda, A., Nowaczewski, J., Trebinski, R. and ation of explosions in vented vessels, Proc. 5th Int. Coll. Dust
Sliz, J. Influence of the composition of gas-dust mixture on the Explosions, Pultusk near Warsaw, Poland, 19-22 April 1993,
content of aluminium oxide in explosion products, Proc. 5th Int. pp42430
Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April 78 Krause, U. Numerical investigation of the influence of velocity
1993, pp 279-285 fluctuation on ventina of vessels. Proc. 6th Int. Coll. Dust
54 Fan, B. C., Ding, D. M. and Tang, M. J. An aluminium dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
explosion in a spherical closed vessel, Proc. (Suppl.) 5th Int. Deng Xufan and Wolanski, P., pp 441-452
Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April 79 Alexander, C. G., Harbaugh, A. S., Kauffman, C. W., Li, Y. C.,
1993, ~~21-31 Cybulski, K., Dyduch, Z., Lebecki, K., Sliz, J., Klemens, R., Wol-
55 Mintz, K. J. Combusr. Flame 1993, 34, 125-130 anski, P. and Zalesinski, M. The establishment of dust deton-
56 Poletaev, N. L. and Korolchenko, A. Ya. A note on the relation- ations. Proc. 5th Int. Coll. Dust Explosions, Pultusk near War-
ship between the lower explosibihty limit of dust, and particle saw. Poland. 19-22 April 1993. DD365-381
size, Proceedings of Joint Meeting of the Russian and Japanese 80 Sichel, M. and Kauffman, C. W. Transition from deflagration to
Sections of the Combustion Institute, Chemogolovka, Moscow detonation in layered dust explosions, Proc. 28th AIChE Ann.
Region, 2-5 October 1993, pp 116117 Loss Prev. Symp., Section No. 13 on Dust Explosions, Atlanta,
57 Hertzberg, M., Zlochower, 1. A. and Cashdollar, K. L. Metal dust USA, 17-21 April 1994
combustion: Explosion limits, pressures, and temperatures, Proc. 81 Khomik, S. V., Gelfand, B. E. and Knyazev, M. V. On the criti-
24th Symp. (Int.) on Combustion, The Combustion Institute, Pitts- cal diameter of detonation propagation in dust suspensions, Proc.
burgh, USA, 1992, pp 1827-1835 Joint Meeting of the Russian and Japanese Sections of the Com-
58 Hertzberg, M., Cashdollar, K. L., Zlochower, 1. A. and Green, G. bustion Institute. Chernogolovka, Moscow Region. 2-5 October
M. Explosives dust cloud combustion, Proc. 24th Symp. (Int.) 1993, p 188
on Combustion, The Combustion Institute, Pittsburgh, USA, 82 Khomik, S. V., Gelfand, B. E. and Knyazev, M. V. Experimental
1992,pp1837-1843 determination of a critical diameter of detonation propagation in
18 Prevention and mitigation of d lti exprosions: R.K. Eckhoff
dust suspensions, Proc. 6th Int. Coll. Dust Explosions, Shenyang, 102 Deng Xufan and He Jicheng A brief review of dust explosion
P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, reaction engineering, Proc. 6th Int. Coll. Dust Explosions, Shen-
P., pp 315-319 yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and
83 Korobeinikov, V. P. The analysis of basic parameters for deton- Wolanski, P., pp 96-115
ation of dusty gases, Proc. 5th Int. Coll. Dust Explosions, Pultusk 103 Siwek, R. Latest development in explosion protection tech-
near Warsaw, Poland, 19-22 April 1993, pp 351-364 nology, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R.
84 Markov, V. V. A new numerical method for two-phase flows, China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P.,
Proc. 5th Int. Coll. Dust Explosions, Pultusk near Warsaw, pp 35-60
Poland, 19-22 April 1993, pp 383-390 104 Eckhoff, R. K. Dust explosion hazards in the ferro-alloys indus-
85 Ding Hua and Huang Wanli On the Chapman-Jouguet condition try, Proc. 52nd Electric Fumace Conference, 13-16 November,
of dust detonation, Proc. 6th Int. Coll. Dust Explosions, Shen- bon and Steel Society, Inc., Warrendale, PA, USA, 1994
yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and 105 Gibson, N. Precautions against fires and explosions in drying
Wolanski, P., pp 341-348 operations, Paper given at the European Summer School on Dust
86 Tulis, A. J., Sumida, W. K., Heberlein, D. C., Patel, D. L. and Explosion Hazards: Their Assessment and Control, Cambridge,
Egghart, H. Detonation tube studies of particle size and RDX UK, organized by IBC Technical Services Ltd in association with
sensitization of aluminium powder-air with regard to spinning BMHB and IELG, 1993
and/or multiple-front detonations, Proc. 5th Int. Coll. Dust 106 Zockoll, C. Preventive protection against fire and explosion of
Explosions, Pultusk near Warsaw, Poland, 19-22 April 1993, organic processes by example of spray driers in the milk industry,
pp 391400 Preprints for seminar on Explosion Satety and Related Risk Con-
87 Paplinski, A. and Wlodarczyk, E. About estimation of the effec- trol, Gent, Belgium, 23-24 March 1994, organized by Techno-
tive energy for detonation initiation in combustible gaseous sus- logica1 Institute-KVIV, in cooperation with EuropEx (Kontich,
pensions, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. Belgium), pp 181-193
China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P., 107 Xu Bo-Wen, Cui Yun-Shen and Xu Wan-Qing Research on fee-
pp 320-328 ble electrical spark igniting deposited fibre dust, Proc. 5th Int.
87 Klemens, R., Teodorczyk, A., Wolanski, P. and Wolinski, M. Coll. Dust Explosions, Pultusk near Warsaw, Poland, 19-22 April
Detonation parameters of hybrid mixtures containing grain 1993, pp 59968
dusts, Proc. <suppl.) 5th Int. Coll. Dust Explosions, Pultusk near 108 Rogers, R. L. Fire, explosion and electrostatic hazards during the
Warsaw. Poland. 19-22 A~til 1993, DD5769 use of big bags for powder handling, Preprints for seminar on
88 WirknerlBott, 1.; Schumann, S. anditock, M. Dust explosion Explosion Safety and Related Risk Control, Gent, Belgium, 23-
venting: investigation of the secondary explosion, Paper pre- 24 March 1994, organized by Technological Institute-KVIV, in
sented at 7th Int. Symp. Loss Prev. & Safety Prom. Proc. Ind., cooperation with EuropEx (Kontich, Belgium), pp 169-179
Italy, 4-8 May 1992. Copy received from Battelle Europe, Frank- 109 Dahn, C. J., Reyes, B. N. and Kashani, A. Static electricity haz-
furt - Main, Germany ards of flexible intermediate bulk containers, Proc. 28th AIChE
89 Schumann, S. Wirkner-Bott, 1. Dust explosion venting: Second- Ann. Loss Prev. Symp., Session No. 12 on Electrostatic Hazards,
ary explosion for vessel volumes up to 250 m3, EuropEx News- Atlanta, USA, 17-21 April 1994
letter, Edition 22, September 1993, pp 2-5 110 Pratt, T. H. Static electricity in pneumatic transport systems.
90 van Wingerden, K. J. L.oss Prev. Proc. Ind. 1993, 6, 241-249 Three case histories, Proc. 28th AIChE Ann. Loss Prev. Symp.,
91 Gelfand, B., Medvedev, S., Polenov, A. and Bartenev, A. Com- Session No. 12 on Electrostatic Hazards, Atlanta, USA, 17-21
busrion Explosion Shock Waves 1990, 3, 85-91 Aoril 1994
92 Medvedv, S. P., Polenov, A. N. and Gelfand, B. E. Blast waves 111 Stut Keping. Investigation on statics explosion-proof of aldehyde
induced by sudden expansion of pressurized dusty systems, Proc. resin powder in pneumatic pipe, Proc. 6th Int. Coll. Dust
6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.- Explosions, Shenyang, P.R. China, 29. Aug.-2 Sept. 1994, eds
2 Sept. 1994, eds De& Xufan and Wolanski, P., pp 289-296 Deng Xufan and Wolanski, P., pp 42@425
93 Medvedev. S. P.. Polenov. A. N. and Gelfand. B. E. On the 112 Wang Mingzhong and Lou Renjie The main electrostatic hazards
amplification of blast wave transmitted through a dust deposit, and its elimination in powder pneumatic conveying system, Proc.
Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 6th Int. Coll. Dust Explosions, Shenyang, 29 Aug.-2 Sept. 1994,
Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 297- eds Denp Xufan and Wolanski, P., pp 426-433 __
301 113 Kleinschmidt, H.-P. Private communication, Fagus-GreCon
94 Smimov, N. N.. Kuksenko, B. V. and Chen Dongqing Mathemat- Greten GmbH & Co., Postfach 1243, D-W-3220 Alfeld-Hann-
ical modelling of shock induced explosion in piydispersed dust over, Germany, 1992
suspended in the oxidizer, Proc 6th Int. Coll. Dust, Explosions, 114 Janssen, L. Private communication, Firefly AB, Huddinge,
Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan Sweden, 1993
and Wolanski, P., pp 306-308 115 T&B Electronic Spark detection and extinguishing systems.
95 Boiko, V. M., Kiselyov, V. P., Kiselyov, S. P. et al. On gas Detailed technical description of system features and appli-
parameter profiles at-non-stationary shck wave interaction with cations, T&B Electronic GmbH, 31061 Alfeld, Germany, 1994
dust cloud. Proc. 6th Int. Coll. Dust Exolosions, Shenvang, P.R. 116 Lloyd, F. C. Electrical equipment in plant handling combustible
China, 29 ug.-2nd Sept. 1994, eds Deng Xufan andwoianski, powder, Paper given at the European Summer School on Dust
P., pp 336-340 Explosion Hazards: Their Assessment and Control, Cambridge,
96 Gelfand, B. E., Khomik, S. V. and Polenov, A. Quenching of UK, organized by IBC Technical Services Ltd in association with
shock waves in dusty medium, Proc. 6th Int. Coll. Dust BMHB and IELG, 1993
Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds 117 van Laar, G. F. M.Area classification for dust explosion hazard-
Deng Xufan and Wolanski, P., pp 478483 ous environments, Preprints for seminar on Explosion Safety
97 Harmanny, A. Explosion effects, Proc. 1st World Seminar on and Related Risk Control, Gent, Belgium, 23-24 March 1994.
the Explosion Phenomenon and on the Application of Explosion organized by Technological Institute-KVIV, in cooperation with
Protection Techniques in Practice. Arranged by EuropEx, Brus- EurooEx (Kontich. Beleium), po 127-141
sels, 17-21 February 1992 118 Wol&ski,M. and Hayashi, T.Explosibility of powders of rare
98 Mercx, W. P. M. Critical appraisal of damage criteria for build- earth metal alloys, Proc. 5th Int. Coll. Dust Exolosions, Pultusk
ings and process equipment, Proc. 1st World Seminar on the near Warsaw, Poland, 19-22 April 1993, pp 225-236
Explosion Phenomenon and on the Application of Explosion Pro- 119 Krause, U., Weinert, D. and Wohm, P. Diagrams for the determi-
tection Techniques in Practice. Arranged by EuropEx, Brussels, nation of the limiting oxygen concentration of dust/air mixtures,
17-21 February 1992 Proc. 5th Int. Coll. Dust Explosions, Pultusk near Warsaw,
99 LAbb, R. J. Explosion effects on people, Proc. 1st World Sem- Poland, 19-22 April 1993, pp 257-266
inar on the Explosion Phenomenon and on the Application of 120 Maddison, N. Avoidance of flammable atmosoheres: use of inert
Explosion Protection Techniques in Practice. Arranged by Euro- gas in powder handling plant, Paper given at ihe European Sum-
pEx, Brussels, 17-21 February 1992 mer School on Dust Explosion Hazards: Their Assessment and
100 Britan, A. B., Levin, V. A., Mitichkin, S. Y. et al. Investigation Control, Cambridge, UK, organized by IBC Technical Services
of relaxation zone under interaction of shock wave with protective Ltd in association with BMHB and IELG, 1993 .__
aqueous foam layer, Proc. 6th Int. Coll. Dust Explosions, Sheny- 121 Gao Guangchun, Chen Zhandong, Tang Qinghua and Luo Ruquan
ang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wol- New technique and equipment for fire and explosion-protection
anski, P., pp 36&365 of blast fumace bituminous injection, Proc. 6th Int. Coll. Dust
101 Britan, A. B., Levin, V. A., Mitichkin, S. Y. et al. Influence of Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, eds
aqueous foam screen upon propagation of shock waves, Proc. Deng Xufan and Wolanski, P., pp 407-411
6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.- 122 Wang Junyi, Zhang Jingjie and Liu Jingxian Prevention of coal
2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 472477 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 of dust explosions: R.K. Eckhoff
168 Wagner, K. A case study of applied risk analvsis in dust mum explosive concentration of coal dusts, Proc. 6th Int. Coll.
explosion hazardous envirnments; Preprints for - seminar on Dust Explosions, Shenvana. P.R. China. 29 Aua-2 Seot. 1994.
Explosion Safetv and Related Risk Control. Gent. Belaium. 23- eds Deng Xufan and Wolanski, P., pp 227-236 1
24 March 1994,-organized by Technological Insbtute-iVV:m 180 Xu Tianrui, Ding Banqing, Gong Youcheng and Wang Xingqun
cooperation with EuropEx (Kontich, Belgium), pp 155-167 Thoughts on some dust explosibility test methods in 20 1appar-
169 Mittal, M. Dust explosion risk analysis of pneumatic transport atus, Proc. 6th Int. Coll. Dust Exnlosions. Shenvang. 29 Auz.-
systems, Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. 2 Sept. 1994, eds Deng Xufan and Wolanski, P.,*pp278-283
China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski, P., 181 Pu, Y. K., Jarosinski, J., Johnson, V. G. and Kauffman, C. W.
pp 381-390 Turbulente effects on dust explosions in the 20-litre spherical
170 Lin Jingxian and Wang Junyi The DSS to prevent and control vessel, Proc. 23rd Symp. (Int.) on Combustion, The Combustion
coal dust explosion, Proc. 6th Int. Coll. Dust Explosions, Shen- Institute, Pittsburgh, PA, USA, pp 843-849
yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and 182 Wang Baoming and Zhang Jinhn Investigation on the process of
Wolanski, P., pp 434440 TNT dust explosion and explosion vent systems of TNT pro-
171 Moore, P. E. and Frehill, R. M. Dust explosion protection - the duction process, Proc. 6th Int. Coll. Dust Explosions, Shenyang,
choices, Proc. 6th Int. Coll. Dust Exposions, Shenyang, P.R. P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and Wolanski,
China, 29 Aua.-2 Sent. 1994. eds Dena Xufan and Wolanski. P.. P., pp 391-402
pp453471 - - 183 von Haefen, E. and Schecker, H.-G. DUSTEXPERT - an expert
172 Liu Wenxin, Sheny Zhongquan, Deng Xufan et al. Measurement system for the assessment of explosion hazards and the selection
of turbulente in the Hartmann bomb, Proc. 6th Int. Coll. Dust of explosion protection methods for dust handling plants, Proc.
Exnlosions, Shenvane, P.R. China. 29 Aua.-2 Sent. 1994. eds 5th Int. Coll. Dust Explosions, Puhusk near Warsaw, Poland,
Ding Xufan and Fvoanski, P., pp 81-95 - * 19-22 April 1993, pp 487496
173 Siwek, R. and Cesana, C. Ignition behaviour of combustible 184 Wach, J. Expert system for designing protection measures against
dusts, Proc. 28th AIChE AnnLoss Prev. Symp., Session No. 12 accidental dust explosions, Proc. 5th Int. Coll. Dust Explosions,
on Electrostatic Hazards. Atlanta. USA. 17-21 Aoril 1994 Pultusk near Warsaw, Poland, 19-22 April 1993, pp 497-502
174 Zhou Benmou, Tan Fenggui, Yu Xuesheng et al. A new type of 185 Tyldesley, A. Dust-Expert - Much more than just a programme
movable electrode electrostatic ignition energy apparatus, Proc. for running the vent size calculations, Paper given at the Euro-
6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29 Aug.- pean Summer School on Dust Explosion Hazards: Their Assess-
2 Sept. 1994, eds Deng Xufan and Wolanski, P., pp 257-262 ment and Control, Cambridge, UK, organized by IBC Technical
175 Cashdollar, K. L., Weiss, E. S., Greninger, N. B. and Chatrathi, Services Ltd in association with BMHB and IELG. 1993
K. Plant/Operations Progr. 1992, 11, 247-255 186 Femando, D. Dust explosion hazards: the human element, Paper
176 Cashdollar, K. L. and Chatrathi, K. Combus. Sci. Technol. 1992, given at the European Summer School on Dust Explosion Haz-
87, 157-171 ards: Their Assessment and Control, Cambridge, IJ-K, organized
177 Zhou Congzhang, Zhang Ruiping and Yu Yongfang On the cri- bv IBC Technical Services Ltd in association with BMHB and
terion of lower explosion limit of dust cloud, Proc. 6th Int. Coll. ILG, 1993
Dust Explosions, Shenyang, P.R. China, 29 Aug.-2 Sept. 1994, 187 Middleton, P. G. Private communication, British Materials Hand-
eds Dene Xufan and Wolanski. P.. OD220-226 ling Board, Index House, Ascot, Berkshire SL5 7EU, UK, 1992
178 Matsuda: T. and Itagaki, H. ffectsof ignition energy on dust 188 Gibson, N. A review of dust explosion research projects for
explosion parameters, Proc. 6th Int. Coll. Dust Explosions, Shen- BMHB research project panel. An interim report, November
yang, P.R. China, 29 Aug.-2 Sept. 1994, eds Deng Xufan and 1991
Wolanski, P., pp 245-256 189 Wang Dongyan Hazards and control countermeasures in China,
179 Tian Renqu, He Chaoyuan and Zhang Yinghe Effect of different Proc. 6th Int. Coll. Dust Explosions, Shenyang, P.R. China, 29
ignition energies and the existente of methane gas on the mini- 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
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 The substances are tested using an unambiguously
imperative to know the dangerous properties of the pro- defined sample following a preparation tailored to the
duct. The reliable way to characterize the combustible requirements of the selected test method. This usually
and explosive properties of a product is to subject a sam- includes careful drying of the product (e.g. 50C in vac-
ple of the product to various tests and classify the results uum or 75C at standard pressure) and preparation of a
in accordance with the technical safety characteristics. specified particle size fraction of < 250 Pm for the dust
The determination of these characteristics of combust- layer or < 63 Pm for suspended dus?. Tests are perfor-
ible dusts is often complicated, even more complicated med under initial atmospheric conditions.
than with gases or vapours, because, in contrast to gases
and vapours dusts are particulate solids and therefore Dust layers
have additional characteristics: particle size distribution,
ageing degree, humidity and density. These character- Ignitability
istics can influence the reaction behaviour and therefore Dust layers are recognized as combustible dusts if they
complicate the classification. can be ignited with a foreign ignition source and the
In order to get a sufficient genera1 view of the local fire wil1 propagate sufficiently after the outside
danger of a product, it is necessary to carry out a large source is taken away.
number of tests. The mosaic-like nature of these tests For determination of the genera1 ignition and com-
results in a more or less extensive picture of a given bustion behaviour of dust layers, dust deposits are
problem. exposed in pre-tests to the effect of different ignition
The extent of the investigations is determined by sources. These sources play an important part in indus-
the problem. This depends on the goal of the protection trial incidents.
system, i.e. the chosen protection measures, whether The ignition sources include: sparks, mechanica1
against the beginning of combustion or explosion or sparks (cerium iron sparks), combustible matches, glow-
against the explosion effects. This decision is made ing cigarettes and gas flames (Figure 1).
cooperatively between the tester and the user. In the fol-
lowing work, some chosen test methods and their Burning behaviour (burning test)
resulting safety technical characteristics are described, This test involves investigation of whether and to what
with their effective magnitudes (if they are sufficiently extent a fire initiated by an external ignition source can
known and relevant for the particular practice). Th propagate in a dust deposit of width ca. 2 cm and length
analysis wil1 distinguish between dust layers and 4 cm (Figure 2, left). The test is carried out at room
dust suspensions. temperature (20C) and at elevated product temperature
21
22 Hazard evaluation of dusts: R. Siwek
Figure 2 Test apparatus to determine the burning behaviour (left) and the burning velocity (right)
Hazard evaluation of dusts: R. Siwek 23
Relative decomposition temperature, RDT (test for Decomposition temperature, DT (heat accumulation
exothermic reaction in an open cup) experiments in a Dewar jlask)
This is defined as the lowest temperature at which the The substance under test is stored in a Dewar flask (heat
temperature of the test substance starts to exceed the accumulation tests) together with a reference substance
temperature of a thermally stable reference substance in a thermostated box or drying oven at a constant tem-
when both are heated simultaneously with exclusion of perature of the surroundings. Usually, Dewar flasks of
air. 200-2000 ml content are used. Self-heating of the pro-
In addition to differential thermal analysis (DTA), duct is shown by the temperature of the test substance
the method described by Ltolf3 has wel1 proved its exceeding that of the reference substance. The test is
worth as it allows not only determination of the relative repeated with fresh samples and a temperature decrease
decomposition temperature but also assessment of the of 10 K each time until exothermic behaviour can no
amount of gases released in the exothermic reaction and longer be observed (AT < 5C) over a period of at least
whether they are flammable (Figure 3). It is used for 72 h. As the determined induction time of the decompo-
products or reaction masses which are exposed to an sition (time between attainment of the oven temperature
elevated temperature in part in the absente of fresh air by the sample and the maximum sample temperature)
(e.g. in vacuum ovens, fluidized-bed dryers). is generally linearly dependent on the reciprocal of the
An approximately 2 g sample and a reference subst- absolute temperature, a reliable storage temperature for
ante (graphite) are heated continuously up to 350C at a specified storage time can be determined.
a rate of 2.5 K min-. If an exothermic reaction occurs,
a fresh sample is tested at constant temperature. The test Spontaneous decomposability, SD
is repeated, each time with fresh sample, at decreasing A substance is regarded as spontaneously decomposable
temperature intervals of 10 K until the substance no if it undergoes complete decomposition with a progress-
longer exhibits exothermic behaviour over a period of at ive reaction zone after local initiation by a sufficiently
least 8 h. powerful ignition source, also without the presence of
The relative decomposition temperature by the (atmospheric) oxygen3.
The test substance (5-10 ml) is placed in a crys-
Ltolf method is the temperature at which, in the exper-
tallizing dish in a Witt jar (V= ca. 1 litre) and ignited
iment at a constant oven temperature, the substance just
with the glowing, integral platinum wire (Figure 4). If
fails to show an increase in temperature compared with
the sample burns, the jar is evacuated. If the product
the reference substance over an 8 h period.
continues to react in a vacuum, it is said to undergo
The results of these investigations can also be use-
spontaneous decomposition. The test is repeated, but this
ful in an assessment of the consequences of decompo-
time ignition is not initiated until the vessel has been
sition reactions in installations as they allow estimation
evacuated. If the product reacts in vacuum and
of the possible formation of readily ignitable and very
undergoes a propagating decomposition reaction, spon-
violently reacting hybrid mixtures.
taneous decomposition is said to have occurred. The test
For reasons of safety, at least 50C are subtracted
is conducted at room temperature (20C) and at an elev-
from the experimentally determined relative decompo-
ated product temperature (e.g. 100C).
sition temperature and the temperature so obtained is
A spontaneous decomposition can be initiated by,
specified as the maximum admissible medium heating for instance, a heated foreign body. However, it can also
temperature. For a more accurate determination of the result from a fire initiated locally in which the fire
admissible temperature exposure, heat accumulation changes into a spontaneous decomposition owing to an
tests must be performed in a Dewar vessel at lower tem- insufficient supply of air. A spontaneous decomposition
peratures if the measured relative decomposition tem- can be neither prevented by inerting nor stopped by
perature is less than 200C. 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
to recorder
to qas meter+
3 ,Nz/ventilation
evacuoted
glans pot
I-e
--r
vacuum
IWittscher pot)
t-
platinum wire
id=
(-1000C 1
test sample
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
Table 1
-KV
Relation between K,,,,, and dust explosion class St
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
8
1
a6
-4
E
2
0 -
0 2% soo 7iitl IOW 1FBO
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)
Figure 13 MIKE 3 apparatus for determination of the minimum ignition energy (MIE) of dlxts
28 Hazard evaluation of dusts: R. Siwek
Hybrid mixtures
Products which contain flammable solvents can form
MIE with L [ mJ ]
hybrid mixtures. These are easier to ignite and explode
Figure16 Influence of the inductance L on the minimum with greater severity than solvent-free dust-air mixtures.
ignition energy (MIE)
This increased hazard can become important when the
solvent vapour concentration exceeds 20% of the lower
Inductance L causes the capacitor to discharge in a explosion limit. With products which contain no more
protracted manner. As Figure 16 shows, dust-air mix- than 0.5 wt% flammable solvents, this is not anticipated.
tures are more readily ignited by time-extended dis- Dusts with a higher solvent content require detailed
charges than by purely capacitive discharges8~o-2. investigation3. The special features of hybrid mixtures
If the MIE is used for assessment of ignition haz- are that
ards resulting from electrostatic charges, it must be Non-explosible dust-air mixtures and non-explosible
determined without an inductance in the discharge cir- flammable gas (solvent vapour)-air mixtures can
cuit. Only in this manner is a realistic assessment of the form explosible hybrid mixtures. The LEL of the
incendivity of electrostatic discharges towards dust-air hybrid mixture lies between the LEL of the combust-
mixtures possibles*o-2. ible dust and the LEL of the flammable gas, e.g. pro-
pane (Figure 18).
Minimum ignition temperature, MIT (ignition test for While addition of flammable gas to the combustible
airborne dwt) atmosphere raises P,,, to some extent (Figure 19,
The minimum ignition temperature, MIT, is defined as left), the explosion constant K,,,,, of the combustible
the lowest temperature of a heated surface at which the dust is increased considerably, even though its con-
most readily ignitable mixture of a dust with air just centration is below the LEL. When the optimum gas
ignites. It provides information on the ignition behaviour concentration is present, al1 dusts in question wil1 be
of a dust suspension when quickly passing over a hot classified into the dust explosion class St 3 (Figure
surface such as in spray and fluid-bed dryers,. 19, right). The K,,, of the hybrid mixtures lies
In Europe, the MIT is usually determined in the between the K,,,,, of the combustible dust and the
BAM furnace2,. This apparatus is enjoying increasing K max of the flammable gas, e.g. propane (turbulent
popularity (Figure 17). condition).
The BAM furnace is heated to a maximum tem- 3. The MIE and the LOC of the hybrid mixture are
perature of ca. 600C. The furnace temperature is determined by the combustible substance with the
allowed to fa11 and a dust sample (1-2 ml) is blown lowest limit value (Figures 20 and 21).
against the test plate axially at intervals of 50 K.
This is repeated until ignition is no longer found. Influencing factors
The furnace is then reheated to the next higher tempera-
ture stage. The test is repeated, again with decreasing Dust accumulation
furnace temperature, but this time at temperature inter- The self-ignition temperature of a specific dust not only
vals of 10 K until no ignition (flame appearance, audible depends on the type of the dust (chemical nature, particle
reaction) is found in at least three experiments. The size, moisture content, etc.) but also on the size of the
appearance of glowing dust particles alone or ignition pile (Figure 22).
later than 5 s after dispersion of the dust is ignored.
Heating
Test plak
40
0
0 0.3 0.6 0.8 1.2 1.5
Figure 17 BAM furnace Figure 16 Lower explosion limit (LEL) of hybrid mixtures
Hazard evaluation of dusts: R. Siwek 29
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
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)
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
10
0
0 0.2 0.4 08 a0 1 1.2 1.4 1.6 3.8 2
inMalpmumPI [bar,abe.l
To assess the safety of a unit operation, the test 12 Siwek, R. and Cesana, Ch. Assessment of the Fire & Explosion
Hazard of Combustible Products for Unit Ouerations. Proceedinns
laboratory must know not only the content of possible of the International Conference and Exhibibon on Safety, Healyh
solvents, but also the usual product properties such as and Loss Prevention in the Oil, Chemical and Process Industries,
empirical formula, melting point and toxicity. Singapore, 1993, Butterworth-Heinemann, London
13 ISO/DIS 61840: Explosion protection systems - Part 1: Determi-
A test must be repeated if a change in the hazard nation of explosion indices of combustible dusts in air, Inter-
situation is anticipated, for instance as a result of national Organization for Standardization, 1985
a change in the production process, including work- 14 ISO/DIS 6184/2: Explosion protection systems - Part 2: Determi-
nation of explosion indices of combustible gases in air, Inter-
UP, national Organization for Standardization, 1985
a change in the specifications or origin of starting 15 ISO/DIS 6184/3: Explosion protection systems - Part 3: Determi-
materials and additives, nation of explosion indices of fuel/air mixtures other than dust/air
and gas/air mixtures, International Organization for Standardiz-
irregularities in the production, or ation, 1985
production relocation. 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
References Annual Loss Prevention Symposium, Atlanta, Georgia, USA, 17-
21 April 1994
1 VDI Guideline 2263: Dust Fires and Dust Explosions, Hazards- 18 Siwek, R. Latest development in explosion protection technology,
Assessment-Protective Measures, Beuth, Berlin und Kln, May Proceedings of the 6th International Colloquium of Dust
1992 Explosions, Shenyang, PR China, 29 Aug.-2 Sept. 1994
2 VDI Guideline 2263, Part 1: Dust Fires and Dust Explosions, Haz- 19 Siwek, R. and Cesana, Ch. Ignition behavior of dusts: meaning and
ards-Assessment-Protective Measures, Test Methods for the interpretation. Process Safety Progress, April 1995
Determination of the Safety Characteristics of Dusts, Beuth, Berlin, 20 Bartknecht, W. and Siwek, R. Meaning of Ignition Energy and
May 1990 Minimum Ignition Temperature for the Assessment of the Ignit-
3 Bartknecht, W. Explosion Protection, Basics and Application, ability of Ignition Sources in Dust-Air Mixtures, Part 1: Minimum
Springer, Berlin, 1993 (in German) Ignition Energy and Minimum Ignition Temperature, Staub-Rein-
4 Guideline 84/449/EC. OfJicial J. Eur. Communiry, L251, 27 A.10 haltung der Luft 54 (1994) Springer, Berlin, 1994 (in German),
1984 pp 325-330
5 UN Recommendations on the Transport of Dangerous Goods, 21 Bartknecht, W. and Siwek, R. Meaning of Ignition Energy and
1986, Chapter 14, p 3 Minimum Ignition Temperature for the Assessment of the Ignit-
6 Bowes, P. C. Self-heating: Evaluating and Controlling the Haz- ability of Ignition Sources in Dust/Air-Mixtures, Part 2: Assess-
ards, Her Majestys Statinery Office, London, 1984 - ment of the ignition efficacy, Staub-Reinhaltung der Luft 54
7 Bartknecht. W. Dust Exulosions. Course Prevention. Protection. (1994) Springer, Berlin, 1994 (in German), pp 325-330
Springer, erlin, 1989 L 22 IEC. Electrical apparatus for use in the presence of combustible
8 Siwek, R. Reliable Determination of the Safety Characteristics in dust, Part 2, Test methods, Section 3: Method for determining mini-
the 20-1 Apparatus, Proceedings of the Conference Flammable mum ignition energy of dustfair mixtures (draft), 1992
Dust Exolosion. St. Louis. Missouri. USA. 2-4 November 1988 23 Cesana, Ch. and Siwek, R. MIKE 3 Minimum Ignition Apparatus,
9 Cesana, Ch. and Siwek, R. KSEP 332 - Measurement and Control Khner, Birsfelden, 1992
System for the 20-1 Apparatus, Khner, Birsfelden, 1991 24 IEC. Electrical apparatus for use in the presence of ignitable dust,
10 Siwek, R. Explosion Indices, Khner, Birsfelden, 1991 Part 2, Test methods, Sheet 2-2: Method for determining the mini-
11 Cesana, Ch. and Siwek, R. Screening Tests: Test Procedures, mum ignition temperature of dusts - Dust cloud in a furnace at
Khner, Birsfelden, 1991 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
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.
More than 70% of dusts or powders processed in indus- However, this guidance usually applies to isolated
try are combustible, and a dust explosion hazard may vessels, whereas, in industry, plant items are connected
arise in any plant handling these dusts. It has been esti- by pipelines or conveyors and an explosion may propa-
mated that there is one dust explosion on average every- gate from the first vessel to the second causing a second-
day in Europe, in the wood, food, chemicals, fuel and ary explosion. The conditions under which this second-
pharmaceuticals industries. Recent data from the USA ary explosion takes place are generally unknown, but
for grain dust explosions show 13 in 1993, and 154 from there is enough evidente from the literature on gas
1984 onwards. The relative frequency of incidents in explosions to show that in some circumstances there can
each industry depends very much upon the source of be a significant increase in the rate of combustion and
statistical data3. violente of the explosion, and a significant increase in
Dust explosions cause a rapid increase in pressure the explosion pressures7-9. Similar effects have also been
when confined in vessels or other process equipment. noted with dust explosionstOst .
Such pressures cannot be contained by most plant and The reasons for this increase in the rate of combus-
measures have to be taken either to prevent the explosion tion are the generation of turbulente as the explosion
or to protect the plant against the destructive effects. passes through the interconnecting pipe and the entry
A number of systematic studies of dust explosions into the second vessel of a relatively large jet flame
have been carried out and the results used to produce ignition source. If protection techniques such as venting
guidance on the prevention and mitigation of dust or suppression are applied to these secondary explosions,
explosion hazards. This guidance is published as VDI the degree of protection calculated on the basis of a
36734, NFPA 6gs and the IChemE guide? Dust single, isolated vessel may be insufficient to cape with
Explosion Prevention and Protection Parts 1 to 3. a more violent secondary explosion.
91
92 Dust explosions in interconnected vented vessels: P. Holbrow et al.
If dust-handling plant does consist of several con- 150 and 250 mm diameter pipes connected to a 1 m3
nected items of equipment, consideration must be given vessel15. Wel1 demarcated, symmetrical zones of lean
to isolating the various units to prevent transmission of and rich coal dust were observed. Upon arrival of the
flame or other ignition sources, to strengthening the plant flame front, the flame intensity in the rich zones was
to cape with high pressures and to increasing the leve1 greater. The effect seemed to be closely related to the
of explosion protection. However, there is little work on flow conditions in the pipe, and hence the explosion
the pressure changes that can occur, and guidance for severity in the vessel. Pineau and Ronchail15 have
the protection of such systems - apart from the instal- assumed that the structure of the dust cloud and the
lation of isolation devices - is sparse. flame just at the end of the pipe is representative of that
The British Materials Handling Board (BMBH) has inside the pipe, so that the extent to which the explosion
sponsored a project with the Health and Safety Executive propagates down the pipe is dependent upon the local
to investigate the behaviour of dust explosions in sys- dust concentrations within it.
tems of connected vessels. The main purpose of the pro- When an explosion in a vessel enters a pipe, the
ject has been to measure the pressure changes that occur change in geometry causes a rapid acceleration of the
when a dust explosion is ignited in one vessel and the flame front16. Depending on the length and diameter of
flame transmits into the other. The aim is to provide the pipe, this flame front may lose its momentum and
industry with guidance as to when, and to what degree, peter out, or may continue to accelerate, possibly to the
additional venting is required on linked vessel systems point of transition to detonation. Where heat losses are
and whether alternative precautions are required. great, e.g. when the flame propagates along a narrow
pipe, the flame may eventually be quenched, and for a
Literature review given dust and experimental conditions a critical quench-
There are a number of accepted methods for calculating ing diameter exists .* . Where the pipe is above this
the area of vent required for effective and safe explosion critical diameter, the flame wil1 accelerate due to wal1
venting4-6. friction turbulence19 and may reach a steady velocity of
Generally, the wel1 known calculation methods the order of the speed of sound in the combustion pro-
such as the K,, nomograph approach produce vent areas ducts. If the pipe is long enough, deflagration to deton-
on the conservative side, i.e. the vent area is larger than ation transition (DDT) may occur. Changes in pipe diam-
is strictly necessary to protect the plant from the real eter along its length promote flame acceleration as do
explosions likely to occur inside it. Nevertheless, there bends which allow pressure piling.
are conditions and circumstances in which intense Oscillatory flame behaviour is reported under cer-
explosions could occur and with which the normal tain conditions for gas mixtures in pipes. These oscil-
degree of venting could not cape. lations of the flame front are triggered by the first contact
The intensity of a dust cloud explosion is affected between the flame and the pipe walls, when rapid cool-
by the turbulente. The more turbulente, the more rapid ing takes place, setting up hydrodynamic instabilities in
the combustion rate, and the more violent the explosion. the flame. Changes in the burning rate of the flame front,
This .becomes particularly apparent when explosions flame-front instability and longitudinal acoustic vibra-
propagate along pipes and ducts where the confinement tions al1 influence oscillatory combustion. Oscillatory
greatly enhances turbulente in the dust cloud ahead of combustion has also been reported for vented gas
the flame and within the flame front itself. This wil1 explosions from single vessels, particularly when the
increase the rate of pressure rise, and therefore more vent area is large *J* . Harrison and Eyre* found that the
venting is required to protect the plant. For dust oscillatory flames always began when the pressure of the
explosions in pipes connected to vessels, whilst the external explosion peaked, often with a consequent rise
flame is stil1 within the vessel the pressure gradually in pressure within the vessel (the external explosion
increases in the pipe as dust-laden air is pushed into it. invariably peaked before that inside the vessel).
Any dust previously deposited in the pipe wil1 be raised Studies of gas explosions have been carried out in
into suspension. Once the flame enters the pipe, the rate interconnected totally enclosed vessels7-9,23. Where the
of pressure rise wil1 be faster than that in the vessel13. vessels are of the same volume, not only can the
This could be especially important where two ves- maximum explosion pressure in the second vessel be
sels are interconnected. An explosion in the first vessel, higher than that in the first, but the rate of pressure rise
by driving material through the interconnecting pipe, in both vessels can be significantly increased. This is
wil1 not only enhance the turbulent conditions in the particularly apparent in the second vessel. The diameter
second, but produce a large energetic jet flame ignition and to a lesser degree the length of the interconnecting
source there. pipe were the important variables16.
However, turbulent flow in dust-air mixtures is If vessels of different sizes were used, the
likely to generate strong concentration gradients, giving maximum pressures and rates of pressure rise depended
rise to rich and lean zones. These may have considerable on the vessel in which the primary ignition took place.
influences on the propagation mechanisms of turbulent When this vessel was smaller than the second, the values
dust explosions. The dust particles cannot always follow of the maximum rate of pressure rise were similar in
the streamlines of a turbulent flow, especially in eddies, each vessel and increased as pipe diameter increased.
even at low eddy velocities. The higher the Reynolds The maximum explosion pressure also increased in both
number and turbulent flow velocity, the more pro- vessels as pipe diameters increased, but for smal1 pipe
nounced this effect becomes14. This effect has been seen diameters this pressure was lower in the second vessel
in pulverized coal dust clouds emerging from 20 m long 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
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.
r - Pressure Pressure
Heasurement Measurement
1
1
2nd dispersers
measurement
Ignition
position
J I
1 Ven
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
, , 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
I , I I I
200 400 600 800 1000
Time (msec)
Pressure - Time Traces
could be attributed to experimental scatter. The impli- be below unity, any value below 1 has been included in
cation is that the explosion venting requirements for Table 6 as unity.
interconnected systems generally exceed those required
for single vessels. Effect of pipe length. Pipe length had an influence on
Tuble 4 shows in more detail the extent to which the the degree of pressure enhancement and the maximum
explosion pressure enhancement for the interconnected reduced pressure. The 5 m long pipe produced the high-
vessel system increases the venting requirements for est enhancement in most cases. However, this was not
reduced explosion pressures of 0.2, 0.5, 1.0 and 1.5 bar. always the case; 10 and 15 m pipes produced high
Experimental data from the tests were used and the ratio enhanced pressures in some tests. It was not possible to
of the experimental to predicted areas, A,,,#AKS,, was determine a single optimum pipe length for enhancing
calculated. This ratio indicates the extent to which the the explosion violente.
interconnected system required increased venting com- The number of transmissions were few when ves-
pared with predictions by the single-vessel K,, nomo- sels were connected by the 0.15 m diameter pipe and no
graph method. As Pred increases, there is a progressive transmissions were achieved with the 0.15 m diameter
increase in the ratio A,,,~AvKs,. The predicted values of X 15 m pipe with any of the vessels or explosible dusts.
AVIV for single vessels are shown in Table 5. Pipe length had an effect on the pressure profile and
Table 6 shows in summary the enhancement factor the peak pressures recorded. The longer the pipe length,
for various configurations of vessel volume and pipe the greater the separation of the pressure peaks.
diameter. For each configuration, the worst-case result The effect of backflow into Vessel 1 decreased with
has been used for the tests with different pipe length. increasing pipe length. There are two likely reasons for
Because in practice this factor would not be allowed to 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
1.80-y I I I I I
0 0.399 0.799 1.198 1.597 1.996
sec
VESSEL 2
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
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 Effect of pipe diameter. The tests involving the 0.5 m
time between primary ignition and arrival of the diameter pipe generally produced the highest pressure
returning pressure pulse at Vessel 1 increases, the less enhancement, and, as the pipe diameter was decreased,
likely it is that this backflow wil1 act to increase the rate mis enhancement decreased also. This is demonstrated
of combustion in the primary explosion. When pipes are in Table 4 when similar vessel configurations but with
short, the back flow can occur before the primary different pipe diameters are compared. Figures 13 and
explosion is completed and influence the final stages of 14 demonstrate the lower explosion pressure that arises
combustion in Vessel 1. 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
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
,&,
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
-,
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
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 virtually al1 the configurations for the coal dust
4 show that the A,/AvKst ratio is generally greater when explosion tests.
the explosion is ignited in the larger vessel of a particular The scale effect was tested at a volume ratio of 1: 1
configuration. As an example, the 20-2 m3 configuration with 2 m3 vessels and 20 m3 vessels using coal dust. The
with a 0.5 m diameter connecting pipe generated the results are shown in Figures 15 and 16. The pressures
highest absolute pressures and pressure enhancement of 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
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
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. -
Dust Pred = 0.2 bar P re3 = 0.5 bar Pred = 1 bar Pred = 1.5 bar
Table 6 Vent areas required for interconnected systems (constant X K., value)
Vessels Volume ratio Pred 0.2 bar Pred 0.5 bar Pred 1.0 bar Pred 1.5 bar
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
(iii) the dust explosibility, 9 Singh, J. Gas explosions in interconnected vessels: pressure piling.
Presented at.iChem E Hazards X11 Svmoosium.
, I 19-21 A~til
. 1994.
(iv) the vent areas themselves. UMIST, M&chester, UK
10 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,
Under some circumstances, when the connecting pipe Brussels, 1992
is relatively long and has a smal1 diameter, and the 11 Van Wingerden, K. and Alfert, F. Dust Explosion Propagation in
degree of interconnectedness is low, the K,, nomo- Cannected Vessels. VDI Betjchte, Nr. 975, -5@,1992
12 Gardner, B. R., WintexR. J. and Moore, M. J. Explosion develop-
graph approach for vent sizing of single vessels can ment and deflagration-to-detonation transition in coal dust/air sus-
be applied to the vessel in which the initial ignition pensions. 2lst Symp. Int. on Combustion, The Combustion Insti-
occurs. When the degree of interconnectedness is tute, 1986, pp 355-343
13 Pineau, J. P. Dust Explnsions in Vessels Connected to Ducts. CpI-
high, both vessels wil1 require the vent areas to be Berichte Nr 494; 1984
increased compared to the nomograph approach. 14 Bond, F., Fresko, M., Knystautus, R. and Lee, J. H. S. Influence
Pressure enhancements in linked vented vessel sys- of turbulente on dust explosions. 1st Int. Colloq. on Explosibility
of Industrial Dusts, 1984, pp 37-50
tem are generally low when the reduced explosion Pineau, J. P. and Ronchail, G. Propagation of Coal Dust
,l5
pressure does not exceed 0.2 bar g. 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
16 Bouhard, F., Veyssiere, B., Leyer, J.-C. and Chaineaux, J,
References Explosion in a vented vessel connected to a duct.Prog. Astronaut
Aekonaut 134. AIAA A. L. Kuhl et al. Eds Dynamics of Deton-
1 Gibson, N. Dust explosion hazards: the problem and the options ations and Exolosions: Exolosion Phenomena. 12th ICDERS. Ann
for safety, European Summer School on Dust Explosion Hazards: Arbor. MI, U:S.A, July 1989
Their Assessment and Control, 21-25 September 1992, Cam- 17 Proust, C. Dubt explosions in pipelines. EUROPEX European Sum-
bridge, UK mer School on Dust Explosions Hazards - Their Prevention and
2 Schoeff, R. W. News release and summary for 1993. Dust Control, Cambridge, UK, September 1992
exulosion statistics - USA. Eurooex Newslett. March 1994. 24 18 Zhang, F. and Gronig, H. Transition to detonation in torn starch
3 Proust, Ch. and Pineau, J. P. Dusiexplosions: risk assessment. Dust dust oxygen and air mixtures. Combust. Flame 1991, 86, 21-32
Explosion Protection, Europex Int. Symp, Antwerp, 1989 19 Eckhoff, R. K. Dust Exptosions in the Process Industries. But-
4 VDI Guideline 3673, Pressure lease of dust explosions. VDI- terworth Heinemann, Oxford, 199 1
Commission Reinhaltung der Luft. Beuth-Verlag, W. Germany, 20 Phylaktou, H. N., Andrews, G. E. and Herath, P. Fast flame speeds
1984 and rates of pressure rise in the initial period of gas explosions in
5 National Fire Protection Association Guide for venting of defla- large L/D cylindrical vessels. J. Loss Prev. Process Ind. 1990, 3
grations. NFPA 68-1988. Quincy, MA, USA, 1988 21 McCann, D. P. J., Thomas, G. 0. and Edwards, D. H. Gas dynam-
6 Lunn, G. Dust Explosion Prevention and Protection. Part 1 - Vent- ics of vented explosions. Part 1: experimental studies. Combusf.
ing. 2nd edn. IChemE, Rugby, UK, 1992 Flame 1985. 59, 233-250
7 Abdullin, R. K., Babkin, V. S. and Senachin, P. K. Combustion of 22 Harrison, A. J. and Eyre, J. A. External explosions as a result of
gas in connected vessels. Comb. Expl Shock Waves 1988, 24, 123 explosion venting. Combust. Sci. Technol. 1987. 52, 91-106
8 Phylaktou, H. and Andrews, G. E. Gas explosions in linked vessels. 23 Bartknecht, W. Explosions. Course, Prevention, Protection.
Presented at 13th ICDERS Meeting, Japan, 1991 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
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
1.20
1.15
2
..
.j
pI 1.10
.s
2
1.os
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.
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.
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
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.
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%).
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.
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.
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
Rolf K. Eckhoff
University of Bergen and Christian Michelsen Research AS, Fantoftveien 38,
IV-5036 Fantoft-Bergen (Norway)
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
Dust cloud formation processes Dust cloud ignition processes Flame propagation processes in Blast waves generated by
dust clouds burning dust clouds
l Inter-particle forces l General ignition theory. 0 Single-particle ignition l Blast wave properties as
in dust deposits Ignition of single and combustion in a a function of properties
(cohesion). particles and clouds. flame front. of burning dust clouds.
l Entrainment of particles from l Ignition by: 0 Laminar and turbulent l Effects of blast waves
dust deposits by turbulent - Smouldering combustion in flames in dust clouds. on humans and
gas flows. dust layers/deposits. Mechanisms of heat transfer mechanical structures.
- Hot surfaces. (conduction, convection,
- Flying burning metal particles. radiation).
- Electric discharges.
- Hot gas jets.
- Shock waves.
- Focused light-beam
hot spots.
l Transport of dust particles in l Influences of dust cloud a Limit conditions for flame l Ability of blast waves
turbulent gas flows. properties on ignition propagation through dust to transform dust layers
sensitivity (composition, clouds (particle properties, dust into dust clouds.
size and shape of particles, concentration, oxygen
turbulence, dust concentration, concentration, geometry).
gas phase composition).
l Measurement and 0 Flame acceleration processes
characterization of state of in dust clouds. Influence of
turbulence of dust clouds. 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
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,
l Laminar flames
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.
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.
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.
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
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)
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
References
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
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
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.
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).
- 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.)
0.6 -
05 -
04 -
03 r
02 -
01 _
00 I J
0 2 i 6 8 10 12 1L
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.
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
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
EXPERIMENTAL
MAX. EXPLOSION
ABSENCE OF
TURBULENT JET
1
0 02
t
001 I
0 5 10 15
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
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
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).
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).
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
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 simulations at T > 0 K with such potentials are left out. In
University Press, Cambridge, 1991, 391 pp., $69.50/27.95 particular, the phenomenon of surface melting is not even
mentioned in the book.
On its back cover this book claims to be a complete Concerning impurity effects, there is a good and rather
course . . . to understand . . , crystallization. And indeed, a complete discussion of adsorption and desorption and the
well accessible and elegant treatment of many complicated influence on the step velocity. The kinematic wave theory is
subjects in crystal growth is given. But, of course, it cannot shown to give insight into the phenomenon of step bunching.
be really complete. By now well-established results in the literature that this
The treatment of the equilibriwn and growth form is good theory cannot explain the firmntion of macrosteps, but
and several effects arc treated which are often neglected, like merely their shape, as well as several alternative models for
surface stress, dislocations and reconstruction. What is step bunching are, however, hardly mentioned.
lackin& on the other hand, is some modern theory and S-a&in& this is a nice and useful book as an in-
experiments on surface roughening (only mean field theory, troduction to the understanding of several phenomena re-
which is now known to give several incorrect predictions, is lated to crystal growth. The fact that it does not cover all
treated in detail). modem developments is not unexpected and probably un-
A nice discussion is devoted to the description of interface avoidable. Taking this point of view, I can advise this book
shapes, e.g. the shape of the triple junction solid/liquid/vapor to scientists who intend to be alerted to general trends in
in crystal pulling systems. Attention is paid to the effect of crystal growth, without being anxious to be well-informed
anisotropy of the surface free energy as well as of the surface about all recent developments. A further positive point is
kinetics. EspeciaIly, the discussion of an atomically smooth that the author has brought together experimental and
surface orientation, like Si( 11 l), near interface junctions and theoretical evidences in a book whose size is not excessive.
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 J. P. VAN DER EERDEN
missed a discussion of the Burton, Cabrera and Frank Lab VCWT
Grensvlakken en Thermydynamica
surface diffusion model for step growth. P~~~~OUPI a
The stabilization at T = 0 K of different crystal and sur- 3584 CH Utrecht
face structures by different interaction potentials is given, but l&e Netherlands
Dnst Exploaict~ in the Procerrs Indmt&% By RQLF K. First is a welcome examination of the formation mech-
FXKHOFF. Butterworth Heinemann. 1991,590 pp., $75 anisms of dust clouds, which takes the reader through the
relevant aspects of particle dynamics in air flows, and how
This is a good book which will rightfully claim its place as a these are able to provide the concentration gradients so
classic treatise on dust explosions alongside the earlier works necessary for flammable and explosive behaviour. This is
of Palmer (1973) and Field (1982). It is comprehensive in followed by a look at flame propagation, again fundamental
coverage, and well set out in its treatment of the principle to those reaction mechanisms which control the generation
parameters of dust cloud formation, ignition and flame of disruptive energies in ignited dust clouds. In these
propagation. Treatment of the subject does tend to ret&t chapters Prof. Eckhoff both describes and explains the ex-
the approach of the researcher and academic, but I was perimental work and the mathematical modelling needed in
pleased to find that Prof. Eckhoff also laced his text with its application, and I was pleased to see his critical remarks
much nseful advice. for plant designers and process engineers. about some of the current methodology in this area.
The book starts by giving an overview of the subject, The nw steps from laboratory to plant are then
although in these first 150 pages the author is able to provide explored in more depth, particularly with respect to how
a valuable insight into the main factors which inllued~e dusts may be ignited, and the parameters which influence
ignitability and explosion violence, followed, importantly, by safe venting an& hence, its sizing. Most of this material is
a useful guide to prevention and mitigation and how these well up-to-date, and indeed the author speculates where it
might be achieved in practice. This is most helpful, and will might lead in the future. This is important because it demon-
steer students, researchers and designers alike to the later strates that our understanding of the science is still develop-
more detailed chapters on these issues. ing, via both laboratory-scale and pilot trials (an area where
As if to punch home the basic messages from the overview, my own organisation under Lunn is making an active
Eckhoff then follows with an interesting and well-chosen set contribution using plant size equipment at Buxton).
of case histories from several countries. These translate the Finally, there is a comprehensive examination of laborat-
theory into real life situations, and so form a valuable ory methods for assessing dust ignitability, explosibility and
platform from which to launch the more detailed treatment related properties, and again the authors critical views are
of the science in the chapters which follow. informative. This is especially helpful in an area where the
629
630 Book Reviews
variability of results can often say as much about the test designs for industrial process plant. I refer to the dimensions
method as the parameter under evaluation. of cost and practicability of some of the solutions, since these
All in all, a worthwhile and valuable book, which through- are as likely to inform the actual selection of methods and
out contains, what appear to be, exhaustive topic biblio- plant as the sophisticated experimental data. A small criti-
graphies for those who want to delve deeper into any cism of a major new text book perhaps, but one that is
particular aspect of the subject. Inevitably in an area as probably realistic nevertheless.
involved as this, some material gets more comprehensive
treatment than others, but I was not left with the feeling that I. G. JONES
anything significant to the science had been left out. How- Health & Safety Executive
ever, one line could use.fully have been more deeply explored, St Annes House
and this perhaps illustrates the difference between the Stanley Precinct
worldly academic and the practical engineer doing real Bootle, Merseyside L20 3MF, U.K.
Liqoid-Vapor Pbaae-ehange Phenomena. By VAN P. CAREY. it would have been helpful to have in the first part of the
Hemisphere Publishing Corporation, Washington, book a simple example, say of bubble growth in a uniformly
Philadelphia, London, 1992, 645 pp., El07 (E41 paperback) superheated mixture, illustrating the effect of mass dif-
fusional resistances.
Professor Carey has written a substantial and wide-ranging The standard models for heat transfer across continuous
book on a topic that is so broad, and of which our under- films of liquid or vapour are given in full. Idealized models
standing is still so incomplete, that compromises in coverage for more complicated situations like nucleate boiling and
are inevitable. The book is subtitled An Introduction to the critical heat flux are sketched out at intermediate levels
Thermophysics of Vaporization and Condensation Pro- broadly representing the state of understanding 5-10 years
eesses in Heat Transfer Equipment. The first 200 pages ago. For elaboration of these models readers are given
introduce the basic physics (thermodynamic equilibrium, references to follow up and make their own judgments.
surface tension and wetting, interfacial boundary conditions Fifteen per cent of the references are to publications in the
and thermal resistances, waves, homogeneous nucleation) to decade immediately preceding the publication of this book.
be used in the following 400 pages on the modelling and Some substantial bodies of work, e;g. in Germany on nu-
correlation of nucleate and film pool boiling, condensation cleate boiling and in Japan on film boiling, are not re-
on plates and cylinders, two-phase flow in round tubes, ferenced. Models often originate from photographic obser-
condensation and boiling inside tubes. This treatment is vations and students should be aware of the often significant
mainly confined to pure, single-component fluids. A final differences between primary data and model. The photo-
section on special topics provides brief discussions of flow genic nature of two-phase flow is one of its attractions and it
instabilities in tubes, boiling and condensation of binary is disappointing that the book contains only three photo-
mixtures and in finned passages. Little is done to relate the graphs, two of which (showing static sessile drops with high
calculations of heat transfer and pressure drop to the design and low contact angles) were apparently chosen to avoid
of heat transfer equipment. The reader will not learn from over-exciting readers. Professor Carey draws attention to the
this book what industrial boilers and condensers look like weaknesses and contradictions sometimes present in models.
and may wonder why anyone cares about film boiling on He provides, by worked examples, useful guides to the
cylinders and spheres; convective two-phase heat transfer on application of empirical correlations that must then supply
the shell side of tube bundles and practical problems of the answers required by designers.
corrosion and fouling are not discussed. I would have wel- The book is intended for graduate-level instruction and
comed an introductory chapter to put the subject in its could be supplemented by texts such as Smith, R. A., Popor-
engineering context. isers (1986). Longman 8cimtifrc and Technical, to provide
The general plan to proceed from basic phenomena to the the engineering context. Problems are provided at the end of
modelling of heat transfer is to be applauded, although the each chapter, in addition to the worked examples in the text.
present incomplete understanding of multiphase heat trans- Research students would find the book a reasonable starting
fer means that it cannot always be carried through. Thus, point but would have to do a lot of additional reading,
ultra-thin films (a welcome addition to the list of topics in starting perhaps from the critical and topical reviews-in the
two-phase textbooks), interfacial resistances and Marangoni proceedings of the quadrennial International Heat Transfer
flows all appear in the first part of the book but do not Conferences. For undergraduate instruction there are less
contribute much to subsequent applications. In part, this is a detailed but more readable alternatives.
consequence of the emphasis on single-component fluids
with local equilibrium and consequently uniform state at the D. B. R. KENNING
liquid-vapour interface. The influence of an additional com- Department of Engineering Science
ponent on the interfacial temperature slips is unheralded in a Oxford University
section on the evaporation of a liquid tilm into an air stream. Parks Road
Given the practical importance of mixtures of volatile fluids, Oxford OX1 3PJ, U.K.
Safety Science, 15 (1992) 207-208 207
Elsevier
Book Reviews
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
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
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.
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
Fig. 1. The University of Michigan wind tunnel facility for the study of the smoldering combustion of dust.
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
Dnmeter=1.22m ; Volumc=O.95 d
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
pressure transducer is used to monitor the pressure Fig. 6. Effect of turbulence on P,,,, for different dusts, and
variation during the explosion, required to find the different moisture content of cornstarch dust.
combustion velocity and ultimately to normalize the
burning velocity to atmospheric conditions. A more To calculate the convective flow velocity, the flame
detailed description of the apparatus and experimental front position, the pressure, and the rate of pressure
techniques can be found in [28]. rise should be known. Then the burning velocity of the
This experimental apparatus allows measurement of dust flame can be calculated [21, 29, 301. Such cal-
the influence of turbulence intensity up to u -4.2 m culations allow one to obtain the burning velocity at
s-l on the explosion parameters. Much data for different the given pressure. Since the pressure in the sphere
dust mixtures was obtained. Typical data obtained are is continuously changing, measurement of the burning
shown in Figs. 6 to 10. Increased turbulence has a velocity at different pressures is necessary for the ex-
relatively small effect on the maximum explosion pres- trapolation of results to atmospheric conditions. Fig.
sure (Fig. 6), but a rather significant effect on the rate 8 shows the burning velocity corrected to atmospheric
of pressure rise (Fig. 7). Since the rate of pressure pressure for different dusts and different turbulence
rise depends directly on the combustion velocity there intensities.
is a similar influence here. In the sphere, the flame From the measured burning velocities obtained at
spreading velocity is measured by the ionization probes. different turbulence intensities, the laminar burning
This velocity is the result of the burning velocity and velocity for dust mixtures can be obtained by extrap-
of the convective flow velocity generated ahead of the olation. Corrected laminar burning velocities for a few
flame front by the expanding combustion products. So different dust mixtures are presented in Fig. 9. Iaminar
to get the burning velocity the convective component or quasi-laminar burning velocities cannot be measured
should be subtracted from the measured flame velocity. 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)
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.
125
4 -I
-b 6 12 18 a4 30 36
x lnitiai shock
//
0
I
Fig. 16. Typical pressure histories at the different tube locations, Fig. 17. Mach number of flow at different tube locations and
and the flame front position for the case of flame acceleration the flame front trajectory for corn a dust layer explosion. The
to a quasi-detonation in a medium corn dust layer with 1% same condition as in the previous figure.
addition of Cabosil. Corn concentration in the layer: 500 g mV3,
layer length: 51.5 m.
from the layer as indicated by the decrease in the
initiation of the H,+ l/20, detonation in the auxiliary intensity of the reflected (diRused) light (t2). From
detonation tube. The explosion of a hydrogen-oxygen Fig. 13 one can see that the precursor shock causes
mixture provides the precursor shock which is respon- initial suspension of the dust into the tube center. At
sible for the slight initial dispersion of the dust layer this stage the dust concentration in the tube center is
and the initiation of the primary dust explosion (com- relatively low, but it increases dramatically just before
bustion of the dust predispersed in the first 3.66 m of flame arrival, and then the dust concentration decreases
the tube). As a result of this primary explosion the due to bum out of the dust particles.
flame accelerates in the tube. Static and dynamic pressures and temperature are
Dust dispersion is monitored by a concentration measured at eight different locations along the tube.
probe, and a high speed camera can be used for the Additionally the temperature of burning dust particles
qualitative observation of the variation of dust dispersion is measured by a four wavelength pyrometer. At the
with time. Direct pictures give general information about end of the tube a special exhaust section provides for
the dust dispersion process while the optical probe the ventilation of the explosion. The whole system is
measures the dust concentration variation in the tube controlled by an IBM 9000 microcomputer. The suitably
center. Quantitatively information such as rate of en- amplified outputs from all instruments are fed into a
trainment may be obtained from backscattered and 32-channel analog-digital converter. After each run all
transmitted laser light as layered dust behaves optically data is processed by the microcomputer. A detailed
differently from suspended dust [33]. The interaction description of the whole system can be found in [34-351.
of a 4.0 mm dust layer with an M- 1.4 shock wave is In the early stage of research, the tube had a length
shown in Fig. 12. The relaxation time for the dust layer of 36.6 m. Some results obtained from these studies
after the passage of the pressure increase can be clearly are presented in Figs. 14 and 15. In Fig. 14 the maximum
noted (tl) as well as the steady erosion of the dust flame velocity is presented as a function of the nominal
127
* Baghouse
+ Bees wings
+ Mira Gel
1C
0 20 40 60 60
VOID LENGTH (tube diameters)
Fig. 19. Effect of voids on the tube exit pressure as a function of dimensionless void length (normalized to tube diameter).
then slow down and the shock will again disperse dust The effect of a void on layered dust explosions has
ahead of the flame. The flame will accelerate again been studied for three different kinds of dust: baghouse,
and the quasi-detonation (or detonation) from will be bees wings, and Miragel. Baghouse dust is the dust
reestablished. So in very long tubes one can expect literally removed from the baghouse of a dust collection
conditions similar to a galloping detonation. system at a grain elevator which handled only corn,
Another interesting study carried out in the FAT bees wings are a red colored flake type material which
tube concerns the influence of a bimodal particle size comes from corn cobs, and Mirage1 is a commercial
distribution and also the effect of a dust void on the starch product. Details are given in [38]. Despite the
explosion deveIopment [36]. For the bimodal particle differences in the explosibility of these dusts, the general
influence of the void on the pressure measured at the
experiments, small corn cob material with a diameter
end of the tube (exit pressure) was the same. It can
of less than 75 pm and large particle corn cob material
be seen in Fig. 19 that the exit pressure decreases
with diameters between 425-800 pm were used. For
continuously as the void length increases. When the
the total dust concentration of 0.5 kg me3 a different
void reaches approximately 65 tube diameters the ex-
ratio between the small and large particles was tested. plosion is no longer transmitted through the void area.
Also separate experiments for the small and for the From this data it is difficult to conclude whether the
large particles were carried out. It was found that the linear distance or number of tube diameters is important
highest propagation velocity and maximum pressure in isolating the primary explosion from the secondary
rise occur when the small particles comprise 60% of explosion of the layered dust. It will probably be a
the mixture (Fig. 18). However, the addition of Cabosil stronger function of the linear distance. The safe void
to the mixture shifts the optimum value to 80% of the distance will depend on the time of burning of the
small particles and significantly increases the maximum dust particles and the velocity of the induced flow. So,
explosion pressure. This can be explained by the in- if burning particles (or still sufficiently hot burned gases)
creased dispersive properties of the dust with the ad- will reach the fresh dust dispersed from the next layer,
dition of the larger particles or the Cabosil. 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
u c
6.8 4
0.9 1.0 1.1 1.2 1.3
100WTEMPERATURE
@)
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.
T tps1
RpfI)
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-
Y I R-16.6 / / / I 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
I t-15 t-5 combustion in tubes. Flame acceleration due to layered
R-24.6 Ft.:
L dust combustion depends on the dust arrangement in
t.00 0.20 0.40 0.60 0.60
RROIRL POSITION IN PRRTICLEk/R) the layer but even more strongly on the dust properties.
Very fine and reactive dust can produce overpressures
Fig. 28. Numerically calculated temperature variations within the
burning particles in the detonation reaction zone.
of up to 26 bars, and even transition to quasi-detonative
combustion is possible.
Quasi-detonative and detonative combustion of some
mixture which could not be detonated without the dust-air mixtures can propagate with a velocity ranging
added component. It was also found that the addition from 1000 m s- 1 (for quasi-detonation) to nearly 1550
of nonreactive particles to the gaseous mixture can m s- (for detonation), and with the addition of oxygen
widen the lean detonability limit [14, 46, 471, but the or in pure oxygen the propagation velocity can be even
addition of inert particles to the gaseous mixtures always higher. The maximum pressure rise across the deto-
decreases the detonation velocity. The influence of nation front is usually much higher than in a deflagrative
nonreactive particles is always proportional to the spe- explosion, but the transition to detonation creates a
cific volumetric particle surface area. A more detailed higher pressure rise than the detonation itself.
description of such effects can be found in [47] or in In hybrid mixtures or in gaseous mixtures with the
the recent survey paper on dust-related detonations addition of nonreactive particles a widening of the
WI. detonability range, and a more rapid acceleration to
Despite significant progress made here and in other detonation can be observed. However, the addition of
laboratories, more experimental and theoretical re- inert or slowly reacting particles to gaseous mixtures
search is necessary for better understanding of dust decreases the detonation velocity, and they can even-
detonation phenomena. tually quench propagation.
Research on dust explosions carried out at the Uni-
versity of Michigan has improved understanding of many
Conclusions fundamental aspects of dust ignition, smoldering com-
bustion, turbulent combustion, layered dust combustion,
At the University of Michigan a comprehensive re- and detonations. Dust explosions are a very complex
search effort concerning dust explosions has been carried problem, and not all aspects of their initiation and
out for almost two decades. It covers the most important propagation are completely understood, so further and
aspects of dust explosions, ranging from dust ignition, more detailed experimental and theoretical studies of
through smoldering combustion and turbulent com- this problem are in progress in our laboratories.
134
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.
/
/
I
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,/ ,/
Flame propagation mechanism In the case of Ti(r) =Ti:,,, a constant temperature, the
Flame propagation from burning particles to the next above solution is reduced to
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
Tg(r9
6)=(Tr To) ; (b)e
rfc(s)+L
ignition point. Substituting eqn. (8) or (9) into eqn. (6), the particle
Suppose the flame has just propagated to particles temperature T. can be calculated.
on the (n-1)th spherical surface, whose radius is (n - l)L. Let At,, be the time when T, reaches Ts2, the ignition
Then, the flame front radius b is assumed to be temperature. At, is the time interval required for the
b=(n-l)L+R, (3) flame to propagate from the (n - 1)th spherical surface
to the nth one. Figure 2 shows At,, decreasing with
where Rb is the flame radius inherent to a single burning increasing n to reach a constant value. The trend can
particle. For liquid droplets, Miesse [lo] proposed 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-
where k* is a material constant ranging from 1 to 10 creasing n, accelerating the flame propagation at the
cm for liquid. Since no data for solids have been earlier stage of the process. On the other hand, the
available, similar values of k* are assumed for solid acceleration shortens the time to warm the particles
particles. and the gas in the outer region, resulting in an asymptotic
For the temperature of unburnt particles in the nth variation of At,,. Denoting the asymptote value as At,
surface at time 0, T,, a heat balance is taken on the At is related with dust concentration, C, in Fig. 3.
particles: The flame propagation velocity is calculated to be
LlAt,, or LIAt. The case of one-dimensional flame prop-
=rDP2[h(Tg-T,)+ 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.
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(6)=m,
[
l-
(
1
-$I (11) Ignition temperature is a measure of the ignition
sensitivity of a combustible dust. Cassel and Liebman
192
0 xl0 0 500
D1A.h 1 0IA.b)
Fig. 7. Lower explosible limits of dust concentration.
J/l
dF
-
Thus, the upper limit at a given oxygen concentration 0 =K3Z (20)
is calculated as a maximal dust concentration at which [
dt 1max
T, at 8= 6, in eqn. (6) is equal to T,,, the ignition
point. where P=P/P,,, the dimensionless pressure, PO is the
Although no exact measurements of the upper limit initial pressure and KGc is a constant. However, no
for atmospheric oxygen content were made, values for theoretical background was proposed for this empirical
rather dilute oxygen contents have been obtained [16, law. In addition, for non-spherical vessels, no reliable
171 and are shown by the dotted lines in Fig. 8 where scale-up laws were found relating the vessel configu-
the full lines are the calculated results. Oxygen contents ration to the maximum rate of pressure rise and other
less than about 10% would support no flame propagation variables concerned.
and such critical oxygen contents should be the basis In vented explosions, most of the data obtained were
for inerting. 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
Pressure development in closed and vented vessels for designing vent areas required for various scales of
equipments.
One of the important explosion characteristics is the This section deals with pressure development in closed
rate of pressure rise in a closed vessel. In particular, 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
IO
Time (set)
Fig. 9. Variation of pressure with time in a closed spherical
vessel. Fig. 11. Cylindrical vessel.
195
&
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.
dF = rpl
_ - l/m(~m~/Yu- 1) Wt)
-
[
dt_($p-l/wpyu~)
+(:)I
1
(29)
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
References
Fig. 14. Comparison of predicted S/Vom with experiment. 1 K. N. Palmer, Dust Explosion and Fire, Chapman and Hall,
London, 1973.
Substitution of the cubic law into eqn. (30) leads to 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)
(31) 384.
4 S. Nomura and T. Tanaka, ffigaku Kogaku Ronbunshu, 4
(1978) 634.
where KG is KGc or KG,, depending upon the type of
5 S. Nomura and T. Tanaka, Kagaku Kogaku Ronbunshy 5
vessel. (1979) 47.
Thus, a new dimensionless scale up factor is proposed 6 S. Nomura and T. Tanaka, I&EC Proc. Des. Dev., 19 (1980)
for designing the area ofvent, instead of the conventional 451.
vent ratio. S/Yom is estimated from KG value for a 7 S. Nomura and T. Tanaka, J. Chem. Eng. Jpn., 23 (1980)
309.
given dust and P,, the predetermined maximum pressure
8 S. Nomura and T. Tanaka, Z&EC Proc. Des. Dev., 23 (1984)
related directly to the strength of the plant. 420.
With respect to vented explosions, many experiments 9 S. Nomura and T. G. Callcott, Powder Technol., 45 (1986)
were carried out with various scales of test apparatus 145.
[22-241. A set of the reported P, and KG values are 10 C. C. Miesse, 6th Symp. (Int.) Cornbust., New Haven, Cr,
substituted into eqn. (31) to obtain the theoretical S/ 1957, p. 732.
11 S. Yagi and D. Kunii, 5th Symp. (Znt.) Cornbust., Pittsburgh,
V, which is plotted against the experimental SwO= PA, 1955, p. 231.
in Fig. 14. The calculated values are a little above the 12 R. H. Essenhigh and I. Fells, Dircuss Faraday Sot., 30 (1960)
corresponding straight line as the theory takes account 208.
of the safe design in eqn. (30). In this sense, the overall 13 S. Nomura, Trans. Iron Steel Inst. Jpn., 27 (1987) 552.
14 H. M. Cassel and I. Liebman, Combust. Flame, 3 (1959) 467.
agreement with experiment is satisfactory not only in
15 I. Hartmann, M. Jacobson and R. P. Williams, Invest. No.
the order of magnitude but also in the trend of the 5052, U.S. Bur. Mines Rep., 1954.
straight line. 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.
Conclusions 18 W. Bartknecht, mZ Ber., I65 (1971) 24.
19 W. Ishihama and H. Enomoto, Anzen Kogaky 14 (1975) 243.
For such complicated phenomena as dust explosions, 20 J. Nagy, E. C. Seiler, J. W. Corm and H. C. Verakis, Invest.
our treatment seems to be rather idealized, simplifying No. 7507, U;S. Bur. Mines Rep., 1971.
21 I-X.J. Heim-ich and R. Kowall, M>I Ber., 165 (1971) 53.
the state of dust clouds and the mechanism of flame 22 I. Hartmann and J. Nagy, Ind Eng. Chem., 49 (1957) 1743.
propagation as well as including some severe assump- 23 C. Donat, Loss Rev., II (1977) 87.
tions. Nevertheless the derived results are believed to 24 R. F. Schwab and D. F. Othmer, Chem. Process. Eng., April
be quite acceptable as a first approximation. The overall (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 powder injected into the combustion vessel
based on bicarbonate, phosphate, and C a C O 3 in during explosion.
dust explosion suppression is compared. The early
(1957-1973) researches on the effectiveness of In experiment 1 the powder-dust mixture was
extinguishing powders established the ranking of prepared and the standard procedure of LEL
effectiveness in which the phosphate was far be- evaluation was adopted [6]. In experiment 2 the
hind bicarbonates [1]. However, in 1984, extinguishing powder was put into a 140-mm-
Hertzberg et al. [2] showed the ranking almost long, 16-mm-i.d. tube (Fig. 1). One end was
reversed for suppression of methane and coal dust attached to the explosion sphere and had a nozzle.
explosions. These observations also were con- At the other end was a 1.5-kJ chemical match that
firmed by Choi et al. [3] for the extinguishing of was activated by the ignition delay unit. The
coal flames. The high efficiency of the phosphate delay time was selected to give an explosion
as an explosion inhibitor is difficult to explain but pressure of 0.05 MPa in the sphere when the
may result from the testing method in which the extinguishing powder was dispersed from the tube
lower explosion limit (LEL) of a powder-dust by combustion gases. In both experiments an
mixture was measured [2]. These conditions are initiation of ignition was performed by the 1.5-1d
different from those in which the explosion sup- chemical match, and the explosion pressure was
pression system operates [4]. Here a comparison recorded by a Kistler 601A transducer.
has been drawn between the inhibitor efficiencies The following four explosion dusts were used:
measured by the LEL method and the method of
direct injection of inhibitors by a miniature sup- melamine;
pression system. wood dust;
wheat flour;
EXPERIMENTAL DETAILS
coal dust So~nica.
The combustion vessel was a 22-L sphere used in
investigation of explosion parameters of dusts All dusts wer sieved on a 80-#m mesh. The dusts
[5]. Two types of experiments were performed: concentration was 0.5 kg/m 3 in experiment 2.
1. measurements of the LEL against the extin- Commercially available extinguishing powders
guishing powder content in the powder-dust representing phosphates, bicarbonates, and inert
mixture; and inhibitors were investigated and are characterized
2. measurements of the dust explosion overpres- in Table 1. The results of the experiments are
sures against the amount of extinguishing presented against the inhibitor content in the sum-
Copyright 1991 by The Combustion Institute
Published by Elsevier Science Publishing Co., Inc. 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
o
mary weight o f the inhibitor and dust introduced
into the vessel.
Fig. 2. Effect of added inhibitors on the flammability limits:
(a) melamine; (b) wood dust; (c) wheat flour; and (d) coal.
RESULTS Vq, CaCO; A, LB-2, O, LB-3.
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- volatility, 11% ash, Q = 6400 k c a l / k g ) and a
surements have confirmed the effectiveness order weaker ignition source was used.
predicted earlier [2]. The most effective extin- Figure 3 presents the L E L o f the explosion
guishing powder in the suppression o f dust explo- overpressure as a function o f the content o f extin-
sions appeared to be the suppressant based on guishing powder which was directly injected into
N H 4 H 2 P O 4 (LB-3). Some o f these results should the vessel during explosion. These measurements
be compared with the data in [2, 7] because the contradict the results from experiment 1. O f the
size distributions for the present Sonica coal dust four dusts investigated the most effective suppres-
and for the Pittsburgh bituminous coal Pgh-1 sant appeared to be the extinguishing powder
were similar. The inerting levels o f N H 4 H 2 P O 4 base on N a H C O 3 - L B - 2 at the low and medium
and C a C O 3 for coal dust were 1.5 times lower ( < 50%) content.of the inhibitor. However, for
than the corresponding levels found by Hertzberg higher inhibitor concentration the phosphate and
et al. [2]. Perhaps, these differences result from bicarbonate curves converge towards one an-
the fact that So~nica coal was less explosive (28% 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%
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 CONTENT [ X]
REFERENCES
Fig. 3. Effect of inhibitors injected into the vessel on the
explosion overpressure; (a) melamine; (b) wood dust; (c) 1. Hastie, J. W., High Temperature Vapors, Academic,
wheat flour; (d) coal. 0, CaCO,; A, LB-2; 0, LB-3. New York, 1975, pp. 344-347.
2. Hertzberg, M., et al., Twentieth Symposium (Interna-
tional) on Combustion, The Combustion Institute,
Jn both types of experiments, CaCO, appeared Pittsburgh, 1984, p. 1691.
to be the least effective suppressant, which sup- 3. Choi, T., Rahiman, S., and Essenhigh, E. G., Twenty-
ports the hypothesis on the chemical character of First Symposium (International) on Combustion,
The Combustion Institute, Pittsburgh, 1986.
explosion inhibition by phosphates and carbon-
4. Moore, P. E., Shenyang International Symposium on
ates. Dust Explosions, Shenyang, Republic of China, Septem-
her 14-16, 1987.
5. Jodkowski, W., et al., First International Colloquium on
CONCLUSIONS
Explosibility of Industrial Dusts, Baranow, November
8-10, 1984.
It was found that the effectiveness of extinguish-
Polish norm PN-84 C-01200/7 (Polish Standard).
ing powders in dust explosion suppression de- Cashdollar, K. L., and Hertzberg, M., Private commu-
pends on the method of testing. We confirmed nication.
that the LEL method strongly favors NH 4H 2PO,. Polak-Kuncerowicz, B., Ph.D. thesis (in Polish),
However, the method of direct injection of the Krakow, 1987.
Hertzberg, M., Zlochower, I. A., and Cashdollar, K.
extinguishing powder led to different results. The
L., Twenty-First Symposium (International) on
study provides evidence that at moderate concen- Combustion. The Combustion Institute, Pittsburgh,
trations of inhibitors NaHCO, is more effective 1988, pp. 325-333.
than NH,H, PO4 and at higher concentrations of 10. Sloane, T. M., Combust. Sci. Technol. 63:287 (1987).
inhibitors the efficiencies of both extinguishers
converge. These observations are in agreement Received 22 February 1990; revked 25 August 1990
Sizing of dust explosion vents in the
process industries
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.
Several kinds of precautions may be taken to prevent an approach takes into account the industrial-process-
damage to humans and property by dust explosions in dependent variables that significantly influence the dust
industry. The importance of fighting potential ignition cloud combustion rate, and the risk-analytical aspects.
sources has been stressed frequently, and this of course The purpose of this paper is to analyse existing
still remains a major issue. However, it is normally evidence on dust explosion vent sizing and indicate a
considered that trying to eliminate ignition sources general approach for the present and near future by
alone does not provide sufficient protection. It is also which adequate vent sizes may be assessed.
necessary, in case an explosion should nevertheless be
initiated, to ensure that destructive effects are mini- Nature of the dust explosion venting process
mized.
Venting has become the most widely used means The maximum overpressure Pred in a vented enclosure
of additional protection, because of its favourable in which a dust explosion takes place is the nett result of
two simultaneous, competing processes: burning of the
technical and economical features. It is therefore
unfortunate that the technical literature on venting of dust, causing release of heat and thus pressure increase;
dust explosions is still to a considerable extent contra- and flow of unburnt and burning dust cloud and
dictory. However, signs of a slowly developing clar- combustion products through the vent opening, causing
the pressure to decrease. In a given enclosure provided
ification are detectable. Useful summaries of existing
with a given vent, the influence of the burning rate of
knowledge up to 1984 have been produced by
the dust on Prrd will be as illustrated in Figure I.
Schofield and Lunn*. Other sources of summarized
Clearly, predicting the rate of heat release that will
information are the books by Field3 and Bartknecht4J.
One particularly important achievement is the actually occur in a specific situation is the key issue.
increasing appreciation of the need for a more differ- However, adequate theories are lacking, and it is
necessary to rely on realistic experiments. The rate of
entiated approach to dust explosion vent sizing6. Such
heat release depends on a number of factors:
l chemical composition of the dust, including mois-
Received I4 August I989 ture;
0960-4230/90/030268-12
@ 1990 Butterworth-Heinemann Ltd
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 ignition point was close to the silo bottom. In all Swiss/German one is only 20 m3, whereas the Norwe-
experiments but one, dust injection was terminated a gian silo has a volume of 236 m3. RadandtZo seems to
few seconds before ignition, allowing the dust cloud to assume that effects observed in the comparatively small
become comparatively quiescent. The experimental 20 m3 silo are also applicable to larger silos. However,
results for wheat grain dust as well as maize starch results from the 236 m3 silo experiments23*24 suggest
explosions are shown in Figure 2, together with predic- that this is not a valid assumption in general. A more
tions based on different vent sizing methods. The vent detailed discussion of this problem is given elsewhereX.
cover used in the experiments was a sheet of plastic One set of results for which the vent areas are
with a low static opening pressure of the order fairly close to geometric similarity, is given in Figure 3.
0.01-0.02 bar(g). The final, exceptional maize starch The hatched envelope of all experimental data for the
explosion named turbulent jet was so violent that the 236 m3 silo shows a dramatic decrease of the maximum
silo wall (nailed steel plates) ruptured at about explosion pressure when the ignition point is moved
0.6 bar(g). In this way an additional vent of 50 m2 was upwards along the silo axis. However, the hatched strip
produced, preventing the pressure from rising even enveloping the data for the 20 m3 silo, shows very little
further. In this exceptional experiment 300 kg of starch of this effect. The discrepancy must be due mainly to
had been charged to the dust feeding system, and dust the different ways in which the dust clouds were
was still being injected at full rate when the ignition generated in the silos26 and to the difference between
source. was activated. This special configuration, blow- the silo heights (10 m and 22 m).
ing dust upwards in the silo from a point about half-way In the case of the 236 m3 silo, the dust cloud was
up on the silo axis, is probably not a realistid situation essentially quiescent at the moment of ignition. There
in an industrial silo. However, it does demonstrate that is little doubt that the marked increase in the maximum
even the conservative VDI 3673 recommendations13 explosion pressure as the ignition point was shifted
may for some very special situations specify vent downwards in the silo, is caused by flame acceleration
openings that are too small. due to expansion-induced flow. This is in complete
All the other experimental results in Figure2 show accordance with previous results from numerous dust
that in the case of a large enclosure of 500 m3 and and gas explosion experiments in one-end-open tubes,
L,fD = 4, VDI 3673 and NFPA 68 oversize the vent by ducts and galleries. In the case of the 20 m3 silo, the
at least a factor of 2-3 if the K,, value (115 bar m s-l for strong, turbulent dust jet, extending from the top of the
maize starch containing 11% moisture) is used, and at silo and several metres downwards, was maintained
least by a factor of 5 if the common St 1 nomograph is during the ignition and explosion process. This most
applied. The Rust method9 agrees well with the wheat probably caused very rapid propagation of any flame
grain dust experiments (150 bars-l in the Hartmann initiated at the top of the silo, to the central parts. This
bomb), but oversizes the vents for the maize starch would explain why top ignition and central ignition
(300 bars-l) by a factor of 3. The vent ratio method gave almost the same explosion pressures. In the case
which simply requires half the silo top as vent for of bottom ignition, the initial flame propagation was
keeping Prcdequal to maximum 0.14 bar(g), oversizes probably comparatively slow. As soon as the flame
the vent in this particular case by a factor of 2 to 3. front reached the turbulent dust zone in the central
Radandts equation gives exceptionally large vents, but parts of the silo, a much more rapid flame propagation
L/D = 4 may be too small to make Radandts con- pattern, similar to that generated by central ignition,
stants for silos applicable. probably developed.
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 Switzerland
= 1.0 L/D = 6 25,
deterministic problem6. The 5QOm3 silo explosion vent orea= 0.5 mz
.Z
experiments attempted to create a worst-case situation b
e
in terms of both dust concentration in the silo, and
location of the ignition point. pe
0.5
vent area = 3.5 m2
In Figure 4, predictions made by Equation (3) are generated by discharge of dust from the cyclone bottom
compared with the results of the experiments in the via a rotary lock, was neither well dispersed, nor
Norwegian 236 m3 silo. The dotted curves are obtained turbulent. The published result.+30 using this method
by means of Equation (3) for 20 m3 and 236 m3 are limited to one data point for maize starch, shown in
volumes, using the constants5 given for dust class St 1. Figure 5. For a vent area of 1.3 m2, the maximum
The upper limit of St 1 is K, = 200 bar m s-r, and it is explosion pressure generated by dropping the dust
assumed that the dotted curves apply exactly to this K, from a cyclone at the silo top via a rotary lock, was only
value. In the absence of further information, it is then 0.2 bar(g). Direct pneumatic injection gave about
assumed that the ratios AA, between the vent areas for 0.5 bar(g), and the traditional, artificial VDI-method
K,, = 200 bar m s-l and K,, = 100 bar msl for the (discharge of dust from pressurized bottles) gave about
20 m3 silo also apply to the 236 m3 silo (geometrical 0.75 bar(g).
similarity). This makes it possible to estimate theoretic- Figure 5 illustrates the need to apply a diierenti-
al Radandt-predictions for 236 m3 and K, = ated approach to sizing of dust explosion vents6 and for
100 harms-, by shifting the Radandt curve for further full-scale experiments under realistic industrial
200 bar ms-l and 236 m3 to the left by AA (which conditions27. The pneumatic injection experiments in
varies with pressure). the 20 m3 silo gave substantial scatter in the ex-
The actual experimental*3 maximum explosion perimental maximum explosion pressure@. The P,&
pressures found in the 236 m3 silo of L/D = 6 are data in Figure 5 are the highest values obtained for
partly considerably higher, and partly considerably each vent area, which means that the curve for direct
lower, than the estimated Radandt-value for pneumatic injection is quite conservative. This kind of
K,, = 100 harms-l. On the other hand, even the scatter, which is also apparent in Figure 2, emphasizes
highest experimental pressure of 1.2 bar(g) is signifi- the relevance of applying risk-analytical considerations
cantly lower than the Radandt value of 1.75 bar(g) for in vent sizing6.
K,, = 200 bar m s-l (St 1 dusts). It must be concluded
that so far the relevance of, and experimental and Recent realistic experiments in Switzerland in vessels
theoretical basis for, the simple scaling law suggested of small LjD
by Radandt has not been fully clarified. The German/Swiss school of dust explosion venting
From the start of the large-scale silo explosion research has now started to use industrial pneumatic
experiments in Norway*,*s, the dust clouds were pipeline transport systems for generating dust clouds
generated by pneumatic pipeline injection, in accord- even in vessels of small L/D. Recently, Siwek3r
ance with typical industrial practice and the basic discussed a series of explosion experiments in which the
philosophy adopted for these experiments27. Following dust clouds were generated in this way in vented
the first international presentation of results*s, Bart- enclosures of 10, 25 and 250 m3 respectively. Some
knecht and Radandts.2s replaced the traditional VDI results for maize starch are reproduced in Figure 6,
dust injection method 2g by pneumatic pipeline injec- together with the current VDI 3673 recommend-
tion even in their 20 m3 silo experiments. This was a ations13. Experimental results for two quite high dust
significant decision, reflecting the need to conduct cloud injection velocities are given. The dust entered
experiments in accordance with reality in industry. In the vessel through a 90 mm diameter nozzle at the
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, \,
\ \
z
3
b
e
lqnitiin at silo bottom P
9
In addition to these realistic dynamic explosion vent area, assuming St 1 dusts, gives an expected
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 the same order as the highest pressures of 1.5 bar(g)
art&al static dust cloud generation method very measured by the artifical static dust dispersion
similar to that used in the VDI 3673 experiments4. The method, and much higher than the pressures measured
dust feeding duct was then blocked at the entrance to in the realistic experiments.
the cyclone, which reduced the effective vent area However, the new NFPA 68 (Ref. 17) also in-
slightly, to 0.032 m2. A system of two pressurized dust cludes an alternative nomograph which covers all St 1
reservoirs and perforated tube dispersion nozzles were dusts that do not yield higher P,, in standard closed
employed for generating the dust clouds. The S kJ bomb tests than 9 bar(g). This nomograph gives
ignition source was located inside the cyclone, half way substantially lower Pled values than the standard
up on the cyclone axis. The source was fired about nomograph, especially for small volumes. In the case of
100 ms after onset of dust dispersion. the 0.32 m3 cyclone with a 0.032 m* vent, the alterna-
Envelopes embracing the results of both series of tive nomograph gives Prcd equal to 0.50 bar(g), which
experiments are given in Figure 7. Two polymer dusts is in fact close to the realistic experimental values. This
were used, namely an ABS resin dust of median alternative nomograph originates from Bartknecht5,
particle size 180 pm, and an ethylene-vinyl acetate and represents a considerable liberalization, by a factor
copolymer dust (EVA) of median particle size 40 pm. of about two, of the vent area requirements for most
The artificial static method of dust dispersion gave St 1 and St 2 dusts. However, the scientific and
much higher maximum explosion pressures in the technical basis for this liberalization does not seem to
cyclone than the realistic dynamic method. This is in have been fully disclosed in the literature, and the
accordance with the results of the earlier realistic situation appears somewhat confusing.
cyclone experiments 6,32. The static results in Figure 7 When applied to the 0.32 m3 cyclone, the method
may be compared with predictions made using presently used in Norway1pL2 gives a Prcd value of
VDI 3673 (Ref. 13). A slight extrapolation to 0.032 m2 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
I normal operation of the filter, which was of the
r air-pulsed, self-cleaning type. Four different dusts were
0
;I used, and their K,, values were determined according
x to VDI 3673 (Ref. 13). The ignition source was located
in the hopper below the filter bag section, and the vent
2 was in the roof of the filter housing. Therefore, to get
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-
C rized in Figure 8, together with the corresponding
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
0.5 obtained. Lunn and Cairns35 also reported on a series
Ncminol dust concmtmtion (kg ni?
of dust explosion experiments in a generously vented
8.6 m3 empty horizontal cylindrical vessel of L/D = 6.
Figure 7 Results from vented dust explosions in a 0.32 m3 The same dusts were used as in the filter experiments,
cyclone using two different polymer dusts and two different
methods of dust cloud generation, and 0.03 and 0.04 m* open but the dust clouds were generated artificially by
vents with ducts. Data from Hayashi and MatsudaJ injection from pressurized reservoirs as in the standard
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%
in the actual industrial enclosure (see Figure 5). the case of mills generating dust clouds that are less
turbulent and less well dispersed, it should be possible
Intermediate (lo-25 m3) enclosures of small L/D to ease the vent area requirements, depending on the
The experimental basis is the original VDI-3673 actual circumstances.
experiments4 with highly homogeneous, well-dispersed
and turbulent dust clouds, and the results for the recent Elongated enclosures of very large L/D
experiments31 with much less homogeneous and well- This enclosure group includes galleries in large build-
dispersed clouds (Figure 6). The vent area require- ings, pneumatic transport pipes, dust extraction ducts
ments identified by these two sets of experiments differ and bucket elevators. In such enclosures severe flame
by a factor of up to 5. Adequate vent sizing therefore acceleration can take place because of the turbulence
requires that the conditions of turbulence, dust disper- produced by expansion-generated flow in the dust
sion and level and homogeneity of dust concentration cloud ahead of the flame. In extreme cases transition to
for the actual enclosure be evaluated in each specific detonation may occur. The generally accepted main
case. principles for venting of such systems should be
followed. Either the enclosure must be made sufflcient-
Cyclones ly strong to be able to sustain even a detonation, and
Two realistic investigations have been traced32,34, and furnished with vents at one or both ends, or a sufficient
both suggest a significant vent area reduction, in number of vents have to be installed along the length of
relation to VDI 3673. Thus, the early investigation by the enclosure to prevent severe flame acceleration.
Tonkin and Berlemont32, using a cyclone of 1.2 m3, Further details can be found in the literature17.
indicates a reduction factor of 0.2. The recent investiga- Important new evidence of how dust explosions prop-
tion by Hayashi and Matsuda, using a smaller cyclone agate in long ducts under realistic process conditions
of 0.32 m3, indicates an area reduction factor of about has recently been presented by RadandtU.
0.5. The situation for cyclones is similar to that for bag
filters. For organic dusts belonging to group St 1, there Scaling of vent areas to other enclosure volumes and
seems to be room for a considerable vent area shapes, other Pred and other dusts
reduction as compared with VDI 3673, in the range of The number of realistic vented dust explosion experi-
factors of 0.5-0.2. However, for metal dusts such as ments is still limited. It may therefore be difficult to
silicon, although there is no direct evidence from find an experiment described in the literature that
cyclone explosions with such dusts, VDI 3673 require- corresponds exactly to the required case. A procedure
ments should probably be followed as in the case of for scaling is therefore needed. NFPA 68 (Ref. 17)
filters. presents the following simple equation intended for
scaling of vent areas for weak structures of
Bag filters P,, 6 0.1 bar(g):
The experimental basis is the evidence in Figures S-11 CA.
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 here A is the vent area, A, is the internal surface area
Xilter and no significant pressure piling prior to igni- of the enclosure and Pred is the maximum explosion
tion), the vent area requirements of VDI 3673 for St 1 pressure (gauge). C is an empirical constant expressing
dusts can be reduced by at least a factor of 0.5. If the the explosion violence, based on experimental evi-
dust concentration in the feeding duct to the filter is dence. By using the internal surface area as the scaling
lower than the minimum explosible concentration, the parameter for the enclosure size, the enclosure shape
vent area may be reduced even more. However, in the is accounted for such that an elongated enclosure of a
case of some metal dusts such as silicon, primary given volume gets a larger vent than a sphere of the
ignition in the filter itself may be less probable and same volume.
ignition will be accomplished by a flame jet entering the Equation (4) is intended for the low-pressure
filter from elsewhere. In this case the vent area regime only, but its form presents no such limitations.
requirements of VDI 3673 should be followed This equation may be adopted even for P,, >
(Figures 20 and II). 0.1 bar(g), and used for first approximation scaling of
vent areas from any specific realistic experiment, to
Mills other enclosure sizes and shapes, other P,, and other
The level of turbulence and the degree of dust dusts. At the outset the constant C should be derived
dispersion in mills vary with the type of mill. The most from the result of the closest realistic experiment, from
severe states of turbulence and dust dispersion prob- which data are available. Subsequent adjustment of C
ably occur in air jet mills. The experimental technique should be based on additional evidence/indications
for dust cloud generation used in the experiments on concerning influence of dust type and turbulence. This
which VDI 3673 is based, is likely to generate dust approach will usually imply extrapolation of ex-
clouds similar to those in an air jet mill. For this reason perimental results, which is always associated with
it seems reasonable that VDI 3673 be used without uncertainty. Therefore, the efforts to conduct further
modifications for sizing vents for this type of mill. In realistic experiments should be continued.
M. Guirao) University of Waterloo Press, 1982, pp. 657-678 37 Cassel, H. M. US Bureau of Mines, Rep. Inv. 6551,1964
28 Eckhoff, R. K. The Control and Prevention of Dust Explosions, 38 Wiemann, W., Bauer, R. and Moller, F. VDI-Berichte No 701,
organized by OYEZ IBC, Basle, Switzerland, 16-17 September Verein Deutscher Ingenieure, Dusseldorf, FRG, 1989
1982 39 Siwek, R. and Skov, 0. VDI-Berichte No. 701, Verein Deutscher
29 Radandt, S. Staubexplosionen in Silos, Untersuchungsergebnis- Ingenieure, Diisseldorf, FRG, 1989 .
se, Teil2, Symposium No. 12, Berufsgenosseschaft Nahrungsmit- 40 Lunn, G. A. in Guide to Dust Explosion Prevention and
teland Gaststiitten, Mannheim, FRG, November 1983 ProtectiowPart 3: Venting of Weak Enclosures and Effect of
30 Bartknecht, W. Bundesministerium fiir Forschung und Technolo- Vent Ducts, IChemE, Rugby, UK, 1988
gie, Abschlussbericht Project No. 01 HK 664 AO, 1988 41 Alfert. F. and Fuhre, K. Report No. 89/25820-l, Chr. Michelsen
31 Siwek, R. VDI-Bericht No. 701, Verein Deutscher Ingcnieure, Institute, Bergen, Norway, 1989
Dusseldorf, FRG, 1989 42 Alfert, F., Eckhoff, R. K. and Fuhre, K. VDI-Berichte No. 701,
32 Tonkin. P. S. and Berlemont, F. J. Fire Research Note No. 942, Verein Deutscher Ingenieure, Dusseldorf, FRG, 1989
Fire Research Station, UK, July 1972 43 Eckhoff, R. K. Special Techn. Publ. 958, Am. Sot. Test. Mat.,
33 Palmer, K.N. in Dust Explosions and Fires, Chapman and Hail, USA. 1987, p. 265
London, UK, 1973 44 Radandt, S. VDI-Berichte No 701, Verein Deutscher Ingenieure,
34 Hayashi, T. and Matsuda, T. Archivum Combustionis submitted Dtisseldorf, FRG, 1989
for publication 45 Eckhoff, R. K. Staub-Reinhalt. Luft 1977,37,110
35 Lunn, G. A. and Cairns, F. J. Haz. Mar. 19X5,12,87 46 ASTM Designations: E789-86 (Hartmann bomb) and E1226-88
36 Eckhoff, R. K., Alfert, F. and Fuhre, K. VDI-Berichte No. 701, (20 I sphere), Annual Book of ASTM Standards, 1988, volume
Verein Deutscher Ingenieure, Dusseldorf, FRG, 1989 14.02
GARYF.BENNETT
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
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.
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)
processes are presented by Fawcett and Wood Commonly encountered failure events
[6]. include those involving mechanical device
After a discussion of integration of differ- failures due to extremes in process variables
ent scenarios of hazardous and other undesir- {e.g., pressure, temperature, concentration),
able events, we concentrate on aspects which control/safety system failure due to instru-
we feel have been somewhat neglected, mentation failure and/or operator 'failure'.
namely, sensors, measurement, actuators, Preventive maintenance must also receive
control, alarms and prediction of events. top priority. This implies inspection, repair,
Sensors are central to the provision of process replacement, and determining the probable
information for hazard prevention and con- frequency of failure of equipment or a com-
trol. Fault diagnosis of sensors, including that ponent [7]. Timing and appropriate action
of actuators, is discussed. Suggestions are put are important and early fault detection should
forward for some new sensors and signal pro- be encouraged [8]. Operator and management
cessing for some known process designs. ;kills have to be utilized.
Whereas preventive maintenance is the
?hilosophy behind safety systems, post-
2. GENERAL CONSIDERATIONS AND accident events also must receive attention in
SCENARIOS a complete assessment to aid in the preven-
tion of future accidents. As stated earlier, we
Process systems in chemical, petrochemical, shall emphasize the former.
food and nuclear technology tend to be com- Plamping and Andow [9] present a pro-
plex. This complexity is largely attributed to cedure for information flow for an alarm
the many interacting subsystems or processes. system. This procedure is divided into nine
Process systems usually have associated safety stages:
systems which include alarm and protection (1) determine the failure events of the
facilities. Besides considerations of equip- plant;
ment, man-machine interactions, environ- (2) assess their credibility;
mental and legal considerations, safety (3) define the modes of operation;
analysis should include instrumentation/con- (4) determine the response pattern of the
trol aspects as well. process variable;
Hazard prevention involves reliability of (5) determine the corrective actions avail-
both processes and instrumentation. Safety able to the operator;
systems can be improved by redundancy (6) determine the level of diagnosis
{spares), upgrading components, control required;
structure/system modification, and perhaps (7) analyse effect patterns and corrective
distributed control. While higher reliability actions to group similar events;
usually implies higher costs, the benefits may (8) select the set of process variables;
be significant. (9) assess the alarms selected and if unsatis-
Most processes of significant complexity factory repeat stages 6 through 8.
have computer-based control and safety sys- This sequential procedure allows an appre-
tems. Failure and fault analysis should include ciation of the dependence of measurement
the following considerations: and states on failure and corrective action.
(a) hardware reliability Various sets of states may be defined for
(b} software reliability alarms given some measurable limits. These
(c) system integrity limits may be dictated by rules such as the
(d) maintenance following [9] :
(e) sensor/instrumentation reliability (a) for those states defined to safeguard
(f) actuator reliability e q u i p m e n t - use absolute limits;
(g) man-machine interface including (b) for those states defined to detect when
operator training the process variable deviates b e y o n d a band
(h) information flow to management. around a set-point value -- use deviation
Ideally, all the above should be fault-tolerant. limits;
This may require some compensation for one (c) for those states defined to detect instru-
or more failed element, service or activity. ment failure -- use zero or full-scale limits;
169
(d) for those states defined to detect pro- Specific design details of a process or unit
cess variables -- use rate-of-change limits. are of significance n o t only with regard to the
Measurement is essential to determine all general development of models but also pecu-
such limits. The multi-disciplinary aspects liarities as far as possibilities of leaks, exces-
of any safety procedure, such as that of sive pressure fluctuations, etc., are concerned.
Plampling and Andow [9], must be recog- Intermediate time-scales would involve
nized. proper measurement followed by appropriate
Hazards are connected with substances alarm and action such as system shutdown.
(gases, liquids, solids or mixtures) having This action must be rapid enough to minimize
properties such as toxicity, corrosivity, flam- the risk of a hazard or the extent of it. Ele-
mability, explosivity, reactivity with other mental time-scales involve appropriate mea-
substances, and radioactivity. Most hazards surement and control. Both these time-scales
can be classified as: involve measurement sensors and prediction
(1) explosions including chemical and dust may benefit from some measurement. There-
explosions; fore, we shall concentrate hereafter on mea-
(2) fires; surement and fault detection of instrumenta-
(3) leaks, which eventually may lead to an tion and actuators. Actuators, especially
explosion or fire or a health hazard; involving moving parts, are subject to failure
(4) radiation. and leaks.
The discussion t h a t follows will be based on
these four major causes.
4. SENSORS IN THE TECHNOLOGIES
Recent discussions on high pressures and One of these developments is based on laser
temperatures are provided by Pegg [12], and beams [19]. This can be online and responsive
Anderson and KoUie [13], respectively. to low gas concentrations, is fail-safe and can
Gas detection devices can be classified as be self-calibrated. These characteristics are
pellistors, infrared cells, electrochemical cells, particularly important for some fugitive emis-
galvanic cells, semiconductor devices and sions.
acoustic devices. Some of the problems Hazardous substance detection and mea-
involved are: surement can be based on that of fugitive gas
(a) Flammable gas sensors such as pellistors detection systems [19]. Figure 1 shows a
may be non-specific and the catalyst gets representation of the important methods
poisoned. Semiconductive devices degrade to available at present. Another representation
the extent that they are not routinely used. for fugitive gas detection is shown in Fig. 2.
Further performance may be limited by asso- The portable instruments enable detection at
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].
REMOTESENSING.
(b) Infrared sensors have m a n y virtues but
METHODS
are not cheap.
(c) Electrochemical cells are not very sensi-
i. ELECIRO~ETRIC
tive to low concentrations usually encoun- SENSORS
various locations such as valves and fans. where the sensor performance is n o t corre-
Valves are notorious for leaks of gases and lated with system dynamics. These include
liquids [20] and it is n o t generally feasible to like-sensor comparison, limit checking, instru-
have a sensor for each valve. However, this ment-loop integrity checking, signal clipping,
should be seriously considered when highly calibration, etc. In the dynamic approach the
dangerous substances are involved. sensor validation is performed using signal
All hardware, including sensors, degrade analysis in conjunction with a physical or an
and experience other functional problems. empirical system model. The latter approach
Temperature and pressure sensors need cali- is more versatile because of its application
bration checks and response time testing [21]. during plant operation.
Some other problems with instrumentation A noninvasive test, such as random data
were pointed o u t above. Analytical instrum- analysis, can be used in some applications
mentation has t w o major problems [22] : w i t h o u t the need for a physical model. Static
(1) leaks in piping/sample lines; and empirical methods are easy to implement
(2) contamination of the measuring unit and are cost-effective. It is possible that most
during sampling. of the sensors can be monitored b y this
Leaks in sample lines cause dilution, those approach. Often methods are developed that
in the hydrogen supply system in a flame are suitable for monitoring a specific detector.
ionization detector m a y n o t be dangerous b u t In this philosophy a separate m e t h o d must be
may make the sensor non-functional. available for each type of sensor. A third
Maintenance of processes including instru- approach for sensor validation is to use signals
mentation, actuators and c o m p u t e r systems that are acquired from dissimilar sensors. This
must be present especially for high-risk provides a diversity of measurement thus
situations. System design should allow for overcoming the problem of common-cause
relatively easy maintenance. Analytical failures. In this Section we review dynamic
instruments, for example, should have a few sensor validation methods and state a few
indicators and allow easy checking of examples.
plumbing and easy spotting of leaks. This
would enable simple techniques such as 5.2. Statement o f sensor failure detection
'soaping' for leak detection for instruments problem
at r o o m temperatures. Soap bubbles from an The methods developed for sensor fault
application o f a soap solution would indicate analysis must be able to distinguish between
a leak [23]. Valve packing should be period- sensor failure and changes in the system
ically checked and if necessary, the packing behavior. For a general solution o f the prob-
replaced [20]. lem, a system dynamic equation and a mea-
Maintenance and proper design are neces- surement equation are necessary. The linear
sary for the prevention of hazardous accidents form of these equations is given by
and protection, b u t are not sufficient. Fault
detection and its analysis is also a most urgent ~((t) = A X ( t ) + BU(t) + W(t) (continuous time)
issue. A number of procedures and techniques (5.1)
exist for the detection of faults in processes,
sensors and actuators. Since sensors and actu- Y(k) = H X ( k ) + V(k) + SF(k) (discrete time)
ators are central to the prevention, control of
and protection from accidents, their fault (5.2)
detection is discussed in the next Section. where
X is an n-dimensional state vector
5. REVIEW OF INSTRUMENT AND ACTUATOR U is a p-dimensional perturbation vector
FAULT DETECTION TECHNIQUES Y is an m-dimensional measurement vector
SF is a vector of sensor errors
5.1. Introduction k denotes the current sampled data index
The methods used for monitoring and veri- W and V are process and measurement noise
fication of process instrumentation and actua- components
tion may be classified into "static" and Sensor failures can be modeled as changes
" d y n a m i c " [24]. Static methods are those in the measurement matrix H, a bias in the
172
sensor output, and additional sensor noise. been reported. A pair of thermocouples [32]
Other fault estimation must be performed by or a pair of 16N detectors [33] may be corre-
first isolating the sensor and using appropri- lated to determine the fluid transit time and
ate fault estimation algorithms. Empirical and hence the velocity. This provides an indepen-
physical model descriptions must also be dent m e t h o d of verifying flow meters and
considered. thus provides diversity against common mode
failure of redundant sensors.
5.3. Dynamic methods for sensor fault detec- The choice of a particular approach for sig-
tion and estimation nal validation and instrument fault detection
The fundamental considerations in this depends on the implementational aspect of
approach are: the m e t h o d and the available measurements.
(1)generate hardware and/or analytical Distinction must also be made between slow
redundancy of the measurements; degradation and transient failures.
(2) a comparison logic for isolating the The above discussion on fault detection
sensor error from system behavior; and identification concentrated on signal pro-
(3) quantitative estimation of sensor errors. cessing. However, other procedures are also
Analytical redundancy methods using to be encouraged such as physical/chemical
Kalman filtering techniques [25] or dedicated effects on component aging, use of mechan-
observers [26] have been applied in the aero- istic models for system behavior and historical
space industry [27] and in nuclear power records of components.
plant models [28]. The Kalman filter is appli-
cable to both steady-state and transient
conditions, and for linear and nonlinear 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
dynamics. OF ACCIDENTS
A generalized likelihood ratio method of
comparing various fault conditions was Prevention of and protection from acci-
applied to a pressurized water reactor (PWR) dents can be visualized as decisions and
pressurizer level detection system [29]. This actions made at various hierarchical levels.
approach is computationally slow. A more The highest of these levels would be at the
general method of isolating faulty sensors national level (such as the EPA in the United
using a parity-space representation was used States) and lowest level would be a plant
for a boiling water reactor (BWR) suppression operator. Generally speaking, as one ascends
pool signal validation [30]. All these algo- such a hierarchy, data becomes somewhat
rithms make use of physical models requiring imprecise, infrequent and incomplete result-
validation. Such models may fail in a transient ing in a greater or lesser degree of risk associ-
situation. ated with decisions and actions.
A data-dependent modeling strategy and A serial procedure for accident prevention
estimation of sensor errors by appropriate is shown in Fig. 3. The first step in Fig. 3
measurement model has recently been devel- attempts to classify a system, region or space
oped [31] and applied for faulty estimation involving a hazardous material (solid, liquid,
of core-exit thermocouples in the loss-of-fluid gas or mixed phase). Factors which are to be
test (LOFT) reactor. A subsystem of the considered in such a grading of flammability
reactor (in this case the reactor core) was con- and toxicity include the following [34] : com-
sidered for analysis a n d a dual-hypotheses ponents present in the space, volume of space,
testing procedure was applied to the model relative quantity of each component present,
generated from the in-core neutron detector phase of component, pressure, temperature.
and core-exit thermocouple signals. Sensor Some substances can easily be classified as
errors such as bias, excess noise, changes in having a high hazard rating, for example,
response time can be estimated using proper methyl isocyanate.
measurement models. Suitable action(s) will be required for sub-
Utilization of a pair of signals not related stances having a high hazard rating. Further
to the sensor under study could provide infor- measures or quantification may be carried
mation about sensor validation. Such a tech- out (e.g., see the Appendix in ref. 34).
nique for monitoring fluid flow velocities has A literature search of all known physical,
173
TOXIC/FLA~ABLE GAS
ABSORBER ORFLARE
CONTROL AND
PROTECTO
I N SYSTEM VA,~r
.... ~ ~
I- ~TOXIC/F~MMABLE
DETECTOR
~ SAMPLE LINE
I CARBON
MONOXI
CONTAI G DSTREAM
E IMPURITY
NINHYDROGEN
P
Z
Fig. 4. Prevention and protection for a carbon Fig. 5. Provision for leak detection, containment and
monoxide stream containing hydrogen as impurity. protection.
175
reading threshold values, alarms for personnel process, f o o d and other systems. The follow-
and surrounding residential areas. All sensors ing factors aid in making difficult the task of
should be regularly analysed for faults either accident prevention:
by methods described in Section 5 of b y (a) incomplete, and imprecise knowledge of
direct testing. Site personnel should carry por- processes involved;
table gas detectors. The frequency of inspec- (b) human errors;
tion of valves, seals and connections for immi- (c) c o m p o n e n t failure;
nent failures and leaks should reflect the (d) lack and/or absence of sensors at poten-
design, k n o w n failure rates of the c o m p o n e n t tial sources of accidents;
and damage due to process operational char- (e) unsatisfactory man-machine interfaces;
acteristics or internal forces. Further, this fre- (f) lack of proper/effective communication
quency should increase with age of the com- and action within a plant, and between the
ponent. Plant shutdown does n o t imply that plant and the public.
the alarm, control and protection systems are There is room for progress in all these
to be disabled. Seveso should be a lesson that factors. Post-accidents must be used to pro-
an emission can take place after a plant vide information for analysis and improve-
shutdown. ment of safety systems. Post-accident studies
Processes with highly flammable substances such as that of Theofanous [46] are of value
must also be provided with fire and explosion and should be used for further analysis and/or
protection. These must respond quickly. New decision-making in connection with accident
technology should be assessed and be utilized prevention and protection.
depending u p o n the costs and benefits. Micro- The most severe accidents have shown that
computer-based fire and security systems have the following were involved: c o m p o n e n t fail-
the following benefits [43] ; ure or valve opening due to excess pressure,
(a) enhanced response, such as voice mes- lack of early detection due to absence or non-
sages tailored to the location of fire and zones availability of appropriate sensors, and
of safety; absence of alarms and appropriate action.
(b) relative reduction in hardware; Disasters involve short time-scales as discussed
(c) relative reduction in system cost; in Section 3. Proper system and c o m p o n e n t
(d) easily upgraded. design, personnel training, early fault/leak
Dust explosions have also received atten- detection followed by prompt, suitable action
tion due to increased incidence, especially in is therefore most important. Reliability of
connection with food dusts and polymeric various components in a safety system,
powders [44, 45]. Such explosions involve human errors and communication is also
dust particles o f appropriate sizes, air or oxy- important. The e c o n o m y of microprocessors
gen environment, and an ignition source. does allow hardware redundancy of computer
Minimizing the incidence requires consider- support thus providing standby processors in
ations such as: emergency situations.
(a) equipment design Recommendations which can be provided
(b) careful operation such as avoiding any are:
naked flame. (1) computerized control system with
As for the other potential accident situa- manual override;
tions, protection must be provided. Methods (2) components should have a manual over-
are available and discussed b y Palmer [45]. ride due to, for example, power failure or loss
These are based on minimizing cloud forma- of compressed air to pneumatic valves;
tion, containment, separation of plant, vent- (3) fixed fugitive gas sensors, with hard-
ing, inerting and automatic suppression. As ware redundancy in critical situations, at
Palmer points out, it may be wise to use these probable locations of gas in case of release.
methods in combination rather than singly. Periodic testing to be provided;
(4) alarm systems which should be suffi-
cient but not give intolerable false alarms;
7. CONCLUSION (5) system desig n should include intrinsic
The present paper has attempted to inte- safety capabilities [47]. For highly explosive
grate the various aspects related to safety in media, this should also apply to power-
176
consuming devices; design of solenoid valves sensors for gas detection, Measur. Control., 16
for intrinsically safe systems is discussed by (1983) 3 5 6 - 357.
Johnson and Colony [48]. 15 Gas monitoring, Chem. Indust., (July) (1984) 27.
16 J. N. Driscoll, Flue Gas Monitoring Techniques,
All processes involving highly hazardous Ann Arbor Science Publishers, 1974.
materials should be flexible enough for tech- 17 J. Fishetti, Report in IEEE Spectrum, (Jan.)
nology updating. This could involve, for (1983).
example, improved design valve installation 18 T. Wangsmo and H. Riddervild, Improved safety
and software addition to the computer in fire and gas]ESD systems, Proc. ISA Conf.
Advances in Instrumentation, Vol. 38, part 2,
support. Action for implementation during pp. 1053 - 64, 1983.
and immediately after a potential accident 19 R. W. Gerstle, Source sampling, analysis and
must be thoroughly examined and period- monitoring of gaseous emissions, in W. Strauss
ically assessed. All preventive and protective (ed.), Air Pollution Control, part 3, Wiley, 1978,
measures should be considered for safety of Ch. 1.
20 A. L. Jones, Fugitive emissions of volatile hydro-
the plant, personnel and the public. 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,
REFERENCES
TN, September, 1984, University of Tennessee,
Knoxville, paper 20.01.
1 F. T. Bodurtha, Industrial Explosion Prevention 22 G. Flanagan, Selecting a volatile organic chemical
and Protection, McGraw-Hill, 1980. detector, Chem. Eng. Prog., (Sept.) (1984) 3 7 -
2 M. D. LaGrega and L. K. Hendrian (eds.), Toxic 39.
and Hazardous Waste, Butterworths, Boston, 23 T. W. Rhoads, D. G. Cole and R. L. Norton, Opti-
1983. mizing fugitive VOC emission control, Chem.
3 J. Rasmussen and W. B. Rouse (eds.), Human Eng. Prog., (Sept.) (1984) 39 - 44.
Detection and Diagnosis o f System Failures, 24 B. R. Upadhyaya, Sensor fault detection and esti-
Plenum Press, New York, 1981. mation, to appear in Nucl. Safety (1985).
4 D. M. Himmelblau, Fault Detection and Diagnosis 25 A. Gelb (ed.), Applied Optimal Estimation, The
in Chemical and Petrochemical Processes, Else- MIT Press, Cambridge, MA, 1974.
vier, Amsterdam, 1978. 26 R. N. Clark, Instrument fault detection, IEEE
5 A. C. Barrell and D. C. Thomas, Assessing Man- Trans. Aerosp. Electron. Syst., Aes-14 (1978)
agement in Major Hazard Operations, (Jubilee) 456 - 465.
Symp. Series No. 73, Inst. Chem. Eng., 1982, 27 J. C. Deckert et al., F-8 DFBW sensor failure
pp. B9 - B17. identification using analytic redundancy, IEEE
6 H. H. Fawcett and W. S. Wood, Safety and Acci- Trans. Auto. Cont., AC-22 (1977) 795 - 803.
dent Prevention in Chemical Operations, Wiley, 28 R. N. Clark and B. Campbell, Instrument fault
2nd edn., 1982. detection in a pressurized water reactor pres-
7 D. T. Smith, Maintenance of chemical equip- surizer, Nucl. Tech., 56 (1982) 23.
ment, in H. H. Fawcett and W. S. Wood (eds.), 29 J. L. Tylee, A generalized likelihood ratio
Safety and Accident Prevention in Chemical approach to detecting and identifying failures in
Operations, Wiley, 2nd edn., 1982, Ch. 13. pressurizer instrumentation, Nucl. Tech., 56
8 R. Isermann, Process fault detection based on (1982) 489 - 492.
modeling and estimation methods, Automatica, 30 Demonstration o f a PWR Suppression Pool Signal
20 (1984) 387 - 405. Validation Program, Report NP-3641, EPRI,
9 K. Plamping and P. K. Andow, The design of California, July, 1984.
process alarm systems, Trans. Inst. Mess. Control, 31 B. R. Upadhyaya and M. Skorska, Sensor fault
5 (1983) 161 - 168. analysis using decision theory and data-driven
10 M. G. Morgan, Probing the question of technol- modeling of pressurized water reactor subsystem,
ogy-induced risk, IEEE Spectrum, 18 (Nov.) Nucl. Tech., 64 (1984) 70 - 77.
(1981) 5 8 - 64. 32 T. Tsunoda, Application of temperature fluctua-
11 Three Mile Island and the future of nuclear tion to anomaly detection in fuel assemblies, J.
power, IEEE Spectrum, 16 (Special Issue) (11) Nucl. Sci. and Tech. (Japan), 13 ( 1 9 7 6 ) 1 0 3 -
(Nov.) (1979). 110.
12 G. N. Pegg (ed.), High Pressure Measurement 33 W. Reimche and D. Stegemann, Velocity profile
Techniques, Applied Science Publ., Barking, U.K., and flow measurements in liquid sodium by signal
1983. correlation of fast intrinsic thermocouples and
13 R. L. Anderson and T. G. Kollie, Problems in high electromagnetic flowmeters, Proc. SMORN-IV,
temperature thermometry (Critical Review), Reactor Noise, Dijon, France, October, 1984.
Anal. Chem., 6 (1976) 171 - 221. 34 A. S. Goldfarb et al., Organic Chemicals Manufac-
14 L. R. Cooper, M. J. Gibson and M. J. Willett, turing Hazards, Ann Arbor Science Publishers,
Institute meetings report on workshop on new Michigan, 1981.
177
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.
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
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
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
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
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.
Table 5 Comparisons of predictions from extended nomograph (figure 18 in Ref. 12) with vent ratio approach (P,d = 1 .I4 bar a)
-1
avF
at w
(16)
VI RTdCd
P,Cd.
T0.S P SIX
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 \
Table 6 Calculations of K by Harts equations from the & nomographs (Pw~ in bar al
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
to 0.1 bar g. (P,,, (bar g) should be set equal to 0.5 P,,J Rust, E. A. Chem. Eng., 1979, 102
(bar g) in this range). With the appropriate constants Hart, R. L. Plant/Operations Progress 1985,4, 1
Lunn, G. A., Brookes, D. E. and Nicol, A. J. Loss Prev.
detailed in this paper, the simple venting equation Process Ind. 1988,1,123
A, = CAs/AP recommended in NFPA, and the ver- National Fire Protection Association, Guide for Venting of
sion for cubical enclosures presented by Grolmes and Deflagrations NFPA 68 Quincy, MA, USA, 1988
Simpson, L. L. Plant/Operation Progress 1986,5,49
Fauske, A /V*l3 = 6C[P,,/AP]@, can be used to Grolmes, M. A. and Fauske, H. K. Dust Explosion Technology
calculate vent areas that are in close agreement with the Applications, IBC Flammable Dust Explosion Conference, 2-4
extended K,, nomographs. Venting models by Rust and November 1988. IBC, St Louis, MO, USA
Epstein, M. eral. Combustion and Flame 1986.66, 1
Hart have been shown to differ from Heinrichs model Swift, I. and Epstein. M. Plant/Operation Progress 1987.6.98
by only small amounts when Pred is less than 0.2 bar g. 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
References prescribing the size of explosion relief vents - a preliminary
studv. Buildine Research Establishment Note N81/81,198t
1 VDI 3673. Guidelines on Venting of Dust Explosions. Verein
15 Heinrich. H. J: Chemie-lng. Techn. 1966.38, 1125
Deutscher Ingenieure, Dusseldorf, FRG, 1979
16 Chappell, W. G. Pressure/Time Diagram for Explosion Vented
2 Bartknecht, W. Explosions. Course, Prevention, Protection.
Space, AIChE Loss Prevention Symposium 1977, p. 76
Springer-Verlag, Berlin, FRG, 1981
3 Schofield. C. Guide to Dust Explosion Prevention and Protec-
tion. Part 1, Venting, IChemE, Rugby, UK, 1984 rQCrown Copyright
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.
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.
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
Igniter
/
Vent timer/
pressure tran
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-
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.
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
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
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
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
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 0.385 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
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:
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
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
.
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
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
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
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
(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)$pu2 for circular ducting,
duct depends on the relative values of the ejected
where CYis the angle of the bend. When several bends are volume of dust and the volume of the duct, VD,, where
present, the pressure loss can sometimes be obtained by vu = AVL. If VE > VD, then the volume of dust to be
simple summation. burnt equals vu, if VE < VDu, then the volume of dust
Thus the equation for the pressure loss along the duct to be burnt equals VE.
when bends are present is The net rate of volume generation depends on the rate
of combustion of this dust cloud compared with the rate
of outflow from the duct to the open air. However, the
rate of combustion depends on many factors that make
where N is the number of bends of a given angle; FF is an accurate assessment difficult.
the relevant loss term; and the summation sign implies A simple empirical model utilizing the K,, value of the
bends of different angles. dust and the cube-root law has been used throughout
Using dmldt = pAu and the equation of state gives a these calculations. A simple formulation for the rate of
variation of Equation (12) expansion due to combustion, Rr, is
-$ (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
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.
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.
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
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
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
References
The effect of vent ducts on the reduced explosion
pressures of vented dust explosions has been investigated 1 Schofield, C., in Guide to Dust Explosion Prevention and Protec-
tion, Part l-Venting. IChemE Rugby, UK. 1985
and a simple empirical model has been shown to calculate 2 Bartknecht, W., in Explosions Course Prevention Protection,
adequately the pressure increases that result, except in Springer Verlag, Berlin, 1981
some examples with vents of low bursting pressure or 3 Heinrich, H. .I. Chemie-lng-Tech. 1966, 38, 1125
4 VDI 3673, Pressure Release of Dot Explosions, Verein Deutscher
open vents. The obvious use of this data and model is to Ingeneiure, 1983
provide practical guidance for industrial dust handling 5 Bartknecht, W. Plant/Operations Progress 19%. 5 (4), 1%
equipment. Because of the experimental conditions used 6 Lunn, C., J. Loss Prev. Process Ind. 1988, 1, 123
7 Operating Manual for 20 I Spherical Explosion Vessel
in this series of tests, this guidance could be directly 8 British Standards Institute, BS 6713, Part 1, Explosion Protection
associated with the K,, nomograph method for estimating Systems, London, 1986
venting requirements. It has been demonstrated that such 9 National Fire Protection Association. Guide for Explosion
Venting, NFPA 68 Boston, MA, USA, Draft 1987
guidance would agree with that already available at the 10 Eckhoff, R. K. Bulk Solids Handling 1986, 6 (5), 913
top of the st2 dust explosibility Group I1 Eckhoff, R. K., and Fuhre, K. Dust Explosion Experiments in a
(IL = 300 bar m s- ), but would enable users to take into vented 500 m3 vessel, Fourth International Symposium on Loss
Prevention and Safety Promotion in the Processing Industries,
account a wide range of variables and so design for lower 1983, IChemE Symposium Hazards Series No 82
explosion pressures than at present for K,, values less than I2 The Efficient Use of Fuel, HMSO, 1958
300 harms-. 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
Acknowledgements Pressure Venting: paper presented at the Europex Seminar Course
1986. Dust Explosion Venting, Amsterdam, 17-19 November,
The extensive contributions to this project of D. E. 1986
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.
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.
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
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
icontinoedl
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
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
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
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
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
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
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:
Method Comments
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
belting, meeting the specifications of the US Bureau of Some damage will occur even with recommended pro-
Mines, etc., were ignited in tests conducted at FM tection in service.
Research Corporation and, once ignited, burned to
completion with high heat release rate. However, use of Bucket elevators
these will reduce fire frequency because they are more Bucket elevators are often subject to explosions and
difficult to ignite. fires, and numerous examples have occurred in the
Belt conveyors can also overheat either because of a past, particularly with vegetable dusts, for which these
jammed idler roller or, if the belt jams, as a result of elevators have customarily been used3. The design of
drive rollers continuing to run. The belt can also these elevators lead to dust clouds being continuously
generate static electricity, and should therefore be of present during working, particularly at the head and the
anti-static material. boot of the elevator. The buckets are also regularly
Belt conveyors used for dusty systems should be subject to impact and the belt supporting the buckets
enclosed. The free volume within the enclosure is likely can slip on the pulleys and generate frictional heat. As
to be much greater than with a drag link conveyor and a result, a source of ignition and a dust suspension can
the explosion hazard is thereby increased. The drag link be present simultaneously, causing explosion or fire.
type should be used in preference. If a belt conveyor is Modern high capacity elevators, with separate delivery
chosen, then full protection against explosion should be and return legs, have a reduced risk because of the
provided in accordance with NFPA 68. reduced volume per unit weight of dust conveyed. In the
Factory Mutual recommends a number of protective general case, with explosible dusts, other types of
measures for belt conveyors4: Automatic sprinklers or elevator are preferable, and are particularly necessary
water spray protection should be installed for all com- for dusts of more severe explosibility, e.g., those giving
bustible conveyor belts. The belt drive should be inter- maximum rates of pressure rise in excess of
locked to shut down on sprinkler water flow. 5000 lb in- s (35000 kN m-s) in small scale test appa-
ratus. Use of elevators should also be avoided for
0 The system should be hydraulically designed for the dusts known to be readily ignited by friction, e.g.,
operation of ten automatic sprinklers and two small sulphur.
hose streams (sprinkler spacing 100 ft per head). Where the use of bucket elevators is unavoidable,
The system should be designed for a flow pressure of their positioning should be carefully considered and
10 psi on the end sprinkler. (This will result in about regular maintenance is essential. The elevators should
200 gpm for sprinklers and the same for small hose.) be mounted outside the building, e.g., supported by an
In a conveyor enclosure less than 15 ft wide, this can outside wall, and the intake and delivery points should
usually be accomplished by installing a single con- preferably be isolated from the rest of the dust handling
tinuous supply line with an unlimited number of plant by means of chokes. The elevator casing should be
sprinklers. a fire resistant construction, sufficient to retain a fire,
0 Water supplies should be adequate for one hour of dust-tight and of sufficient strength not to rupture in the
use. event of an explosion. To meet the strength require-
l High initial water pressures should be used to over- ments, the casing should be provided either with auto-
come friction and elevation loss. A constant dis- matic suppression or with explosion relief at the head
charge pressure system would be one method of and the boot, with vent areas calculated per NFPA 68.
providing such initial high pressure. Long elevators, say more than 6 m (20 ft), may require
additional relief at intervals along the casing to ensure
A small hose or equivalent should be provided at that no point is too remote from a vent.
suitable intervals with sufficient FM-approved li in hose Particular care should be taken to ensure that flame
at all locations to reach any part of the conveyor system. burning dust, etc., discharged from the vents during an
Small hose connections can be made to the sprinkler explosion cannot injure operators or damage nearby
piping. A suitable alternative to this is the provision of plant. Provision of ducting or deflectors over the vents
hydrants and easy access to the conveyor by fire fighting may be required. Where it is unavoidable to site a
equipment. All weeds, brush and trees should be cleared bucket elevator inside a building it is desirable for the
from underneath and at least 25 ft from both sides of internal pressure to be slightly below atmospheric to
conveyor supports. Unprotected combustible buildings minimize leakage of dust. Discharge of combustion
and similar exposures should be removed. Each con- products from vents to the outside of the building is
veyor belt system should be provided with tamperproof essential, and the design requirements of ducting from
devices arranged to automatically shut off driving vents should be in accordance with NFPA 68. The need
power in the event of greater than 20% belt slowdown for dust tightness and adequate casing strength and fire
or misalignment of belts. In addition, interlocking resistance should be rigorously met.
devices should be arranged to shut off power on all Steps should also be taken when designing bucket
contributing conveyors. Where conveyor belts are elevators to minimize the generation of ignition sources.
critical to plant operations, spare belts in quantity These steps may include the provision of strong fixing
consistent with expected fire damage (with recom- for the buckets and strong bearings for all shafts,
mended protection in service) should be kept on hand. external to the casing, provided with detectors for
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
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.
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
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
18
Maize starch explosions in a 236 m3 experimental silo with vents in the silo wall: R. K. Edhoff et al.
19
Maize starch exdosions in a 236 m3 experimental silo with vents in the silo wall: R. K. Eckhoff et al.
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.
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
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
0 1 2 3
larger 5.7 m2 roof vent. The pressure versus time histor-
Time (s)
ies generated by the explosions vented through the silo
Figure 8 Pressure ~ersos time history of explosion in the 236 ms
wall were, apart from a general reduction of the peak silo with a 4.4 m2 vent in the silo wall. Ignition 13.5 m above silo
pressures, similar to those found with roof vents. bottom. Pressure transducer located 3 m above silo bottom
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 pressure rise. This corresponds closely to the 17-18 Hz
3-6 Hz (depending on gas temperature) are very similar oscillations observed previously with ignition at 4.5 m
to those observed in roof venting experiments with above the silo bottom using a roof vent. The exact
ignition at 13.5 rnlm3. The frequency corresponds to the nature of these oscillations has still not been resolved.
first harmonic of a standing wave in a one-end-open However, as indicated previously2S 3, one possible
pipe of length 22 m. explanation could be generation of some kind of
Figure 9 shows a comparison of two pressure versus Helmholtz oscillations across the vent opening during a
time histories generated in the wall vented silo in two short phase of the explosion, when the dust cloud
apparently identical experiments, with ignition at 4.5 m outside the vent explodes. This will occur just after the
above the silo bottom. The upper trace exhibits a flame front inside the silo has reached the vent opening,
distinct spiky, oscillatory pattern after the first main 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.
0.40
0.30 -
z
2
jJ 0.20 -
j 0.10 -
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)
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 97
HENRY WECKMAN
Technical Research Centre of Finland, Fire Technology Laboratory, SF-02150 Espoo (Finland)
(Received December 11, 1986; in revised form January 26, 1987)
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 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 ](}
"\
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.
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
< 2 0 0 pm) peat dust, the maximum explosion the morphology of the particles. The influ-
pressure and the g s t value can be estimated ence of these is, however, small compared
with a reasonable accuracy from the moisture with the moisture content and mean particle
content and mean particle size (with, for ex- size. For eqns. (1) and (2) it was found that
ample, eqns. (1) and (2) or similar); R 2 was 89.1% and 90.9%. If the equations
- - f o r homogeneous moist (moisture content are computed for the twelve samples of
>35% W/W) and/or coarse (mean particle Table 2 prepared from one sample of milled
size ~ 2 0 0 gm) peat dust, it is often possible peat, it is found that R 2 is 92.2% and 93.2%.
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- REFERENCES
ilar), and
1 K. Mutanen et al., Safe Production and Use o f
- - f o r peat mixtures composed of dusts with Domestic Fuels, Part 1, A Preliminary Study,
very different moisture contents and/or mean Research Notes 323, Technical Research Centre
particle sizes, it is not possible to estimate the of Finland, Espoo, 1984, 118 pp. (in Finnish).
maximum explosion pressure and the K s t 2 E.-W. Scholl, Explosionsverhalten yon Torf, Torf-
forschung GmbH, Bad Zwischenahn, 1978, 21 pp.
value reliably, or even to predict whether the
3 ISO 6184/1, Explosion Protection Systems. Part
mixture is explosible or not, from the mean 1: Determination o f Explosion Indices o f Com-
moisture c o n t e n t and the mean particle size bustible Dusts in Air, International Organization
of the mixture. for Standardization, Geneva, 1985, 5 pp.
The explosion properties of peat dust are 4 R. Siwek, Development o f a 20-1itre Laboratory
Apparatus and its Application for the Investiga-
also affected by other properties than the
tion o f Combustible Dusts, Ciba-Geigy Ltd.,
moisture content and mean particle size. Such Basle, 1985, 189 pp.
properties are, for instance, the degree of hu- 5 W. Bartknecht, Explosionen, Springer, Berlin,
mification, the botanical composition and 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
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.
*Paper presented at the 5th International Symposium on Loss Prevention and Safety Promotion
in the Process Industries, Cannes, France, September 15-19,1986.
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).
Thesilo
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 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
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
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.
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
Fig. 3. Example of jagged pressure pulse resulting from ignition 3-6 m above silo bottom.
I
I ,
1
8 I
I
I
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
2 3 4 5 6 7
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
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
- WALL
r
I I 7 n
0 0.2 0.L 0.6 0.8 1.0 1.2
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
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.
CONCLUSIONS
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 - 164 153
TABLE ]
Percentage of dust explosions in the grain, feed and food industry assumed t,) be initiated i)y 'friction sparks" or
unknown sources
1 535 20 46 66 1860-1973
1 128 17 27 44 1949-1973
2 91 54 18 72 1941 -1945
3 137 9 62 71 1958 -1975
4 83 28 5 30-35 1965 -1980
The ability of metal sparks or hot spots [9, 10]. The specific objective has been to
from grinding and cutting to ignite dust clouds answer the following question: under what
has been demonstrated by several researchers. circumstances of impact energy, impacting
The experiments by Leuschke and Zehr [5], velocity and type of moving object and anvil,
published in 1962, are probably the first ones can single impacts between solid bodies ini-
in which dust clouds were ignited by grinding tiate a dust explosion in a cloud of grain dust
wheel metal sparks. However, no clouds of in air?
organic dusts ignited. Zuzuki et al. [6] ignited
different coal dusts using both sparks and hot
2. EXPERIMENTAL APPARATUS
spots from a piece of steel in contact with a
grinding wheel rotating at 23 - 47 m/s peri- 2.1. Impact generator
pheral velocity, Later Allen and Calcote [7] After having considered various principles
conducted similar experiments in which metal for constructing a single-impact generator
sparks were generated by pressing a steel rod suitable for laboratory work, it was decided
against a rotating grinding wheel. By retarding to build the apparatus shown in Fig. 1. The
and focusing the spark stream, it was possible length of the rigid, spring-loaded arm, the
to ignite clouds of natural organic dusts such 'hammer', from the centre of rotation to the
as maize starch and wheat grain dusts. An- tip of the test object, can be varied by means
other recent, and very comprehensive study is of an adjustment screw. The condition for
Ritter's [8] investigation. This work confirm- obtaining impact is that the arm is slightly
ed that metal sparks generated by grinding, longer than the perpendicular distance from
rubbing and repeated impacts can ignite the arm axis to the anvil. By careful adjust-
clouds of several dusts, including cellulose and ment of this small difference it is possible to
lycopodium. generate impacts of different impact energies
The second situation, in which metal sparks without changing the impact velocity. The
may be generated in industry, is accidental cylindrical tip of the test objects that impact-
single impacts between solid bodies. In this ed on the anvil material had a reduced diam-
case relevant evidence of the ignition poten- eter of 4 mm. Details of the end of the arm
tial has been lacking. However, as long as the are shown in Fig. 2. Springs of different
necessary conditions for initiation of dust strengths were used for generating a range of
explosions by such impacts remain unidenti- impacting velocities of approximately 8 - 25
fied, one is forced to maintain the hypothesis m/s. Fine adjustment of the velocity was pos-
that such sparks may be hazardous in general. sible by adjusting the spring tension.
This in turn forces industry to take precau- When performing an impact test, the arm is
tions that may be superfluous, and causes fear first twisted to the start position, whereby the
that may be unnecessary. helical tension spring is stretched. In this
This unsatisfactory state of affairs created position the arm is locked by a pneumatically
the motivation for the present work which operated device that can be released electri-
has been reported in greater detail elsewhere cally. After having been released, the arm
155
probe in various locations, the concentration temperature of the body ff the spectral emis-
development of the dust cloud with time, at sivity is approximately constant in the range
any desired location, could be determined. of the spectrum of wavelengths used in the
measurement.
2.4. Timing system
A microprocessor-based 4-channel timer 2.8. Determination of minimum electric spark
was used to control the system. The timer energies for ignition of dust clouds
controlled the triggering of the impacting arm, These energies were determined by means
dispersion of the dust, as well as high-speed of the method described by Eckhoff [15].
movie camera and other photographic equip- This method, which is based on using electric
ment. sparks of known energies (integration of
power-versus-time curves), is essentially iden-
2.5. Equipment for measuring the velocity of tical with the new method proposed recently
the arm and the impact energy in W. Germany [16].
The kinetic energy of the arm at any instant
was calculated from its mass distribution and
instantaneous velocity. By measuring the velo- 3. DUSTS USED IN THE EXPERIMENTS
city of the arm immediately before and after
the impact, i.e., immediately before and after Since maize starch is one of the agricultural
'6 o'clock' position, the kinetic energy lost in materials that have often been involved in
the impact could be calculated. Two sets of dust explosions in the U.S.A. and is one of
reflective opto-switches were used for this the most sensitive to ignition, it had been de-
purpose. The opto-switches were mounted on cided to use this material in the experiments.
the wall of the test rig, with narrow clearance Native maize starch supplied from Amylum,
to the passing arm. A small mirror was fixed Aalst, Belgium, was used. On receipt the
to the arm, and an electrical signal was gener- starch contained 10 - 11% moisture, which
ated when the mirror passed the opto-switches. increased the minimum ignition energy con-
siderably compared with that of dried starch.
2.6. Photographic equipment In order to ensure maximum sensitivity to
Different kinds of photographic equipment, ignition, the starch used in the majority of
including ordinary single-frame cameras, video the present experiments was first dried for
camera, 16 mm movie camera and 16 mm four hours at about 105 C to remove the
high-speed movie camera, were used. The high- moisture.
speed camera (Hycam 4043020) was primarily A few other dusts of different minimum
used to make black and white negative film electric ignition energies (see Table 2), were
recordings at 1000 and 2000 frames per sec- used in order to investigate how the ignition
ond. The films were developed and examined probability in impact ignition tests ~}aried
at CMI immediately after each experiment. with the minimum electric spark ignition
energy.
2.7. Temperature measurements by a four-
wavelength optical pyrometer
In order to measure the temperature of the 4. CONCENTRATION DISTRIBUTION AND
sparks, a four-wavelength optical pyrometer ELECTRIC SPARK IGNITION SENSITIVITY OF
was used [14] (see 'Acknowledgements'). THE EXPERIMENTAL DUST CLOUDS
The method is based on the determination of
the relative magnitude of the radiation inten- 4.1. Dust cloud concentration distribution
sities from the investigated solid body at four The first phase of the experiments concen-
different wavelengths. The intensity ratios are trated on formation of a suitable experimen-
compared with the radiation intensity ratios tal dust cloud of sufficient volume and low
from black bodies of various temperatures. By minimum ignition energy. This was achieved
identifying the black body temperature that by dispersing 8 g of dried maize starch at 2 bar
gives the best fit, a best estimate of the 'colour overpressure in the compressed air line. With
temperature' of the solid body is obtained. 8 g of dust, the theoretical average dust con-
The 'colour temperature' is close to the real 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)
1.0
Temperatures of titanium sparks were also
0.9 measured 140 mm and 240 mm downstream
I 5m/,~ of the impact point. The arm velocity just
O.B
prior to impact was 24 m/s and the net impact
0.'/ energy about 20 J in all the experiments. In
~ 0.6 the case of grinding-wheel sparks, tempera-
tures were measured 40 mm downstream of
E, o.s
>,- the contact point between the steel rod and
O.l. the grinding wheel. The peripheral velocity of
g o.~ the grinding wheel was 29 m/s.
c)
The results are summarized in Table 3. Ap-
R~ 0.2
proximately 10 tests were carried out for each
0.1 combination of test object and anvil material,
using different combinations of detecting
0.1 I 10 100 1000 wavelengths in the pyrometer.
SPARK ENERGY , mJ
Fig. 6. Effect o f i m p a c t velocity o n ignition sensi- TABLE 3
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- 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
gering o f t h e electric spark is 0.3 - 10 ms. E n v e l o p e s the four-wavelength pyrometer
enclose t h e e x p e r i m e n t a l points.
Test o b j e c t Anvil material Spark
the peripheral velocity of the arm was 22 m/s, temperatures
the electric spark energy necessary to ignite C~c)
the turbulent dust cloud was two orders of Titanium R u s t y steel 2030 - 2900
magnitude higher than the 4.5 mJ necessary
Mild steel, St37 R u s t y steel 1730 - 2 3 8 0
to ignite the quiescent dust cloud. At 9.5 m/s
the electric spark energy required for ignition Alloy steel R u s t y steel 1530 - 1980
was approximately one order of magnitude Mild steel, S t 3 7 Concrete 2530 - 2730
higher. This relative ranking of the ignition Mild steel, St37 Rotating 1600 - 1730
sensitivities of the clouds only applies when grinding wheel
ignition occurs close to the point of impact.
For spark~luster ignition away from the
impact point, it would be necessary to perform The results show that the highest tempera-
separate electric spark ignition tests with the tures were measured on sparks produced from
spark gap at the location where ignition takes impacts of titanium against rusty steel, and
place, and with the appropriate delay between mild steel against concrete. The temperatures
passing the impacting arm and firing the elec- of titanium sparks at 140 - 240 mm down-
tric spark. stream of the impact point were generally
higher than at 40 mm. There was a clear ten-
dency of increasing temperature with increas-
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 ing length of travel, i.e., increasing burning
MECHANICAL IMPACTS AND GRINDING time.
The measured temperatures of sparks from
Temperatures of single-impact sparks from mild steel against concrete were also high. At
ordinary mild steel (St37), alloy steel and tita- 40 mm from the point of impact, some steel
nium were measured by means of the four- spark temperatures were even higher than
wave optical pyrometer. The anvil materials some titanium spark temperatures. However,
used were rusty steel and concrete. For the boiling point of titanium is higher than
comparison, temperatures of sparks from mild that of iron, indicating that on average the
steel in contact with a rotating grinding wheel titanium sparks should attain higher tempera-
were also measured. In the case of the single tures than sparks from steel. This was also
impacts the temperatures were measured on observed.
sparks 40 ram 'downstream' of the impact The results from the tests with mild steel
point and 10 mm above the anvil surface. as spark-producing material showed a large
159
variation in the spark temperatures, depend- temperature becomes constant and indepen-
ing u p o n the anvil material. The lowest spark dent of the pressure. A similar observation
temperatures were obtained in the tests with was made b y Allen and Calcote [7] who
the rotating grinding wheel. The observed found that low pressure sparks were relatively
temperatures of 1600 - 1730 C were very small and irregular in shape. The irregular
close to those for mild steels measured b y shape indicated that the sparks had not
Hardt [17 ], and also in agreement with Wahl's reached the melting point. The high pressure
[18] t w o ~ o l o u r p y r o m e t e r measurement of sparks were, however, spherical and coated
temperatures of steel sparks from grinding with oxide, which means that the particles
wheels. had melted and burnt.
The temperature measurements showed In the present investigation, the maximum
that impact sparks from mild steel attained contact pressure was estimated b y measuring
higher temperatures when concrete was used the static force exerted on the anvil by the tip
as wall material than when rusty steel was of the impacting arm, when the arm was left in
used. This may in part be due to the heat the vertical position. The measurement was
conductivity of concrete being much lower accomplished b y mounting a mechanical load
than that of steel, causing more of the fric- cell between the tip of the arm and the anvil.
tional heat generated during an impact to be The force was distributed over a circular area
retained in the particles. This m a y provide of diameter 4 ram. The static pressure at the
more favourable initial conditions for com- contact area as a function of excess in arm
plete combustion of the particles and hence length is given in Fig. 7.
contribute indirectly to a significantly higher
Pressure.
spark temperature. [MPa)
The observation that the temperatures of 200
quency of ignition of maize starch by tita- very few sparks compared with titanium im-
nium sparks, was studied. The peripheral pacts. Typical numbers of sparks produced
velocity of the arm was varied using different from mild steel were 5 - 50. Immediately after
springs, whereas the net impact energy was impact, no or very few sparks were visible.
varied by adjusting the excess arm length. But 3 - 5 ms after impact, some sparks became
The results are given in Fig. 8. The frequen- visible, and after 10 - 20 ms the number of
cy of ignition for each combination of arm sparks visible on the f i l l was at maximum.
velocity and impact energy is based on obser- Impacts with steel never generated a sufficient
vations from 10 or 20 tests. The scatter (+1 std. number of sparks to ignite clouds of dried
dev.) is indicated. At constant velocity of maize starch.
approach, the frequency of ignition decreases The films of ignition by sparks from tita-
systematically with decreasing impact energy, nium against rusty steel showed that the dust
as would be expected. Low-energy impacts clouds could be ignited in two different ways.
produce fewer sparks than high-energy im- (Reference 9 contains reproductions of some
pacts, and if the number of sparks is low, the of the film pictures.) First, ignition frequently
ignition probability is also low. occurred very close to the point of impact,
immediately (i.e., 0 - 5 ms) after the impact
10 had taken place. At that m o m e n t an extremely
luminous hemispherical 'pocket' close to the
09
impact point, with a radius of a few centi-
08, meters and containing 100 - 1000 sparks, was
07
observed. In addition to the great number of
discrete particles, a 'thermite' flash probably
c06 also contributed to the formation of the lu-
o
~ m/s minous pocket. Impacts of titanium against
c,
concrete also generated high spark densities
~ 0~
close to the point of impact, but not the addi-
03 tional continuous, luminous phase as was seen
with impacts against rust. Flame propagation
02
started somewhere in the region of the 'lumi-
0t nous pocket', but could neither be referred
specifically to one single metal particle nor to
I i i i i i i t i i
2 z. 6 8 10 12 ~z. 16 18 ~0 the hot spot generated at the point of impact.
Impact energy All the high-speed films of ignition of maize
Fig. 8. F r e q u e n c y o f ignition o f clouds of maize starch by titanium impacts against rusty steel
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 showed that ignition and flame propagation
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.
started in this region.
The second mode of ignition was only ob-
The impacts of similar net energies, but of served in some of the experiments with lyco-
different velocities of approach, i.e., different podium. In this case ignition occurred 1 0 - 50
turbulence levels, gave decreasing frequency ms after impact at a distance of 10 - 30 cm
of ignition with increasing turbulence. This is downstream of the point of impact. In this
in agreement with the results from the experi- region the number of sparks per unit volume
ments where the minimum electric spark ener- of dust cloud was considerably lower than in
gy for ignition was determined when the arm the 'luminous pocket' close to the point of
was moving through the dust cloud at the impact. But still it was not possible to trace
m o m e n t of spark discharge (Section 4.2). the ignition back to any single metal spark.
'Explosion' and fragmentation of single metal
sparks were observed frequently, but never
8. H I G H - S P E E D F I L M I N G seemed to cause ignition of the dust cloud.
The high-speed films suggest that a cluster
The 1000 - 2000 frames per second high- of a large number of sparks is necessary to
speed films confirmed that impacts with mild cause ignition of a cloud of grain dust. Im-
steel against rusty steel or concrete produced pacts with titanium as spark-producing
161
material generated such conditions, and usual- the spark f o r m a t i o n process. Impacts by car-
ly resulted in ignition. 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
9. RESULTS FROM IGNITION TRIALS WITH was removed, so t h a t the impact t o o k place
DIFFERENT DUST CLOUDS AND IMPACTING on t he gravel or crushed stone fraction of t he
MATERIALS concrete, visible sparks were produced. For
A series o f impact tests with various mate- this reason experiments using concrete as
rials was carried o u t in clouds o f dried maize anvil material were always carried o u t after
starch in order to identify combinations o f removal of the cem ent layer.
impacting materials t hat would cause ignition. Ignition experiments with impacts between
The anvil materials used comprised typical titanium and several anvil materials were also
c o n s t r u ctio n steels with either fresh or rusty carried out. T he combinations o f materials
surfaces, and different concrete ~ m p l e s . Sup- used, and t he results f r o m the tests, are given
p l e men tar y tests were c o n d u c t e d with anvils in Table 5. Dust clouds o f dried maize starch
o f different minerals and grinding wheels. were used in all tests.
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 TABLE 5
n o t available, b u t t h e y are likely to be repre- Results from the experiments with impacts between
sentative o f the range of steels in c o m m o n titanium and different anvil materials
use. Th e combinations o f steel test objects
and anvil materials investigated, are given in Titanium against Ignition ?
Table 4.
-- rusty, mild steel Yes
-- concrete Yes
TABLE 4 -- sandstone Yes
Combinations of steel test objects and anvil materials - granite, gneiss, quartz
- No
used in the experiments - grinding wheel
- No
- grinding wheel
-
T he n u m b e r o f sparks produced d e p e n d e d
Acid resistant steel, alloyed steel, hardened steel or
stainless steel against u p o n the surface t e x t u r e of t he anvil material.
-- mild steel A rough surface p r o d u c e d m ore sparks t han a
-- concrete s m o o t h surface, and ignitions were onl y ob-
-- granite tained f r o m impacts against anvil materials
-- grinding wheel having a relatively rough surface.
T he i m port ance of t he surface properties
T he peripheral velocity of t he arm in the of t he anvil material was emphasized during
test rig just b ef o r e impact was 24 m/s. T he the experiments with titanium against rust y
calculated average impact energy was a b o u t steel. Surfaces apparently covered by t he
20 J. After each test, t he anvil was displaced same kind o f rust, would in some cases pro-
slightly, so th at every impact t o o k place on duce large numbers o f powerful sparks t h a t
fresh anvil surface. ignited the dust cloud, whereas in o t h e r cases
T he experiments revealed t h a t neither mild considerably fewer sparks were produced and
steel against concrete, nor any o t h e r combina- no ignition occurred.
tions o f steel objects and anvil materials, pro- A few experiments with zirconium as spark-
duced an y sign o f ignition o f t he dried maize producing material were also carried out.
starch cloud. Rust y steel was used as anvil material. The
F r o m i n t r o d u c t o r y tests with concrete as impacts produced m ore sparks than titanium
anvil material, it was f o u n d t hat the nature against rusty steel, and ignition of clouds of
o f the surface was of decisive i m por t ance for dried maize starch were easily obtained.
162
(5) In most cases, ignition by titanium During his stay at CMI in the a u t u m n of
sparks was observed very close to the point 1985, Dr. R. Klemens, Technical University
of impact. However, ignition was occasionally of Warsaw, gave invaluable help, both theo-
observed 10 - 30 cm downstream of the im- retically and experimentally, with the mea-
pact point. Ignition by one single metal spark surement of metal spark temperatures by a
was never observed. A fairly dense cluster of four-wavelength optical pyrometer borrowed
sparks seemed to be necessary for igniting from TU of Warsaw.
the clouds of maize starch.
(6) A n y moving object in the dust cloud REFERENCES
reduces the ignition sensitivity o f the cloud in
the vicinity of the object by inducing turbu- 1 H. de Goijer, Th. M. Groothhuizen and M. E.
Reinders, Literature Investigation into the Dust
lence. Experiments with titanium against rusty Explosion Danger in Industries Storing and Pro-
steel showed that at a given net impact energy, cessing Cereals and Flour, Report from Tech-
the ignition frequency dropped when the im- nological Laboratory TNO, October, 1975.
pact velocity increased. Thus, at a given net 2 D. J. Rasbash, Fires Involving Dusts, Technical
impact energy, objects generating low turbu- Booklet No. 5, Fire Protection Association,
London, 1949.
lence represent a greater ignition hazard than 3 M. Verkade and P. Chiotti, Literature Survey of
objects generating high turbulence. Dust Explosions in Grain Handling Facilities:
(7) At a given impact velocity, the frequen- Causes and Prevention, Energy and Mineral Re-
cy of ignition by titanium sparks decreased sources Institute, Iowa State University, Ames,
systematically with the impact energy, as IA, 1976.
4 H. Beck and A. Jeske, Documentation Staub-
would be expected. explosionen, Analyse und Einzelfallderstellung,
(8) The microscopic nature of the anvil sur- Report No. 4/82, Berufsgenossenschaftliches
face is decisive for the spark formation pro- Institut fiir Arbeitssicherheit, St. Augustin,
cess. For example, impacts against the gravel F.R.G., 1982.
or crushed stone fraction of concrete pro- 5 G. Leuschke and J. Zehr, Ziindung yon Staub-
lagerungen und Staub/LuftGemischen dutch me-
duced considerably more sparks than impacts chanisch erzeugte Funken, Arbeitsschutz, 6
against a fresh concrete surface covered with (1962) 146.
cement. 6 T. Zuzuki, S. Takaoka and S. Fujii, The ignition
(9) The overall practical conclusion of the of coal dust by rubbing, frictional heat and sparks,
present investigation is that up to net impact Proc. Restricted Int. Conf. of Directors of Safety
in Mines Research, July 1965, Safety in Mines
energies of 20 J, tangential accidental single Research Establishment, Sheffield, England.
impacts between various types of steel, and 7 J. Allen and H. F. Calcote, Grain Dust Ignition
between steel and rusty steel or concrete, are by Friction Sparks, SMS-81-049, National Grain
unable to ignite clouds of grain and feed dust, and Feed Association, Washington, DC 20005,
or flour, even if the dusts are dry. Impacts of 1981.
8 K. Ritter, Die Ziindwirksamkeit mechanisch
standard quality aluminium against rusty steel erzeugter Funken gegeniiber Gas/Luft- und
did not even generate any visible sparks. In Staub/Luft-Gemischen, Dr.-Ing. Dissertation, Uni-
the case of titanium or zirconium, the sparks versit/it Fridericiana, Karlsruhe, 1984.
produced may initiate explosions in clouds of 9 G. H. Pedersen, Initiation of Dust Explosions by
dried dusts, but not in clouds of dust contain- Heat Generated during Single Impact between
Solid Bodies, Fire and Explosion Research Re-
ing 10% moisture or more.
port No. SMS-86-055, National Grain and Feed
Association, Washington, DC, 1986. (Prepared by
Chr. Michelsen Institute, Report No 833310-2,
ACKNOWLEDGEMENTS December 1985.)
10 G. H. Pedersen and R. K. Eckhoff, Initiation of
dust explosions by heat generated during single
CMI wishes to express its sincere gratitude impact between solid bodies, Proc. 2nd Int.
to the National Grain and Feed Association Colloquium in Dust Explosions, Jadwisin, Poland,
(NGFA), Washington, DC, U.S.A., for the November, 1986, Report No. 863351-1, CMI,
substantial financial support rendering this 1986.
work possible, and for unfailing constructive 11 R. K. Eckhoff, Use of (alP/dr)max from closed-
bomb tests for predicting violence of accidental
advice and support throughout. This paper is dust explosions in industrial plants, Fire Safety
a condensed version o f a comprehensive J., 8 (1984/85) 159. (Paper presented at the First
report prepared by CMI for NGFA [9]. Int. Colloquium on Explosibility of Industrial
164
Dusts in Baranow, Poland, November 8 - 10, 1984. 14 K. L. Cashdollar and M. Hertzberg, Infrared
12 K. L. CashdoUar, I. Liebman and R. S. Conti, Pyrometers for measuring dust explosion temper-
Three Bureau o f Mines Optical Dust Probes, Re- atures, Optical Eng., 21 (1) {1982) 82 - 86.
port o f Investigations No. 8542, U.S. Bureau of 15 R. K. Eckhoff, Towards absolute minimum igni-
Mines, 1981. tion energies for dust clouds?, Combust. Flame,
13 R. K. Eckhoff, K. Fuhre and G. tt. Pedersen, 24 (1975) 53 - 64.
Vented Maize Starch Explosions in a 236 m 3 16 W. Berthold, Bestimmung der Mindestziindenergie
Experimental Silo, Fire and Explosions Research yon Staub/Luft-Gemischen, Progress Report No.
Report No. ESV-86-070, National Grain and 134, VDI-Verlag, Diisseldorf, 1987.
Feed Association, Washington, DC, 1986. (Pre- 17 L. Hardt, Temperaturme~ungen an Schleiffun-
pared by Chr. Michelsen Institute, Report No. ken, Arbeitsschutz, 15 (1954) 430 - 436.
843307-2, December, 1985.) 18 H. Wahl, Z. Angew. Phys., 12 {1960) 60 - 62.
Erratum
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.
Basic features
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 (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
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.
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
Flarie detector 2
iressure nrohe C
Fla:~e detector i
Bolted-steel-vlat~
silo wall
11111tlon sourci' -.
xLorlcrete silo
Sottoll
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
TABLE 1
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
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.
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
100
0
0 1 2 3 4
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.
\
1+1: 103 m3 experiments
I I I I I I I I I I I I a
0 5 10 l!5
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).
z 0.3
2 0.2
2 0.1
5;
Z 0
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
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
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.
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
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
D,-m
D,--0
which cause
.%2,X,--
---cm
2*,X1
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
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
tars. The wheat dust and corn dust samples one because each individually did not provide
were collected from systems that employed enough material for an esplosion test.
cyclones to separate dust from air, and the The coarse fraction from the 250 mesh
grain sorghum dust samples were collected sieving of each dust sample was also further
from a system that used a baghouse to make divided into size fractions with a series of
the separation_ Each sample was 2 to 3 kg in sieves having Tyler mesh numbers of 65,115,
weight, and a sieve with a 1.0 mm mesh 150, li0, and 200 (corresponding openings of
opening was used to remove very large trash. 208, 124, 105, 88, and 74 pm). These sis
One hundred pounds of cornstarch were fractions were accumulated from O-100 kg
obtained in bulk from a mill (General Mills, portions, which were sieved for 15 min
Minneapolis, MN). on a Ro-Tap shaker (W_ S_ Tyler Company,
Cleveland, OH)_ The fraction on top of the
Separation of dust samples into size fractions 65 mesh sieve was considered trash and
A 250 Tyler mesh sieve was used to initial- discarded because it coniiained a wide range of
ly divide each dust sample into a coarse particle sizes. Thus, each type of dust was
fraction (having particle diameters approxi- separated into 11 size fractions by utilizing
mately greater than 61 pm) and a fine frac- the same air velocities for air classification
tion_ This separation was performed because and the same series of sieves for sieve classifi-
the series 6000 Microparticle Classifier used cation Note that size fractions 1 through 6
to further separate the fine fraction of each were from the air classification and size
sample could not effectively classify grain fractions 7 through 11 were from the sieve
dust particles with diameters larger than classification_ During sieve classification,
61 pm.. a large degree of carry-over of particles
The series 6000 Microparticle Classifier with diameters smaller than the sieve aper-
(manufactured by A.E. Bahco in Sweden and tures occurred unless the sieving was done
distributed by Harry W. Dietert Co_, Detroit, carefully_
MI) was employed to separate grain dust
particles with diameter less than 61 ,xm_ The Determination of particle size distribution
classifier used the combined effects of cen- The particle size distribut.ion in each of the
trifugation and elutriation to separate O-02 kg eleven size fractions was determined by the
of dust into a fine fraction and a coarse AACC method SO-10 [ 3 3, namely, the Whitby
fraction. The dust particles were subjected to sedimentation method which classifies parri-
a centrifugal force which was opposed by a cle size hydrodynamically_ By centrifugal
current of air. The fine fraction, composed of sedimentation, dust is allowed to settle in a
dust particles with a terminal velocity less capillary tube filled with a liquid termed the
than the air velocity, was blown into a collec- sedimentation liquid_ The diameter obt.ained
tor_ The remaining dust, the coarse fraction, corresponds to the diameter of a sphere that
was thrown by centrifugal force into another falls with the same velocity as the real parti-
collector. cles_ Though the physical dimension of the
The fine fraction from the 250 Tyler mesh particles obtained by this method might be
sieve of the wheat dust, corn dust, graili different from those of the real particles in
sorghum dust, or cornstarch was further many cases, the distribution is one for spheres
separated into eight size fractions using the that behave hydrodynamically and is obtained
microparticle classifier_ Size fractions were by measuring the cumulat.ive volume of dust.
obtained by performing a series of separa- Benzene was used as the sedimentation
tions, each with a progressively higher air liquid for the corn, grain sorghum, and wheat
velocity_ The finest size fraction was fist dust; isopropyl alcohol was used for the
separated out of the entire fine fraction by cornstarch. The dust was initially dispersed in
the lowest air velocity. After increasing the air a feed solution consisting of the sedimenta-
velocity, the next fraction was divided again tion liquid and naptha, which was then placed
into a fine fraction and a coarse fraction. This on top of the sedimentation liquid in the
was repeated until eight size fractions were tube_ To decrease the settling time, the
obtained; however, the three size fractions particles were centrifuged for increasing
with the finest particies were combined into lengths of time at 600,1200, and 1800 rpm.
268
The weight percent of the total dust sample Whitby sedimentation method a limit on the
that had settled out was determined by maximum particle diameter that can be
measuring the height of the settled dust measured accurately. The first cumulative
column. The diameter of the largest particles weight percent measurement is recorded at
that settled out was determined from the rpm the moment when particles with a diameter
of the centrifuge and the length of time the equal to the maximum diameter have settled
sample was centrifuged. to the bottom of the capillary tube. Particles
with diameters larger than the maximum
Determination of composition dianreter will have already accumulated in the
The composition of each size fraction was tube before the first reading, and this could
characterized by determining its content of result in the first cumulative weight percent
moisture, ash, protein, and starch and fiber. reading of less than 100%.
The weight fraction of moisture, ash, and In this case, it is more critical to have data
protein was determined by the AACC meth- in the lower, rather than upper, end of a
ods 44-40,08-01, and 46-10, respectively [3]. weight distribution, because the number of
The weight fraction of starch and fiber was particles per unit weight of dust is larger at
obtained from the difference. the lower end of the distribution than at the
upper end. All of the distributions, except for
the size fraction of cornstarch, contain data
RESULTS AND DISCUSSION with cumulative weight percents less than
14.0%.
Particle size The parameter of the log normal approxi-
Particle size distribution . mation for each distribution for data with
The geometric mean diameter D,q3 and the values of 121 < 2 was recorded. Only data
natural logarithm of the geometric standard with 121 > 2 were discarded because the
deviation, In (z&~,of the log normal approxima- Whitby sedimentation method does not give
tion of the actual particle size distribution for accurate results for data in this region [4].
data with 121 < 2 were calculated. The Note that only 16 of the 43 distributions are
coefficient of determination R2 and the 95% from size fractions that were sieve classified
confidence interval for the population correla- and the remainder are from size fractions that
tion coefficient pr are also calculated (data were air classified. The four sieve-classified
not shown). Twenty-one of the 43 particle size fractions are from corn dust. The twelve
size distributions are not complete; they do aircIassified size fractions consist of two corn
not contain data at the two extreme cumula- dust fractions, six cornstarch fractions, two
tive weight percentages, 0% and 100%. How- wheat dust fractions, and two grain sorghum
ever, 15 out of these 21 distributions are dust fractions. Four of the six fractions of
essentially complete because each of them cornstarch, whose size distributions are
spans more than 95% of the total range of essentially log normal, were from non-freeze-
cumulative weight percents. Five of the dried samples (this is 2/3 of the total number
remaining incomplete distributions contain no of non-freeze-dried size fractions) and the
data between the cumulative weight percents remaining two are fractions that were freeze
of 95% and 100%. Among these five distribu- dried (this is l/3 of the total number of
tions, four are from size fractions of grain freeze-dried size fractions). Figure 2 repre-
sorghum dust with maximum cumulative sents the particle size distribution of size
weight percents of 93.5, 91.0, 76.5, and fraction No. 4 of grain sorghum dust, Fig. 3,
4&O%, respectively, and one is from a size fraction No. 2 of cornstarch, and Fig. 4,
fraction of corn dust with a maximum cumu- fraction No. 1 of cornstarch by freeze drying.
lative weight percent of 84.2%. Only one Notice that without freeze drying, cornstarch
contains no data with cumulative weight can be separated into fraction 1, since parti-
percents between 0% and 5%. This is from the cles in the small fraction tend to agglomerate
size fraction of cornstarch having a minimum when freeze dried. After freeze drying, starch
weight percent of 5.4%. granules were broken up as indicated in
Data at the upper extreme of the distribu- Fig. 4. The diameter is plotted on a loga-
tions are missing because there exists in the 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.
TABLE 1
Amount of weight in particles with diameters less than the lower boundary sieve apertures
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
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 boundary sieve aperture and the weight of
portion of the distribution is the result of a particles with diameters larger than the upper
range of diameters for a relatively large boundary sieve aperture can be estimated.
number of particles. Correspondingly, the However, to compare methods of sizing
shape of the upper portion of the distribution particles, the shape of the particles should be
is the result of a range of diameters with considered_ Irani and Callis [ 51 have reported
relatively small number of particles followed the value of the shape factor relating the sieve
by a range of larger diameters for a relatively aperture to a stokes diameter to be approsi-
large number of particles_ mately 0.9. Sieve openings, transformed to
The deviations from the shape predicted by equivalent stokes diameters utilizing this
Herdan et al. 123 in the upper end of the shape factor, are shown in Tables 1 and 2. For
distribution in Figs. 2 - 4 are more pro- each sieve-classified size fraction, Table 1
nounced for the air-classified size fractions gives the lower boundary sieve aperture, da=,
than for the sieveclassified fractions. The used to obtain the size fraction and the
,-deviations in the lower region of the distribu- equivalent stokes diameter, dsL, of the aper-
tion are more pronounced for the sieve- ture; the value of dsL was obtained with the
classified size fractions than for the air- use of the shape factor of 0.9 [5]_ In addi-
classified size fractions except for the fifth air- tion, the fraction of the total weight of dust
classified size fraction. Also, the deviations in the size fraction that is contained in
that occur in the corn dust size fractions are particles with diameters less than dsL is given.
less than those in the wheat and the grain In Table 2. the aperture of the upper bound-
sorghum dust size fractions. ary sieve dau of each size fraction is given
When particles are sized by sieving, the with its equivalent stokes diameter dsu. The
range of particles with diameters in each size fraction of the total weight of dust in the size
fraction should fall within the apertures of fraction contained in particles with diameters
the bounding sieves. Therefore, the weight of greater than dsu is also presented. Note that
particles with diameters less than the lower 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
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
ranges from 80% to 100% in a size fraction The results in Table 3 show that significant
that is contained in particles with diaroeters differences at the 1% level do exist both
less than the stokes equivalent diameter of the among the types of dust and among the size
lower sieve aperture_ However, the fraction of fractions; however, the interactions are also
the total weight of the dust ranges from only significant at the 1% level_ The significant
0.01% to 20% in the size fraction that is interactions indicate that the type of dust
contained in particles with diameters larger has different effects on the coefficients of
than the stokes equivalent diameter of the variability for various size fractions_ Also. the
upper boundary sieve aperture_ effect that the size fraction has on the coeffi-
Small balls of particles could be seen in the cient of variability is not the same for every
size fractions of wheat dust and grain sorghum type of dust. However, a difference can still
dust. Martin [ 63 found that wheat dust contains exist between air-classified size fractions and
particles called tricombs which range from 50 to sieveclassified size fractions_ Figure 6 shows
200 firn in length and 10 to 30 pm in diam- that the values of the coefficient of variabilit-y
eter, and which have a large length to diam- for all the air-classified size fractions, except
eter ratio in the range of 5 to IO. These for wheat dust, are consistently lower than
tricombs can trap large quantities of smah those for the sieve-classified size fractions_
particles during the sieving operation and Figure 7 presents the expected particle size
prevent them from passing through the sieve. distributions from the perfect sieve or air
Martin [6] also found that grain sorghum dust classification of a sample of dust originally
contains hair-like projections_ During the having a log normal particle size distribution,
sieving operation, these projections can curve A. Three types of classification are
capture small diameter particles to form a ball presented I
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 Fig_ 6_ Correlation between the coefficient of vari-
were the type of dust and the size fraction_ ability and the size fraction_
272
TABLE 4
Comparison of the geometric mean diameter of the boundary sieve apertures to mass mean diameters
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 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
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 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
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 n o r m a l d i s t r i b u t i o n s can have the same
dist~_buted a m o n g the particles. The mass 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
mean d i a m e t e r does not characterize the 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
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 og.
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 Composition
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 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,
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 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
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 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
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
and fibercontentwas obtained by subtracting fiber cont.ent from 10.0% to 14.69o. The
thesum oftheweightpercents ofmoisture, standard deviation of the starch and fiber
ash, and protein from lOO%, since the dust content in cornstarch, 3_4%, is less than that
was assumed to contain only those com- for the other kinds of dust. In contrast, the
ponents_ standard deviation of 2.8% among the average
Table 6 shows the average values of mois- values of the ash content for wheat dust,
ture, ash, protein, and starch and fiber con- grain sorghum dust, and corn dust, and that
tent for each dust. The standard deviation of of 3.7% for t.he starch and fiber content
the content of each component among the indicate only a slight variability among the
size fractions of each kind of dust is also given dust types.
as we3 as the coefficient of variability of each Table 7 contains the average values of all
component. The standard deviations of types of dust, the standard deviation among
protein content among the size fractions of these average values, and the coefficient of
each test dust range from 1_3% to 2_7% and variability among these average values For -
those for the moisture content from 1.3% to exh component_ The standard deviation of
1.5%. This indicates that the protein content the average value for the moisture content of
and the moisture content vary only slightly all dust samples is 1.8% and that for protein
among the size fractions of the same type of content is 0.9%. The variability of protein
dust. The standard deviation of 3_4% for content and moist.ure content among the
moisture content among the size fractions of different kinds of dust is approximately the
cornstarch indicates that their variability is same as or smaller than the variability among
larger than those of other types of dust_ The size fractions within each kind of dust.
larger variability for cornstarch is the result In Fig. 8, the moisture content of each t-est
of freeze drying some of the size fractions to dust is plotted against the size fraction_ The
improve their dispersibility. results correlating ash, protein, and starch and
The ash content and the starch and fiber fiber content with size fraction are presented
content of each type of dust show more in Figs. 9 - 11_
variability bet-ween size fractions than do MartW and Lai [7] have shown that air
the moisture and the protein content. The classification of grain dust results in a large
standard deviations of the ash content among ash content (approximately 40%) -in the
size fractions of each type of dust range from residue size fraction. Figure 9 indicates a
10.6% to 18.3% and those for the starch and 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
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 fiber content. For these two quantities. the
fractions correspond to the sixth size fraction_ fourth, fifth, and sixth size fractions exhibit
The fifth size fraction of each dust contains lower values than those of the remaining
large weight percentages of ash (approxi- size fractions. The correlation coefficients
mately 25%). For the wheat dust and the between the ash content and each of those
gram sorghum dust, the fourth size fraction quantities are significant at the 1% level
also contains a relatively large ash content (Table 8). Table 8 contains the standard
(approximately 11%) when compared with deviations of compositional content among
the ash content of the remaining size fractions the size fractions of each test dust that results
(approximately 4%). when the data from size fractions 6,5, and 4
In Figs. 8 and 11, a similar trend-is noted for are removed one at a time. The highly signi-
the moisture content and for the starch and cant inverse correlation between the ash and
TABLE 8
Simple correlation coefficients for the correlation between compositional componerits
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
moisture content results from the ash material less than l%_ The standard deviations of the
bc-ing less hygroscopic than the organic grain protein content among the size fractions
dust. The correlations between the starch exhibit essentially no change.
and fiber content and the ash content are The correlat-ion coefficients in Table 8
significant. indicate three additional significant correla-
The large standard deviations of the ash tions. The correlation between the moisture
content and the starch and fiber content content and the starch and fiber content is
among the size fractions of each test dust significant for all types of dust at the 1%
have resulted from the large ash content in level. The correlation between the protein
size fractions 4, 5, and 6. The standard content and either the ash content or the
deviations of the compositional contents of starch and fiber content is significant for only
the remaining size fractions are given in the wheat dust at the 5% level_ The correla-
Table 9. For all three types of dust, the stan- tion between the moisture content and the
dard deviations of the ash content decrease starch and fiber content is the result of the
to less than 2%, and those of the starch and difference between the moisture cont.ent and
fiber content to 4.2%. For the moisture the ash content and that between the ash
content, the standard deviations decrease to content and the starch and fiber content.
278
The partial correlation coefficient of -0.70 dust size fractions that contain the lowest
with the effect of ash removed between the moisture content also contain the highest ash
moisture content and the starch and fiber content and happen to have the largest mass
content indicates no significant correlation at mean diameters_ Again, grain sorghum dust
the 5% level. and corn dust do not exhibit these trends.
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
A - Unvented explosion
B - Small vent area
C - Large vent area
Design strength
Time
- Air
227
TABLE 1
Vent ratios for dusts having maximum rates of pressure rise determined in the Hartmann
bomb
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).
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
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
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)
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.
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.
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
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
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
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
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.
References
15 R.
F.S. LAI
U.S. Grain Marketing Research Labomtoxy. Science an< Education~Administmtion, U.S. Department of Agrhxd-
ture, 15I5 College Avenue, Manhattan, KA 66502 (l7S.A.)
D_ W_ GARRETT and L. T. FAN
Department of Chemical Engineering, Kanscrs State Uniuersity. Mnrrhattan. KA 66506 (U.S.A.)
(Received October 13,198l)
..
Research on dust explosions has been con- Many of the limited number of publications
ducted since 1917 but the progress has been on particle size and composition, as they affect
painfully slow El] _ Limited work has been the four explosion characteristics, present
carried out to systematically explore and results of experimental studies graphically
quantify the effects of basic parameters and/or tabularly with minimal analysis. The
governing the dust explosions_ A dust explo- majority of such publications are from the
sion consists of a very rapid combustion U.S. Bureau of Mines [9,11,153 _ Other pub-
reaction that liberates heat and results in the lications include empirical correlations of data
expansion of gases surrounding the dust. IS, 143 -
When this expansion occurs in a vessel or en-
closure, the pressure in it rises rapidly, which 2.1. Particle size
may lead to structural damage to the vessel or Price [221 separated oat and corn dust,
enclosure. Two of the basic param eters zue wheat flour, wheat flour dust and potato starch
the diameter and composition of the gram into four different size fractions by elutria-
dust particles. tion. Particles in a sample from each fraction
The overall objective of this work, cuhni- were microscopically sized. The average
nating in a series of articles, was to study the particle diameter for each fraction was the
effects of the particle diameter and composi- arithmetic average of the maximum particle
tion on the explosibility of grain dust_ Specify diameter in the fraction and that in the next
ically, the explosibihty was characterized by smaller fraction_ The average particle diameters
four parameters from Hartmann bomb tests ranged from 10 to 100 pm. Price used the
[ 5]_ These four parameters were the maximum standard Bureau of Mines apparatus for in-
explosion pressure, P,,, the maximum rate flammability to perform experiments. The
of pressure rise, (dP/dt)-, the average rate standard procedure was altered by using
of pressure rise, (dP/dt),,,, and the minimum 0.000075 kg (75 mg) of the dust sample
explosible concentration, C,,_ The first in instead of 0.0001 kg (100 mg), and-by dis-
this series of articles contains a comprehensive persing the dust sample with pressurized air
and critical survey of publications related to instead of oxygen_ The maximum explosion
the effects of particle size and composition of pressure measurements from the tests were
gram dust on its explosion characteristics. plotted against the average particle diameter_
This survey is concerned mainly with dust The maximum explosion pressure increased
from grains, including wheat, corn and grain with decreasing average particle diameters;
s-orghum. however, for wheat flour and wheat flour
dust with particle sizes less than approxima- because of the log-norm&y of the particle-
tely 35pm, the maximum explosion pressure size distributions [lo] ; The average particle
decreased with decreasing particle size_ It was diameters of these fractions ranged from 18
concluded that particle size was an extremely to 171 pm. Jacobson et uL [14] plotted the
important factor in governing the explosibility results of the experiments for all eleven types
of thedust. . of dust on a single graph. Specifically, they
&rtmann and Nagy [9] used a series of plotted the ratio of explosion characteristics
sieves to separate dust into four size ranges: against the ratio of average particle diameters.
20 - 65,65 - 100,100 - 200, and greater than The ratio of explosion characteristics (e.g. the
200 mesh_ They calculated the average particle maximum explosion pressure) is defined as
diameter for each of the size ranges, Le. the the explosion characteristics of a given dust
arithmetic mean of the particle diameter divided by that of a through No_ --. 200 sieve
corresponding to the sieves used to determine sample of the same type of dust. The maxi-
the upper and lower bounds of each range. mum explosion pressure, as one of the explo-
The geometric mean, however, should have sion characteristics, showed a slight increase
been used instead of the arithmetic mean as the ratio of average particle diameters
because particle size distributions are charac- decreased_ The ratio of minimum explosible
teristically log-normal (Herdan et al. [lo] ). concentrations decreased with a decrease in
According to Herdan et aL, the geometric the ratio of average particle diameters for
mean better represents the particle size than coarse dust (, 74 pm); however, for fine
the arithmetic mean for a log-normal distribu- dusts, there was only a slight change with the
tion. For each range, Hartmann and Nagy [9] ratio of average particle diameters. Jacobson
plotted the minimum explosible concentra- et al. [14] empirically correlated the ratio of
tion, the maximum pressure rise, and several explosibility indexes, In, with the ratio of
other quantities against the average particle average particle diameter, DR_ The hazard of
diameter_ They observed that the minimum a dust is related to its ease of ignition and to
explosible concentration decreased with the severity of the ensuing explosion. Among
decreasing particle size; however, for several other parameters the ease of ignition may be
kinds of dust, it changed very little when the considered as a function of the ignition
particle sizes were less than approximately temperature, minimum ignition energy, and
102 pm (O-004 in_)_ For example, for cellulose minimum explosible concentration_ The
acetate molding powder with a dust cloud severity of an explosion is related to the
concentration of 0.005 kg/m3 (0.005 oz/cu. pressure and the rate of pressure rise. They
ft.), the maximum pressure increased with defined the explosibility index as follows:
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
(1)
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 where Tigp = ignition temperature, Ed =
of the fractions and calculated the average minimum ignition energy, P,, = maximum
particle diameters with a m.ethod described by explosion pressure, (dP/dt),, = maximum
Dallavalle [4] _ The arithmetic mean of the rate of pressure rise, and C,, = minimum
respective size openings of a pair in the series explosible concentration
of sieves was multiplied. by .the fnction of The relationship obtained by them is:
dust retained on the finer of the.two sieves_ :
IR =DR-3
The average particle_~diameterw& the summa- (2)
diameter for -es@. sample, Jacobson.: et a& .-_ not depend markedly on the particle size.
[15] used dust samples of-~various.metals However, at higb dust concentrations (O-4 -
including zinc, iron .and cobalt, that. .were O-5 kg/m3), these quantities did increase some-
obtainedfiom.different processes, e.g. milling, what with decreasing particle diameter_
stamping, and atomizing. The effects of aver- Eckhoff [S] determined tbe maximum
age particle,diameter on the minimum explo- explosion pressures and the maximum rates of
sible concentration, maximum pressure rise, pressure rise of several types of dust from
maximum rate of pressure rise, and average agricultural grams and feedstuffs (e.g. corn,
rate of pressure rise for atomized aluminum wheat, rye, fish powder, soya meal and potato
dust were similar to those reported by starch). The dust samples were obtained from
Jacobson et uL [ 14]_ elevators in different countries such as the
The explosion of coal dust-air mixtures United States and Norway, and from various
was studied by Hertzberg et al. [ll] . Dust locations in the elevators, such as the bottom
was separated into nine size ranges with aver- of a b-u&et elevator and the dust filter_ TO
age particle diameters from 2.7 to 65 pm. Six determine the particle size distribution,
of the nine size ranges were classified by a Eckhoff [6] used a series of sieves for the
Donaldson centrifugal classifier, two by sieve- coarse (> 74 pm) end of the distriioution and
classification, and one was unclassified pulver- a Coulter Counter analyzer for the fine end. -4
ized coal_ The surface mean diameter was specific surface area (the envelope surface
determined for each size range by a Coulter area) was approximated for each sample. The
size analyzer. The experiments were performed shape of the dust particle approximated by a
to determine the minimum explosible con- sphere, and the corresponding mass average
centration, the maximum pressure, and the particle diameter were used as the radius in
maximum rate of pressure rise in a modified determining the particle size distibution.
Hartmann bomb_ The volume of the explosion B.E.T. measurements were conducted to
chamber was 7.8 liters instead of the standard obtain another approximate average surface
1.2 liters and the inside configuration of the area The Hartmann apparatus was used to
chamber was altered to achieve a more uniform determine the maximum explosion pressure
dust dispersion than in the original Hartmann and maximum rate of pressure rise for each
bomb. The standard criterion for determining sample. A relation was obtained between the
the minimum explosible concentration is the maximum rate of pressure rise and the specific
lowest concentration of dust that when ignited envelope surface area by means of a least-
will cause a paper diaphram to rupture [ 5]_ squares regression line for starch and protein,
The criterion used by Hertzberg et al. [ll] Le.,
was the lowest concentration of dust that
when ignited would create a maximum
pressure that reached a value twice that of the =KS
initial pressure just prior to ignition_ They
altered the standard procedure by partially
evacuating the explosion chamber to a pressure where (dP/dt),, = maximum rate of pressure
of 0.2 - O-3 atm so that the pressure reached rise, K = a constant, and S = specific surface
approximately 1 atm after the air for dis- area
persing the dust was injected into the chamber. Each regression line was forced through the
Instead of the standard procedure employing origin and two data points were used to
a continuous electric spark, initiated prior to determine the slope. Two different sizes of
the injection of the dust, they used electric fish powder represented protein, and two
matches ignited after a spatially. uniform dust different types of starch, potato and maize,
cloud was achieved. Hertzberg et al. [ll] represented starch. Eckhoff obtained a good
observed the minimum explosible concentra- correlation with slope values of 450 bar-g/
tion to be virtually independent of particle (s m2) for protein and 1250 bar-g/(s m2) for
size in the size range 2.7 to 65 pm. At dust starch.
concentrations close to the minimum explo- Using two different apparatuses Enomoto
sible concentration, the maximum pressure [ 81 generated sufficiently uniform dust clouds
rise and maximum rate of pressure rise did 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
197
where n = t&al number of all components, Ki The ignition delay (the time interval
= the value of K for pure component i, and Xi between the dispersion and the ignition of the
= the weight fraction of component i. dust) was taken as an indicator of the degree
The value of Sri-was experimentally deter- of turbulence. For example, if the ignition
mined for starch using wheat starch and corn delays of two explosions were equal, the
starch and for protein using fine and coarse degree of turbulence was considered to be the
fish powder. The value of Ki for starch WAS same. Therefore, to eliminate the effect of
considered applicable to the star& and fiber turbulence on the correlation of the maxi-
fractions because of the similarity between mum rate of pressure rise wi& weight percent
their molecular structures_ The value of Ki for moisture, only those data from the explosion
the fish powder was considered applicable to tests with similar ignition delays were com-
the rest of the components: protein, fat, and pared [6] _ This correlation was significantly
nitrogen-free soluble organic compounds different from the correlation of the maxi-
other than starch. The resultant correlation is mum rate of pressure rise with weight percent
moisture where the differences in ignition
(W Starch & Fiber)
K = (SOO- + 450), delay were not considered_ At short ignition
100 delay times (approximately 0.005 set) there
were significant maximum rates of pressure
kp- g/(cm2- s- m2) (5) rise for moisture contents considerably higher
than 30%. When the delay time was not con-
Eckhoff [S] also correlated the maximum sidered, there were few significant maximum
rate of pressure rise with the weight percent rates of pressure rise for moisture contents
moisture_ He selected four types of dusts: higher than 22%_
maize starch, cellulose powder, potato starch The particle size and composition of grain
and Norwegian oat dust, and prepared several dust affect the explosion characteristics_ The
samples from each with varying weight percents exact relation is, however, difficult to quanti-
of moisture_ He obtained an approximate fy_ Most of the experimental work attempted
correlation given by to correlate the data and presented the dat,
graphically and/or tabuiarly with minimum
(6) analysis. To improve the experimental studies.
theoretical analyses are needed. Hartmann
where y = wt.% of moisture, and ymin = the and Nagy [9] separated dust into several size
wt.% of moisture above which no explosion ranges and measured the explosibility cha-
can occur- ymin was assumed to be approxi- racteristics of the dust in each size range.
mately 15% for all types of dusts_ They used the arithmetic mean of particle dia-
To study the effect of moisture on the rate meter, which has been shown to be an
of pressure rise in the Hartmann bomb tests, incorrect method of representing the particle
Eckhoff and Mathisen ]7] attempted to elimi- size distribution [ lOI_ Jacobson et al. [ 143
nate the effects of moisture on the quality of improved their experimental methods.
the dust dispersion, the effective particle size However, the characterization of particle size
distribution and the turbulence_ Corn starch and composition showed little progress_
was the only sample used. In a test apparatus Eckhoff [ 63 performed the test of explosibility
similar to the Hartmann, the effective particle of several grain dusts and obtained the correla-
size distribution was measured. The dispersion tion with the moisture, ash, protein and fiber
system and the shape of the apparatus are the contents. In summary, only limited work has
same as those in the Hartmann apparatus; been conducted on this subject and many
however, there is no ignition system and the areas of work are still needed, especially in
tube is fitted with a piston Double adhesive the characterization of particle size distribu-
tape was attached to the bottom of the piston. tion_
When the dust was dispersed, the tape captured
a sample of the dust, and an effective particle
size distribution was determined by micro- 3_ THEORETICAL STUDIES
scopic examination. It was concluded that
moisture does not affect the effective particle There are few publications that attempt a
size distribution greatly. theoretical prediction of the explosion chamc-
198
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
1+-
Ti(9)
pressure and rate of pressure rise.
To predict the minimum explosible con- (L-b)
centration, Jaeckel [16] performed a heat 2&zP5
balance on the dust-air system. He considered
the minimum explosible concentration, Cd, where DP = particle diameter, 11= t/T, T =
to be the smallest concentration of dust time required to completely burn the particle,
capable of producing enough heat. when com- &, = proportionality constant between 2 and
pletely cornbusted, to elevate the temperature Df, Tf = temperature of the flame, and z =
of the dust cloud and the air to the ignition initial temperature of the system.
temperature; the following expression was An energy balance was performed around the
obtained_ second dust particle to determine the tem-
perature of the particle, TaL, as follows:
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. where ps = density of the dust particle, C,, =
This model depicts the system as being a specific heat capacity of the dust particle,
spatiahy uniform dust cloud throughout the E, = emissivity of the particle, Ef = emissivity
entire expIosion and does not take into. of the flame, u = Stephan-Boltzmann constant,
account the dependence of C,, on the dust F = shape factor, h = heat transfer coefficient,
particle size- Jaeckel [ 161 reported that eqn. 2K/D$, and k = thermal conductivity of the
(6) predicted a minimum explosible concentra- air.
tion of 0.022 kg/m3 for sugar powder; in The resultant equation is
contrast, Jacobson et a.L [14] reported an
experimental value of O-045 kg/m3.
Tanaka [23] developed a model to predict T,,(o) = 1
e_., Cl$FCcT&n) + D-G3 dn +
the minimum expiosible concentration. A
particle at the center of. the dust cloud was + Tao] (11)
ignited by an outside source. The ignited par-
ticle heated its near neighbor, a second par- where C = IUr,k/p, -C,,, D = 3KDEPEfFD,/
ticle, at a distance L_ The minimum concen- @SC,,), and Tdo = initial temperature of the
n&ion, C,,, is the smallest concentration particle.
that allov-7 the flame to propagate from the The value of p, i.e. t/T, at which Td,
ignited par,%cle to its neare& neigbbcr. In the reaches the ignition temperature, Ti, can be
model, the equation for calculating the flame calculated from eqns. (9) and (11) by selecting
height, b, was taken from a publication on the a distance L. The minimum distance, Lo, is
burning of liquid droplets by Miesse [lS] _ the distance L which allows the second particle
Tanaka 1231 has assumed that the gas phase t.o reach the ignition mmpeE&We, Tan = Ti*
between the *_o particles is stagnant. He then at the moment that the initial particle is com-
estimated the gas temperature at-the gas-solid pletely consumed. 9 = l_ The minimum explo-
interface- of the second particle,. SF,&, by sible concentration is calculated-from-l,, by
solving the conduction equation the following equation.
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
Tanaka 123 J has .definedthe ignition tem- d.P SvK,T;eP:g
-= (P& - P$)13 x
perature, Ti, differently from that experimen-
dt aT2Pzmu
r
tally determined by he Godbert-Greenwald
2
furnace. The ignition temperature defined in
p3vs
the model is the temperature of the dust par- (15)
ticle that causes it to ignite; however, the
experimentally determined ignition tempera- In either case, the expression for pressure
ture is the minimum wall temperature that as a function of time -SC
obtained by integration
causes a particle immersed in it to ignite [5] _ of its equation for the rate of pressure rise,
Nagy et al. [ 191 have derived an expression eqn_ (14) or (15).Note that neither eqn. (14)
to predict the rate of pressure rise as a function nor eqn (15) is an explicit function of particle
of time. They have assumed that the pressure size. However, they could have an implicit
developed is affected by changes in tempera- size dependence through K,_
ture and increases in the total number of The expressions for the rate of pressure rise,
moles of gas in the explosion chamber. They eqns. (14 j and (15), are explicit functions of
have baaed their derivation on an equation neither concentration nor composition. Nagy
empirically derived from gaseous explosions, et aI_ [19] compared their models with exper-
which is imental results involving several gases and
corn starch. They obtained good agreement,
except when the explosion was nearly com-
(13) plete and the heat loss to the walls of the
vessel became significant.
Normura and Tanaka [21] have theoreti-
where A = surface area of the flame front, Vu
cally derived the experimentally determined
= volume of unburned fuel and gas, If, =
cubic law used by Bartlmecht [ 2 J , which is
burning velocity reference at 1.0013 bar (1
atm) and 298 K (537 X), T, = reference tem-
perature, 298 K (537 R), R, = reference Vp3 (16)
_L
pressure, 1.013 bar (I atm), andp = aconstant.
They have assumed that the air and combus- = maximum rate of pres-
where (dP/dt),,
tion gases behave ideally and that any solid sure rise, and V, = enclosuz volume (constant
present has negligible heat capacity and with time).
volume. At any time during the explosion, the Note that Nagy et al. [ZO] also obtained
combustion chamber is assumed to be divided this result. Equation (15) implies
into two zones, the burned and unburned
zones, by a very thin combustion zone. d.P 1 1
Two cases on the rate of pressure rise have
-lx-=-
v--u3
dt a
been considered by Nagy et al. [ZO] . In one
case the system is assumed to be isothermal, Normura and Tanaka [21] have assumed
and in the other it is assumed to be adiabatic. the dust cloud to be stagnant with the par-
For the isothermal system, the rate of pressure ticles uniformly spaced in a fixed geometry.
riseis 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
(14)
particles ignite. The flame continues to pro-
2GG
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- 13 W. Ishihama and H_ F.nomoto, Com&st. F&me, . -.
massnahmen. Springer Verlag, Berlin, 1978. ZI(l973) 177 - 186.
3 J_ Bautling. fir Gasbeleuchiung, 57 (1918) 32 - 14 M. Jacobson, J. Nagy, A Ri Cooper and F. J. Bi,
55. US. Bur. Mine-$ Rep. Invest; ~RI-5753. 1961:
4 J. M. Dallavalle, The Techhnoiogy of Fi.ze ParticZes. 15 M. Jacobson, k R. Cooper. and J. Nagy, U.S. Bur.-
Pitman, New York, 2nd edn., 1948, Chap_ 5, p_ Mines, Rep. Invest. RI-6516.1964.
113. 16 G. Jaeckel. 2. Tech; Phys. 3 (1922) 67 - 78.
5 H. G. Dorsett, Jr., M. Jacobson, J. Nagy and R_ P. 17 0. Levenspiel, Chemical Reaction Engineering,
Williams, U.S. Bur. Mines, Rep. Invest. RI-5624. Wiley, New York, 1962.
1960_ 18 C. C. M&se, 6th Symp. on Combustion, 1957. $.
6 R_ K. Eckhoff, Fire Res.. I (1977/78) 71 - 85. 732.
7 R. K. Eckhoff and K_ P_ Mathisen, Fire Res.. I 19 J. Nagy, J. W. Conn and H. C. Verakis, U.S. Bur.
(1977/78) 273 - 280_ Mines, Rep. Invest_, RI-7279 1969.
s H. Enomoto, Proc_ Int. Symp. on Grain Dust Ex- 29 J. Nagy, E. C. Seiter, J. W_ Corm and H. C.
plosions, Grain Elevator and Processing Society, Verakia, U.S. Bur_ Mines. Rep. Invest.. R&7507.
Minneapolis, MN, 1977, pp. 143 - 170. 1971.
9 I. Hartmann and J_ Nagy, US. Bur. Mines, Rep. 21 S. Normura and T. Tanaka, Kagcku Kogcku.
Invest., RI-3751. 1944. (1979) 601- 608_
10 G. Herdan, MI_ L_ Smith, W_ H_ Hardwick and P. 22 D. J. Price and H. H. Brown, Dust Explosions.
Connor, Small PartkZe Statistics, Butterworths, National Fire and Protection Association, Boston,
London, 2nd edn., 1960. MA, 1922.
11 M. Hertzberg, K. L. Cashdollar and J. Operman, 23 T_ Tanaka, Proc In t. Symp. on Grain Dust ExpZo-
U.S. Bur. Mines, Rep. Invest.; RI-8360. sions, Grain Elevator and Processing Society,
1979. Minneapolis, MN, 1977, pp_ 79 - 99.
12 W. Ishihama, II th Int_ Conf. ofDirectors ofSafety 24 S. Yagi and D. Kunii, 5th Int. Symp. on Combus-
in Mines Research. Warsaw. Poland. I961_ 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 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
Flammability (% in air)
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
Cumulative
volume x GRAIN DUST PARTICLE SIZE DISTRIBUTION
loo -
SO-
SO-
70 -
Sn-
SO-
4D-
30 -
Diameter in Microns
[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.
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
(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
TABLE 4
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)
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
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.