EPA-650/2-75-002
JANUARY 1975
Environmental Protection Technology Series
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environ-
mental Protection Agency, have been grouped into series. These broad
categories were established to facilitate further development and applica-
tion of environmental technology. Elimination of traditional grouping was
consciously planned to foster technology transfer and maximum interface
in related fields. These 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
9. MISCELLANEOUS
This report has been assigned to the ENVIRONMENTAL PROTECTION
TECHNOLOGY series. This series describes research performed to
develop and demonstrate instrumentation, equipment and methodology
to repair or prevent environmental degradation from point and non-
point sources of pollution. This work provides the new or improved
technology required for the control and treatment of pollution sources
to meet environmental quality standards.
-------
EPA-650/2-75-002
INFLUENCE OF FIBER CHARACTERISTICS
ON PARTICULATE FILTRATION
by
B. Miller, G.E.R. Lamb, and P. Costanza
Textile Research Institute
P.O. Box 625
Princeton, N. J. 08540
Grant No. R-800042
ROAP No. 21ADL-022
Program Element No. 1AB012
EPA Project Officer: J.H.Turner
Control Systems Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
January 1975
-------
EPA REVIEW NOTICE
This report has been reviewed by the National Environmental Research
Center - Research Triangle Park, Office of Research and Development,
EPA, and approved for publication. Approval does not signify that the
contents necessarily reflect the views and policies of the Environmental
Protection Agency, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use.
This document is available to the public for sale through the National
Technical Information Service, Springfield, Virginia 22161.
ii
-------
CONTENTS
Page
List of Figures iv
List of Tables vi
Acknowledgments vii
Sections
I Conclusions • 1
II Recommendations 2
III Introduction 3
IV Apparatus 7
V Filtration Parameters 9
VI Fabric Filter Formation 11
VII Experimental Results 18
VIII EDP Scanning Microscope • 33
IX Epitropic Fibers 35
X References 42
XI Glossary 44
XII Nomenclature 46
iii
-------
FIGURES
No. Page
1A Apparatus for measuring filter performance 8
B Humidity control system 8
2 Scanning electron micrographs of selected fibers
used in the main experiment 13
A. Round, smooth (0.1% TiO2) (3000X)
B. Round, rough (2.0% Ti02) (3000X)
C. Trilobal, smooth (0*1% Ti02) (1000X)
D. Trilobal, rough (2.0% Ti02) (1000X)
3 Relationship between air permeability and fabric
density 14
4 Relationship between air permeability and latex
content 14
5 Half-normal plot for E(l) 20
6 Half-normal plot for E(10) 20
7 Half-normal plot for APe/V 20
8 Half-normal plot for APf/V 21
9 Half-normal plot for K ' 21
10 Half-normal plot for C (10) 21
11 Half-normal plot for E(2.5y) 21
12 Three-dimensional plots of the effect of crimp
x length on E(10) and CQ(10) 25
13 Three-dimensional plot of the effect of shape
x length on E(10) 25
14 Three-dimensional plot of the effect of linear
density x shape on E(10) 25
15 Efficiency distributions for samples 1 A-D 27
16 Efficiency distributions for samples 2 A-D 27
17 Efficiency distributions for samples 3 A-D 27
18 Efficiency distributions for samples 4 A-D 27
iv
-------
FIGURES (continued)
No. Page
19 Efficiency distributions for samples 5 A-D 28
20 Efficiency distributions for samples 6 A-D 28
21 Efficiency distributions for samples 7 A-D 28
22 Efficiency distributions for samples 8 A-D 28
23 Scanning electron micrographs of filters 4C and
6A (150X) 30
24 Optical density contour maps of filter samples 34
A. Low density sample #4C (5.5X)
B. Low density sample #4A (5.5X)
C. High density sample #4C (5.5X)
D. High density sample #4A (5.5X)
25 Micrograph of epitropic fiber surface (from
Ellis, V. S., Reference 13) 36
26 Micrograph of epitropic fiber cross section
(from Ellis, V. S., Reference 13) 36
27 Efficiency distribution curves for 100%
polyester and for 50% epitropic/50% polyester
filter samples 38
28 Diagram of filtration apparatus modification
for electrification trials 39
29 Effect on E(l) of high voltage applied to
100% polyester and to 50% epitropic/50%
polyester filters 39
30 Effect pn E(l) of high voltage applied to 100%
polyester (needled) with and without grounding
center of sample 41
-------
TABLES
No. Page
1 Filtration Parameters 10
2 Description of Experimental Fiber Samples 12
3 Fabric Properties of Main Experiment Samples 16
4 Measurements of Filtration Performance Responses
of Main Experiment Samples 17
5 Summary of Yates Analysis at 95% Confidence 23
6 Physical Properties of Epitropic/Polyester Filters 38
VI
-------
ACKNOWLEDGEMENTS
The authors wish to express their appreciation to Dr. J. H.
Turner of the Environmental Protection Agency for his advice
and encouragement throughout the period of this research.
The authors also wish to acknowledge the assistance pro-
vided by Professor John C. Whitwell and Dr. Charles J.
Shimalla with the statistical analyses.
VI1
-------
SECTION I
CONCLUSIONS
Of the five fiber parameters whose effects on filtration per-
formance have been studied, three have been shown to have
significant effects. Efficiencies are shown to be improved
by the use of trilobal rather than round cross-section fibers,
3-denier rather than 6-denier fibers, and crimped rather than
uncrimped fibers. Pressure drops are also improved by the
use of crimped fibers. The improvement in efficiency found
with low linear density fibers is obtained at the cost of a
greater pressure drop. Surface roughness appears to have no
effect at the levels studied. A nonstatistical examination
of the results, however, seems to indicate that rough fibers
are more efficient in removing the smallest particles. At
the 90% confidence level, longer staple fibers give improved
efficiency.
Significant interactions occurred between crimp and length,
shape and length, and linear density and shape. These are
difficult to interpret but are tentatively attributed to
fabric formation effects. Confirmation of this assumption
would require measurements of density fluctuations and pore
size distributions in webs made from the different fibers.
Particle size analysis of the dust passing through the
filters has shown that improvements in overall efficiency
are accompanied by even greater improvements in the effi-
ciency at the small particle end of the distribution curve.
It follows that studies of these geometric parameters may
be useful in developing filters with improved capacity for
removing very small particles.
Experiments with filters made from epitropic fibers did not
show significant improvements in efficiency resulting from
the greater surface roughness of these fibers. It is not
established whether the conductivity of these fibers was
active in opposing the effect of roughness.
Application of high D. C. voltages to the filter fabrics
show important improvements in efficiency with filters made
of 100% non-conducting poly(ethylene terephthalate) (PET)
fibers. With PET filters incorporating 50% epitropic
fibers, the improvement was much smaller.
-------
SECTION II
RECOMMENDATIONS
The results of this study suggest that further work could be
profitably undertaken in the following areas:
1. Extension of the measurements of filtration performance
over wider ranges of the fiber parameters. Only two
levels of each parameter were examined in this study.
While these levels were chosen to be within a practical
working range for many textile fabrics, the variables
were far from the limit of capabilities of fiber tech-
nology. In order to obtain optimum performance, it
would clearly be useful to have information on the
dependence of filtration responses over a wide range of
fiber parameters.
2. Changes in fiber geometry appear to affect filtration
performance not only through their own interaction with
the dust-laden air stream but also through changes in
the structure of the nonwoven filter. A study of
density fluctuations and pore size distributions as they
are affected by fiber geometry should therefore be made
in order to gain an estimate of the relative importance
of these effects.
3. The short study reported herein on the effects of
electric charges applied to the filter fabrics suggests
a considerable potential for improving the performance
of fabric filters. This might be achieved by relatively
simple modifications of existing equipment. Profitable
avenues of research would be the application of electri-
cal voltages, to affect, besides the efficiency, also
the cleaning of filters. Optimal results would pre-
sumably be obtained by combining such improvements with
optimization of fiber geometry.
4. The performance of filter bags made of some of the fabrics
used in the present work should be evaluated in order to
verify that the results obtained with four inch diameter
patch filters are valid for full scale application.
-------
SECTION III
INTRODUCTION
Fabric filters have been used in industrial applications for
over a century. Originally introduced as a means of recover-
ing valuable products from gas streams, they still perform
this service today. Currently, they are being employed more
and more for the cleaning of stack gases in order to reduce
levels of air pollution. Properly operated bag filters can
remove more than 99.9% of the dust from a stream of gas, and
will do so less expensively than many other available devices.
The first baghouse filters were made of wool or cotton, these
being the only fibers available at the time. More recently,
synthetic fibers have been used, mainly for their higher
temperature resistance, a valuable advantage in many applica-
tions.
Filter bags have been made from a variety of fabrics. The
most frequently used are woven or felted cloth. The former,
as the name implies, is fabric produced by conventional
weaving, some weaves such as satin and twill being preferred.
Woven fabrics are durable but on a microscopic scale they
present inhomogeneities which reduce their effectiveness.
The yarns composing a woven fabric tend to be compact bundles
of filaments which do not utilize their maximum filtration
potential. The spaces between yarns also tend to form voids
or at least present a lower density of filaments to the on-
coming dust laden air. Felted cloth represents an attempt
to eliminate these shortcomings. Woven wool cloth can be
felted by agitating in hot water (precisely the same process
that must be avoided when washing woolen garments). Shrink-
age and fiber rearrangement result in a more compact, more
uniform fabric with some of the characteristics of felt,
hence the name. Similar structures are made from synthetic
non-felting fibers by combining an open scrim (a very open
woven cloth) with a looce mat of fibers and consolidating
the whole by some compacting process, usually needle punch-
ing. The term nonwoven cover's any fabric made without
recourse to a weaving step. Nonwovens are made by a variety
of processes. They are usually formed from a loose mat of
fibers which may be a card web or may be laid from air or
liquid suspension. The web is then consolidated by needle
punching or by bonding with latex, with a thermoplastic
binder fiber or by a chemical process.
In a more recent process, "spunbonded" fabrics are made by
spinning, drawing and then blowing a continuous filament
onto a moving belt. The filaments form a mat which is then
bonded by some suitable means. This process is attractive
because it is continuous from polymer to fabric. In a
different form, melted polymer from an extruder is converted
-------
by an air jet into many fine filaments. These are blown
directly onto a moving screen where they form a fabric-like
layer which can be further consolidated by heat and pressure.
Almost any one of the fabrics described above can be used for
filtering air, but the selection of one or another for commer-
cial filtration purposes has, in the past/ usually been
dictated by a multitude of considerations among which are
durability, dust loading capacity, cleanability, efficiency,
drag characteristics and ability to resist blinding.
Until recently, there has been little application of complete-
ly nonwoven fabrics (except paper) to filtration uses. Since
nonwovens are potentially better filters than woven fabrics,
and since they are also cheaper to produce, there is now con-
siderable interest in the development of completely nonwoven
filter bags. Such bags have already been shown to give
satisfactory performance in extended trials [1]. Other
improvements may be possible with nonwoven filters because
methods of fabrication allow a greater range of some proper-
ties than can be achieved with woven structures. One example
of such a property is fabric density which in the case of
nonwovens may be controlled by varying the degree of inter-
fiber bonding.
In a nonwoven, single fiber characteristics assume a dominant
role, since the effects of weave patterns, yarn twist, weave
density, etc., are absent and the single fiber, rather than the
yarn, is the filtering element of the structure. Single fibers
may affect and control filtration performance through their
geometric properties, surface finish, electrical properties,
hardness, and other mechanical properties.
The mechanism of capture of a particle by a single fiber has
received the attention of many workers, who have examined the
role of diameter [2-3], shape [5,6], surface [7,8], modulus [9],
and hardness [9], as well as the ambient relative humidity [10]
and the electrical charge on the fiber [11]. When a gas stream
is passed through a filter medium, there are three basic mechan-
isms of particle capture- d_ ^ct interception, inertial depo-
sition, and diffusion. Tae i^rst occurs with relatively large
particles, which collide with the fibers even when carried
along the streamlines of the carrier gas. Inertial deposition
occurs when the viscous force of the gas is insufficient to
keep the particle following the fluid streamline around the
obstacle; inertial forces then increase the probability of
impact with the obstacle. Deposition by diffusion occurs signi-
ficantly only at low gas velocities or long path lengths and
with particles of the order of O.ly and below. Here the
Brownian motion of the particle carries it off the streamline
and brings about collision with the obstacle. In all cases, it
is assumed that the particle adheres to the obstacle on contact.
-------
Dahneke [9], considering the conditions which would lead to a
particle bouncing off an obstacle rather than adhering to it,
concluded that for maximum capture ability, fibers should have
small diameters and be made of material with a low Young's
modulus. Reducing fiber diameter has a two-fold influence on
the capture of large particles: (a) it lowers the velocity
range in which inertial impaction is effective, and (b) it
raises the velocity at which the onset of bouncing occurs.
Dahneke's treatment focused attention on the changes in the
coefficient of restitution due to the work absorbed in flexural
deformation resulting from impact. The coefficient of resti-
tution is the ratio of the velocity of the rebounding particle
to the velocity of the impinging particle, both at the moment
of impact. Dahneke further examined the effect of the depth
of indentation of the particle on the obstacle. The greater
the depth, the greater the surface-particle potential well.
This increases the limiting velocity for rebound and in effect
means that soft surfaces make capture easier than hard ones.
This agrees with experimental findings of Zimon and Lazarev [7]
and is also intuitively acceptable.
The question of obstacle shape was reviewed by Ranz [5] who
showed that, in general, the blunter the body, the higher the
impaction efficiency. Thus, a ribbon with its flat side facing
the oncoming particles is about 46% more efficient than a
cylinder of equal width, while a cylinder is more efficient
than a ribbon of elliptical cross section with its thin side
facing the oncoming stream. More efficient than a ribbon is a
recessed collector, presenting a concave surface to the in-
coming stream. As will be shown, this is supported by the
higher efficiences found in the present work with filters made
from trilobal fibers.
The possible effects of obstacle surface roughness on capture
efficiency appear not to have been extensively studied, pro-
bably because of the difficulty of defining roughness and of
introducing a roughness parameter into aerodynamic expressions.
Leva [8] found no dependent j of permeability on the surface
roughness of granules composing a filter bed. His observa-
tions were made at low flow rates where the Reynolds number
was less than 10. It was assumed that roughness is only
important to the extent that it determines the onset of tur-
bulence at higher Reynolds numbers (^103). However, it would
be surprising if roughness were to have no effect at all on
capture efficiency, especially when the protuberances are of
the same order of magnitude as the particles.
The above remarks illustrate the nature of the considerations
involved in this study of the role of single fibers in filtra-
tion. Although a great deal of theoretical work has been
devoted to the various collection mechanisms, practical testing
of the theories appears to have been neglected. This may be
-------
because of the assumption that, since a filter actually oper-
ates with an accumulation of dust particles on each fiber,
the physical properties of the fibers will be insignificant
factors. The work presented here was undertaken to test this
assumption.
6
-------
SECTION IV
APPARATUS
As shown in Figure lf the filtration test apparatus included:
(1) humidity and temperature control devices (see IB for
greater detail, (2) a unit for controlled feeding of particu-
late matter, (3) a test chamber with provisions for convenient
placement of a "patch" filter, (4) a sampling filter to
capture particles passing through the test filter, (5) a pump
mounted on the exhaust end, (6) devices for cleaning of the
test filter by shaking and reverse air flow, (7) means for
continuous monitoring of the pressure difference across the
test filter, and (8) a flow meter located between the pump and
the test filter.
A sequence timer was also incorporated in the apparatus. This
consisted of a bank of ten microswitches actuated by cams
driven by a common shaft. The microswitches were connected to
the various valves and motors so that the full cycle, consist-
ing of preset filtering and cleaning periods, was performed
automatically and could be repeated as often as desired. In
the present study, one cycle consisted of five minutes of
filtering followed by two minutes of reverse air cleaning.
These time periods were arbitrarily chosen.
The following operating conditions were kept constant:
Face velocity: 12.4 cni/sec (24.5 ft/min)
Dust loading (material to air ratio): 5.51 g/m3
(2.4 grains/ft3)
Volumetric flow rate: 1040 cm3/sec (2.2 ft3/min)
Area of filter: 81.07 cm2 (0.0873/ft2)
Relative humidity: 30 + 2% R. H.
Reverse air velocity: 16.3 cm/sec (32 ft/min)
-------
MOTORIZED
VALVE
PRESSURE
RELIEF VALVE
J TRANSO.
FILTER
SOL.
VALVE
TEST FILTER
CLEAN-OUT
SOLENOID
VALVE
MANUAL
TWO-WAV
VALVE
AIR
BLEED
AIR PUMP
(ROOTE)
HUMIDITY AND
TEMPERATURE]
CONTROL
GELMAN
SAMPLING
FILTER
t^^—INLET AIR
Fig. 1A Apparatus for measuring filter performance.
ROOM AIR
INLET
HUMIDISTAT
COMPRESSED
INLET TO
FILTRATION
UNIT
SOLENOID
VALVE
SMOOTHING
CHAMBER
Fig. IB Humidity control system.
8
-------
SECTION V
FILTRATION PARAMETERS
In the filtration experiments, the pressure drop was automa-
tically registered on a time base recorder, while mass
efficiency was obtained by weighing the amount of fly ash on
the main filter and on a sampling filter (0.45p pore size)
through which was passed the full flow of air issuing from
the main filter. The fly ash contained particles with dia-
meters up to 40u. The measurable lower limit was 2.5y with
the instrumentation used. Ten filtering-cleaning cycles
were run for each web and efficiencies were measured at the
first and tenth cycle.
The filter drag is defined as the pressure drop, AP, divided
by the face velocity, V. The face velocity is given by
v _ Volumetric flow rate through the filter (Q)
Area of filter (A)
The effective drag, A?e/V, is defined as the drag after the
filter has been stabilized and is at the point in the filtra-
tion cycle where a cake has been established and the change
in pressure drop with time becomes a straight line function.
In these experiments, it was measured at the beginning of the
ninth cycle. The terminal drag, APf/V, was measured at the
end of the ninth filtration cycle, just before cleaning. The
specific cake resistance, K, may be written ^f'dW, where S
is the drag and W is the mass of cake per unit area. It was
measured for the tenth cycle.
The outlet concentration, Qj , is the ratio of the mass of
dust passed by the filter to the volume of gas passed during
a filtration cycle. It may be expressed as GO = mp/Qtc,
where tc is the time for one cycle and mp is the mass passed
by the main filter. Outle. concentration was evaluated for
the tenth cycle. These parameters are summarized in Table 1.
In addition to the collection of data related to mass effici-
ency and pressure drop, particle size analysis of fly ash
which passed through the main filter was performed using a
Coulter Counter. The distribution was then compared to that
of the fly ash fed to the main filter.
9
-------
Table 1
FILTRATION PARAMETERS
c
1. % Efficiency = m ° m x 100
c p
2. Effective Drag = AP /V
Terminal Drag = AP,/V
AP = Initial Pressure Drop
e
APf = Final Pressure Drop
V . Pace verity =
AP-/V - AP /V
£ 6
3. Specific Cake Resistance, K = m /A
4. Outlet Concentration, C = m /Qt
o pc
10
-------
SECTION VI
FABRIC FILTER FORMATION
Fabrics were made from a set of 32 samples of polyester fiber
in which two levels each of linear density, cross-sectional
shape, surface roughness - obtained by varying titanium
dioxide (TiO2) content - crimp, and staple length were repre-
sented. The samples and their code numbers are shown in
Table 2. Some of the fiber characteristics are illustrated
in the scanning electron micrographs shown in Figure 2.
Various methods of forming the filter fabrics were considered,
including needle punching, bonding with low-melting fibers,
and bonding with latex. Needle punching was rejected because
of the difficulty in maintaining the same fabric density for
all fibers. Figure 3 shows the strong dependence of air
permeability on fabric density, and illustrates the need for
avoiding density variations from sample to sample. Some
fabrics were made with low-melting binder fibers, but it was
difficult to obtain a good dispersion of these fibers among
the base polyester fibers. Latex bonding was finally adopted
as the most suitable method.
The problem of different fibers picking up different amounts
of latex was examined. The effect of latex content on air
permeability was investigated in a separate trial in which the
base fiber was kept constant (3 den, 1.5-inch, round, crimped
polyester). The filter fabrics used in this trial were formed
from cross-laid card webs which were dipped in latex, dried,
and consolidated by heating under pressure. The samples were
allowed to vary considerably in area density (oz/yd2) and in
percentage latex add-on. (Although it is EPA's policy to use
metric units in documents it produces, areadensity [or
"weight"] is expressed in non-metric units in this document
for the convenience of readers accustomed to textile common
usage. The reader may use ^.he conversion factor 1 oz/yd2 =
33.91 g/m2.) The latex used, designated Resyn 25-2853, and
supplied by the National Stanch Company, contained 45% vinyl
acetate-acrylic copolymer solids.
The results of air permeability measurements are shown in
Figure 4. These are expressed as air permeability (yd/min)
multiplied by bulk density (oz/yd3) and area density or weight
(oz/yd2). The normalized values plotted in Figure 4 are in-
tended to eliminate the effects of density variations as well
as those due to any variations in weight. The normalized
values thus should reflect only the presence of the latex
polymer, which, since it must intrude on the pores in the
structure, should cause a decline in air permeability with
increasing add-on. This is confirmed by the results shown in
11
-------
Table 2
DESCRIPTION OF EXPERIMENTAL FIBER SAMPLES
Sample
Batch
Code
1A
IB
1C
ID
2A
2B
2C
ID
3A
3B
3C
3D
4A
4B
4C
4D
5A
5B
5C
5D
6A
6B
6C
6D
7A
7B
7C
7D
8A
8B
8C
3D
TiO
Content
(%)
0.1
0.1
0.1
0.1
2.0
2.0
2.0
2.0
0.1
0.1
0.1
0.1
2.0
2.0
2.0
2.0
0.1
0.1
0.1
0.1
2.0
2.0
2.0
2.0
0.1
0.1
0.1
0.1
2.0
2.0
2.0
2.0
Shape of
Cross
Section
round
round
round
round
round
round
round
round
round
round
round
round
round
round
round
round
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
trilobal
Linear
Density
(den)
3.0
3.0
3.0
3.0
2.7
2.7
2.7
2.7
5.9
5.9
5.6
5.7
6.6
6.6
6.6
6.6
3.2
3.2
3.2
3.2
3.3
3.2
3.2
3.2
6.2
6.2
6.2
6.2
5.7
5.7
5.7
5.7
Crimp
Frequency
(no. /in. )
11-12
11-12
none
none
11-12
11-12
none
none
11-12
11-12
none
none
9-10
9-10
none
none
11-12
11-12
none
none
11-12
11-12
none
none
11-12
11-12
none
none
11-12
11-12
none
none
Nominal
Staple
Length
(in.)
3
6
3
6
3
6
3
6
3
6
3
6
3
6
3
6
3
6
3
6
3
6
3
6
3
6
3
6
3
6
3
6
12
-------
A. Round smooth (0.1% Ti02)
(3000X)
B. Round rough (2.0% Ti02)
(3000X)
Trilobal smooth (0.1% Ti02 ) D. Trilobal rough (2.0% Ti02 )
(1000X) (1000X)
Fig. 2 Scanning electron micrographs of selected fibers
used in the main experiment.
13
-------
zoo
e
§
3<50
100-
50
0.35 040 0.45 0.50 055 060 0.65 0.70
DENSITY (g/etn3)
Fig. 3 Relationship between air permeability and fabric
density.
»
15
0 10 20 30 40 50 60
% ADD-ON
Fig. 4 Relationship between air permeability and latex
content.
14
-------
Figure 4 for a wide range of latex add-on. In the main exper-
ment, latex add-on was maintained around 5%. Figure 4 shows
that the changes in permeability corresponding to small fluctu-
ations in binder content within this working range can be
considered negligible.
Fabric density has a greater effect on air permeability than
latex add-on. The need for maintaining constant density in
filter fabrics is illustrated by the air permeability data in
a somewhat different form (Figure 3). Here air permeability
is normalized only for weight and plotted as a function of
density. It can be seen that a relatively rapid increase
occurs with decreasing density.
All filter webs were made from card webs cross laid in four
alternate layers at right angles to one another. Pieces,
5 inches square, were cut from these and their weight adjusted
to three grams (corresponding to approximately 5.5 oz/yd ).
These pieces were then immersed in a latex bath, squeeze-rolled
four times, and allowed to dry overnight. They were then
pressed between Teflon® sheets at 135°C for one minute.
Since the main purpose of this study was to measure the effects
of fiber parameters, it was clearly important to minimize the
influence of fabric construction. To this purpose, formation
pressures were adjusted so as to produce nearly constant fabric
densities for all samples. Measured percentages of latex add-
on and fabric densities for all the samples in the main experi-
ment are listed in Table 3.
Although efforts were made to keep add-ons, thicknesses and
weights within a narrow range, some fluctuations occurred, as
shown in the table. Correlation coefficients calculated
between these values and the outlet concentrations listed in
Table 4 were found to be less than 0.1 and therefore insigni-
ficant. Correlation coefficients with respect to terminal drag
were slightly higher. However, even the highest (0.335 for
percent add-on) was low enough to indicate negligible effect.
15
-------
Table 3
FABRIC PROPERTIES OF MAIN EXPERIMENT SAMPLES
1A
2A
3A
4A
5A
6A
7A
8A
1C
2C
3C
4C
5C
6C
7C
8C
IB
2B
3B
4B
5B
6B
7B
8B
ID
2D
3D
4D
5D
6D
7D
8D
% Add-on
4.29
3.69
3.71
3.51
3.19
2.26
2.75
2.79
3.68
3.36
4.56
3.93
6.39
5.18
6.60
5.17
4.82
4.46
4.71
4.31
5.90
5.46
5.43
5.62
4.01
5.24
4.27
4.57
5.71
6.28
5.40
5.12
Density
(g/cm3)
0.213
0.228
0.217
0.208
0.207
0.210
0.229
0.201
0.200
0.213
0.220
0.202
0.223
0.200
0.204
0.209
0.204
0.231
0.223
0.238
0.234
0.224
0.202
0.213
0.230
0.227
0.235
0.226
0.204
0.224
0.227
0.229
Weight
(oz/yd2)
5.57
5.44
5.58
5.44
5.48
5.52
5.52
5.55
5.51
5.63
5.57
5.62
4.99
5.42
5.40
6.40
6.16
6.19
6.14
5.85
6.01
5.94
5.92
5.83
6.15
6.00
5.98
6.04
5.78
5.85
6.05
5.98
Thickness
(in.)
0.035
0.032
0.034
0.035
0.035
0.035
0.032 .
0.037
0.037
0.035
0.034
0.037
0.030
0.036
0.035
0.041
0.040
0.036
0.037
0.033
0.034
0.035
0.039
0.036
0.036
0.035
0.034
0.036
0.038
0.035
0.035
0.035
16
-------
Table 4
MEASUREMENTS OF FILTRATION PERFORMANCE RESPONSES
OF MAIN EXPERIMENT SAMPLES
E(l) E(10)
1A
2A
3A
4A
5A
6A
7A
8A
1C
2C
3C
4C
5C
6C
7C
8C
IB
2B
3B
4B
5B
6B
7B
8B
ID
2D
3D
4D
5D
6D
7D
8D
99.8
99.6
99.1
99.1
99.5
99.9
98.0
97.2
99.6
99.7
98.5
98.3
99.8
99.9
98.9
98.3
99.6
96.5
98.1
98.3
99.4
99.6
98.9
97.6
99.1
99.7
98.5
97.7
98.9
99.7
98.8
98.6
97.
99.
81.
76.
94.
99.
98.
98.
69.
87.
70.
75.
95.
99.
84.
89.
97.
97.
81.
81.
98.
98.
91.
94.
93.
99.
85.
82.
99.
99.
90.
88.
8
6
3
9
6
9
2
8
0
5
7
9
3
2
1
7
9
8
4
0
4
6
7
1
1
4
0
3
2
3
8
9
APe/V APf/V K
(dyn sec/cm3) (dyn sec/g cm)
174
195
160
153
181
49
139
83
292
389
195
181
195
160
195
209
160
125
125
160
160
146
125
111
195
139
153
174
146
174
132
137
786
563
452
320
452
139
348
243
765
1377
522
431
1064
835
522
626
542
512
431
494
800
849
362
327
855
751
591
584
542
1008
459
403
3
3
1
1
2
2
2
1
2
4
2
1
3
5
2
2
3
3
1
2
5
5
1
1
4
4
2
2
2
5
2
1
.49
.46
.28
.00
.47
.21
.03
.20
.82
.27
.10
.70
.92
.39
.07
.25
.19
.00
.66
.49
.19
.33
.91
.24
.99
.74
.67
.96
.48
.85
.48
.92
0
0
5
7
1
0
0
0
15
6
12
9
1
0
6
3
0
0
15
10
0
0
3
2
2
0
6
7
0
0
3
4
C0(10) E(10)(2.5y)
(g/m3) (%)
.69xlO"7
.10
.63
.63
.04
.03 -
.28
.45
.63
.28
.40
.86
.74
.24
.63
.89
.87
.87
.48
.36
.63
.56
.53
.34
.94
.24
.80
.42
.32
.28
.97
.96
93.
97.
0
0
58.
99.
68.
90.
0
67.
0
0
89.
94.
56.
79.
72.
72.
0
0
79.
86.
0
31.
10.
94.
0
0
84.
90.
10.
0
0
0
0
0
0
0
0
0
5
0
5
5
0
0
8
6
0
0
8
0
7
17
-------
SECTION VII
EXPERIMENTAL RESULTS
A. Statistical Analysis of Filtration Data
Determination of efficiency and drag characteristics were
made in the manner described above. The results were exam-
ined by using the so-called Yates algorithm, a statistical
analysis which provides a quantitative assessment of the
relative significance of each fiber variable on each filtra-
tion parameter, and, in addition, supplies estimates of
.interactions between variables.
The purpose of the standard Yates method [12] is to determine
the effects of a number (k) of variables (XjJ upon a response
This is accomplished by estimating the coefficients of the
following equation, which is a model of the system:
Ye = CQ + Z CiX. + Z c-.X.X. + E cljtX1XjX£ + ...
where Ye is the estimated value of the response Y, co is the
zero-order coefficient, ci are the first-order coefficients,
cij and GXJ& are the second and third-order coefficients, and
so on. The method yields k first-order effects (main effects) ,
k(k-l)/2 second-order effects (two-factor interactions),
k(k-l) (k-2)/2 three-factor interactions, and so on. This
method is applicable only when there are 2k observations
(k an integer) as in the present case.
The significance of each coefficient may be determined by means
of a half-normal plot. To prepare such a plot, the coeffici-
ents are listed in order of increasing absolute value and
assigned rank numbers (from 1 to 31 for 2s factorial). The
absolute values of the coefficients are then plotted on a
linear ordinate scale. The values plotted on the abscissa,
which' is a probability scale, are given by
50 + |£ (n - i)
where N is the total number of coefficients (31 for 2s factor-
ialj, and n is the rank number. The origin of the plot occurs
at the 50% point, which is the reason it is called half-
normal. On such a plot, most of the points will fall on a
straight line. However, if a coefficient is large enough, it
may deviate from the line, and the extent of the deviation is
a measure of its significance. A quantitative estimate of
the significance may be obtained by plotting "guardrails." A
18
-------
guardrail is drawn from the point where the line crosses 84.2%
probability to a point obtained as follows: (1) Multiply the
ordinate value at 84.2% probability by a number that depends
on the confidence level desired. In this analysis, the value
of 0.94 corresponding to the 95% confidence level was used.
(2) Add the product to the ordinate value corresponding to
the highest rank number (99.2% probability for 2 factorial).
Guardrails are shown as dashed lines in Figures 5-11. Points
lying above these are accepted as representing nonzero effects,
and are accented and labeled in Figures 5-11. Points lying
above the guardrail and having the highest Yates coefficients
carry the greatest significance. As the coefficients decrease,
on point will fall below the guardrail. All points with co-
efficients lower than this one are not significant even though
some may fall above the guardrail.
The above method was applied to the data shown in Table 4.
Inspection of this table shows that in most cases tenth-cycle
efficiency E(10) is lower than first-cycle efficiency E(l).
This is contrary to normal experience in baghouse operation,
where efficiency gradually improves with increased dust load-
ing. This effect is due to the high face velocity used in
the present study. The 24.5 ft/min velocity is almost an order
of magnitude greater than those commonly employed in baghouses
(3-4 ft/min). Under these conditions, a greater degree of
seepage occurs. These severe conditions were chosen to magni-
fy differences in performance between fabrics.
19
-------
1.6 r
1.4
3
§1.2
£
I 1.0
u Q8
u
t
Q6
0.4
Q2
Linear Density
50 60 70 80 90 95 98 99 99.8
PROBABILITY
Fig. 5 Half-normal plot for E(l).
40 r
He
o
U
8
en 2
Linear density
Shape •
/ 95 %
/,90%
Crimp—i / /
Crimp-length -A /
Shape-Denier-crimpAV^
. Length ./"/x
Shope-Denier
Shape length
~ 35
3
1
30
< 25
020
t
til
Sis
50 60 70 80 90 95 98 99 99.8
PROBABILITY
50607080 90 95 9899 998
PROBABILITY
Fig. 6 Half-normal plot for E(10).
Fig. 7 Half-normal plot for AP /V.
20
-------
160
- 140
o>
I
£ 120
5 100
S
U an
E *
(3
60
40
20
Linear density
Crimp •
50 60 70 80 90 95 98 99 99S
PROBABILITY
Fig. 8 Half-normal plot for APf/v.
4r
3
0>
o
JS
o
y 2
o
u.
u_
§
I
Linear Density
50 60 70 80 90 95 98 99 99.8 99.9
PROBABILITY
Fig. 10 Half-normal plot for CQ(10j.
Q8r
0.7
I-
8
5 OS
a o.4
t
§0.3
Q
§ 0.2
0.1
Linear
Density
50607080 90 95 9899 99.8
PROBABILITY
Fig. 9 Half-normal plot for K.
o>
§30
£
25
2 9Q
u
uj 15
I10
Linear density
506070 80 90 95 9899 99.8
PROBABILITY
Fig. 11 Half-normal plot for E(2.5M)
21
-------
In Table 4 the six dependent variables or responses were:
efficiency in the first cycle E(l)f efficiency in the tenth
cycle E(10)f outlet concentration in the tenth cycle Co(10),
effective drag APe/V, terminal drag APf/V, and specific cake
resistance K. For each of these the effects of the five inde-
pendent variables were determined by the standard Yates algor-
ithm. The half-normal plots are shown in Figures 5-10, and
some of the results of this analysis are given in Table 5,
which contains only significant first order effects.
The values shown in the table represent the average response
associated with the variable level. For example, the effici-
ency at the tenth cycle, E(10), improves from 86.0 to 95.0% if
the filter is made of trilobal rather than round fibers. The
absence of a number indicates no effect.
The following conclusions may be drawn from the 95% confidence
level data in the table:
1. Cross-sectional shape: Use of trilobal rather than round
fibers improves efficiency with no detrimental effect on
drag.
2. Surface roughness: No effect at the levels examined.
3. Linear density: Use of 3-denier rather than 6-denier fibers
improves the efficiency but at the cost of increased drag.
4. Crimp level: Use of crimped rather than uncrimped fibers
improves drag characteristics.
5. Fiber length: No effect at the levels examined.
In addition to the above effects, Figure 10 indicates that a
significant crimp-length interaction occurred with the Co(10)
response at the 95% confidence level. It should be noted that
neither crimp nor length alone have significant effects on
Co(10) or on E(10) at this level. To understand this inter-
action, the graphical presi ->tation shown in Figure 12 was used.
The presence of an interact- n is indicated when a response
depends on two or more variables simultaneously. Such depend-
ence can be represented, as shown, by a surface in a three-
dimensional plot for a two-factor interaction. In the original
data analysis, an implicit though unstated assumption was that
relationships between variables and responses were linear. The
plot in Figure 12 illustrates that this is highly unlikely.
Also, it is not possible to determine the true shape of the
surface from only the four available points. From this graph
it can be seen that for short fibers the presence of crimp
reduces the outlet concentration considerably, but not for long
fibers. Also it can be seen that for uncrimped fibers greater
22
-------
Table 5
SUMMARY OF YATES ANALYSIS AT 95% CONFIDENCE (32 POINT EXPERIMENT)
Shape Roughness Linear Density Crimp Level Fiber Length
Round Trilobal Smooth Rough 3 den 6 den 0 12 cpi 3 in. 6 in.
E(l)
E(10)
86.0 95.0
CO
AP /V
(dyn sec/cm3 )
APf/V
K
(dyn sec/g cm)
C (10)
(g/m3)
E(2.5y)
0.645 0.193
32.0 64.0
99.6
95.4
740
3.93
73.0
444 708 476
1.94
0.203 0.622
21.0
98.4
85.6 88.1 93.0* 88.6 92.4*
90% Confidence
-------
length produces greater efficiency. The reverse is true for
crimped fibers. Therefore, both crimp and length are important
parameters but their effect is not linear over the entire 32
point experiment.
Although the 90% confidence level is generally not considered
in a 2s factorial, for the case of E(10) it is interesting to
examine effects at this level. First of all, among these
effects are crimp and length alone. Imparting crimp to fibers
improves E(10) overall from 88.1 to 93.0%. Increasing fiber
length from 3 to 6 inches produces an overall improvement in
E(10) from 88.51 to 92.4%. At the 90% confidence level, the
crimp-length interaction also appears as represented in
Figure 12. The plot is similar to that for C (10) at the
95% confidence level. °
In addition to the crimp-length interaction, two more two-
factor interactions and one three-factor interaction have sig-
nificant effects on E(10) at 90% confidence.
Three-factor interactions cannot be illustrated graphically in
one diagram. Moreover, as the number of interacting variables
increases, each point on a particular diagram represents an
average of fewer data points, and is therefore less reliable.
For this reason, the three-factor interaction found to be sig-
nificant at 90% (shape x linear density x crimp),-will not be
considered.
The two other two-factor interactions (shape-length and linear
density-shape) are presented in Figures 13 and 14. Explana-
tions of all first-order and second-order effects will be
given in the last part of this section.
B. Particle Size Analysis
In addition to efficiency and pressure drop characteristics,
the particle size distribution of the dust in the outlet air
stream is an important aspect of filtration performance. A
Coulter* Counter was used 'o obtain distributions for the
original fly ash (i.e., ^or sample taken from the ash bin)
and for the fractions that passed through the main filter.
From these distributions it was possible to obtain the filter
efficiency distribution, or the efficiency of the filter at
each particle size level.
Coulter Electronics, Inc., Hialeah, Fla.
24
-------
CRIMP-LENGTH INTERACTION
E(IO)
Fig. 12 Three-dimensional plots of the effect of
crimp x length on E(10) and C (10).
SHAPE-LENGTH INTERACTION
SHAPE-LINEAR DENSITY INTERACTION
Fig. 13 Three-dimensional plot of the effect of
shape x length on E(10).
Fig. 14 Three-dimensional plot of the effect of linear
density y. shape on E(10).
25
-------
The efficiency distribution function E (d) was calculated from
the expression
E(d) - 1 - U-E(10)]f'(d)
E(d) - 1 -
where E(10) is the overall efficiency (at the tenth cycle),
f ' (d) is the weight fraction of passed particles at
a particular diameter, and
f (d) is the weight fraction of fed particles at a
particular diameter.
The diameter, d, ranged from a low value of 2.52y to the max-
imum diameter, with each increment taken corresponding to a
doubling of the volume of a spherical particle. Efficiency
distributions have been plotted in Figures 15-22. Examination
of the plots reveals that most of the improvement in efficiency
occurs at the small particle end of the distribution.
A Yates analysis was performed using the tenth-cycle efficiency
at the 2.5y level as a response (see Table 4, last column).
The results are included in Table 5 and show that removal of
these smallest particles is improved by use of trilobal fibers
of low linear density. Although the Yates analysis did not
show crimp to be an important parameter in this case, it
appears from Figures 15-22 that for the "A" and "C" series
(short fibers) , crimp brings a visible improvement in the fine
particle removal.
Observation of the efficiency distribution curves also reveals
a difference between the curves on the left and right hand side
of each pair corresponding to smoother and rougher fibers
respectively. In a paired comparison of these curves at the
low end of the distribution, rough fibers were more efficient
than smoother fibers in 11 cases, less efficient in 2 cases,
and equally efficient in 3 cases. The Yates coefficient shows
that average efficiencies for rough fibers in this region
(i.e., 2.5y) are 45% greater than that for smoother fibers,
but only at the 60% confide ~e level. Considering that scan-
ning electron micrographs (F ure 2) have shown only a small
difference between smooth and rough fibers in this study, the
above observations suggest that significant effects might be
found with fibers exhibiting greater differences in roughness.
By contrast, the average improvement due to surface roughness
on the overall efficiency E(10) for all particles from 40 to
2.5y is only 2.8%. It seems, therefore, that surface rough-
ness becomes more important as particle sizes become smaller.
Another interesting observation may be made from Figures 15-22.
Curves for samples 1,2,5, and 6(A,B,C, and D) clearly reflect
the higher efficiency due to lower deniers.
26
-------
100
80
60
UJ
£40
UJ
20
2 345 678910 15 20 30 40
PARTICLE DIAMETER, /X
100
80
UJ
u
t 40
UJ
8?
20
• -A
A-B
0-C
A-D
345 678910 15 20 30 40
PARTICLE DIAMETER, fJL
Fig. 15 Efficiency distributions for samples 1 A-D.
Fig. 16 Efficiency distributions for samples 2 A-D.
100 r
80
^ 60
UJ
y
u.
u.
" 40
85
20
2 345 6789O 15 20 3040
PARTICLE DIAMETER, /J.
100 r
80
60
UJ
o
il
40
5*
20
•*-o-o
.-A
A- B
0-C
A-D
2 345 678910 15 20 30 40
PARTICLE DIAMETER, fJL
Fig. 17 Efficiency distributions for samples 5 A-D.
Fig. 18 Efficiency distributions for samples 6 A-D.
27
-------
100
80
60
UJ
O
tt 40
Ui
20
2 345 678910 15 20 3040
PARTICLE DIAMETER,/!
100
80
60
UJ
o
It 40
UJ
20
234 5678910 15 20 30 40
PARTICLE DIAMETER, fJL
Fig. 19 Efficiency distributions for samples 3 A-D.
Fig. 20 Efficiency distributions for samples 4 A-D.
z
UJ
O
u.
100 r
80
60
40
20
100 h
80
UJ
u
E
20
345 678910 15 20 30 40
PARTICLE DIAMETER,/I
3 4 5678910 15 20 3040
PARTICLE DIAMETER, U
Fig. 21 Efficiency distributions for samples 7 A-D.
Fig. 22 Efficiency distributions for samples 8 A-D.
28
-------
C. Physical Interpretation of the Results
Although theoretical expressions have not been derived for the
relations between filtration responses and fiber parameters,
qualitative explanations may be offered.
Surface Roughness - Micrographs presented in Figure 2 essen-
tially explain the lack of effect on the major responses due
to surface roughness since no great differences in roughness
could be seen. The indication of improved efficiency at the
small particle end of the distribution bears further investi-
gation. It is possible that a relationship exists between the
size of surface asperities and the size of particles to be
captured.
Linear Density - Two explanations can be advanced. First, the
projected area of a constant mass of fibers is inversely pro-
portional to the square root of the linear density. It follows
that with fibers of lower linear density, the probability of
impact is increased. The second effect of decreasing linear
density at constant fiber mass is an increase in the number of
fibers. This in turn reduces the interfiber distances and
facilitates "bridging." Similarly, the increased projected
area and decreased pore size would be expected to produce
higher drag characteristics.
Cross-Sectional Shape - A similar argument applies to the case
of trilobal fibers. The 3-denier trilobal fibers used in this
study have a 25% greater projected area than the 3-denier
round fibers. The probability of impact increases proportion-
ally. It is difficult to explain why the greater projected
area of trilobal fibers does not cause increased drag. The
increased projected area alone appears not to cause as much of
an increase in drag as would an increase in the number of
fibers due to a decrease in linear density, which also de-
creases the average interfiber spacing. It can be seen in
Figure 23 which shows vie*, i of the downstream side of the fiber,
that there appears to be a Crapping mechanism peculiar to tri-
lobal fibers, where particles lodge in the concave region of
the fiber.
Crimp - This parameter improves both efficiency and drag char-
acteristics. The reason for the reduced drag and higher
efficiency with crimped fibers may be found in Figure 23B. It
can be seen that straight fibers seem to form groups of two or
more where the fibers run close together for a considerable
length. The space between them becomes clogged with filtered
particles, and the group then acts as a single wide flat fiber
with a higher resistance to air flow. Efficiency decreases
because of the larger spaces between these groups. None of
these groups is visible in the photograph of crimped fibers
(Figure 23A) which maintain an open structure.
90
-------
a) Sample 6A (3-den, trilobal,
crimped)
b) Sample 4C (6-den, round,
uncrimped)
Fig. 23 Scanning electron micrographs of filters
4C and 6A (150X)
30
-------
Length - The reason for the improvement in efficiency with
greater fiber length is not obvious, but becomes more under-
standable as interactions at the 90% level are examined.
Interpretation of two-factor interactions is difficult. Inter-
actions are likely to be a result of structural changes in the
fabric due to variables, and not simply inherent in the
variables themselves. Consequently, they must remain subject
to conjecture, with few exceptions (as in the case where some
structure difference could be observed between micrographs of
filters made from crimped and uncrimped fibers). With this
caution in mind, the following tentative explanations
are offered:
Crimp-Length Interaction - This interaction is most significant
since its effect on E(10) is at the 90% level and on Co(10) is
at the 95% level. Two physical interpretations appear to be
involved. First, it is proposed that in the carding process,
the crimp was removed from the longer fibers by a stretching
action. Both 3-in. and 6-in. fibers were processed on the same
card and it is likely that this unit, not being optimized for
the longer fibers, subjected the latter to severe elongation
and removed much of the crimp. This explains the lack of effect
of crimp in long fibers and why short crimped fibers were more
efficient (lower CQ) than long crimped fibers. This does not
explain the large difference between short and long uncrimped
fibers. A second physical interpretation must also be offered.
While crimp does indeed "open" the fiber bundles and thus
improves efficiency by creating a more uniform distribution of
fibers (and lower drag), it also appears that staple length may
have the same effect as crimp. Having a better chance of being
caught by the card, long staple fibers result in a more even
distribution of fibers even in the absence of crimp. This ex-
plains the occurrence of lower efficiency in the case of short
uncrimped fibers which have the advantage of neither length nor
crimp. The same effect is mirrored in the crimp-length inter-
action found in the outlet concentration response, which shows
high values only with short uncrimped fibers.
Shape-Length Interaction - A similar mechanism appears to be in
effect for the cross-sectional shape - fiber length interaction.
A trilobal shape has the same effect as crimp in promoting fiber
separation and thus uniform distribution. Accordingly, effici-
ency is poorer for short round fibers than for the other three
combinations. The improvement in E(10)due to greater length
for both round and uncrimped fibers causes an overall increase
in the average E(10) due to length. This explains the appear-
ance of fiber length at the 90% confidence level.
Linear Density-Shape Interaction - This interaction reflects
merely that in going from round to trilobal fibers, efficiency
increases less for 3-denier than for 6-denier fibers. This
31
-------
occurs because 3-denier filters have a high efficiency even
with round fibers, and have therefore a lesser possibility of
rising further. It is possible that if efficiency were
plotted on a logarithmic scale, this interaction would not
appear.
The above observations lead to the conclusions that inter-
actions arise from differences in fiber arrangement in the
nonwoven fabric. It follows that a different set of inter-
actions might well be observed if the filter fabrics were
fabricated by a different process, e.g., a different card or
an air-laid random-webber.
32
-------
SECTION VIII
EDP SCANNING MICROSCOPE
In the previous section it has been shown that significant in-
teractions may occur as a result of differences in fiber
arrangement within the nonwoven structure. A preliminary
investigation of variations in fiber distribution from web to
web was made using an EDP Scanning Microscope.* This instru-
ment provides an optical density contour map of a specimen as
it is spirally scanned from the center outward by a trans-
mitted beam of light 200 microns in diameter. The light
transmitted by the specimen passes to a photocell and is
converted into an electrical signal proportional to the amount
of light transmitted. This signal charges the stylus of a
facsimile printer which is mechanically coupled to the scan-
ning stage, thereby producing the contour map as the instrument
is operating. The contours may consist of a maximum of sixteen
distinguishable shades of darkness. The density distribution
can therefore be quantitatively mapped. The radial velocity
of the stylus is greater than that of the light beam by an
adjustable factor of 1 to 50. The contour map is therefore a
magnified image. Figures 24A and B are 5.5X magnifications of
samples #4C and #4A. The density of these samples is 0.06 g/cm3
and their weight is 0.8 oz/yd2. As can be seen in Table 2
these samples differ in crimp level only. In Figure 24A, the
uncrimped sample shows a large number of random lines of width
equivalent to 4 to 6 fiber diameters. These lines represent
fiber bundles. No such lines are visible in Figure 24B de-
picting the crimped sample. These results essentially verify
the assumptions made in the previous section with regard to the
effect of crimp.
Denser and heavier filter webs of these same two samples were
also scanned. Figures 24C and D show the uncrimped sample and
the crimped sample, respectively, at 5.5X magnification. The
density of these webs is 0 14 g/cm3 and their weight is
6.4 oz/yd2. Light intensity was increased to compensate for
greater density and weight. Individual fiber bundles are no
longer visible in the uncrimped sample. However, the dark and
light contours are seen to be less evenly distributed and
larger than those for the crimped sample.
These preliminary results obtained with the EDP Scanning Micro-
scope indicate that it is a promising method for further inves-
tigation of fiber distributions within the web structure. The
time available in the period covered by this report was not
sufficient for a complete study using these techniques.
*
Manufactured by Photometries, Inc., Lexington, Mass.
33
-------
*7»
>^v- ^,
Uncrimped Crimped
Fig. 24 Optical density contour maps of filter samples (5.5X)
A. Low density sample #4C
B. Low density sample #4A
C. High density sample #4C
D. High density sample #4A
34
-------
SECTION IX
EPITROPIC FIBERS
A. Surface Roughness
In the principal study of the present project, one of the var-
iables was fiber surface roughness. Fibers with different
surface roughness were obtained by addition of two levels of
Ti02 (0.1% and 2.0%). However, subsequent microscopical exam-
ination showed that even at the 2.0% TiOa level the degree of
roughness was not pronounced. This was consistent with the
observation that, in general, no significant effects of rough-
ness on filtration performance were found with these samples.
However, at the small end of the particle size distribution,
some increase in efficiency was apparent from visual examina-
tion of efficiency distribution curves and also by a score-
card method. This suggested that significant effects might
be found using fibers with higher degrees of roughness. These
could be obtained by addition of greater amounts of TiO2, but
it is known that increases in filler content would lead to
problems in fiber formation and to deterioration of fiber
tensile properties.
Epitropic (surface-modified) fibers are a recent development
of ICI Fibres Ltd., (England)[13], and represent a means of
introducing high levels of solid additives without encounter-
ing the above difficulties. Using an as yet unpublished
technique, a central fiber core is encased in an outer layer
of lower melting polymer containing a very high percentage of
filler. Figures 25 and 26 shov; micrographs of such a fiber
and its cross section; the substantial surface concentration
of the filler is evident.
Since such fibers might be suitable for a further study of
roughness effects on filtration, a quantity was obtained in
the form of 3-denier, 3-inch, uncrimped staple. The core
polymer and low melting point outer polymer was polyester,
and the filler was carbon black particles smaller than 5y.
The specific gravity of these fibers is 0.30 g/cc, which is
very close to that of polyester fibers.
Card webs were then made from a 50/50 blend of epitropic fibers
and polyester fibers selected from the main experiment. The
particular fiber sample chosen (#1-C) was 3-den., 3-in. , and
uncrimped, and therefore similar to the epitropic fibers in
all properties except roughness. As a control, a card web of
100% polyester (sample #1-C) was also made.
Filter fabrics were prepared from these webs according to the
procedure described in Section VI. The properties of these
fabrics are given in Table 6.
35
-------
Fig. 25 Micrograph of epitropic fiber surface
(from Ellis, V. S., Reference 13).
Fig. 26 Micrograph of epitropic fiber cross section
(from Ellis, V. S., Reference 13).
36
-------
Experiments were conducted with a finer fly ash (<10 ym) than
previously used. It was obtained by cyclone separation down-
stream from the feeder. The 50/50 epitropic/polyester sample
and the 100% polyester sample were measured for first-cycle
efficiency E(l). The pressure drop across both samples was
too low to be measured, since no appreciable cake buildup
occurred. Two replicate runs for each sample were used to
determine particle size distribution resulting in the effi-
ciency distribution curves shown in Figure 27. The higher
efficiency for the polyester sample is reflected in the
efficiency distribution curves especially at smaller particle
sizes (4y and below). The overall average E(l) was 94.3 for
the polyester sample and 89.8 for the epitropic/polyester
blend. This result was unexpected since the fiber with in-
creased surface roughness was apparently less efficient than
the smooth fiber.
An explanation of this anomalous result may lie in the conduc-
tivity of the carbon black on the surface of the epitropic
fibers. If there is a tendency for charge buildup on a filter
while it is operating, with the epitropic fibers such charges
probably would leak away to the metallic filter holder. This
would act against any electrostatic aggregation of particles
with its associated increase in collection efficiency. How-
ever, the 100% polyester fibers, being essentially nonconduc-
tive, would retain most of the induced charges which would
then be available to assist in the aggregation process. This
electrostatic effect on efficiency is presumably greater in
this case than any opposite effect due to surface roughness.
B. Electrostatic Effects
The conductivity of epitropic fibers is considerably higher
than that of polyester fibers. The resistance of a 50/50
epitropic fiber/polyester nonwoven was found to be about 10**
ohms while that of a polyester nonwoven was about 1010 ohms.
It was decided to measure the effect of this property on
filtration efficiency as voltages are applied to the filter
fabric. The filtration appa -atus was modified as shown in
Figure 28. The aluminum fil.or holder was insulated from the
filter chamber wall by a rubber ring. The ducts on either
side of the filter holder were converted to plastic tubing
for further insulation. Connections were made from a high
voltage power supply to the filter holder and from the filter
chamber wall to ground. With the power supplies available,
either positive or negative charges could be generated on the
filter holder and filter in the range of -2 Kv to + 30 Kv.
The upper voltage level for positive polarity was limited to
10 Kv since audible discharging occurred above this level.
37
-------
Table 6
PHYSICAL PROPERTIES OF EPITROPIC/POLYESTER FILTERS
50%/50%
Epitropic/
Polyester
100%
Polyester
1
2
1
2
Wt.
(oz/yd2)
5.49
5.63
5.54
5.52
Dens. % Latex
(g/cm3) Add-on
0.148
0.150
0.148
0.152
4.87
4.65
3.36
4.13.
lOOr
90-
80-
70-
60-
. SO-
I 40-
30-
3D-
10 •
100% POLYESTER
0.1 0.2
50% EPITROPIC/
50% POLYESTER
05 I 2 5
MICRONS DIAMETER
10
Fig. 27 Efficiency distribution curves for 100% polyester
and for 50% epitropic/50% polyester filter samples.
38
-------
FILTER
CHAMBER
COVER
Fig. 28 Diagram of filtration apparatus modification for
electrification trials.
100
99
98
97
(l16
95
94
93
92
9!
'K>
89
88
100% POLYESTER
50 7. EPITROPIC /
50 % POLYESTER
-2-1 01 2 3456789 10 KILOVOLTS
Fig. 29 Effect on E(l) of high voltage applied to 100% poly-
ester and to 50% epitropic/50% polyester filters.
39
-------
An epitropic/polyester sample and a polyester control sample
were then subjected to high voltage while filtration was going
on tinder the same conditions as before. Only one specimen was
used for each sample so that construction variability would be
eliminated as voltage was varied. Since the deposit of fly
ash after each filtration was small, it was possible to vacuum-
clean almost all deposit from the samples after each filtra-
tion. Vacuum cleaning was conducted through a metallic screen
to avoid damage to the filter. To further reduce the possible
effect of the residual dust, voltages of -2, -1, 0, + 5, and
+ 10 kilovolts) were applied to both samples and two or three
repeat measurements were taken at each voltage level, all in
random order. The averages of first cycle efficiency results
have been plotted in Figure 29.
Filtration efficiencies for the two samples, while about the
same under zero charge conditions, become significantly differ-
ent when charges were applied. The epitropic/polyester filter
was affected little or not at all by charging either positively
or negatively. The polyester filter, however, showed a defin-
ite increase in E(l) as the positive voltage increased. The
increase in E(l) from 92.5 at 0 KV to 98.8 at + 10 KV was
equivalent to a reduction in the outlet concentration of almost
90%. With negative voltage, the polyester sample also showed
a comparable increase to the limit of -2 KV.
The improvement in efficiency obtained by charging the 100%
polyester samples is due to the potential gradient developed
across the filter. It may be assumed that small leakage cur-
rents allow a lowering of potential in the central portion of
the filter; when the fibers are conductive no such potential
gradient may be established. The gradient may be increased
with polyester fibers by grounding the center of the filter.
Figure 30 shows first cycle efficiency for a needle-punched
nonwoven fabric. The improvement following the application of
a voltage to the outer edge is significant. If the center of
the sample is grounded there is a large further increase in
efficiency. Since the radius of the filter was about 5 cm, the
maximum potential gradient «. .stablished (at 10 KV) was about
2 KV/cm.
The principle has been described previously by Rivers [14]. It
appears not to have found application in baghouse filter
fabrics, but a study of the savings in energy that could be
derived from use of low drag filters rendered more efficient
by application of electric voltages would seem worthwhile.
40
-------
lOO
90
82
GROUNDED
CENTER
UNGROUNDED
CENTER
-2-1012349678910
KILOVOLTS
Fig. 30 Effect on E(l) of high voltage applied.to 100% poly-
ester (needled) with and without grounding center
of sample.
41
-------
SECTION X
REFERENCES
1. Turner, J. H. , "Performance of Nonwoven Nylon Filter Bags,"
Paper No. 73-300, APCA Annual Meeting, June 1973.
2. Ranz, W. E., "The Impaction of Aerosol Particles on Cylin-
drical and Spherical Collectors." Technical Report No. 3,
Contract No. AT(30-3)20 SO 1004, (1951).
3. Davies, C. H., "Separation of Airborne Oust and Particles,"
Institute Mechanical Engineers (London), Proceedings (B)
IB, No. 5, pp 185-213, (1952).
4. LaMer, V. K., "Studies on Filtration of Monodisperse
Aerosols." U. S. Atomic Energy Commission, Report NYO 512,
Contract No. AT(30-1)-651, (1951).
5. Ranz, W. E., "Principles of Inertial Impaction." Engineer-
ing Research Bulletin B-66, College of Engineering and
Architecture, Pennsylvania State University, (1956).
6. Ranz, W. E. and Wong, J. B., "Impaction of Dust and Smoke
Particles on Surface and Body Collector." Industrial and
Engineering Chemistry 44, pp 1371-1381, (1952).
7. Zimon, A. D. and Lazarev, K. A., Kollidnyi Zhurnal 31,
No. 2, pp 214-219, March 1969.
8. Leva, M., "Fluid Flow Through Packed and Fluidized Systems.1
Bureau of Mines Bulletin 504, U. S. Government Printing
Office, (1951).
9. Dahneke, B., "Capture of Aerosol Particles by Surfaces,"
Journal of Colloid & Interface Science 37, No. 2,
pp 342-353, (1971).
10. Durham, J. R. and Harrington, R. W., AICHE 63rd Annual
Meeting, Chicago, Illinois, November 1970.
11. Rodebush, W. H., et al, Report No. 2050, PB 32203,
November 24, 1932.
12. Daniel, C., "Use of Half-Normal Plots in Interpreting
Factorial Two-Level Experiments." Technometrics l_f No. 4,
pp 311-341, (1959).
42
-------
13. Ellis, V. S. , "Epitropics-Third Generation Conductive
Fibers," Textile Manufacturer 101, No. 1193, pp 19-23
July 1974.
14. Rivers, R. D., "Operating Principles of Non-Ionizing
Electrostatic Air Filters," ASHRAE Journal, pp 37-40,
February 1962.
43
-------
SECTION XI
GLOSSARY
Card - A textile processing machine which separates fibers
from each other, lays them parallel, and forms them into a
thin web.
Crimp - (1) The waviness of a textile fiber
(2) An individual wave in a textile fiber
Denier - A unit of linear density corresponding to the weight
in grams of 9000 meters of a filament or yarn.
Drag - The pressure drop across a filter divided by the face
velocity (volumetric flow rate normalized for filter area).
Effective Drag - The drag after the filter has been established
and is at the point in the filtration cycle where a cake has
been established and the change in pressure drop with time
becomes a straight line function.
Efficiency - The percentage of the total weight of dust im-
pinging on a filter that is collected by the filter.
Epitropic Fiber - A fiber whose surface contains embedded
particles which modify one or more of the fiber properties.
(From the Greek epi meaning upon and tropaios to change).
Fly Ash - A product of coal burning consisting of spherical
particles ranging in diameter from several hundred to below
one micron.
Interaction - A combination of two or more independent varia-
bles (a second or higher order term) that acts as a single
variable.
Yates Algorithm - A statistical method for calculating the
coefficients in the linear model representing any two-level
factorial experimental design.
Outlet Concentration - The weight of dust per volume of air
that passes through a filter.
Specific Cake Resistance - The change in drag per mass of dust
cake per unit filter area.
Staple - Fiber, cut into short pieces, that can be processed
on a card.
44
-------
Terminal Drag - The drag at the end of a filtration cycle just
before cleaning.
Trilobal - A fiber cross-sectional shape with three rounded
projections.
45
-------
SECTION XII
NOMENCLATURE
A area of filter
C outlet concentration
o
C (10) outlet concentration of tenth cycle
ci' cij' cijfc ...Yates coefficients
d diameter of particle
E(l) efficiency at the first cycle
E(10) efficiency at the tenth cycle
E(10) (2.5y) efficiency at the tenth cycle for particles
2.5y diameter
E(d) efficiency for particles of diameter d
f (d) weight fraction of passed particles at a
particular diameter
f ' (d) weight fraction of fed particles at a
particular diameter
k number of variables in Yates analysis
K specific cake resistance
m mass of dust captured
C
m mass of dust passed
n rank number of Yates coefficient
N total numbo of Yates coefficients
AP effective pressure drop
APf final pressure drop
Q volumetric flow of air through the filter
C
S drag
46
-------
t time period for one cycle
c
V face velocity
W mass of dust per unit area of filter
Xi' Xij' XijA ... Yates variables
Y estimated value of response in Yates
analysis
47
-------
TECHNICAL REPORT DATA
I'Please read /attractions on the reverie before completing)
1. REPORT NO
RPA-650/2-75-QQ2
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Influence of Fiber Characteristics on Participate
Filtration
5. REPORT DATE
January 1975
6. PERFORMING ORGANIZATION CODE
7 AUTHOR(S)
B. Miller, G.E.R. Lamb, and P. Costanza
8. PERFORMING ORGANIZATION REPORT NO,
9. PERFORMING OR6ANIZATION NAME'AND ADDRESS
10. PROGRAM ELEMENT NO.
Textile Research Institute
P.O. Box 625
Princeton, NJ 08540
1AB012; ROAP 21ADL-022
11. CONTRACT/GRANT NO.
R-800042
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
NERC-RTP, Control Systems Laboratory
Research Triangle Park, NC 27711
PERIOD COVERED
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACTThe repopf gives results of an evaluation of the influence of five fiber para-
meters (cross-sectional shape, linear density, surface roughness, crimp, and
staple length) on the filtration performance of model nonwoven fabrics made from
the fibers. Nonwoven fabrics made from 32 polyester fiber samples were used to
filter fly-ash particles from a stream of air. Filter performance was assessed
by measuring: pressure drop across the filter, collection efficiency, and particle
size distribution. Statistically, at 95% confidence: efficiency was improved by using
trilobal (rather than round cross-section) fibers with no detrimental effect on drag;
efficiency and drag were improved by using crimped (rather than uncrimped) fibers;
and efficiency was improved by using 3 (rather than 6) denier fibers, but at the cost
of greater drag. These efficiency improvements were especially pronounced for
fine particles (approximately 2. 5 microns). Non-statistically, except for epitropic
fibers with very rough carbon-embedded surfaces, rougher fibers appeared more
efficient in removing fine particles. Applying d. c. voltages to 100% non-conducting
polyester filters showed considerable increases in efficiency; again, no such
effects were seen with polyester fikurt 'ncorporating 50% epitropic fibers.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air Pollution
Dust
Filtration
Fibers
Nonwoven Fabrics
Particle Size Distribution
Air Pollution Control
Stationary Sources
Particulates
Collection Efficiency
13B
11G
07D
HE
14B
8 DISTRIBUTION STATEMENT
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
55
Unlimited
20 SECURITY CLASS (This page)
Unclassified
22 PRICE
EPA Form 2220-1 (9-73)
48
------- |