074149-Filtration Magnétique

Magnetic ,.eparmion News, 1984, Vol. 1, pp.
67-99
073 I- 3632/84/0 02-0067 $25.00/0 (C) Gordon and Breach Science Publishers, Inc., 1984
Printed in the United Kingdom
HIGH GRADIENT MAGNETIC GAS FILTRATION
AIK CHONG LUA AND ROBERT F. BOUCHER

Department of Mechanical Engineering, University of Sheffield, Sheffield, S1 3JD, England.
Highly successful tests have shown .hat high Abstract gradient, magnetic filtration can provide a viable alternative to conventional filtration to curb the amount of stack gas particulate emissions from iron-based industries, especially steelmaking processes. Using basic oxygen furnace dust in our work at Sheffield University, filtration efficiencies of 99% and greater were obtained for submicron particles down to an optically measurable 0.24 m diameter; for particles of um and above, 100% filtration efficiency was achieved. High gas throughput, together with low pressure losses, low applied magnetic fields and good filter matrix loadability all further indicate the potential commercial practicality of HGMF in the steelmaking industry.
I. INTRODUCTION
Since the first commercial production of kaolin clays using high
gradient magnetic separation (HGMS)
in
1969, an increasingly
growing number of various separation processes using HGMS are
reported in the literature.
These applications include as diverse
as coal desulphurization, mineral beneficiation and the separation

of red blood cells from whole blood.
A major potential application of dry high gradient magnetic filtration (HGMF) relates to the removal of all fluid-borne particulate matter from gas streams and,
in particular, its use as an air-pollution-abatement device for
magnetic particles.
The role of gaseous magnetic filtration is to
67
68
supplement
A. C. LUA AND R. F. BOUCHER

the existing particulate control devices to curb
large-scale emissions of particles
from industrial
particularly the submcron particle size range.
sources, These very fine
submcron particles are especially harmful as they constitute the
greatest inhalation risk.

region of the lungs.
airborne
They are responsible for numerous
pulmonary diseases as they can penetrate into the respiratory

Also submcron particles tend to remain
the
atmosphere
for
long
periods
of
time,
thus
contributing to atmospheric haze or smog.
Thus, there
a need to
develop more economical and better dust control methods for the
fine particle size range.
Stack gases arising from steelmaking processes contain vast

quantities of particulate matter which need to be removed before
the air
released to the atmosphere.
The particulate matter
these stack gases has a high iron content and
thus susceptible
to HGMF.
The current, methods used for stack gas clean-up from

There are certain important
steelmakng processes are mainly electrostatic precipitation, wet.

scrubbing and fabric filtration.
advantages of HGMF over
these methods.
In most industrial
electrostatic precipitators, the negative corona mode of operation
For successful operation, electron-absorbing gases have to be present, of which sulphur dioxide is one of the best. Thus, the increasing drive to eliminate sulphurous gas emissions leads to difficultles in the
usually used to obtain a higher corona current.
operation of electrostatic precipitators, and HGMF may offer the
alternative solution. In the wet scrubbing process, the problem of
atmospheric pollution has been turned into a waste water pollution
problem.
The venturi scrubber which is chiefly used in this
process, is highly energy consumptive and thus its operating costs are high. On the other hand, the loaded matrices of a dry HGMF system can be cleaned using a dry technique, e.g. mechanical
rapping, infrasonic cleaningl.
Lastly, fabric filtration suffers
from temperature limitations (the maximum operating temperature for
69
cotton media is about 80oc) and high maintenance costs as filter media have to be replaced regularly. Gas temperatures close to the Curie temperature of the filter matrix are possible with the HGMF process. For ferritic stainless steel wool (AISI Type 430), the Curie temperature is 610oc2. The gas velocity through the fabric filter is generally low; in the range 0.01 to 0.6 m/s whilst gas velocities up to 13.6 m/s were reported by Lua and Boucher3 for HGMF tests on basic oxygen furnace dust. Hence, a baghouse system is area extensive whilst an HGMF unit is compact for the same
quantities of gas to be cleaned.
Dry HGMF has potential applications in other processes where

weakly magnetic or paramagnetic particles are to be removed.
In
the nuclear industry, HGMF can be used to remove radioactive
particles, consisting mainly of various oxides of iron, in the

primary cooling circuits of gas-cooled reactors. Also in nuclear
fuel reprocessing plants, radioactive paramagnetic particles are
present as a result of certain shearing

techniques.
and
laser cutting
processes; these gases or fumes can be cleaned by means of HGMF
HGMF devices are attractive to the nuclear industry because they can withstand high temperatures, pressurization, fires
and explosions much better than the high efficiency particulate air
filters presently used; further HGMF matrices are cleanable thus
reducing active-waste disposal problems.
Other possible
dry HGMF
and HGMS applications are the removal of iron oxide particles in

gas-carrying pipelines; the recovery of magnetite from coal fly
ash; the recovery of glass from municipal incinerator residues by

removing the iron-bearing compounds; the beneficiation of iron
ores; and the capture of

environments.
respirable asbestos fibres in working
2. HGMF OF FINE PARTICLES FROM STEELMAKING STACK GASES
Many of the steelmaking processes produce large emissions of fumes containing substantial amounts of iron-bearing dust particles which
70
have to be
removed before the stack gases are discharged to
the
atmosphere.
The total iron concentration from several of these
processes 4 is as follows- basic oxygen furnace (55- 70%); open hearth furnace (55 70%) scarfing process (50 70%); blast furnace (35 -50%) and sintering process (25 50%). In 1975, the Research Triangle Institute (RTI), North Carolina, U.S.A., carried out an investigation for the U.S. Environmental Protection Agency to determine if HGMF technology
could be
applied
to
control
fine
particle
emissions
from
A laboratory pilot plant was constructed and tests were carried out for two types of dust viz. basic oxygen furnace (BOF) dust and electric arc furnace (EAF) dust. The
steelmaking
schematic diagram of the laboratory pilot plant is shown in Fig. I. The dust is dispersed into the air stream of the 61-cm diameter
wind tunnel using a fluidized-bed dust generator.
processesS.
A slipstream of
the dust-laden air is drawn off the wind tunnel into the HGMS device. Before the HGMF filter, the stream passes through a
FAN
FIGURE
Schematic diagram of the RTI HGMF laboratory pilot plant (after Gooding et al, ref. 5)

cyclone (a 50-percent cut point of about 5m) to remove large
As the air stream flows upwards through the cyclone, a baffle is installed to break up the vortex. Upstream and downstream
particles.
of the filter, dust samples are extracted to determine particle

sizes and concentrations.
After the downstream sample ports, the
clean air stream passes through a blast gate which is used for flow
control, through a booster fan and back into the wind tunnel loop. A pitot tube is located between the downstream sample ports and the
blast gate to determine the average downstream flow velocity.
All
particle sizes and concentrations were determined using MRI Model
502 inertial cascade impactors (manufactured by Meteorology Research, Inc., California, U.S.A.) except for the two long term matrix loadability tests. In these two tests, an optical particle
counter, Climet Model 208A (Climet Instruments Co., California, U.S.A) was used, together with a dilution system to avoid particle AISI 30 ribbon-shaped coincidence losses in the counter.
stainless steel wires were used.
The magnetization curves for both the BOF and EAF dusts are
shown in Fig. 2.
Filter penetration values were computed from
particle sizes and concentrations at filter inlet and outlet.
Figures 3 and 4 show the matrix 1oadabilities for the BOF and EAF dusts respectively, starting from clean matrices in both tests. The
BOF data shows no deterioration in efficiency during the test period even though the matrix has collected about twice its own
mass of dust.
specific
On the other hand, the EAF dust which has a lower

than
the
magnetization
BOF
dust,
shows
gradual
deterioration in collection efficiency with loadin from the start. At the end of the EAF test, the matrix had collected about three
times its own mass of dust.
However, extreme deviations from
isokinetic sampling are apparent in the way Gooding et
a15
carried
out the two 1oadin tests for Figs. 3 and4. The air velocity just Instead a ahead of the sampling nozzle was not determined. constant sampling velocity of approximately 5.0 m/s was assumed and
72
BOF DUST
40-
1-0
2"0
3.0
4.0
5.0
APPLIED FIELD, kG
FIGURE 2
Magnetization curves of BOF and EAF dusts (after Gooding et al, ref. 5)
in
used
throughout
both
tests even though these tests were
conducted at a constant superficial air velocity of 8.2 m/s. Also, the sampled dust aerosols were passed through a small cyclone (with
a 50-percent cut point of about
m) and then into a dilution
system before any particle measurements were carried out using the optical particle counter. Such measurements of particle sizes and
concentrations would surely be vastly different from those actually
entering and leaving the HGMF filter, even for small particle
sizes.
Gooding et
a15
also investigated the effects of various
parameters viz. matrix packing density, magnetic field and gas

velocity on filter performance.
Increasing the matrix length shows

100
73
MAGNETIC FIELD =0"4T FILTER LENGTH :15m SUPERFICIAL AIR VEL. 8-2 PACKING FRACTION 0-005
.. "0.7-1-Oum o.s-o.
0.3-0.51m
0 100 200 300 400 500 600
TIME, MINUTES
FIGURE 3
Matrix loadability for BOF dust (after Gooding et al,
ref. 5)
0-5-0-71,I
p
O
20-
,o-
IIGNETIC FIELD O-4T FILTER LENGTH 15m
SUPERFICIAL AIR VEL. 8-Z PCKIN FRACTION 0-005
oo
.1.. 100
200
30O
400
500
=,00
TIME, MINUTES
FIGURE 4
Matrix loadability for ref. 5)
EAF dust (after Gooding et al,
74
a general trend of decreasing filter penetration (Fig. 5). Penetration is defined as the ratio of particle concentration at
filter outlet to that at inlet. Theoretical consideration of the
filter penetration equation (i.e. P
f1ter penetration, F
a volume packing fraction of
exp.
4FLRc -z (i-F) }where LP == matrix,
filter length, R c : dimensionless capture radius of single wire, a = wire radius) suggests that doubling the matrix length should
result in squaring the penetration. The improvement with the longer

matrix
in Fig.
5 is not as significant as should have been

The effect of matrix
predicted by the penetration equation.
packing density is shown in Fig. 6.

reduces the penetration.
Increasing the packing density
The improvement in efficiency achieved by
doubling the packing density is greater than predicted by the

filter penetration equation.
The effects of applied magnetic
10
O5
02
l-
z
0 02 0 02
001
0-002
PARTICLE 01AMETER, Im
PARTICLE DIAMETER, Im
FIGURE 5
Effect of matrix length on filter penetration for BOF dust (after Gooding et al, ref. 5)
FIGURE 6
Effect of matrix packing density on filter penetration for BOF dust (after Gooding et al, ref.
75
fields on filter penetration for the BOF and EAF dusts are shown in
Figs.
and
8 respectively.
The
random
scatters in the
experimental results for both dusts do not show any definite trend.
Theoretically, increasing the magnetic field (before both the wire
and
particle
saturate)
increases
the
wire
and
particle
magnetizations which enhance particle capture,
if the other
the
parameters remain unchanged.
Figure
shows
effect
of
superficial air velocity. Increasing velocity does not seem to have
a strong effect on the smaller particles but detrimental to larger

particles.
The generally wide scatters in the experimental results of
Gooding et
a15
viz. Figs.5
9, could be attributed partly to the

the inertial cascade
choice of the particle measurement method
impactor.
Its use is associated with many sources of
error6.
The
05
./
l-
\.\
\
O
I-,-
z
002
z
\
0 02
00
FITTER LENETI, =15m

IMKIII FRCrlON :O-OO
1,,,
0OO5
kMNTIC tMSOLS R[LO(T) VEL(m/)

o,,,...o
"r
o
oio
0.21
47
e,--4
O-ZO
0-40
t<O
-I
.4
0-002
0--.4]
oo%,
o-
o,
,o
PARTICLE DIAMETER, Im
PARTICLE DIAMETER, lm
FIGURE 7
Effect of applied magnetic field on filter penetration for BOF dust (after Gooding et el, ref. 5)
FIGURE 8
Effect of applied magnetic field on filter penetration for EAF dust (after Goodlng et el, ref. 5)
76
FIGURE 9
Effect of superficial air velocity on filter penetration for BOF dust (after Goodin et ai, ref. 5)
of
main
sources
errors
in
cascade
impactors
are
particle
depositions on the walls between the stages, on the walls of

nozzles and on perforated plates in multi-jet cascade impactors
particularly the deposition of
larEe
particles on the first
perforated plate in which the inlet pipe acts as a nozzle and the
first perforated plate as an impaction plate.
As particles are collected in localized deposits on the impaction plate and as these
deposits become overloaded, incominK particles will bounce off the
Also, when the dust deposit reaches a sufficient thickness, it can be blown off and be re-entrained in the air stream and thereafter deposits on the lower
previously deposited particles.
3 m particle will appear by Kravimetric analysis to represent 1000 particles of 0.3 m diameter. A number concentration analysis rather than a
impactor
staes causin
spurious results, e.E. a
gravimetric analysis will be more ideal for small particles for
avoidinK lonK samplin period in order to obtain weighable quantity

of dust, especially submicron particle analysis downstream of a
highly efficient filter.

Another possible reason for such spread of experimental
9 could be due to the experimental techniques employed by Gooding et a15. Likewise for these efficiency tests as in the loadability tests (Figs.3 and ), the velocity profiles upstream and downstream of the HGMF filter were not determined prior to samplin6 into the cascade impactor. A sampling velocity
results in Figs. 5 based on the filter downstream average velocity was used throughout
for all the samplin positions.
But,
it must be highly likely that
the velocity profile at the filter upstream sampling position would
be greatly distorted and irregular after the flow had passed
through the cyclone.

by Gooding et
Also the characteristic swirling flow of the
cyclone might not be cured completely by means of a baffle as used
alS.
Thus, the velocity at any sampling position
could be vastly different from the downstream average velocity

value.
Similarly for the downstream side, the velocity profile
would also be highly non-uniform after passing through the random

matrix configuration as was observed in similar HGMF work at
Sheffield
University7. Hence,
anisokinetic sampling
miEht have also

a pilot
contributed to such scatters in Figs. 5
9.
Gooding4
conducted an impressive on-site testing of
HGMF plant on a sinter plant in Pennsylvania, U.S.A.
In the
production of iron, the sintering process is used to combine iron
ore fines with flux in the form of limestone or dolomite and with
other iron-bearing materials such as flue dust, mill scale,
turnings and borings to form a blast furnace feed material of

appropriate composition and size.
In the pilot plant, a slipstream

did not achieve a
from the windbox exhaust of the sinter plant was drawn into and
through the HGMF filter.
However, the pilot plant

efficiency
satisfactory level of filtration
because
of the
extremely low magetization value of the sinter dust that actually
reached the HGMF filter. Samples obtained at the filter inlet
78
A.C. LUA AND R. F. BOUCHER
showed a saturation magnetization oF 2.4 emu/g at 0.3 Tesla
compared
to
emu/g
For
samples
obtained
From
the
plant
precipitator hoppers.
Moreover, the saturation magnetization oF

The larger sinter dust particles were
this sinter dust at the HGMF Filter inlet was only 1/20 that oF the
BOF dust reported earlier.
believed to be higher in iron content but substantial quantities oF
these particles did not reach the HGMF Filter due to gravitational
settling in the slipstream upstream oF the HGMF Filter and also
possibly less large particles were drawn From the windbox exhaust.
3ize measurements and subsequent chemical analyses revealed the
outlet dust after the HGMF Filter contained a high percentage oF
very Fine particles oF alkali chlorides which would not be captured

by a magnetic Filter.
pilot HGMF plant,
Thus, For Future on-site evaluation oF a

industrial
emissions
suitable
have
to be
identified, dust characterization and laboratory HGMF tests have to

be carried out on the actual source prior to setting up the pilot
plant.
Drawing on the set-backs by Gooding et
alS,
a thorough and
careful HGMF investigation on BOF dust was carried out by

his Ph.D dissertation.
Lua7 For
The individual eFFects oF a wide range oF

matrix volume packing Fraction, Filter
operating parameters
length, magnetic Field, gas velocity, wire size and wire aspect
ratio (For rectangular wires) on the Filtration eFFiciency oF the
HGMF Filter were investigated. It was also desirable to determine the matrix dust load capacity or loadability For the BOF dust under diFFerent combinations oF these parameters The schematic set-up oF the HGMF test rig is shown in Fig. 10. Room air is drawn into the 98.4 -.- diameter copper duct by a main
centrifugal Fan and a smalle Fan within the Fabric bag unit.
Air
Flow in the duct is adjusted by 2 Flow dampers.
The pre-Filter and
absolute Filter at the duct entrance clean the incoming air.
3lightly downstream oF the absolute Filter, BOF dust particles are
inected into the main airstream.
The particles are generated by

an aerosol generator which uses the fluidized bed technique to
thorou6hly deaglomerate the dust particles before elutriation. The
aerosol generator has been tested and found to be capable of fully
dispersin the particles as single particles at steady output for
long periods of time, the details and performance of which are

reported
elsewhereS.
Located above the fluidized bed, an aerosol

charges
acquired
neutralizer (Model neutralizes
3054, Thermo-Systems Inc., Minnesota, U.S.A.)

by
the
aerosolized
any static
durin the fluidization process. It contains a sealed source of 10 millicuries of krypton 85 gas. The filter matrix consistin of random stainless steel wires
particles
(AISI
430)
of
rectangular
cross-sections,
is
located
in the
at
approximately
6.5 m downstream of the duct entrance,
113 mm
bore of a water-cooled compensated solenoid magnet. This self-built
31.7 cm lon solenoid magnet,
with an inside notched compensated
coil, has a very homogeneous field distribution a variation of only 1.6 alon a 15 cm length central region of the bore. Wall
static tappings across the matrix are used to determine the filter
FIGURE 10
Schematic set-up of HGMF test rig
8O
resistance.
Upstream and downstream of the matrix, velocity
profiles were determined with a pitot tube and wall static tappings
(in accordance with BSI042"
Part 2A"
Aug.
1973)
to obtain
isokinetic sampling rates as well as the average air velocity in
the duct. Sampling ports are located upstream and downstream of the
matrix.
Also in the downstream section of the duct, a thermometer

After the main
is inserted for reading the air temperature.
centrifugal fan, a filter unit consisting of fabric bags cleans the
air, if necessary, before discharging to atmosphere.

Particle sizes and concentrations upstream and downstream of
the matrix were det.ermined by an optical particle spectrometer,
CSASP-100-HC (Particle Measuring Systems, Inc., Colorado, U.S.A.).

The spectrometer employs a multi-mode laser light source for
particle illumination.
It. eliminates errors associated with
diluting the sample or passing it through a micro-jet before
optical analysis.
retained
The 10 mm diameter flow in the inlet tube is

and
particles
throughout,
are analysed
in a
small The
optically determined area at the centre of this stream.
instrument was calibrated by the manufacturers and checked during
the present investigation using Dow latex particles, and found to

be accurate. It covers the size range
0.24-7.0 m
in
3 overlapping
sub-ranges with 15 sizes in each sub-range. A gravimetric particle size analysis, such as the use of an inertial cascade impactor, was
avoided in anticipation of substantial
errors, especially with
sub-micron particle sizes, as discussed earlier in relation to the
work by Gooding et
alS.
The schematic arrangement of the sampling
train for the optical particle spectrometer analysis is shown in
Fig. I.
For scanning electron microscope examination of particle
characteristics, samples were collected on a

the filter sampling train shown in Fig. 12.
47
diameter
Nuclepore polycarbonate membrane filter, pore size
0.8m, using
The BOF test dust used was collected dry at the stack of a
steelmaking plant.
The dust was milled, dried completely and
81
T Denotes flexible P.V.C.tubing.

Dust stream
Probe. inlet
tube
Probe Outlet
scattering aerosol spectrometer probe (CSASP- I00- Hc ).
valve
valve
Air-tight
Vacuum
PMS data acquisition
Interface
NEC PR-IOOI
fast Iper printer
drum
system,
Gate
Air Iced
system
(DS-32)
FIGURE 11
Schematic arrangement of the sampling train for optical particle spectrometer analysis
Duct
,Sampling
][_. "5mpling Tll tube

47ram di Filter holder
Needle (:) valve
( On-off valve
Vacuum
Air-tight
drum valve
Denotes flexible P.V,C. tubing
Air bleed
FIGURE 12
Schematic arrangement of the filter sampling train
82
sieved to size range below
300
before it was dispersed into the
test duct by the aerosol generator. Scanning electron micrographs of dust particles (Fig. 13) show that the particles are fully
dispersed as single particles.
The particle size distribution of
It shows the particle sizes are below 2 m diameter, mainly in the submicron range. Figure 15 shows
the relationship between the
particle magnetization and
the BOF dust is shown in Fig.
I.
the
magnetic field for the BOg dust.
The particle magnetization tends
to a saturation value at an applied field of about 0.3 Tesla. Consequently, the particle magnetic susceptibility tends to an
asymptotic value for increasing magnetic fields (Fig.
16).
Thus,
these data show that BOF dust is ferromagnetic.
A spectrographic
The other
analysis shows that the iron content of the BOF dust consists of
53
and
by weight of
Fe304
and
Fe203
steel
respectively.
constituents are
ZnO, PbO, MnO, SiO 2 and traces of various other
chemical compounds.
AISI
Type
430
stainless
wool,
consisting
of
ribbon-shaped rectangular wires, was used for the matrix. Four
grades of wires, differing in size and aspect ratio, were tested.
The dimensions of these wires are given in Table I. The hydraulic

radius is defined as a ratio of twice the cross-sectional area to
the wetted perimeter, whilst the aspect ratio is the ratio of width
to thickness.
Table I.
Dimensions of different grades of wire. Width
Thickness
Hydraulic radius
Aspect
ratio
Grade
(m)
112
(m)
33.5
29.5
O0 0
48
165 198
38 69
97
48.5 65.0
2.33 3.55 2.39 2.04
FIGURE 13
Scanning electron micrographs of BOF dust particles: (a) (c) and (d) and (b) 1250X, white bar spacing = 10 2500X, white bar spacing = 10 m
SY MSOL
CONCENTRATION (rag/m3).
226 2109 4017
0.20
o
0.16
0.12
0.3
0.5 0.6 0.7 (18 0.9 1.0 0.4 PARTICLE DIAMETER, m
1.2
1.4 1.6 1.8 2.0
FIGURE 14
Normalized BOF particle size distribution
$0
E
0
10
10
11
12
APPLIED MAGNETIC FLUX DENSITY, 8o(kG)
FIGURE 15
Magnetization curve for BOF dust
85
APPLIED MAGNETIC FIILD, T
FIGURE 16
Magnetic susceptibility of BOF dust
To study the performance of the HGMF filter with dust loading, Dust particles were sampled long-term tests were carried out.
isokinetically and continuously from the centre of the duct into
the optical particle spectrometer,
downstream of the matrix.
Pressure drop across the matrix was recorded at regular intervals.

Sampling was stopped when the penetration of the smaller particles
was becoming excessive. The matrix was removed for gravimetric
Upstream sampling, without the matrix, was carried out both before and after downstream sampling to obtain an average upstream particle concentration for computing filtration
analysis.
efficiencies. Figures 17 ,18 and 19
show the matrix loadability
for magnetic fields of 0.05, 0.4 and 0.8 Tesla respectively, with
all
the other operating
parameters remaining constant.

field
increases
As
expected,increasing magnetic
the
filtration
efficiency for all particle sizes shown before loading effects

become
significant and lead
to deterioration of efficiency,
especially for the smaller particle sizes.
For a very low field of 0.05 Tesla (Fig. 17 ), the Filtration

efficiency for all particle sizes shows only modest deterioration
up to the stage when the matrix has collected about 2 times its own
86
Io0
98
98
++
z
U
0"
93 IJ, m
.72 }.l,
I.I,m
90
X"
,0. 585
0.45 bl,
88
86 84
0.37p.m
82
80
0.31 I.un
78
KEY
72
DIA (p.m)
O. 265 0.315 O- 37 0.4 5 0.565 0.72
0 93
o
A
70
0
x
70 90
1.89
2.43
0.265 JJ.m
IC
150
FIEL0
0.05 T
FILTER LEH I0 cm
7.32 mls GAS VELOCITY PACKING FRACTION 0.01 430 STAINLESS STEEL WOOL GRAO( 0
I00 0
10
30
50
110
LOAOING TIME, rains

0 0.27
0.81 1.35
2.97
(TOTAL MASS OF
OUST COLLECTEd)/(MASS OFWIRE WOOL)
FIGURE 17
Matrix loadability at 0.05 Tesla
mass of dust.
Larger captured BOF particles create their own local

sites for further capture, almost
Kradients thus becomin improved
maintaining filtration efficiency despite such dust loading.
deterioration in the filtration efficiency is clearly evident for

the smallest particles (0.265 pro) thereafter, and for all particles
when the matrix has collected about
3 times
its own mass of dust.
The total amount of dust collected at the end of the test was 162.8
g, a mass ratio of 2.89 (mass of dust collected to mass of wire
wool) or a volume ratio 4.72.
87
At 0.4 Tesla (Fig. 18 ), the higher filtration efficiency is

sustained for all the particle sizes until the matrix holds twice its
own mass of dust.
Thereafter
it
suffers
significant
and volume
deterioration only for the smallest two particle sizes up to the
maximum loading tested of 216.2 g, a mass ratio of raio of
3.84
6.27.
The higher applied
field
thus
endows
good
loadability, providing high efficiency with high dust loading.
At 0.8 Tesla (Fig.19), significant deterioration in filtration efficiency occurs at lower matrix loading (mass ratio 1.5) than at 0.4 Tesla, but from a higher initial level. For loading mass ratios
of 2 and greater, the efficiencies for these two fields are almost
identical for all but the smallest (0.265
it is worse with the higher field.
m) particles, for which
Since the particles approach

(.]
99
O. 37rn
95
KEY
O
A ()
0. 265 O. 31 0.37
0,4
ETIC FIE 4T FILTER LE I0

GAS VELOCITY
7.32 PAI FRAI 0.01 430 STLESS E
L26m
: AOE
0
0.5
0,72
0
0.93
3oo
200
20
40
60
,,,,,, 0.48
(TOT
LOADI
I.
TIME, ns
1.92
80
0.9
,..,
20
DT T)/(M
6 2. )
3.
FIGURE 18
Matrix loadability at
0.4 Tesla
88
100
99
.._;..-....._
O. 25 0.3 0.370 O. 450 O. 565 0.720 0.9
o. 72.
0.45m
0. 315
x
0
MAGTIC FIELD 0.8 T FILTER LEITH I0 m S VELITY 7.32 m/s PAI FRACTION. 0.01 430 STAI NLESS STEEL WOOL GRADE
400
x
0.2m
200
100-
%
0.(2
40
1.24
,"
2.48
:).10
LOADING TIME:, rain.

0
(TOTAL
3.72
4.34
MASS OF OUST
COLLECTED)/(MASS
OF WIRE WOOL)
FIGURE 19
Matrix loadability at 0.8 Tesla
saturation at above 0.2 Tesla, the general conclusion is that
little or no further improvement in loadability is to be obtained
beyond such field levels.

There is
good agreement in the pressure drop versus matrix
mass load ratio of Figs.17-19.
For the heaviest loading of
mass ratio (7.17 volume ratio) the pressure loss has multiplied 3.5 times. Thus, substantial geometrical distortion which results in reduced magnetic gradients, appears to be the primary factor
reducin efficiency.
The reduced efficiency is unlikely to arise
either from re-entrainment (since no large particle aglomerations
were detected by the particle spectrometer) or from particle

bounce-off which produces highly erratic results.
89
Figure 20 shows the matrix loadability without a magnetic

field. Although the initial filtration efficiency for all particles
is high, rapid fall occurs very soon afterwards with hardly any
dust collected in the matrix; the fall is most marked for the
smallest size.
This very low matrix loadability can also be
observed from the flat pressure drop curve. The total amount of
dust collected was only
2.3 g (mass
ratio 0.041 or volume ratio
0.067) thus demonstrating the importance of an applied magnetic

field to both initial and sust.ained particle capture.
The matrix loadability for a lower packing fraction is shown

in Fig. 21.
The filtra.ion efficiency is maintained over all
particle
sizes
up
to
the
maximum
loading
tested
(166.6 g)
80
70
E E
3O
LOAOING TIME. rain

0
(TOTAL MASS OF
0.041
DUST
COLLECTED)/(
MASS OF WIRE: WOOL)
FIGURE 20
Matrix loadability without a field
9O
100,
98
97,
9,
E
0 A
.
o o.
GTIC FIgLD -0.2 T
FILTER LTH- loom GAS VELITY -7.32
PARTICLE
()
ING FRACTION -0.005 STAINLESS STEEL WOOL: A 00
o.2
+
120
O. 315 O. 37 O. 4
_ioo
80
O0
LOADI TIME. rain
0.
FIGURE 21
I.7
2)96
(TOTAL MASS OF ST
COLLECTED)/(MASS
.g
OF WIRE WOOL)
Matrix loadability at
7.32 m/s
corresponding to a mass ratio 5.92 or volume ratio

of the wire.
9.66
times that
This matrix loading (packing fraction 0.005) is
higher than those for Figs. 18
and 19 (packing fraction 0.01). The
reduced loadin for the hiher packin fraction is probably due to increased wire-wire interference which is detrimental to particle
capture3. Furthermore,
despite the extremely high loading in
FiE.
21, the filter resistance doubles slightly, again benefitting from greater wire separation.
Figure 22 shows loading under the same conditions as Fig. 21
except that the velocity is increased from 7.32 to 11.51 m/s.
The

I00
91
gg
98
g7
%-\
.7
g3
gl
31
go
8g
88
KEY
PARTICLE OIA()
0.265 0.315 0.37 O. 45
MA FIELD FILT LEITH
O. 2 T
lOom
3o
300 250 200
-----x-0
40
G VELITY I1.i m/s PACKI FRACTION 430 STAINLE STEEL WL A 00
0.72
I00
80
120
lt50
200
240
LOADING TIME
0
rnins
O.
1.88
2 82
3.76
4.70
(TOTAL MASS OF DUST
COLLECTED)/(MASS
5 64 OF WIRE WOOL)
FIGURE 22
Matrix loadability at 11.51 m/s
filtration efficiency, although lower, is again sustained for the
smaller particles as
the matrix loads
up.
For the larger
particles, however, a secondary effect of particle bounce-off from
the collector surface is thought to occur when the ratio of
magnetic to inertia forces at impact is less than a critical value.

The particle is then carried away if the magnetic forces are too small to overcome the viscous forces in the absence of forward
particle inertia. This is quite distinct from re-entrainment where
captured particles are later torn off the collector surface,

producing agglomerations appearing to be large size particles.
In
Fig. 22, particle bounce-off occurs at the outset and continues for
92
all loadings, producing erratic fluctuations, although good average

efficiency is still maintained.
Despite the excessive velocity, no
large particles outside the range dispersed were detected within

the 7pro capacity of the particle spectrometer, suggesting that
re-entrainment of particle agglomerations played no part in the
fluctuations.
For the HGMF filter efficiency tests, sequential 8-point isokinetic dust samplings were carried out, both upstream and
downstream of the matrix, for 30 rain at each point.
concentrations.
The sample was
analysed by the particle spectrometer for particle sizes and number

The effect of varying one basic parameter at a
time on the filter performance was examined with respect to the
particle size.
The
effect of applied magnetic
field on
the filtration
Increasing
efficiency is shown in Fig. 23 for grade 00 wire wool.
..
1"
U
"
PARTICLE CONCENTRATION BEFORE
26.79
36-46
96-
Z
n-
SYMBOLS
e--.---e
MAGNETIC
FIELD(T)
0-05 02
TEMP(C) FILTER(mg/mo)
19-7
18-
92-
o---.o
m.---e
0-4
18-1
21-27
0-
FILTER LENGTH -15cm RI&CKING FRACTION 0-01 GAS VELOCITY 7-32m/s 430 STAINLESS STEEL WOOL: GRADE O0
0-2
880
FIGURE 23
0-4
0-6
0-8
1-0
1-2
1-4
1-6
1-8
PARTICLE DIAMETER, pm
Effect of efficiency
applied
magnetic
field
on
filtration

magnetic field improves the efficiency.
93
The improvement is only
marginal for substantial increase above 0.2 Tesla where the particle magnetization approaches its saturation value (Fig. 15).
At 0.05 Tesla, it shows a decrease in filtration efficiency for
0.93 m and larger particles.
However, the efficiency should never

This phenomenon again suggests
fall below that of small particles.

particle agglomerates.
particle bounce-off rather than re-entrainment which produces large
None were detected within the 7
capacity
of the spectrometer. The effect of filter length on the filtration efficiency is

shown
in
Fig.
24.
Extra
length
improves
which
efficiency
the
for
is
increasingly
smaller
particles,
above
length
SYMBOLS
m--..-m
PARTICLE CONCENTRATION BEFORE LENGTH(cm) TEMP(eC) FILTER(mIIm3)
FILTER
5
19-7
16-79 12.54 36.46
z 92
o----o e--.--e
10
17-8 18-8
15
o----o
25
20-7
14-53
MAGNETIC FIELD 0-2T PACKING FRACTION -0-01 GAS VELOCITY 7-32rn/s 430 STAINLESS STEEL WOOL:GRADE O0
0-2
0-4
0-6
0-8
1-0
1-2
1-4
1-6
1-8
FIGURE 24
Effect of filter length on filtration efficiency
94
overgenerous.
:
By examining the filter penetration equation i.e. P
}, doubling the filter length should result in squaring the penetration. But, from analysing the curves for the 5
exp.{-
.za(l_F)
4FLRc
cm and 10 cm filter lengths, the reduction in penetration for

doubling the length is less dramatic, thus showing the same general
trend as observed by Gooding et al (Fig. 5).
The effect of volume
packing fraction of the matrix on the filtration efficiency is
shown in Fig. 25 for wire wool grade O.
Increasing the packing
fraction improves the filtration efficiency throughout.
But, the
is
decrease
in
penetration
for
increasing
packing
fraction
generally less than that predicted by the filter penetration

equation.
Thus, increasing packing fraction very often lowers the
effectiveness of individual wires in a filter, due primarily to

increased wire shadowing and magnetic interference
effects3.
pr
,"
:/
//p._._O____--O--
"0
92-
SYMBOLS
88-
PARTICLE CONCENTRATION BEFORE PACKING FRACTION TEMP(C) FILTER(mI/m3)

0"005 0.0075
18-9
= =
o-- --o
16-J
20-2
11-93
e-.--e
0.010 0-015
19-5
19-9
o-----a
13.18
blAGNETIC FIELD 0-2T FILTER LENGTH 15cm VELOClTt 7-52 m/s 450 STAINLESS STEEL WOOL: GRADE 0
70
(2
OA
(6
0-8
14)
1-2
1-4
1-8
PARTICLE DIAMETER, pin

FIGURE 25
Effect
of matrix efficiency
packing
fraction
on
filtration

The
95
effect of increased gas velocity on the filtration
efficiency is shown in Fig. 26.
(In the inertial regime, if the
capture radius is less than the wire radius, particles should have an increased likelihood of capture at higher velocities due to
In Fig. 26, increased velocity generally produces improved capture efficiency, thus demonstrating the significance of particle inertia. At 13.62 m/s between 0.565-1.155
their higher inertia.)
m, the deterioration in efficiency
is another manifestation of
particle bounce-off in which at these higher velocities, the ratio
of magnetic to inertial forces is reduced. Above 1.33m, either

adequate matrix length or the highly favourable filter front face
(which is discussed later) ensures particle capt.ure. The effects of wire size and aspect ratio on the filtration
efficiency are shown in Fig. 27 for a given packing fraction of
0.01. Decreasing wire size with approximately the same aspect ratio
PARTICLE CONCENTRATION
VEL(m/s)
7-32
9.44
SMBOLS
e-----e
TEMP(*C) FILTER(rncg/rn3)
21-5
BEFORE
53-20 35.43
49-4
MAGITIC FIELD -0-2T FILTER LENGTH -15cm

430 STAINLESS STEEL WOOL:GRADE O0
o----o
22.7
1:5-62
ZO-6
0-2
OA
0-8
1<3
T2
1-4
1-6
FIGURE 26
Effect of gas velocity on filtration efficiency

increases the filtration efficiency as expected because finer wires
offer greater surface for the same packing fraction. But, fine wire
(grade 0) aspect ratio 3.55 has lower efficiency than medium wire (grade 2) aspect ratio 2.39. The lower efficiency for the grade 0 wire must probably be due to its higher aspect ratio. This can be
explained as follows. Wires of rectangular cross-sections are easily magnetized if the long side lies in the direction of the
applied magnetic field but then offers a small target area, and
conversely are difficult to magnetize when offering the larger
target area. Thus rectangular wires are less effective than square or round wires. If the front face of the HGMF filter is inspected after
loading and with the magnetic field uninterrupted, the dust has
rJ
9-
Z
e,,
SYMBOLS
92m----e
WIRE GRADE
O0
TEMP(C) FILTER(mg/m
8-8
PARTICLE CONCENTRATION BEFORE ASPECT

6.46
19-5
5-55
2-33 3-55
2.O4,
2
4
20
-8
250
29-06
MAGNETIC FIELD 0-27 FILTER LENGTH- 15cm PACKING FRACTION -0-01 GAS VELOCITY T-32m/s 430 STAINLESS STEEL WOOL
--0
0-2
0-4
0-6
08
10
1-2
1-4
,I I-6
1-8
FIGURE 27
Effects of wire size and aspect ratio on filtration efficiency
97
always collected in the form of spiky structures projecting from

the wire. The dust load carried by these spikes on the front wire
is, by visual inspection alone, obviously much in excess of that on

downstream wires. Although the matrix requires positive cleaning
action to remove its dust load, the spikes simply collapse and fall
away when the field is removed.
By collecting this dust and
scraping carefully any remaining material from the front face, the
ratio of the mass of dust collected on the front face to that on
the rest of the matrix was determined and plotted in Fig. 28 for
varying fields at constant velocity and vice-versa. It can be seen
that for fields of 0.2T and above, at 7.32 m/s, more dust is collected on the front face than in the body of the matrix.
At a
constant field of 0.2T, this effect is even more pronounced at low

velocity but reduces with increasing veloci.y.
VELOCITY = 7.32 rnls
FIELD = 0-2
FILTER LENGTH =lScm. PACKING FRACTION = 0.01 430 STAINLESS STEEL WOOL:
GRADE O0
0
!.
0.I
0-2
MAGNETIC
3.32
FIELD
"/-32
0-3 TESLA.
9-32 mls
0.4
11-3Z
5.32
SUPERFICIAL GAS VELOCITY

FIGURE 28
Effects of magnetic field and gas velocity on particle capture at upstream matrix face
98

The mass is concentrated in the smaller numbers of largest
particles and the probability of capture of these on the front face

is high.
The spiky structures suggest that large particles become
favourable capture sites for other particles by creating stronger
local gradients, thus enhancing the capture radius of the wire and
promoting growth at these favourable sites.
The front face is
additionally favoured by lack of shadowing by or wake effects from
upstream wires.
The results of Fig.
28
were taken during
efficiency tests when the matrix was not heavily loaded. Less
complete results taken during the 1oadability tests (Figs. 17-22)

indicate that increasing proportions of dust collect on the front
face with increasing matrix loading.
3. ECONOMIC COMPARISON
Preliminary economic estimates by Gooding et
a15
indicate that the
capital cost of the HGMF equipment and the power requirements for
magnetic field generation and gas movement can be competitive to
corresponding
air
control
devices
such
as
in
electrostatic
precipitators and wet scrubbers.

the
The HGMF estimates are based on

dusts
which
filtration of highly magnetic
high
flow
velocities
are
feasible
and
low
magnetic
fields
(using
conventional solenoid) would be adequate.
REFERENCES
I. R. F. BOUCHER and K. J. BULMER, I.EEE Trans. Magn., MAG-18, _6, 1665-1667 (1982) 2. CRC Handbook of Chemistr and Ph sics,63rd editlon, edited by R. C. Weast (CRC Press, 1982) 3. A. C. LUA and R. F. BOUCHER, IEEE Trans. Magn., MAG-18, 6_, 169-1661 (1982) C. H. GOODING, Pilot-scale Field Tests of High-gradient Magnetic Filtration, U.S. Environmental Protection Agency Report, EPA-600/7-80-037 (March 1980)
5. C. H. GOODING, T. W. SIGMON and L. K. MONTEITH, Application of High-gradient Magnetic Separation to Fine Particle Control, U.S. Environmental Protection Agency Report, EPA-600/2-77-230 (Nov. 1977) Aerosol Impactors, Fundamentals of 6. FUCHS. N. A.Aerosol Science, edited by D. T. Shaw, (John Wiley & Sons, New York, 1978), Chap.l, pp. 1-83 7. A. C. LUA, An Investigation of High Gradient Magnetic Filtration of Fine Particles from Gas Streams, Ph.D. Thesis, Sheffield University (1982) 8. R. F. BOUCHER and A. C. LUA, J. Aerosol Sci., 13, 499-511 (1982)

074149-Filtration Magnétique

Încărcat de

Informații document

Descriere originală:

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

074149-Filtration Magnétique

Încărcat de

Drepturi de autor:

Formate disponibile

Magnetic ,.eparmion News, 1984, Vol. 1, pp.

HIGH GRADIENT MAGNETIC GAS FILTRATION

AIK CHONG LUA AND ROBERT F. BOUCHER

gradient magnetic separation (HGMS)

growing number of various separation processes using HGMS are

reported in the literature.

These applications include as diverse

as coal desulphurization, mineral beneficiation and the separation

The role of gaseous magnetic filtration is to

A. C. LUA AND R. F. BOUCHER

large-scale emissions of particles

particularly the submcron particle size range.

sources, These very fine

submcron particles are especially harmful as they constitute the

greatest inhalation risk.

They are responsible for numerous

pulmonary diseases as they can penetrate into the respiratory

contributing to atmospheric haze or smog.

Stack gases arising from steelmaking processes contain vast

released to the atmosphere.

The particulate matter

these stack gases has a high iron content and

The current, methods used for stack gas clean-up from

steelmakng processes are mainly electrostatic precipitation, wet.

advantages of HGMF over

electrostatic precipitators, the negative corona mode of operation

alternative solution. In the wet scrubbing process, the problem of

atmospheric pollution has been turned into a waste water pollution

The venturi scrubber which is chiefly used in this

Lastly, fabric filtration suffers

from temperature limitations (the maximum operating temperature for

HIGH GRADIENT MAGNETIC GAS FILTRATION

Dry HGMF has potential applications in other processes where

the nuclear industry, HGMF can be used to remove radioactive

particles, consisting mainly of various oxides of iron, in the

fuel reprocessing plants, radioactive paramagnetic particles are

present as a result of certain shearing

processes; these gases or fumes can be cleaned by means of HGMF

reducing active-waste disposal problems.

and HGMS applications are the removal of iron oxide particles in

ash; the recovery of glass from municipal incinerator residues by

ores; and the capture of

respirable asbestos fibres in working

2. HGMF OF FINE PARTICLES FROM STEELMAKING STACK GASES

A. C. LUA AND R. F. BOUCHER

removed before the stack gases are discharged to

The total iron concentration from several of these

HIGH GRADIENT MAGNETIC GAS FILTRATION

of the filter, dust samples are extracted to determine particle

After the downstream sample ports, the

particle sizes and concentrations were determined using MRI Model

Filter penetration values were computed from

particle sizes and concentrations at filter inlet and outlet.

On the other hand, the EAF dust which has a lower

However, extreme deviations from

isokinetic sampling are apparent in the way Gooding et

A. C. LUA AND R. F. BOUCHER

tests even though these tests were

a 50-percent cut point of about

m) and then into a dilution

also investigated the effects of various

parameters viz. matrix packing density, magnetic field and gas