United States
Environmental Protection
Agency
Industrial Environmental
Research Laboratory
Research Triangle Park NC 27711
Research and Development
EPA-600/S7-84-029 May 1984
Project Summary
Studies of Particulate Removal
from Diesel Exhaust
M. G. Faulkner, J. L. DuBard, and J. R. McDonald
Tests have been conducted to
characterize the collection of participate
emissions from diesel exhaust by
several different methods. The sources
of particulate emissions were 5.7 liter
General Motors (GM) diesel engines.
The control devices that are discussed
include fiber filters, gravel bed filters,
trap/cyclones, and electrostatic
precipitators (ESPs). The overall mass
collection efficiencies, fractional mass
collection efficiencies, and operating
characteristics of these devices were
determined by measuring inlet and
outlet total mass loadings and particle
size distributions.
A device containing a fiber filter was
investigated with three different filter
materials: long-fiber glass wool, batts
derived from fiberglass insulation, and
stainless steel fiber mats. Collection ef-
ficiencies as high as 90% were achieved,
coupled with a quick pressure rise
culminating in gas sneakage. A device
containing a gravel bed of 45.7 cm
diameter and 10.2 cm depth was in-
vestigated with 2 mm steel shot. Effi-
ciencies ranged from 45 to 70% and
increased with increasing system
backpressure.
A two-stage ESP was used to agglom-
erate the primary particulate matter,
resulting in an order-of-magnitude in-
crease in mass median diameter. The
agglomerated particulate is character-
ized. Aerosol sampling data are pre-
sented for the variation in particle size
distribution and the efficiency of trap-
ping the agglomerated particulate in
cyclones, fiber filters, and a granular bed
filter. The ESP/cyclone combination can
provide a collection efficiency of at least
50%. Overall mass removal efficiencies
greater than 80% have been achieved
with an ESP/granular bed filter system
for a duty distance greater than 800 km
(500 mi) at a constant highway speed of
88 km/hr (55 mph). Methods of cleaning
the devices and removing collected par-
ticulate are discussed.
This Project Summary was developed
by EPA's Industrial Environmental Re-
search Laboratory, Research Triangle
Park, NC, to announce key findings of
the research prefect that is fully docu-
mented in a separate report of the same
title (see Project Report ordering infor-
mation at back).
Introduction
The superiority in fuel economy of the
diesel engine compared to a gasoline engine
of the same horsepower has created a wide-
spread market for diesel powered light duty
vehicles due to the increasing cost of fuel.
In addition, automobile manufacturers view
the diesel engine as a possible solution to the
Federal regulation requiring that the average
fuel mileage for each auto manufacturer be
at least 11.6 km/liter (27.5 mi/gal.) by the
year 1985. Unfortunately, one characteristic
of the diesel engine is a high level of com-
bustion by-products, the most obvious of
which is particulate matter (defined by EPA
as "everything collected on a filter at 52 °C
(125°F) after dilution with ambient air in a
dilution tunnel").
The particulate effluent of the diesel
engine consists largely of agglomerates of
very small particles of carbon which have
condensed hydrocarbons adsorbed onto
them. The mass median diameter (mmd) of
the particles ranges from 0.3 to 0.5 pm, with
about 70% occurring at diameters less than
1.0 /^m. The small size of the particulate
places it in a range where the collection ef-
ficiencies of both filtration and electrostatic
devices are reduced. The problem of parti-
-------
cle collection is magnified by the large quan-
tities of paniculate produced, ranging from
0.14 g/km for the Volkswagen Rabbit to 0.53
g/km for the 5.7 liter GM engine. The
ultimate goal for paniculate emissions set by
EPA is 0.125 g/km, which would require that
76% of the paniculate matter from the GM
5.7 liter engine be removed from the exhaust
stream. Over a 8000 km (5000 mi) service in-
terval, this would amount to about 3.2 kg of
particles. Using 120 kg/m3 for the value of
the bulk density of this collected material,
this corresponds to about 27 liters (~1 ft3).
Southern Research Institute (SoRI) has
been studying nonregenerative aftertreat-
ment devices for collection of diesel par-
ticulate. These devices include both those
specifically developed for light duty vehicles
and stationary-source devices which .have
been adapted to vehicular sources. The
devices tested in this study can be classified
as either electrostatic or purely mechanical.
The mechanical devices include a deep bed
filter which can accommodate up to a 15 cm
depth of fibrous filter material, a granular bed
filter, and a barrier filter. Also in this category
are condensation traps which depend on gas
stream cooling to condense particles on a
relatively coarse wire mesh. Electrostatic
devices include an electrified filter, a moving-
belt ESP, and a two-stage ESP used as a
primary collection device and as an ag-
glomerator followed by a mechanical collec-
tor. Data on these devices are presented
following a brief description of the test layout
and particle sampling techniques used.
Test Conditions
The diesel exhaust used in the control
device tests was furnished by 5.7 liter GM
diesel engines. At the SoRI facility in Birm-
ingham, AL, a 1979 Chevrolet diesel pickup
truck mounted on a chasis dynamometer
provided the exhaust. For tests at the
Department of Transportation's (DOT's)
Transportation Systems Center, Cambridge,
MA, an engine mounted in one of the engine
dynamometer tests cells was used.
Most of the tests were run with the truck
operating at a steady road speed of 88 km/hr
(55 mph) to provide a high load for the con-
trol device. At DOT the engine was operated
at 1700 rpm to achieve this same condition.
This provided a gas flow rate of about 7.6
actual nWmin (270 acfm). For lower speed
runs, the truck was operated at 24 or 32
km/hr (15 or 20 mph). These speeds cor-
responded to flow rates of 2.4 and 3.2
m3/min, respectively. The other engine
speed used at DOT was 900 rpm, correspon-
ding to 48 km/hr (30 mph). Total mass and
particle size distribution were measured
before and after each control device to
characterize the particle collection of the
device. The total mass loading information
was obtained by using Gelman 47 mm
stainless steel filter holders with glass fiber
filters.
The particle size distribution data were ob-
tained using modified University of
Washington Mark V cascade impactors hav-
ing seven stages with stainless steel inserts
coated with Apiezon H grease to prevent
particle bounce. These instruments were
used to obtain particle size distributions on
a mass basis over the size range from about
0.2 to 4 um. The impactors were either
mounted in ovens close-coupled to the ex-
haust pipe or wrapped in heating jackets
which allowed them to be operated at the
same temperature as the exhaust gas.
In addition to the cascade impactors, a
Thermo Systems Model 3030 Electrical
Aerosol Size Analyzer (EASA) were used to
determine concentrations and electrical
mobility derived size distributions of particles
in the 0.01 to 0.3 urn size range. A Climet
Model 208 optical particle counter was used
to monitor concentrations in the fine parti-
cle size range. The SoRI SEDS III sample ex-
traction, conditioning, and dilution system
was used as an interface between these in-
struments and the exhaust stream. The
system removes condensable vapors from
the sample gas at elevated temperatures
followed by controlled dilution to particle
concentrations within the operating ranges
of the measurement instruments.
Optical data were not used for detailed siz-
ing information but gave quick indications
of efficiency in several size bands.
Deep Bed Filter
The first device to be tested was the deep
bed filter shown schematically in Figure 1.
Device dimensions are based on a theoretical
study which predicts 82% collection of diesel
paniculate when used with a 10 cm thick
fiber mat of 10 f*m fiber diameter and a
porosity of 0.99. The filter box was fabricated
at SoRI for use in a joint testing program
with the Automotive Research Laboratory of
DOT's Transportation Systems Center in
Cambridge, MA. Fiberglass bans derived
from roll insulation and long-fiber spun glass
(angel hair), which had demonstrated effi-
ciencies in the 80-90% range, were used in
the DOT test. The fiberglass batts had a fiber
diameter of about 15 um and, for the sam-
ple tested, a porosity on the order of 0.99.
However, since the material was very com-
pressible, the porosity was highly variable.
The angel hair had a 20 ^m fiber diameter
and an undefinable porosity since it came
compressed in 0.45 kg (1 Ib) bags and could
easily be pulled apart with the fingers.
The tests were run at engine speeds
equivalent to 48 and 88 km/hr which imply
Exhaust Gas
Outlet Below
Filter
Rectangular
Fibrous Filter
Element
Exhaust Gas
Inlet Above
Filter
4181-317
Figure 1. Deep bed filter.
-------
linear velocities through the filter of 0.17 and
0.27 m/sec. The pressure drop across the
control device was monitored and the testing
was discontinued if the pressure became too
high or rose quickly and then leveled out or
dropped, an indication that the gas was find-
ing an alternate path through or around the
filter material.
Initially, the deep bed filter demonstrated
about 90% collection efficiency at 88 km/hr
(55 mph) and 95% efficiency at 48 km/hr (30
mph) using 15 cm of fiberglass batts for the
filter material. These values are degraded,
however, as the pressure drop across the
filter rose and gas sneakage was induced.
Figure 2 shows the pressure drop across the
filter for one of the tests. In this example the
pressure curve starts low, rises quickly as the
cake builds up, and then levels off. This
pressure behavior proved to be typical of
every run in the series. The front surface of
the filter material was very black and showed
signs of building a cake of paniculate. The
edges were also very black, which implies
the presence of gas sneakage, an effect that
would explain the leveling out of the pressure
curve of Figure 2. Collection efficiencies
measured after the pressure drop curve had
leveled off were quite low. Figure 3 shows
the collection efficiencies calculated from
data taken early and late in the same test
shown in Figure 2. The overall efficiency for
the early data is 89%. The data taken late
in the test, after the backpressure had leveled
off, showed a 6% efficiency.
The behavior of the spun glass in the deep
bed filter was very similar to that of the
fiberglass batts. Initial efficiencies using
about 15 cm of the material were in the
80-90% range for the 88 km/hr equivalent
engine speed. This material appeared to be
less susceptible to sneakage than the
fiberglass batts. This is probably due to the
less rigid form of the spun glass which
should allow the material to flow to fit the
container. Two runs were made with this
material: in the first, the pressure curve
started to break over, an indication of the
onset of sneakage; and in the second, with
spun glass, no break in pressure rise was evi-
dent. This run was terminated due to high
backpressure (23 kPa or 92 in. H20).
During a second series of tests at DOT,
the deep bed filter was run using a stainless
steel fiber mat which has been fabricated
specifically for the test. This material was
composed of 12 pm diameter stainless steel
fibers which were spun into a 0.64 cm (0.25
in.) thick mat which was then sintered on
both sides to give it some rigidity. A 2.5 cm
thick bed (four layers) was used in the test.
The test was run at 88 km/hr (55 mph)
equivalent speed with an upward gas flow
to determine if the soot could be dislodged
I
1
.c
i
Q
1
I
20 40 60 80 100 120 140 160 180 200 220 240 260 280
Time, mm 4181-284
Figure 2. Pressure drop across fiberglass batt.
UL/
10
1
1
i 1 r i i w 1 1 ™ i™~ T """'41 i • i i I 1 1 1-
: A **•••. :
00 \
o o •..
0 0
r Oo000o0 -_
'. o :
O Impactor Before Breakdown OQQ
A EAA Before Breakdown
• Impactors After Breakdown
• EAA After Breakdown
i i i i i I I ll I i I I I 1 1 1 1 i I I 1 I 1 1 1
!T2 10'' 10° 10
u.u
§
o
rcent Efficiency
«?
99.0
i
Particle Diameter, pm
4181-31SA
Figure 3. Efficiencies before and after onset of sneakage for deep bed filter using
fiberglass batts.
-------
from the face of the filter mat by rapping the
filter box. However, the test was cut short
when the backpressure rose to 1.4 kPa (5.6
in. H20) in 25 minutes and then leveled out.
After the test, the filter box was opened and
the filter material was examined: obvious
soot trails on the sides of the fiber bed in-
dicated severe sneakage. There was no
evidence of soot between the layers. The
soot cake on the first surface of the fiber mat
was very dense and tenacious: it could not
be shaken from the surface. Attempts to
remove the cake by using a backflow of
compressed air were only slightly successful,
but the compressed air permanently com-
pressed the filter material. The cake could
be removed only by scaping the face of the
filter. The cake appeared denser here than
it did with the fiberglass batts used in the
previous test, probably due to the smooth
surface of the stainless steel.
Gravel Bed Filter
The next device examined was a gravel
bed filter. It was anticipated that the gravel
bed would agglomerate the paniculate mat-
ter so that it could be collected in a large par-
ticle separator such as a cyclone. A gravel
bed offers the advantages of elimination of
gas sneakage, limited dust capacity (which
should promote fast filling and subsequent
reentrainment), higher linear velocities for
compactness, and rugged filter material
(which will allow stirring or other agitation
of the filter bed to break up clogging and in-
duce loss of collected particles). The device
was built from a 114 liter (30 gal.) drum —
45 cm (18 in.) inside diameter — with a
removable lid. A platform covered with a
steel screen supported the bed, which con-
sisted of industrial grade steel shot. A stirrer,
a rake whose teeth extended 3-4 cm into the
shot bed, was added later in the test.
Initial tests were conducted at the SoRI
laboratories with an upward gas flow to
allow observation of the bed while the truck
was running. The truck was run at 88 km/hr
(55 mph) to supply 7.6 mVmin of exhaust
resulting in a linear velocity of 0.78 m/sec.
The initial collection efficiency using 5 cm
of 2 mm shot was 22%. After 6 hours of
operation, the backpressure dropped and
localized boiling of the shot was observed.
The efficiency was then 5%.
At this point the system was replumbed
to a downdraft. The bed was changed to 10
cm of 2 mm shot and was covered by a
screen. The results of this test are shown in
Figure 4. After 6 hours of operation, the unit
was opened and inspected: some of the shot
had been rearranged due to the direct blast
from the inlet air. Consequently, the screen
was removed, the shot bed was smoothed
4
8
I
30
25
20
15
10
5
0
120
100
«
so
40
20
0
Stirring of Gravel
I
I
I
J I
700
75
50
25
c
•S
Ui
I
456
Time, hr
10
4181-316A
Figure 4. Gravel bed filter efficiency and backpressure.
out, and the screen was recut and replaced.
At this point, although the paniculate had
dispersed itself through the shot bed fairly
evenly, it was also piling up on the top sur-
face. This surface deposit was easily broken
up when the bed was smoothed out and
releveled. This mild stirring of the shot caus-
ed the break in the AP curve in Figure 4.
However, the pressure quickly resumed its
upward trend and reached 27 kPa (108 in.
H20) after 10 hours of operation. Inspection
again showed a thick surface layer of par-
ticulate. The collection efficiency at the end
of the test was 72%.
At this point the system was modified to
remove the screen on top of the bed, install
a baffle so the gas stream would not dig
holes, and install a rake so the top 3 cm of
the shot could be stirred to break up the top
layer. Stirring at 1 hour intervals constrained
the pressure drop to 10-12 kPa (40-50 in.
H20) with efficiencies of about 40%. Emis-
sions increased briefly each time the bed was
stirred. Although the ratio of large to small
particles increased considerably during the
stirring puff, no appreciable agglomeration
effects were observed during the non-stirring
portion of the operation cycle. Therefore the
gravel bed filter must be considered a parti-
cle collector rather than an agglomerator.
Condensation Trap
A condensation trap (the Aut-Ainer) is be-
ing developed specifically as a diesel exhaust
control device by Eikosha Co. in Japan.
Figure 5 is a sketch of one version of the Aut-
Ainer with three bands of stainless steel
mesh followed by a cyclone for soot collec-
tion. This mesh is a coarsely woven belt of
0.2 x 0.4 mm flat wire which has been rolled
around the central tube and sandwiched bet-
ween perforated plates. The device depends
on cooling condensation for agglomeration
and collection of particulate. The cooling is
supplied by the central tube which acts as
a ram air tube. Heat transfer is augmented
by several perforated plates across the gas
flow and connected to the cooling tube. The
exhaust gas flow is expected to tear ag-
glomerates from the wire mesh and swirl
them through the cyclone to a catch bag on
the side of the device, resulting in an effec-
tively self-cleaning device. The only
maintenance required would be periodic
changing of the soot collection bag.
Two models of the Aut-Ainer were tested.
The first was the one shown in Figure 5. The
first two bands of steel mesh are 5 cm thick
and the third is 2.5 cm thick. The mixing of
the cooling air with the exhaust at the end
of the Aut-Ainer initially caused some con-
cern because the effects on the condensable
hydrocarbons in the exhaust stream were
unknown. This difficulty was overcome by
extending the cooling air tube down the
center of the sampling port tube so that un-
diluted exhaust could be sa'mpled. After this
section, a baffle was placed in the exhaust
pipe to promote mixing and another sampl-
ing tube was placed downstream to allow
sampling of the diluted exhaust. The amount
of dilution present was determined by ex-
amining the concentrations of CO, C02, and
NOX present before and after dilution and
deriving a correction factor based on the
assumption that the quantities of these gases
were invariant.
The first Aut-Ainer was examined at DOT.
According to the manufacturer, the Aut-
Ainer requires about 12 hours of running-in
time to obtain maximum efficiency. After 20
hours of running at 1700 rpm (88 km/hr),
which corresponds to a linear velocity
through the filter of 2.9 m/sec, an efficiency
of 12% was obtained with a pressure drop
of 5 kPa (20 in. H20). The manufacturer also
-------
pointed out that the device was designed for
an engine in the 2-3 liter displacement range
rather than the 5.7 liter GM engine being
used. By scaling speeds and displacements,
48 km/hr (900 rpm) on the 5.7 liter engine
can be considered the equivalent of the
desired test speed of 88 km/hr for a 3 liter
engine. Therefore a test was run at 900 rpm
which yielded an efficiency of 32%. The ef-
ficiencies quoted are overall figures derived
from filter and impactor data. Figure 6 shows
size dependent efficiencies for both engine
speeds. Ultrafine data indicate that the
device was condensing but not collecting
large quantities of small particles.
The second Aut-Ainer tested was de-
signed after the DOT test by Eikosha Co.,
which believed that a larger design would be
more suitable for an engine of 5.7 liter
displacement. This device is essentially the
same as the one shown in Figure 5 except
that (1) the three sections of mesh are 18 cm
long for a total of 54 cm, and (2) a bleed-off
pipe is connected to the back of the cyclone.
The outlet used a double pipe which carried
the cooling air past the sampling ports before
allowing mixing. At the suggestion of the
manufacturer the catch bags were enclosed
so that the exhaust through the bags could
be examined. The flow rates at the inlet and
regular outlet were close to 7.6 actual
mVmin (270 acfm) which gives a velocity
through the primary filter of 2.1 m/sec. The
flow through the bags barely registered on
the water manometer used with the pitot
tube. Based on an estimated 0.18 cm (0.05
in.) deflection, a flow rate of 0.9 actual
mVmin (31 acfm) was obtained. This is only
14% of the total flow.
The large capacity Aut-Ainer was tested
at SoRI using the conditions for which it was
designed; i.e., a 5.7 liter vehicle operating
at 88 km/hr. Filter data at the regular gas
outlet indicated that the device was collect-
ing 35% of the particulate after 40 hours of
operation. Electrical aerosol analyzer (EAA)
data taken at this point showed a negative
efficiency for particles of less than 0.7 /^m
diameter, which suggests that this device,
like its smaller counterpart, was condensing
vapors to form small particles.
The filter samples taken from the exhaust
through the catch bags were pale yellow
which implies that only negligible quantities
of soot passed the bags. However, these
filters showed a larger weight gain than the
very black filters from the regular exhaust.
The collection efficiency through the bags
was 22%, with negative efficiency again oc-
curring for the small particles. The
temperature of the gas at this outlet was
84°C (184°F) compared to 132°C (271 °F) at
the regular outlet. Gas chromatography data
from the residue collected on the filters sug-
[V
Air Cooling — — •• 1
\r
l.ll
i !!
r .! i
-•ili-KV
l )'
u •
! i ! ' ' — 7
iil'rf//^
Exhaust Out
Stainless Steel Mesh '
Figure 5. Aut-Ainer.
Soot Container
4181-297A
100
2
«
10
I* i ^ ~rr^ 11| 6
b°
Mll
O 88 km/hr Impactors
£ 88 km/hr EAA
• 48 km/hr Impactors
m 48 km/hr EAA
. i i l l I ll l
I | I I I I 11 I I I I I I I 11.9.90
0.0
u
c
90.0 £
I
fi
10~* W"1 10°
Particle Diameter, urn
Figure 6. Aut-Ainer efficiency for 88 and 48 km/hr speeds.
4181-314A
gest that the weight gain was due to con-
densed hydrocarbons by showing a distinct
shift toward lighter, lower condensation
temperature, organic compounds in the sam-
ple from the catch bag exhaust filter com-
pared to a sample from the regular outlet
filter.
A different condensation trap, designed by
the University of Tokyo, was also tested.
This device has no air pipe but is designed
to maximize its surface area for cooling. The
device is 5 cm thick by 40 cm wide and con-
tains three 27 cm long sections of the same
steel mesh. There is no cyclone in this
device; therefore, any agglomerates which
are reentrained should be discharged
through the exhaust outlet. The device was
built for a 1 liter engine having a maximum
gas flow rate of 3.5 mVmin. Consequently,
the truck was run at 32 km/hr, giving a gas
flow of about 3.2 mVmin. and a filtration
velocity of 2.7 m/sec. After 20 hours of
operation the efficiency was 36% with a
backpressure of less than 3 kPa (12 in. H2O).
EAA data indicate that the device is not ag-
glomerating at all but is producing a conden-
sation fog of very small particles.
Barrier Filter
The barrier filter tested was a ceramic
monolith made by Corning. The device was
run at an engine speed equivalent to 48
km/hr (30 mph) and showed efficiencies of
92% using impactors, and 97% using the
electrical aerosol analyzer. However, its
lifetime is very limited due to the rapidly ris-
ing pressure drop which reached 35 kPa (140
in. H20) in 7 hours.
5
-------
Electrostatic Precipitator
An electrostatic precipitator (ESP) was
developed by SoRI and tested at the SoRI
laboratories. The ESP is a cylindrical, two-
stage device that obtains maximum collec-
ting plate area in a small volume (Figure 7).
It is designed to be wet-flushed periodically
in a vertical orientation. The internal diameter
(ID) of the cylinder is 20 cm (8 in.). A volume
gas flow of 6.2 m3/min (220 acfm) at 88
km/hr (55 mph) gives an average linear gas
flow down through the ESP of 3.2 m/sec.
For the charging section, five star-shaped
corona electrodes are mounted on a vertical
high-voltage rod chosen for mechanical
strength. The corona electrodes were
typically operated at 40 kV (negative) and
about 250 >iA. The product of negative ion
density and gas treatment time in the cor-
ona section is
N0t = 0.6X1013sec/m3,
at an electric field of approximately 4 x 10s
volt/m. The electric field for particle charg-
ing could be increased by using improved
high voltage insulation.
The collector section is a set of concen-
tric cylinders (overlapping length 28 cm),
alternately grounded and biased negative.
The total collecting area is 1.18 m2 (12.7 ft2).
For a volume flow of 6.2 rrWmin (220 acfm),
the specific collection area is about 11.3
m2/(m3/sec) (58 ft2/1000 acfm). The collec-
tor cylinders have a nominal spacing of 0.6
cm. After wet-flushing loose particulate, the
cylinders will hold about 6 kV without arc-
ing. To increase the collecting electric field,
the voltage is increased typically to 15 kV,
with steady arcing carrying about 1.5 mA.
Then the collecting electric field is about
24 x 10s volt/m.
Early tests of the ESP produced little per-
manent collection of diesel particulate
because of reentrainment. Preliminary mass
train sampling (non-isokinetic) indicated an
overall mass removal efficiency of about
30-40%. Later measurements (with cascade
impactors and a backup filter) of the mass
removal efficiency of the ESP gave values
of about 26%. While the ESP achieved only
low collection of diesel soot, there was
substantial agglomeration of the particles.
Electrical aerosol size analyzer data showed
roughly 50-70% removal of particles of
aerodynamic diameter on the order of 0.1
fjm. Moreover, the mmd was raised by an
order of magnitude. This phenomenon was
easily confirmed by visual observation of
some microscopic agglomerates. Micro-
scopic examination of particles at the ESP
outlet showed a prepondrance of very loose
and fluffy agglomerates on the order of 25
jjm in actual size. Settling velocity ex-
periments with large agglomerates indicated
that the aerodynamic diameter was roughly
6
Insulator
Cable to High
Voltage Supply
Spray Nozzles
Inlet
Support Spider for
Grounded Cylinders
Collecting
Section
0.30 m Long
(12 in.)
Cable to High
Voltage Supply
Star-Shaped
Electrodes
(Detail)
To Spray
Nozzles
Support Spider for
High Voltage
Cylinders
Insulator
Support Spider
\m Outlet
Removable
Fiber Filter
Storage Tank
Figure 7. Schematic diagram of the precipitator.
4181-299
10% of the physical particle diameter. This
was consistent with later cascade impactor
measurements of the mmd at the ESP outlet.
Subsequent testing of the ESP emphasized
its role as an agglomerator, with the ESP
followed by some other device for trapping
the agglomerated diesel soot.
The electrical resistivity of the diesel soot
is of concern in regard to surface conduc-
tion through sooty deposits on insulating
standoffs in the ESP. The resistivity of a bulk
sample of diesel soot collected at the ESP
outlet was measured in moist air using an
AS ME PTC-28 test cell. The soot was com-
pacted as little as possible. With temperature
and moisture content comparable to those
of the diesel exhaust gas, the resistivity of
the fluffy agglomerated soot was about 107
ohm-cm.
Several common cleaning fluids were
bench-tested for effectiveness in cleaning
the ESP. They were tested both on bulk
-------
samples of diesel soot and on soot layers im-
pacted on metal parts. Perchloroethylene, a
common dry cleaning fluid, was chosen for
cleaning the ESP because it is effective,
nonflammable, and of relatively high boiling
point and low toxicity.
The ESP was occasionally flushed after
the device had cooled down overnight.
About 4 liters (1 gal.) of perchloroethylene
was pumped directly from a supply container
through four spray nozzles in the top of the
ESP. The dirty fluid flowed directly from an
outlet in the bottom of the ESP into a take-
up container. Teflon insulators were thor-
oughly cleaned, but there always remained
a thin film of soot on metal surfaces which
could not be removed by any cleaning fluid
(without rubbing).
One problem in the aftertreatment of
diesel exhaust is the collection and handling
of the large volume of low bulk density soot
(roughly 120 kg/m3). An advantage of wet-
flushing with perchloroethylene is that the
collected soot is left in a highly compacted
state when the dirty cleaning fluid is
evaporated (to be followed by condensation
in a closed system). The measured bulk den-
sity of dry residue from the wet-flushing pro-
cedure was 1300 kg/m3. Consequently, the
collected paniculate storage volume required
for an 8000 km service interval would be
reduced to 3 liters (~0.1 ft3).
ESP/Cyclone
Three series of tests were conducted on
three different ESP/cyclone combinations,
with the diesel truck running at 88 km/hr (55
mph). First, a Fisher-Klosterman XQ-4 in-
dustrial cyclone was connected in the ex-
haust line after the ESP. The cyclone was
calibrated by the manufacturer under am-
bient conditions to have a DM cutpoint of 2.0
^im for a volume gas flow of 5 m3/min (180
acfm). By Lapple's law, the Dw cutpoint ex-
trapolates to 1.6 ^m under actual operating
conditions: 175°C, 2.5 x 1Q-6 Pa-sec, 6.2
rrvYmin, 36.5 m/sec, and 108 kPa positive
pressure (27 in. H20). There was a pressure
drop of 44kPa (11 in. H20) across the ESP,
and a temperature drop of 5°C. Across the
cyclone, there was a temperature drop of
10°C, but no measurable pressure drop on
a mercury manometer. Performance of the
ESP/cyclone combination was tested by ex-
tracting samples of the exhaust gas through
.cascade impactors. Three test stations were
used to accumulate data simultaneously at
the outlets of the diesel truck, the ESP, and
the cyclone. Average results of several tests
are shown in Figure 8 where the agglomera-
tion of carbon soot by the ESP is
demonstrated: the aerodynamic mmd at the
ESP outlet is increased by roughly an order
of magnitude to about 3.5 pm.
700
80
60
-------
measured. The three cyclones used were
designated 9, 3, and 4, with specifications
as follows:
Cyclone 9 — Large Cyclone
DM = 15 urn at 0.015 mVmin, 22°C
DM = 17 um at 0.015 mVmin, 170°C
CycloneS — Small Cyclone (mounted inside
Cyclone 9)
DSO = 2.5 /jm at 0.015 nWmin, 22°C
DM = 2.9 /Lim at 0.015 m3/min, 170°C
Cyclone 4 — Cyclone 4 of the Five-Stage
Sampling Train
Da, = 0.6 ^m at 0.03 nWmin, 22°C
DM = 1.1 /urn at 0.03 mVmin, 170°C
Results of testing the cyclones at high flow
rates are given in Table 2. These tests in-
dicate that an ESP agglomerator followed by
cyclonic trapping of agglomerated par-
ticulate is a viable concept for aftertreatment
of diesel exhaust. The mass removal effi-
ciency of such a device is at least 50%. The
mass removal efficiency can be improved by
extensive engineering and testing to match
the cyclone performance to the character-
istics of the agglomerated diesel paniculate.
Methods of disposal of the trapped diesel
particulate require further investigation. An
ESP will require cleaning sooner or later, by
wet flushing or air jet scouring. Soot col-
lected in the cyclone catchpot will have to
be removed and/or compacted. During the
tests of the ESP in combination with other
devices, it was necessary to clean the col-
lector cylinders by wet flushing with per-
chloroethylene after 16 to 20 hours of
operation with the diesel truck running at 88
km/hr.
ESP/Gravel Bed Filter
The gravel bed filter was connected in the
exhaust line after the ESP to test the con-
cept of using a granular filter to trap diesel
particulate that had been agglomerated in
the ESP. The ESP/gravel bed filter was
tested at a diesel truck speed of 88 km/hr
(55 mph). The gravel bed consisted of a 10
cm depth of 2 mm steel shot. Cascade im-
pactor data for the ESP/gravel bed combina-
tion are shown in Figure 9. The combination
was tested with and without stirring of the
surface of the gravel with a rake to break up
the cake of carbon soot that forms on the
surface. Stirring resulted in a large puff of
collected carbon soot passing through the
exhaust line. The curve labeled WITH STIR-
RING represents average data for impactor
runs which included one stir in the middle
of each 30 minute data collection run. The
curve labeled WITHOUT STIRRING repre-
sents average data for runs taken over 3
days, with the gravel bed stirred once each
day before the beginning of testing. This
daily stirring reduced the engine back-
8
i
§
100
80
60
| 40
I
20
j I
0.2
Figure 9.
1.0
Impactor Stage D5o Outpoint, um
10
20
4181-304
Impactor testing of ESP/gravel bed combination. Cumulative percent particulate
mass of aerodynamic diameter less than the impactor stage Dso outpoints.
pressure from a maximum value of 17 kPa
(68 in. H20) to about 12 kPa (48 in. H20).
Comparison of data from afternoon impac-
tor runs and runs the next morning, after stir-
ring, indicated that mass removal efficiency
did not vary measurably with the change in
engine backpressure.
The total mass loading of the impactor
stages plus backup filter is summarized in
Table 3. The overall mass removal efficiency,
without stirring the gravel bed, was 86%.
The data in Figure 9 show that the mmd of
the agglomerated diesel soot (about 3.5 ^m)
greatly decreases when the gas stream
passes through the gravel bed. This is the
result of larger particulate being trapped in
the gravel bed. Figure 10 shows electrical
aerosol size analyzer data for the ESP/gravel
bed combination, with the diesel truck run-
ning at 88 km/hr. In the aerodynamic size
range around 0.1 ^im, the ESP alone
achieved 60-70% removal of particulate
mass. The addition of the gravel bed resulted
in only a small improvement in mass removal
in this small size range. The main function
of the gravel bed is to trap the particulate
mass which has been shifted to a larger size
range by the ESP.
Table 1. Cumulative Percent Mass > D*, of Diesel Soot Agglomerated by the ESP
8.3 \*m 3.9 um 2.5 \un 1.1 \un
0.5 um
Cascade Impactors 17%
Five-Stage Cyclone 24%
35%
35%
49%
44%
61%
50%
67%
57%
Table 2. Summary of Testing Cyclones at High Gas Flow Rates
Cyclone
9
9 and 3
9 and 3
4
Flow Rate
rrf/min
0.55
0.55
0.80
0.80
Pressure Drop
kPa (in. //2OJ
1(4)
56(14)
96(24)
96(24)
Mass in Catchpot
%
23
52
52
37
-------
Because the ESP/gravel bed combination
had demonstrated high-efficiency cleaning
of diesel exhaust gas for a practical operating
time, the gravel bed was reduced to a more
practical size. A gravel bed diameter of 20
cm was chosen, to correspond to the ID of
the ESP and to achieve a reduction in filter
area by a factor of 5. Operating the diesel
truck at 88 km/hr results in a filtration veloc-
ity of 3.2 m/sec. The mass removal effi-
ciency was tested by mass train sampling at
the diesel truck outlet and at the gravel bed
outlet. Test results for gravel beds 5 and 10
cm deep are summarized in Table 4.
These tests indicate that a compact gravel
bed filter can be contained in the ESP hous-
ing, directly below the collecting cylinders,
and can achieve mass removal efficiency of
70% or better, at acceptable levels of engine
backpressure. The gravel bed filter can be
cleaned simultaneously with the ESP by wet
flushing with perchloroethylene. The mass
removal efficiency of the ESP/gravel bed
combination is largely independent of the
running speed of the diesel truck. The higher
mass loadings at lower exhaust gas flow
rates are compensated by collection of
greater paniculate mass per unit volume gas
flow by both the ESP and the granular filter.
However, lower exhaust gas flow rates
reduce the scouring of the insulating stand-
offs in the ESP and lead to more frequent
cleaning.
ESP/Fiber Filter
Two types of ESP/fiber filter combinations
were tested: (1) the deep bed filter, pre-
viously tested on primary diesel exhaust, was
loaded with fiberglass batts and connected
in the exhaust line after the ESP; and (2) the
concept of agglomerating and trapping diesel
particulate in a single compact device, with
a removable cartridge filter, was tested with
an electrified filter. This two-stage device
transforms high axial gas flow past a disk
electrode in the corona charging section into
a much lower radial gas flow in the collec-
tor section. The inner and outer sections of
the collector assembly were loaded with
various types of fiberglass filter media. The
device had both electrical and mechanical
forces acting to collect diesel soot.
The performance of these devices ex-
hibited the behavior typical of all fiber filters
tested: high mass removal efficiency
(60-70%) but a lifetime of at most a few
hours due to rising engine backpressure. The
electrified filter also experienced deteriora-
tion of electrification because of diesel soot
buildup in the electrified fiberglass filter. Two
general conclusions derive from a variety of
laboratory tests involving the aftertreatment
of diesel exhaust with fiber filters: (1) high
collection efficiencies and short operating
700
80
8
.g
.o
§
1
1
%
CD
**
O
20
Gravel Bed Outlet
l
0.04
0.10 0.20
Geometric Mean Diameter, fjm
0.40
4181-298
Figure 10. Mass removal efficiency in small size ranges of diesel particulate.
Table 3. Total Mass Loadings Measured for the ESP/Gravel Bed Combination
Sampling Point
Total
Mass Loading
mg/scm
Total
Removal Efficiency
Truck Outlet
ESP Outlet
Gravel Bed Outlet:
With stirring during
impactor runs
Without stirring
37.7
27.7
20.7
5.4
26
45
86
Table 4.
Mass Removal Efficiency of the ESP in Combination with a 20-cm Diameter
Gravel Bed Filter
Gravel Bed
Depth
cm
5
10
Backpressure
kPa fin. H2O)
10 (40) rising to
15 (60) in 3 hours
12 (481 rising to
16 (64) in 3 hours
Outlet Mass Loading
Truck Gravel Bed
mg/scm mg/scm
34.9 15.3
37.9 11.6
Efficiency
%
56
69
-------
periods will result from a dense cake of car-
bon soot forming on the surface of the filter
material and causing a rapidly increasing
pressure drop; and (2) exhaust gas leakage
around the edges of the filter material,
resulting from high pressure drop, will be a
problem in using any replaceable cartridge
filter.
Moving Belt ESP
A moving belt ESP designed for
automotive use was also tested. In this
device, the collection surface is a thin steel
belt which slowly moves so that the collected
material is removed from the area of high gas
flow to reduce reentrainment. The device
was tested in Japan using a 0.9 /jm mmd
soot produced by burning natural gas with
excess air. For a 2.7 nWmin gas flow rate
at 40°C, collection efficiencies of 60% with
a stationary belt and 82% with a moving belt
were measured. At the SoRI laboratory, the
moving belt ESP was tested on diesel ex-
haust which differed from the soot used in
the Japanese test in that the paniculate mmd
was 0.3 ^im at a flow rate of 3.2 nWmin. The
gas temperature was 113°C, resulting in a
drier, more easily reentrained soot. In addi-
tion, current leakage problems necessitated
reducing the operating voltage of the ESP
from 18 to 15 kV. Consequently, the collec-
tion efficiency was reduced to 9% with a sta-
tionary belt and 21 % with a moving belt. The
latter was unreliable: it repeatedly "walked"
off the end of the rollers and jammed the
mechanism. Tests at a lower flow rate (2.0
anWmin) were inconclusive due to conduc-
tive contamination of the insulators.
Conclusion
Use of a fine fiber filter to capture diesel
soot does not appear to be practical due to
its tendency to form a cake on the face of
the filter. This cake of particles reduces gas
flow through the filter which rapidly in-
creases the pressure drop across the device.
The pressure continues to increase until it
either impedes the operation of the engine
or establishes a lower pressure route around
the filter mat. Once a route for gas sneakage
is established, the collection efficiency of the
device drops sharply.
The three condensation traps tested were
unable to achieve acceptable collection ef-
ficiencies. In addition, there was no evidence
that these devices were agglomerating par-
ticulate. Instead, they produced a fog of
small particles as hydrocarbon vapors in the
diesel exhaust were condensed in the con-
trol device.
The granular bed filter may be applicable
to the diesel exhaust problem. Although its
collection efficiency was not as high as that
of a fiber filter, the gravel bed has the im-
portant advantage of being cleanable,
whether by mechanical agitation or wet
flushing. The ceramic monolith filter showed
a high collection efficiency but has a limited
lifetime without in situ regeneration. This
device is being developed by EPA at Ann
Arbor, Ml, and by several automobile
manufacturers to extend its usable lifetime
by direct or catalytic combustion of cakes
of collected particles.
Conventional ESP is effective in agglom-
erating diesel paniculate matter. An ESP can
be used with either cyclones or a granular
filter to obtain mass removal efficiencies of
50-85% at moderate engine backpressure
and for a practical operating time between
maintenance periods. Such a combination
device can be made sufficiently compact to
be used on stationary diesel engines or on
large, heavy duty diesel vehicles engaged in
fleet operation where periodic maintenance
can be performed at a central garage.
However, an ESP will demonstrate too low
a collection efficiency to be used alone.
Recommendations
Fiber filters do not appear suitable for the
capture of diesel soot. Fine fiber filters ex-
hibit high collection efficiency but have short
lifetimes due to filter blinding. The coarse
fiber filters tested, the traps, had inadequate
collection efficiency. The granular bed filter
proved to be usable, but needs to be
redesigned for adaptation to a vehicle and
for easy disposal of collected particles. The
ESPs tested had poor collection efficiencies
but were successful particle agglomerators.
An ESP in combination with a mechanical
collector, such as a cyclone or gravel bed,
may be usable, but further development is
needed to provide a device of practical
safety, size, and durability as well as to
develop the total system for electrode and
bed cleaning.
M. G. Faulkner, J. L DuBard, and J. R. McDonald are with Southern Research
Institute, Birmingham, AL 35255.
Dennis C. Drehmel is the EPA Project Officer (see below).
The complete report, entitled "Studies of Paniculate Removal from Diesel
Exhaust." (Order No. PB 84-168913; Cost: $13.00, subject to change) will be
available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
10
-------
=• (D
£ »
------- |