u-s- Environmental Protection Agency Industrial Environmental Research
Office of Research and Development Laboratory
Research Triangle Park, North Carolina 27711
EPA-600/7-77-095a
August 1977
EPA FABRIC FILTRATION STUDIES:
4. BAG AGING EFFECTS
Interagency
Energy-Environment
Research and Development
Program Report
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into seven series. These seven broad categories
were established to facilitate further development and application of environmental
technology. Elimination of traditional grouping was consciously planned to foster
technology transfer and a maximum interface in related fields. The seven series
are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
This report has been assigned to the INTERAGENCY ENERGY-ENVIRONMENT
RESEARCH AND DEVELOPMENT series. Reports in this series result from the effort
funded under the 17-agency Federal Energy/Environment Research and Development
Program. These studies relate to EPA's mission to protect the public health and welfare
from adverse effects of pollutants associated with energy systems. The goal of the
Program is to assure the rapid development of domestic energy supplies in an environ-
mentally-compatible manner by providing the necessary environmental data and
control technology. Investigations include analyses of the transport of energy-related
pollutants and their health and ecological effects; assessments of, and development
of, control technologies for energy systems; and integrated assessments of a wide
range of energy-related environmental issues.
REVIEW NOTICE
This report has been reviewed by the participating Federal Agencies, and approved
for publication. Approval does not signify that the contents necessarily reflect the
views and policies of the Government, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Information
Service, Springfield, Virginia 22161.
-------
-------
ABSTRACT
Fabric filter life can be divided into three periods: (1) break-in;
2) steady-state; and 3) wear-out. During the break-in period, bag
collection efficiency increases, but pressure drop across the bag also
increases. In the steady-state period, performance parameters remain
relatively constant until the fabric begins to fail catastrophically
(bag efficiency and pressure drop both decrease), signaling the onset of
the wear-out period.
The transition between the break-in and steady-state periods can be
interpreted in terms of dust buildup in or on the fabric, ending in a
steady state in which the dust arriving during a complete filtration
cycle equals that leaving during the same period. Dust permanently
added to the fabric filter during the break-in period is located at
"bulk" sites—sites which are not emptied during the shake-cleaning
step of the cycle. Dust can be emitted or lost from these sites, how-
ever, as is evident from baghouse operation with clean air only and from
performance measurements made after washing the fabric filter in soap
and water. The evidence presented here supports a further subdivision
of the dust trapped at bulk sites into: 1) that which is loosely bound
and which affects bag pressure drop primarily; and 2) that which is more
tightly held and which influences bag collection efficiency primarily.
This preliminary classification of dust trapped at bulk sites requires
more study, being based at present on only one observation of the dramatic
effect of hand-washing a dust-loaded bag in soap and water. Alternative
hypotheses could be equally plausible.
The total quantity of dust trapped at bulk sites varies with bag
operating conditions. The dependence of bag life upon the quantity of
trapped dust in the fabric is not known.
-------
PREFACE
This report is the fourth in a series of reports, entitled EPA
Fabric Filtration Studies, which summarize the results of EPA laboratory
testing of new baghouse fabric materials and present the conclusions of
specialized research studies in fabric filtration. These tests have
been carried out over the past 4 years by the Industrial Environmental
Research Laboratory, Research Triangle Park, N. C. and previously by
predecessor agencies. The purpose of these investigations was to evaluate
the potential of various new fabrics as baghouse filters and to obtain
data for use by the fabric filtration community. The testing consisted
of simulating a baghouse operation in a carefully controlled laboratory
setting that allowed measurement and comparison of bag performance and
endurance. The simulation discussed in this paper covered only a very
narrow range of operating conditions:
1) Redispersed, classified flyash (mass median diameter
about 4.0 vim) entrained in air was the only dust used.
2) All filtering was done at room temperature.
3) Relative humidity was either held at 50 percent or
randomly varied from about 30 to 70 percent.
4) The air to cloth ratio was fixed at 4:1.
5) The inlet dust loading was held in the vicinity of
3 grains/ft3 (6.9 g/m3)*.
*EPA policy is to use metric units only or to list both the common
British unit and its metric equivalent. For convenience and consistency
with existing baghouse conventions, British units are used in this
report. Readers more familiar with the metric system may use the
factors in the Appendix to convert to that system.
in
-------
Extreme caution should be used in extrapolating the results reported
here to the substantially different conditions that occur in all field
applications. The usefulness of the present results is primarily as
either an initial screen of candidate fabrics for baghouse applications
or as an exploration of baghouse phenomena.
The projected EPA Fabric Filtration Studies series consists of the
following reports:
1) Performance of Non-Woven Nylon Filter Bags [Ref. 1].
2) Performance of Non-Woven Polyester Filter Bags [Ref. 2].
3) Performance of Filter Bags Made from Expanded PTFE Laminate
[Ref. 3].
4) Bag Aging Effects (this report).
5) Bag Cleaning Technology.
6) Analysis of Collection Efficiency by Particle Size.
-------
CONTENTS
Abstract ......................... ii
Preface ......................... iii
Figures ......................... vi
List of Abbreviations and Symbols ............ vii
Section
1 Introduction .................. 1
2 Conclusions .................. 2
3 Experimental Procedures ....... ..... 4
4 Results ......... . ......... 7
5 Discussion ................. . . 31
6 References ................... 36
Appendix (Conversion Factors) .............. 38
-------
FIGURES
Number Page
1 Test apparatus 5
2 Performance data under accelerated life test: mass
collection efficiency and outlet concentration ... 8
3 Performance data under accelerated life test:
specific cake resistance and drags 9
4 Traces of pressure drop across the bag during
filtration ...... 10
5 Bag performance data under standard conditions: mass
collection efficiency and outlet concentration ... 12
6 Bag performance data under standard conditions:
specific cake resistance and drags ......... 13
7 Total number density of particles (0.3 urn and above)
in the outlet during filtration 18
8 Relationship between bag pressure drops and length
of shake cleaning period . 20
9 The influence of shake cleaning upon the number
density of outlet particles (0.3 ym and above) ... 21
10 Aging as reflected in the number density of outlet
particles (0.3 ym and above) 22
11 The effect of eliminating the dust feed upon the
number density of outlet particles .... 24
12 Total number density of outlet particles during
sequential cycles 1-4 . . ....... 26
13 Total number density of outlet particles during
sequential cycles 5-8, 12, and 17 27
14 Size distribution of particles at each of the first
6 minutes of the first cycle with no dust feed
(cycle 6, Table 1) ....... 29
15 Size distribution of particles in the first minute
of various cycles following the termination of dust
feed 30
-------
LIST OF ABBREVIATIONS AND SYMBOLS
filtration area of fabric (sq ft)
mass outlet concentration (grains/1000 cu ft)
mass collection efficiency (percent)
true value of specific cake resistance (in. hLO/fpm)/(lb/sq ft)
measured value of specific cake resistance (in. H-O/fpm)/
(lb/sq ft)
pressure drop across bag extrapolated to time zero of filtration
cycle (in. H20)
pressure drop across bag at end of filtration cycle (in. F^O)
effective drag (in. HLO/fpm)
terminal drag (in. hLO/fpm)
air/cloth ratio (fpm)
Vll
-------
SECTION 1
INTRODUCTION
The life of a fabric filter can be divided into three stages:
1) A break-in period in which its measures of performance are
continually changing. Typically collection efficiency
and pressure drop increase from cycle to cycle;
specific cake resistance does not generally change
much but the number of penetrating particles continually
decreases between successive cycles. Residual dust
cake in and on the fabric continually builds up.
2) A lengthy steady-state period characterized by stable
bag performance. Initial pressure drop following
each cleaning cycle remains essentially constant; bag
collection efficiency is uniformly high, and outlet
concentrations are low. Performance parameters cycle
between stable end points before and after cleaning.
3) The wear-out period in which tears and holes appear in
fabric because of physical or chemical wear. Bag
efficiency and pressure drop both decrease
irreversibly (although often erratically) unless repair
is possible and warranted. The onset of this period
signals the end of bag life; scheduling of bag
replacement is the usual response.
In this report the progressions of several woven polyester bags are
followed throughout their life cycles in an attempt to isolate those
variables that hasten, retard, or otherwise influence the bag aging
process.
-------
SECTION 2
CONCLUSIONS
Performance parameters, efficiency, outlet concentration, and drag,
measured during two life tests carried out on woven polyester fabric
filters, displayed the well known "bathtub" curve (or its inverse)
behavior as a function of time. Specific cake resistance was also
measured but showed no obvious time dependence. The two life tests
differed: one was an accelerated test consisting of a large number of
shakes per cycle; the second was a more standard cycle—the filtration
and the shake-cleaning times more closely corresponded to actual field
conditions. A full life cycle was completed only for the accelerated
life test; however, that portion of the life cycle completed under
standard conditions appeared to follow the same general pattern.
A major difference between new and used bags is that the latter
contains many trapped dust particles. The quantity of trapped dust
affects the performance parameters as evident from the differences in
the steady-state values between bags life-tested under accelerated and
standard conditions. The higher dust content of the latter resulted in
not only its higher dust filtration efficiency, but also its higher
pressure drop and drags. The effect of the higher dust content on bag
life was not determinable from these data, however.
What the data do suggest is that dust is trapped by a fabric filter
both at surface sites — the conventional dust cake—and at bulk sites.
The bulk sites are defined simply as those sites retaining dust after
the cleaning cycle; they are presumed to be located primarily within
the fabric, but also include the residual surface dust cake not removed
by the shake-cleaning. These conclusions are based on operation of
shaker-cleaned filter bags used in a single-bag pilot baghouse. Mass,
optical particle size, and pressure drop were the primary measurements.
-------
Little is known about the properties and consequences of the dust
trapped in the fabric. This dust can serve as a source of emitted dust,
as shown by the significant outlet concentration remaining even after
the inlet dust loading was dropped to zero. The dust trapped in bulk
sites, which contributes to a high pressure drop, may not be the same
trapped dust that enhances collection efficiency, as the hand-washing of
the bag suggested. This hypothesis is just one speculative explanation,
however, and is not yet a firmly established conclusion.
The intriguing possibilities uncovered by this work raise many
questions, the answers to which would constitute important pieces in the
fabric filtration puzzle.
-------
SECTION 3
EXPERIMENTAL PROCEDURES
As in previous investigations of this series [Refs. 1-3], the
reported experimental data originated from a laboratory baghouse con-
sisting of a single compartment, housing one bag (Figure 1). The bag
was cleaned by mechanical shaking of the bottom end, which was oscillated
through a 1.75 in. stroke at 4 cycles/sec.
The fabric filters used in these experiments were woven polyester
bags purchased from Lamports* to the following description:
Thread count: 76 x 65
Yarn denier: 250/250
Weave: 3 x 1 twill
Yarn type: Continuous filament
2
The fabric weight was approximately 5 oz/yd , and the filtering area of
2
the mounted bag was 8.5 ft . Flyash of mass median diameter equal to
4 um was the only test dust used. It was fed into the air stream by a
rotary screw mechanism at a constant rate of 3 grains/ft . The air/cloth
ratio was maintained at 4 fpm. Ambient temperature in the laboratory
was between 70 and 80°F, and the relative humidity in the air stream was
adjusted to 50 percent for certain life tests. For other life tests the
humidity was not controlled but varied with the ambient humidity. For
still other supplementing short tests the humidity was varied between 23
and 79 percent.
Testing took place over a 3 year period in which bags were operated
continuously for various lengths of time. During the life tests employing
The Lamports Company, 2303 Hamilton Avenue, Cleveland, OH 44114.
-------
96 In.
HUMIDITY CONTROL
CHAMBER
MECHANICAL
SHAKER
DISPERSION
VENTURi
OPTICAL
PARTICLE
ANALYZER
VARIABLE SPEED DUST
FEEDER ( FLYASH )
MILLIPORE
FILTER SAMPLING
TRAIN
CONTROL VALVE
ROTARY BLOWER
COLLECTION HOPPER
Figure 1. Test apparatus.
-------
the standard (non-accelerated) cycle, some interruptions occurred in
which the bag was removed from the test chamber and later remounted.
These interruptions did influence the measured performance parameters.
The accelerated life test was carried out without interruption over a 1
year period.
The measurements made routinely in evaluating the bag performance
were pressure drops across the bag, relative humidity in both the inlet
and outlet air flows, dust loading in both the inlet and outlet, and
total flow through the system. Measurement technique was the same as
previously reported [Refs. 1,2]:
1) All b;ag pressure drops were measured continuously using a
differential pressure cell.
2) Relative humidity was sampled by a wet-bulb/dry-bulb method.
3) Dust loading in the inlet was periodically sampled by
weighing the dust that was constantly and more or less uni-
formly added to the air flow by a rotary screw mechanism.
4) Dust- loading in the outlet was determined by isokinetically
*
sampling the flow continuously and weighing the particulate
cellected on a 0.45 ym Mi Hi pore* filter.
5) Size distribution of outlet dust was periodically
determined by analyzing the entrained dust with an optical
particle analyzer (the Climet 201)**. The sample flow for
this measurement was constant through the counter;
it was drawn isokinetically by adjusting the dimensions
of the sampling nozzle.
6) Total flow rate through the system was monitored
continuously by measuring the pressure drop across a flow
rate venturi.
*Millipore Corporation, Wiggins Avenue, Bedford, MA 01730.
**Climet Instruments, P. 0. Box 1165, Redlands, CA 92373.
-------
SECTION 4
RESULTS
The major results of the bag life tests are displayed in Figures 2
and 3. These data depict the behavior of the major performance para-
meters throughout the life cycle of a woven dacron bag:
1) An initial "break-in" period (from zero to about 15 to 20
million shakes) during which efficiency increases,
outlet concentration decreases, and drags increase.
2) A period of stabilized performance (15 to 20 million
shakes up to 45 to 50 million shakes) during which all
performance parameters are relatively constant (efficiency,
specific cake resistance, and drags actually decrease slightly
during this period; outlet concentration increases slightly).
3) The terminal period (from 45 to 50 million or more cycles)
in which performance parameters (except specific cake re-
sistance) all swing toward their initial values, completing
the bath tub curve or its inverse.
Specific cake resistance decreases very slightly throughout the bag
life as shown by the fitted curve in the top plot of Figure 3. The
p
correlation coefficient (r ) for this curve is on the order of 0.01,
meaning that the value of 1C is independent of the number of shakes.
The time required to establish a constant change of drag per incremental
dust cake loading changes with time as sketched in Figure 4.
All the curves traced in Figures 2 and 3 are least squares fits to
the data over all or a portion of the bag life. A better fit to the
efficiency and outlet concentration data results when the steady-state
period is divided in two, as shown by the dashed curves in Figure 2.
-------
DD
20 30 /]Q
NUMBER OF SHAKES (106)
50 60
70
Figure 2. Performance data under accelerated life test: mass
collection efficiency and outlet concentration.
-------
Vs1
— E
co u.
u
LU
Q_
LU
CO
oo
CD
a
U-
Li
,3
o o
OJ "i
.1
1.0
.9
,8
o o
o
O O
O Of, QO
0
Figure 3.
10
20 3D
NUFBER OF SHAKES
40
50
Performance data under accelerated life test:
cake resistance and drags.
60
specific
70
-------
a) at 0.5 x 10 shakes (break-in)
20 mln. c 0
b) at 36 x 10 shakes (steady-state)
20 mil).
c) at 58 x 10 shakes (wear-out)
Figure 4. Traces of pressure drop across the bag during filtration
10
-------
The data of Figures 2 and 3 are based on accelerated testing
(i.e., the test cycle used consisted of time increments differing from
those of the "standard" cycle):
Accelerated shake cycle 2 minutes 1 minute 15 minutes
Shake Cleaning
(4 shakes/sec)
Standard cycle 20 minutes 1 minute 2 minutes
At a shake frequency of 4 shakes/sec the accelerated shake cycle
consists of 3600 shakes/cycle rather than the 480 shakes of the standard
cycle. Thus the effects of mechanical shaking could be observed in much
shorter calendar times than if using the standard cycle.
All measurements made during accelerated testing, however, were
made using the standard cycle. The procedure was to shift to the standard
cycle and, after operating for several cycles with the standard cycle,
use the full 20 minute filtration portion of the cycle to collect a mass
sample for efficiency calculations.
While the modified cycle does have the merit of accelerating the
number of shakes accumulated per unit of test time, it departs from
realistic conditions in that the total dust passing through the fabric
and the average dust loading to which the fabric is exposed are lower
than normally encountered in a typical field installation. Not only are
the number of cleaning shakes greater during the accelerated tests but
the filtration time is a factor of 10 less than during the standard
cycle. The combination of accelerated cleaning and reduced dust filtration
could be expected to change the aging properties of the bag.
Performance measurements generated by a woven polyester bag operating
on the standard cycle are plotted as a function of time in Figures 5 and
6. The abscissa in these plots is number of cycles. Ten thousand
cycles represent 4.8 x 10 total shakes, a relatively small number on
the scales of Figures 2 and 3. Total operating time (2.5 standard
cycles per hour) was a little over 4567 hours; but, since the testing
11
-------
o o
10
CYCLE NUMBER (10 )
12
Figure 5.
Bag performance data under standard conditions: mass collection
efficiency and outlet concentration.
12
-------
±L
10
-CV1
8 8
s ^ ^^
O CQ D
O Z
O
LU
Q_
4
-
2.4
.2
i—
I 2.0
LU
00
1 '8
\\ .6
1 z .4
Q s '
LU ' tL
i i.o
u_
LU
.8
O
p 00° cP cP
o
_o
o o o
o o 0
o o o0 ° °
O o °O ° °
0 0
° o o
o
; A&T
D SE
"
-
^ A
AAA A Q AaQ
A & A
* %a ° a
^ Q A
-a a A
* ° A na a
n° a°
A Q
n A acPa D a
a D
i O i i i i
2 4 6 ? 3 10 1
CYCLE NLMBER (1(T)
Figure 6. Bag performance data under standard conditions: specific cake
resistance and drags.
-------
was not continuous, it took over a year to complete. Some of the dis-
continuities in the data plots result from interruptions in operating
time.
Four known or suspected sources of breaks in the data exist:
1) At about cycle 2000 a leak was discovered in the baghouse
which probably caused feed rate errors. Discontinuities in
the plots of E, CQ (Figure 5) and K^ (Figure 6) resulted from
this source. Drags, being only indirectly related to
dust feed rate, do not reflect a major discontinuity (Figure
6).
2) After cycle 3730 the pressure drop across the bag exceeded
10 in. of water, an unacceptably high value. The bag was
removed, hand-washed in soap and water, dried, and remounted.
A major discontinuity appears in the drag plot and anomalously
high readings, of K£ occurred temporarily. C and E do not
show significant breaks.
3) After cyel^e 8380, the bag was again removed for cleaning.
This time it was not washed but merely shaken by hand and then
remounted. A 2-1/2 month delay occurred before the
next test data. An obvious discontinuity exists in the drag
plots; less obvious discontinuities may exist for the other
variables.
4) After cycle 9981 a 4 month break in the test program took
place. The most noticeable effect of this interruption is
seen on the drag plot.
The increase in drags around cycle 6200 is anomalous. A shift in
drag comparable to those occurring at cycles 8380 and 9981 is evident at
cycle 6200 in Figure 6, but no ready explanation exists for it.
More data exist than are plotted in Figures 5 and 6. For clarity
these plots show only every fourth datum point up to cycle 5527.
Beyond this cycle the plots include every other datum point. The curves
shown in Figure 5, however, have been fitted by the least squares method
to all the data; clearly the omission of over half the data points from
the plots does not change the reasonable fit.
14
-------
Aside from the discontinuities evident in Figure 6 (attributable,
as discussed previously, to interruptions, errors, special procedures,
or unknown causes), standard-cycle bag performance parameters appear to
be following the same general aging pattern as delineated by the time
dependence of the accelerated-test-cycle performance parameters and as
plotted in Figures 2 and 3:
1) The efficiency curve (Figure 5) shows a sharp initial rise
and the C curve (Figure 5), a rapid initial decline—
both curves reaching an extremum at about cycle 4400
and subsequently undergoing a gradual reversal in values.
This behavior correlates with the first half of the data
plotted in Figure 2, the corresponding plots for accelerated
life tests (as shown by the dashed curves of Figure 2). In
Figure 2 the extrema occur at about 15 x 10 shakes with the
reversal extending to about 30 x 10 shakes. By comparing
Figures 2 and 5, the life cycle depicted in Figure 5 can be
estimated to be about half that of the full life cycle shown
in Figure 2. The influence of the larger average dust loading
characteristic of the standard cycle reflects itself in higher
efficiencies and lower outlet concentrations for that cycle
(Figure 5). The increased dust loading, in combination with
reduced shake cleaning, apparently produces an increased
residual dust cake in or on the fabric which not only improves
filtration efficiency, but also causes higher drags as discussed
next.
2) The drags measured under the standard cycle (Figure 6) also
show a rapid initial buildup like those observed with the
accelerated cycle (Figure 3). Specific cake resistance in
both cycles shows no marked trend (assuming the K£ data for
cycles less than cycle 2000 (Figure 6) are neglected because
of the feed rate errors discussed previously [see Item 1, p.
14]). The higher average dust load of the standard cycle (2
min dust feed/accelerated cycle vs. 20 min dust feed/standard
cycle) produced intolerably high pressure drops across the bag
15
-------
(the final pressure drop exceeded 10 in. of water) which led
to an unscheduled bag washing after cycle 3730. This washing
succeeded in reducing drags without penalizing bag efficiency--
a very desirable response, more fully discussed in Section 5.
Following the discontinuity at cycle 3730, attributable to bag
washing, drags increase again but level off below the high
values encountered just before the washing. Several smaller
discontinuities, attributable to test interruptions, are also
evident.
The effect of the higher average dust load/reduced shake-cleaning
is to cause higher values of both K^ and the two drags, as can be seen
by comparing the values plotted in Figures 3 and 6. The life cycle of
the bags readily apparent in the drag data of Figure 3, is only in-
completely revealed in Figure 6, being too brief in duration to show the
entire cycle and being obscured by various pronounced discontinuities.
Total dust load—the time-integrated dust feed--and shake-cleaning
together produce significant differences in aging characteristics of
performance parameters of these woven polyester fabric filters. A
"complete" life cycle is assumed to be that portrayed in Figures 2 and
3, based on data recorded during the accelerated life test. An incomplete
life cycle is represented by data plotted in Figures 5 and 6, based on
data taken during the standard operating cycle. By comparing the initial
rise time of the aging curves for efficiency and C^ (and consistent with
6
the other performance data as well) 15 x 10 shakes of the acccelerated
cycle can be equated to about 4700 cycles of the standard cycle (the
maximum value of collection efficiency in Figure 2 is at 15 x 10
shakes; that in Figure 5 is at 4700 cycles). This crude comparison
implies a life acceleration factor of about 6.5 for the accelerated test
cycle over the standard cycle (4700 cycles x 480 shakes/cycle = 2.3 x 10
shakes), only slightly less than the acceleration factor based solely on
the ratio of the number of shakes (3600/480 = 7.5). This reasonable
agreement implies that the accelerated test cycle is a valid method of
generating bag life data in spite of the low average dust loading.
16
-------
More compelling in suggesting that dust filtration time and cleaning
parameters play an important role in fabric filtration are their influence
upon the steady-state values of all performance parameters during the
period of stable bag performance (the flat portion of the curves in
Figures 2 and 5). Comparing Figures 2 and 3 with Figures 5 and 6 shows
that the accelerated test cycle in the steady-state period displays
lower efficiency, drag, and specific cake resistance, but higher outlet
concentration. These observations imply that more dust clings to the
surface of the fabric or is imbedded within it during the standard cycle
than during the accelerated test cycle--even after the steady-state
period has been reached.
To further investigate the presence of this dust and its location
on or within the fabric, additional data were gathered, using the optical
particle counter. All of these data were taken with the basic time
sequences of the standard cycle rather than those of the accelerated
test cycle. Often, however, some modification of one or more of the
time sequences was used.
As discussed previously the standard cycle is:
1 minute
20 minutes
2 minutes
The optical counter measures the size distribution of particles in
the outlet during the 20 minute filtration cycle. Figure 7 is a typical
plot of the optically detected particles in the outlet, using a woven
polyester bag to filter flyash. This dust/fabric system displays the
commonly observed dust cake filtration properties [Ref. 4]. At the
beginning of the filtration period, particle penetration is high (the
peak occurring at 2 minutes in Figure 7 may be an artifact of the
17
-------
CO
+J
q-
o
I—I
I—
cz
o23% RELATIVE HUMIDITY
038% RELATIVE HUMIDITY
o54% RELATIVE HUMIDITY
^70% RELATIVE HUMIDITY
(These curves represent values
averaged over 3 standard cycles)
o
24 6 8 10 12 14 16
TIME FROM START OF FILTRATION PERIOD (MIN)
Figure 7. Total number density of particles (0.3 pm and above) in the
outlet during filtration.
18
-------
measuring technique—the blower starts at time zero of the filtration
period so that any delay in particle penetration and transport through
the baghouse and its feeder lines (Figure 1) could manifest itself a~
lower particle particle penetration during the first samplings of the
filtration period). As the dust cake builds up during the filtration
period, particle penetration decreases rapidly because of particle
capture by the dust cake. The added pressure drop across the dust
cake/fabric filter composite necessitates a shake-cleaning during which
much of the dust cake is removed. When the next filtration cycle beginss
pressure drop across the bag is low but the particle penetration is
again high. As the dust cake rebuilds, these parameters change as
before—pressure drop increases and particle penetration decreases--
until the cycle must be repeated again, as dictated by a predetermined
maximum pressure drop in a field installation. In the standard cycle
employed in the work reported here, the duration of the filtration cycle
is held fixed at a constant 20 minutes rather than being determined by
the pressure drop.
The data in Figure 7 reflect a humidity dependence often seen in
the fabric filtration of flyash—a reduction in mass penetration (C )
with increasing relative humidity [Ref. 5], Figure 7 shows that the
number density of the penetrating flyash particles (0.3 urn and above)
also follows this pattern. The optical counter data to be presented in
this paper are low humidity data (23 percent relative humidity) simply
because these number densities are the largest. Similar behavior occurs
at other relative humidities: relative humidity, while obviously
influencing particle penetration, is assumed not to alter the basic
interactions in any major role.
The first modification of the standard cycle investigated was a
reduction in the shake-cleaning time and hence the total number of
shakes. Figure 8 shows the effect of shortening the cleaning cycle on
bag pressure drop. The pressure drops vary inversely with the log of
the total number of shakes in the cycle. Figure 9 shows that the total
number of particles measured by the optical detector decreases with
decreasing number of shakes and, more importantly, that the time
required to re-establish high efficiency filtering also decreases with
19
-------
5 rA,
O-
<
20
Figure 8.
50 100
NUMBER OF SHAKES/CYCLE
200
500
Relationship between bag pressure drops and length of shake-
cleaning period.
decreasing number of shakes. Not surprisingly, the net effect of reducing
the number of cleaning shakes per cycle is to leave a dirtier bag in the
sense that more dust remains in or on the bag as detected by the higher
pressure drops and the reduced particle penetration during the filtration
cycle. The assumption is that both these changes are attributable to
less complete removal of the dust from the bag.
These same changes in performance take place during the initial
break-in period of a bag. Pressure drops increase with succeeding
cycles, as plotted in Figures 3 and 6, and particle penetration through
the bag decreases, as shown by the C plots of Figures 2 and 5. The
optical counter data also detect less particle penetration after, say 54
cycles, than after 5 cycles (Figure 10). The conclusion is that, during
the initial break-in period, the bag becomes dirtier by receiving and
retaining more particles than it releases—the bag acts as a particle
sink. Figure 10 illustrates the conclusion well known to baghouse
operators that the bag accumulates a residual dust holding during the
20
-------
24
22
20
18
16
12
10
o
O,
O
OJI2 SHAKES/CYCLE
O 40 SHAKES/CYCLE
A 16 SHAKES/CYCLE
\
-
O
^
\
V \
\ \
O
o.
A
O
O
\
o
o.
\.
"o
__»
\.
CX
O.
o.
A
ko
k —A —A
- i - -
TIME FROM START OF FILTRATION PERIOD
Figure 9. The influence of shake cleaning upon the number density
of outlet particles (0.3 urn and above).
21
-------
72.1
20
18
16
\ \
\. °
\
\ \
A C
A
A
O
\O
i:
•.
O L± O
o
o
\
^ C
\ \
\ \
i *
O CYCLE NO. 5
CYCLE NO. 54
I 1 I 1 I I I
2 4 6 8 10 12 , 14 16 18
TIME FROM START OF FILTRATION PERIOD WIN)
Figure 10. Aging as reflected in the number density of outlet
particles (0.3 ym and above).
22
-------
initial bag break-in period. It acts as a sink by retaining more particles
than it releases until reaching a steady-state condition.
During its stable period, the bag presumably operates in a steady
state whereby the total number of particles received per cycle equals
the total number lost. In the steady-state period, however, the bag is
more than just a fabric; it is a fabric loaded or saturated with dust
particles. Moreover, these added dust particles play a useful role as
evidenced by the improved filtration efficiency of the bag in steady
state and the reduced particle penetration through it.
Questions now arise regarding these dust particles acquired by the
fabric filter during its break-in period:
1) Where are they?
2) How are they held?
3) In what kinetic processes do they participate?
4) What are their population densities?
5) What influences their numbers, sizes, and other properties?
The results already reported in this paper provide some answers.
In addition, further procedural modifications were undertaken to learn
more about their properties.
One modification consisted of stopping the dust feed partway through
the filtration cycle, but continuing to monitor the size distribution of
dust in the outlet (with the optical counter). The classical particle
capture processes (interception, inertial impaction, diffusion [Ref. 6])
all assume straight-through particle penetration—the particle is either
caught or it isn't; no allowance is made for multi-stepped delayed
penetration. If these classical mechanisms are all that influence
penetration, then the dust concentration in the outlet should drop to
zero as soon as the inlet dust loading is shifted to zero. What actually
happens is shown in Figure 11--the dust loading drops off gradually
after the dust feed is stopped. The data presented in Figure 11 compare
the dust in the outlet (measured during the fourth standard measuring
cycle at 23 percent relative humidity) with that measured during the
first measuring cycle at that humidity (all bags were stablized at a
given humidity by running the standard cycles for at least 24 hours
prior to making any measurements; the fourth standard measuring cycle is
23
-------
UJ
CJ
1 CYCLE (STANDARD)
th
4cn CYCLE (FEED OFF a 3 HIM,)
23% RELATIVE HUMIDITY
2 -
0
16 18
2 4 6 8 10 12 14
TIME FROM START OF FILTRATION PERIOD (MIN)
Figure 11. The effect of eliminating the dust feed upon the number density of outlet particles.
-------
the fourth cycle of a test sequence). During the 20 minute filtration
period of the fourth cycle, the dust feed was stopped 8 minutes into the
period. The outlet concentration varied as plotted in Figure 11 for the
three most populous dust size-groups. Some drop-off in dust concentration
is evident after the dust feed stops (fourth cycle), but the time dependence
of the outlet dust concentration is not markedly different during the
first and fourth cycles.
After completion of the fourth cycle, the dust feed was returned to
normal for the fifth cycle, but was turned off afterwards and left off
for all succeeding cycles. Total (the sum of all size groupings) particle
concentrations as measured by the optical detector are plotted in Figures
12 and 13. Table 1 summarizes the conditions under which the filtration
step of each cycle was conducted. Curves for cycles 1, 2, 3 and the
first 8 minutes of cycle 4 (Figure 12), illustrate experimental variation
typical of the optical counter data. These curves are taken under
identical conditions but show particle penetration profiles that vary as
illustrated.
Table 1. DUST PENETRATION SERIES, WOVEN POLYESTER BAG
(LAMPORTS STYLE NO. 4027-SHT) AT 23% RELATIVE HUMIDITY
Cycle No. Conditions
1
2
3
4
5
6
7
8
12
17
Standard*
M
II
" (dust feed OFF at i
II
11 (dust feed OFF)
M
II
II
II
3 minutes)
*Air/cloth ratio = 4 fpm; dust feed = 3 gr/ft3).
25
-------
24 -
O CYCLE NO. 1, Dust feed ON
D CYCLE NO. 2, Dust feed ON
O CYCLE NO. 3, Dust feed ON
• CYCLE NO. 4, Dust feed OFF
at 8 minutes
Figure 12.
46 8 10 12 14 16
TIME FROM START OF FILTRATION PRIOD (MIN)
Total number density of outlet particles (0.3 urn and above) during
sequential cycles 1-4.
26
-------
CsL
O
O
24
22
20
18
16
14
12
~ 10
o CYCLE NO. 5, Dust feed ON
o CYCLE NO. 6, Dust feed OFF
a CYCLE NO. 7, Dust feed OFF
A CYCLE NO. 8, Dust feed OFF
CYCLE NO.'
Dust feed OFF
CYCLE NO.17, Dust feed OFF
Figure 13.
16
18
4 6 8 10 12 14
TIME FROM START OF FILTRATION PERIOD (MIN)
Total number density of outlet particles (0.3 ym and above)
during sequential cycles 5-8, 12, and 17.
27
-------
Clearly dust concentrations, significantly above background levels,
continue to exist in the outlet of the bag, even though the inlet dust
feed has been stopped. As in standard cycles, dust concentrations are
greatest at the start of each new cycle and subsequently decline to low
values before the cleaning step. At the start of the next filtration
period, however, dust concentrations have built back up to much higher
values than at the termination of the previous filtration period. This
recovery occurs even though no dust feed has taken place during the
cycle.
This same behavior is presented somewhat differently in Figures 14
and 15. Figure 14 is simply a bar graph plot showing the outlet dust
concentrations during each of the first 6 minutes of the sixth cycle.
This cycle is the first cycle having no dust feed at all (Table 1). The
data labelled 1 minute are compared with the 1 minute data from subsequent
cycles in Figure 15. Dust concentrations are decreasing as the bag
cycles through the sequences with no dust feed. Note, however, that the
minute 1 data of cycle 7 (Figure 15) show much greater dust concentrations
than the minute 6 data of cycle 6 (Figure 14). So do the minute 1 data
of cycles 8, 12, and 17 (Figure 15).
The implication is that a dust source remains in the baghouse even
after the source of external dust feed is removed. Furthermore the
hypothesis is that this dust source is the dust-loaded bag itself which
emits a continually declining quantity of the dust trapped within it to
each succeeding period of air flow. Bag flexure during the stopping and
starting of the air flow probably helps loosen some of the trapped dust.
The optical counter data presented in Figures 11 through 15 have
been taken with air flow at 23 percent relative humidity. At higher
relative humidity--54 to 78 percent—the dust concentrations are not as
great but they exhibit the same general behavior. Relative populations
in the size classes may shift with changing relative humidity, however.
28
-------
10
ro
o
O
uu
_i
o
Size Ranges (ynr)
0.3-0.5
t^MI 0.5-1.0
I I 1.0-2.0
Figure 14.
23456
TIME FROM^START OF FILTRATION PERIOD (MIN)
Size distribution of particles at each of the first 6 minutes of the first
cycle with no dust feed (cycle 6, Table 1).
-------
CO
+->
M-
to
O
UJ
o
o
C_J
o
i—i
fe
e£.
O.
Cycle
Cycle
Cycle
Cycle
Cycle
6 =
7 =
8 =
12 =
17 =
1st cycle after dust feed OFF
2nd cycle after dust feed OFF
3rd cycle after dust feed OFF
7th cycle after dust feed OFF
12th cycle after dust feed OFF
Size Ranges (ym)
0.3-0.5
Cycle 6 Cycle 7 Cycle 8 Cycle 12
Figure 15. Size distribution of particles in the first minute of various
following the termination of dust feed.
Cycle 17
cycles
-------
SECTION 5
DISCUSSION
A used fabric filter is not quite the same as a new fabric filter.
The used fabric filter is a more efficient filter but is also less
permeable to air flow—it presents a higher impedance to air flow than
the new fabric filter.
One obvious explanation—an explanation consistent with the observa-
tions reported here—is that the used filter is a dust-saturated, or at
least a dust-loaded, fabric. While the fiber density and number do not
change from those of the new filter, the used filter contains added dust
particles occupying either (or both) surface sites or bulk positions
within the fabric that are not occupied in the new filter. As reported
elsewhere, the added dust does not directly alter the mechanical strength
of the fabric [Ref. 7], although certain dusts can hasten wear because
of their abrasive properties. Gas permeability of a used filter, however,
is reduced because the dust fills otherwise unoccupied voids.
The interaction between a dust and a fabric filter depend on the
properties of both the dust and the fabric. The following discussion
interprets primarily the observations reported in this paper; certain of
the conclusions could be restricted to just the flyash/polyester system
investigated here.
The picture of the dust/fabric interaction developed in the course
of these investigations is one in which dust penetrates the fabric
filter by both straight-through (no time delay) passages and delayed multi-
stepped paths, perhaps analogous to the seepage and pinhole plug mechanisms
identified by Leith et al. in studying mass transport through dust
cake/fabric composite structures [Ref.8]. The only distinction justified
by the observations reported here is that between immediate penetration
31
-------
(no time delay) and delayed penetration. Presumably the former
corresponds to a straight-through trajectory whereby the particle is
not captured by any fiber of either the fabric or the dust cake. The
delayed penetration is that component of penetration that "works" its
way through the dust cake and fabric to emerge on the clean air side of
the filter at a later time. Germane to the concept of a delayed component
of particle penetration is the assumption of a multistep trajectory
whereby a particle moves through the filter in a series of short hops
from one metastable site to another. The concept implies a sequence of
capture and release processes governing the transport of dust particles
through the filter.
The previous discussion suggests that dust particles can be trapped
or retained by a fabric filter at one of two general sites:
1) a surface site, or
2) an interior bulk site.
Surface sites make up the dust cake. Interior bulk sites extend from
one surface of the fabric through to the opposite surface and consist of
metastable dust-particle-trapping sites which fill when exposed to dust
flow. The filling of the bulk sites takes place during the bag break-
in period. Saturation of these sites or attainment of their maximum
steady-state values of occupancy marks the beginning of the bag's steady-
state period.
Surface sites make up that portion of the dust cake which is removed
by the shake-cleaning of each cycle. Bulk sites are those occupied by
the dust not removed by the shake-cleaning. Some bulk sites may actually
be in the vicinity of the fabric surface but, because of their relative
permanence, they are called bulk sites by the preceding definition.
The influence of surface sites (dust cake) upon bag performance
parameters is well known [Refs.5,6]. Less understood is the role of the
bulk sites. The performance data gathered here show that their influence
can be substantial, too. Intracycle variations of performance parameters
are attributable to dust cake; similarly, at least some of the long term
32
-------
behavior of these parameters is attributable to dust in bulk sites.
The break-in period has been described as the time required to reach a
steady state in which the dust being trapped at bulk sites just equals
the dust being released. For operation at high dust loadings and long
filtration periods, the steady state requires a higher occupancy of bulk
sites. Hence, the dust trapped in the fabric filter is greater when
operating with the standard cycle (Figure 5) than with the accelerated
cycle (Figure 2), as reflected in the efficiencies and outlet con-
centrations of the two cycles as well as in the drags and specific cake
resistance (Figures 3 and 6).
The hypothesis being promulgated here is that one consequence of
the dust trapped in the bulk sites of the fabric is the irreversible
alteration of the performance parameters of the filter (with respect to
the standard shake-cleaning). These changes need not be irreversible to
all cleaning methods: witness the dramatic effect of the hand-washing
with soap carried out on the bag tested under the normal cycle (Figures
5, 6). The hand-washing apparently succeeded in removing significant
dust from the bulk sites of the bag, causing a dramatic drop in the
drags. The fact that this washing step did not cause a corresponding
decrease in the filter efficiency (or, similarly, an increase in the
outlet concentration) allows the additional postulation that the bulk
sites are subdivided into:
1) Sites which influence efficiency and which are sufficiently
tightly bound to remain occupied throughout the wash cycle.
2) Sites which influence pressure drop and which are removed
in the washing step.
Dust bridge orientation offers one possible explanation to make
such a subdivision of bulk sites seem plausible. This model assumes
that the trapped dust which offers the highest resistance to air flow is
itself most susceptible to removal by even higher drag forces such as a
water wash. The following paragraphs amplify this explanation.
From prior EPA work [Ref. 9] and elsewhere [Ref. 10] it is known
that particle deposition on a fiber can occur in chains or filaments,
rather than as a uniform coat--the first particle deposited on a clean
33
-------
fiber adheres at a random site, but the second, third, and all subse-
quent particles tend to adhere to previously deposited particles rather
than to another site on the clean fiber. Consequently the initial
growth of dust cake is in the form of long thin chains or filaments of
deposited particles. These growths extend outward from the surface of
the initial deposition site into the gas flow until the aerodynamic
forces break them apart or until they reach some adjacent point of
support, such as another fiber or another growing particle chain. Once
a particle chain establishes a stable anchor at its growing end, it has
bridged a void and itself becomes a semipermanent surface capable of
particle capture in directions normal to its axis of growth.
In the physical model suggested here, the orientation of these dust
bridges with respect to the gas flow determines whether they will have a
greater or lesser influence on pressure drop. If the dust bridges are
regarded as primarily one-dimensional filaments, those that are aligned
parallel to the gas flow have negligible influence on pressure drop
(pressure drop is assumed to be proportional to surface area perpendicular
to the gas flow across any plane of the filter); those aligned perpendicular
to the gas flow present maximum impedance to gas flow. Most dust bridges
will probably be oriented somewhere between these two extremes, having
an orientation that is partly longitudinal and partly lateral with
respect to the gas flow. The lateral component is that which both makes
the largest contribution to pressure drop because of air drag, and is
most vulnerable to removal by the even higher drag forces associated
with a liquid flow; i.e., a wash. Thus, the effect of the wash cycle is
to increase the ratio of longitudinally oriented dust filaments to those
laterally oriented. Efficiency of dust removal depends on both orientations,
although most likely the dependence varies with particle size. Capture
by diffusion is favored by a longitudinal orientation; lateral orientation
favors interception, inertia! impaction, and sieving. Removal of lateral
sites only would not necessarily reduce efficiency if most collection is
due to diffusion.
The dust cake on the surface of the fabric favors the formation of
lateral dust bridges because it lacks the depth of the fabric. The washing
34
-------
action may just be a more effective removal of residual surface dust
cake, characterized by a high ratio of lateral to longitudinal dust
bridges,while the dust bridges formed within the three-dimensional
fabric, being more longitudinally oriented, remain and preserve a high
collection efficiency.
These thoughts are speculative; no further information clarifying
the properties or location of these two types of bulk sites emerges from
the data. Their existence is postulated to explain only one observation-
the effect of washing. Additional experimental support for the concept
seems warranted (or, alternatively, other explanations of the washing
action) before accepting their existence.
A possible alternative explanation could be found by postulating a
suitable electrostatic interaction attributable to the washing. Under
the right conditions, electrostatic effects have been shown to reduce
pressure drops while maintaining or improving filtration efficiency
[Ref. 11]. This effect is exactly what the hand-washing of the bag
brought about. The highly desirable consequences of the washing make
further investigation and understanding of the interaction important
follow-up work.
The final period of the life cycle of a fabric filter is the
wear-out period. In this period the physical integrity of the bag is
failing and enough leakage paths exist to permit significant dirty air
flow through holes, voids, or non-homogeneous regions of the fabric
filters with significantly different filtration properties. The drags
decrease because of these "holes" and the outlet concentrations increase.
The holes can be covered over by the dust cake, but the time required to
do so becomes an increasingly large fraction of the total filtration
period (Figure 4).
The wear-out is not necessarily abrupt and, if adequately monitored,
provides ample warning of the coming deterioration. At the time the
life cycle depicted in Figures 2 and 3 was declared complete and the bag
classified as worn out, bag performance was still acceptable in an
absolute sense. Even at the end of its life here, the bag's filtration
efficiency was greater than 99 percent—testimony to the performance
standards now routinely expected of fabric filters.
35
-------
SECTION 6
REFERENCES
[1] Turner, J. H., "EPA Fabric Filtration Studies: 1. Performance of
Non-Woven Nylon Filter Bags," EPA-600/2-76-168a, December 1976,
(NTIS No. PB 266271/AS), EPA, Industrial Environmental Research
Laboratory, Research Triangle Park, N. C. 27711.
[2] Ramsey, G. H., R. P- Donovan, B. E. Daniel, and J. H. Turner, "EPA
Fabric Filtration Studies: 2. Performance of Non-Woven Polyester
Filter Bags," EPA-600/2-76-168b, June 1976, (NTIS No. PB 258025/AS),
EPA, Industrial Environmental Research Laboratory, Research Triangle
Park, N. C. 27711.
[3] Donovan, R. P., B. E. Daniel, and J. H. Turner, "EPA Fabric Filtration
Studies: 3. Performance of Filter Bags Made from Expanded PTFE
Laminate," EPA-600/2-76-168c, December 1976, (NTIS No. PB 263132/AS),
EPA, Industrial Environmental Research Laboratory, Research Triangle
Park, N. C. 27711.
[4] Dennis, R., "Collection Efficiency as a Function of Particle Size,
Shape and Density: Theory and Experience," J. Air Poll. Control
Assoc. 24-, Dec. 1974, pp. 1156-1163.
[5] Durham, J. F. and R. E. Harrington, "Influence of Relative Humidity
on Filtration Resistance and Efficiency of Fabric Dust Filters,"
Filtration and Separation, July/August 1971, pp. 389-398.
[6] Bergmann, L., "New Fabrics and Their Potential Application," J.
Air Poll. Control Assoc. 2^, Dec. 1974, pp. 1187-1192.
[7] Cass, R. W. and R. M. Bradway, "Fractional Efficiency of a Utility
Boiler Baghouse—Sunbury Steam-Electric Station," EPA-600/2-76-
077a, March 1976, (NTIS No. PB 253943/AS), GCA/Technology Div.,
Bedford, Mass. 01730.
[8] Leith, D., S. N. Rudnick, and M. W. First, "High-Velocity, High-
Efficiency Aerosol Filtration," EPA-600/2-76-020, January 1976,
(NTIS No. PB 249457/AS), Harvard School of Public Health, 665
Huntington Ave., Boston, Mass. 02115.
36
-------
[9] Draemel, D. C., "Relationship Between Fabric Structure and Filtration
Performance in Dust Filtration" EPA-R2-73-288, July 1973, (NTIS No.
PB 222237), EPA, Control Systems Laboratory, Research Triangle
Park, N. C. 27711.
[10] Billings, C. E. and J. Wilder, "Handbook of Fabric Filter Technology,
Vol. 1, Fabric Filter Systems Study" EPA No. APTD 0690, December
1970, (NTIS No. PB 200648), GCA Technology Division, Bedford,
Mass. 01730.
[11] Penney, G., "Using Electrostatic Forces to Reduce Pressure Drop in
Fabric Filters," Presentation to the 8th Annual Fine Particle
Society Conference, Chicago, 111., August 1976.
37
-------
APPENDIX
CONVERSION FACTORS
To Convert From:
foot?
yard
3
grains/foot ^
grains/1000 ft
Ib (force)
foot
inch
mil
yard
grain
Ib (mass)
inch of2water (60°F)
lb/inch? (psi)
Ib/foot
foot/mi n (fpm)
foot3
inch,
yard
oz/yd
To
r
meter'
meter:
meter^
3
Multiply By:
newton
meter
meter
meter
meter
kilogram
kilogram
2
newton/meter^
newton/meter?
newton/meter
meter/sec
meter.,
meter.,
meter
kg/m3
9.29 x
6.45 x
8.36 x
2.29 x
2.29 x
4.49
3.05 x
2.54 x
2.54 x
9.14 x
6.48 x
4.54 x
2.49 x
6.89 x
4.79 x
5.08 x
2.83 x
1.64 x
7.65 x
3.39 x
oi/ — 3
K- ~ TT
10IJ
10-1
10 '
">:!
10 6
"1
10-5
10 J
10 '
^
10 '
^2o
103
101
TO'3
lo:|
10.1
10 '
io-2
(°F -
38
-------
TECHNICAL REPORT DATA .
(Please read Instructions on the reverse before completing)
, REPORT NO.
EPA-6QO/7-77-Q95a
2.
I. RECIPIENTS ACCESSION-NO.
4. TITLE AND SUBTITLE
EPA Fabric Filtration Studies: 4.
Bag; Aging Effects
5. REPORT DATE
August 1977
6. PERFORMING ORGANIZATION CODE
7.AUTHOR
------- |