EPA-650/2-75-009
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-650/2-75-009
FABRIC FILTER CLEANING STUDIES
by
Richard Dennis and John Wilder
GCA Technology Division
Burlington Road
Bedford. Massachusetts 01730
Contract No. 68-02-0268
ROAPNo. 21ADJ-049
Program Element No. 1AB012
EPA Project Officer: Dale L. Harmon
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 2216L
-ft
ii
-------
CONTENTS
List of Figures
List of Tables
Nomenclature
Conversion Factors for British and Metric Units
Section
Title
II
INTRODUCTION
ROLE AND CAPABILITIES OF FABRIC FILTER SYSTEM
PROBLEMS IN AEROSOL FILTRATION
PROBLEMS IN FABRIC FILTER CLEANING
PROGRAM OBJECTIVES
SELECTION AND SEQUENCE OF STUDY OF FABRIC
CLEANING METHODS
SUMMARIZED CONCLUSIONS AND RECOMMENDATIONS
Conclusions
Recommendations
REFERENCES
MECHANICAL SHAKING STUDY
OBJECTIVES AND APPROACH
BACKGROUND AND THEORY
Adhesion and Removal Mechanisms Including
Acceleration
Bag Motion Theory
Basic Concepts
Waves and Wave Velocity
Tension Changes
Damping and Reflection
Page
xiii
xx i
xxiv
xxv ii
1
1
1
2
3
4
6
6
8
9
11
U
13
14
21
21
24
25
28
iii
-------
CONTENTS (Continued)
Sect loo Title Page
Bag Acceleration 29
Shaking Energy and Power 30
Bag Life . 32
Collection Efficiency 32
APPARATUS, TECHNIQUES AND MATERIALS 36
Shaker Apparatus . 36
Test Fabrics 41
Bag Properties and Measurement Technlquei 41
Lineal Density 45
Bag Weight and Bag Tension 45
Bag Modulus 46
Shaking Energy and JPover 49
Motion of the Shaking Baft 50
Test Dusts 50
Bust Measurements -q
Aerosol Concentration and Particle Size 5^
Dust Filtering Rate 60
Dust Removal 60
Residual Dust 60
Pressure, Air Flow and Air Propertjep
Measurements 62
RESULTS 63
Cleaning Forces and Bag Motion 63
Iv
-------
CONTENTS (Continued)
Section Title ^ Page
Bag Tension versus Shaking Frequency 63
Bag Acceleration 69
Bag Power Consumption 80
Dust Removal 86
Dust Removal versus Number of Shakes 89
Dust Removal versus Shaking Frequency 92
Dust Removal versus Shaking Amplitude 94
Dust Removal versus Acceleration 100
Dust Removal versus Initial Bag Tension 107
Dust Removal versus Cloth Loading 107
Dust Removal versus Fabric Type 113
Dust Removal versus Dust Type 115
Summary of Dust Removal Studies 119
Collection Efficiency and Emission 121
General Observations 122
Effluent Concentration versus Filtration
Time 124
Effluent Concentration versus Shaking
Amplitude 131
Effluent Concentration versus Shaker
Frequency 131
Effluent Concentration versus Bag Life
and Bag Stretch 134
Efficient Concentration versus Fabric
Type 137
-------
CONTENTS (Continued)
Section _____ _ Title _ Page
Efficient Concentration versus Dyat Type 138
Discussion of Particle Emission Studies 144
Filter Resistance 146
Linearity of Resistance-Time Curves
Specific Resistance Coefficient, K
Effective Residual Resistance 154
Discussion of Operating Parameters for
Various Filter Systems 157
Bag Life 158
CONCLUSIONS TO MECHANICAL SHAKING STUDY 167
General Conclusions 167
Particulate Emissions 168
Dust Removal by Mechanical Shaking 171
Filter Resistance and Power Requirements 172
REFERENCES 173
III PULSE CLEANING STUDIES 175
OBJECTIVES AND APPROACH 175
BACKGROUND 176
Applications and Advantages 176
Problem Areas 177
Collector Performance 178
Resistance and Particle Removal 178
Factors Involved in Dust Removal 181
vi
-------
CONTENTS (Continued)
Section Title Page
Adhesion 181
Dislodgement 181
Reverse Air Transport 182
APPARATUS, MATERIALS AND TECHNIQUES 182
Filter Assembly 183
Pulse Jet Cleaning Equipment 185
Test Fabrics 188
Bag Properties and Measurement Techniques 190
Bag Motion 191
Accelerometer 191
Strain Gage 191
High Speed Photography 192
Test Dusts 192
Dust Measurements 192
Pressure and Flow Measurements 194
RESULTS 196
Particulate Emission Characteristics 196
Compressed Air Pressure 197
Cleaning Frequency (Operating Cycle) 204
Cleaning Pulse Duration 210
Felt Type 210
Dust Type 213
Filtration Parameters 215
vii
-------
CONTENTS (Continued)
Section Title Page
Design Modifications 218
Particle Size and Concentration Changes During
Filtration Cycle 224
Residual Filter Resistance 233
Concentration Profiles. Dust Emissions 236
Operating Filter Resistance 237
Compressed Air Pressure 242
Pulse Interval 242
Pulse Duration 245
Felt Type 245
Dust Type 245
Filtration Parameters 248
Design Modifications 250
Control of Operating Resistance 252
Residual Resistance 253
Transition Pressure 268
Operating Resistance 269
Factors in the Selection of Optimum
Operating Conditions 270
CONCLUSIONS TO PULSE JET CLEANING STUDIES 272
General Conclusions 272
Particulate Emissions 273
Dust Removal and Filter Resistance 275
System Design and Operating Factors 276
viii
-------
CONTENTS (Continued)
Section Title Page
REFERENCES 278
IV REVERSE FLOW STUDIES 281
OBJECTIVES AND APPROACHES 281
BACKGROUND 282
Postulated Cleaning Actions 282
Basic Design Concepts 284
APPARATUS, MATERIALS AND TECHNIQUES 285
Modified Mechanical Shaking System 285
Modified Pulse Jet System 288
Test Fabrics 289
Measurements and Instrumentation 289
Test Dusts 290
RESULTS 290
Low Pressure, Reverse Flow with Mechanical
Shaking 290
Low Pressure, Reverse Flow With Pulse Jet
Equipment 295
Slow Inflation Rate, Reverse Flow Only 296
Pressure Rise Rate, Reverse Flow Only 299
Frequency of Cleaning, Reverse Flow Only 302
Reverse Flow Duration ~ Reverse Flow Only 304
Reverse Flow With Reverse Pulse Cleaning 307
CONCLUSIONS TO REVERSE FLOW CLEANING STUDIES 310
General Conclusions 311
ix
-------
Section
CONTENTS (Continued)
Title
Appendix
A
B
Augmentation of Mechanical Shaking by Reverse
Flow
Felt Bags Cleaned by Reverse Flow
REFERENCES
CONCLUSIONS AND RECOMMENDATIONS
CONCLUSIONS
Limits of Data Application
Outlet Versus Inlet Concentrations
Fractional Particle Size Efficiencies
Filter Effluent Concentrations
Filter Cleaning Action
Energy Requirements and System Capabilities
RECOMMENDATIONS
Equipment Application and Operating Param-
eters for Mechanically Shaken Filters
Equipment Applications and Operating Param-
eters for Pulse Jet Filter Systems
Research and Instrumentation Needs
AUTOMATIC FLOW CONTROL SYSTEM
FORCE AND VELOCITY MEASURING INSTRUMENTATION
FORCE AND PRESSURE MEASUREMENTS
VELOCITY MEASUREMENT
PHOTOELECTRIC DETECTION OF SHAKEN BAG MOTION
312
313
314
315
316
316
316
317
317
318
319
319
319
320
321
325
331
331
338
339
-------
CONTENTS (Continued)
Appendix
D
E
F
G
H
I
J
K
L
Title
THEORETICAL AVERAGE WAVE VELOCITY
BAG STRETCH AND TENSION DURING SHAKING
DAMPING AND AVERAGE BAG AMPLITUDE THEORY
ENVELOPE PHOTOGRAPHS OF BAG MOTION
DATA SHEETS FOR MECHANICAL SHAKING STUDIES
DYNAMIC SHAKING TESTS
Dust Removal Tests
Effluent Size Properties
THEORY FOR LATERAL FORCES IN SHAKING
SYSTEM PRESSURE DIFFERENTIAL VERSUS SINGLE ELEMENT
PRESSURE DIFFERENTIAL
CONCLUSIONS
Mathematical Model of a Five-Compartment
Baghouse
Results
DUST TRANSPORT DURING PULSE CLEANING
FABRIC ACCELERATION IN PULSE CLEANING
MECHANICAL PROPERTIES OF THE BAG
Dynamics of Bag Motion
Estimating Fabric Deceleration ,
STIFF, LIGHT FABRICS
Flexible, Heavy Fabrics
Moderately Stiff, Moderately Heavy -Fabrics
Velocity of Removed Agglomerates
Page
341
345
349
351
359
359
360
367
373
379
385
386
387
389
397
398
400
401
401
402
403
404
xi
-------
CONTENTS (Continued)
Appendix Title p
M SUPPLEMENTARY DATA ON PRECISION OF MEASUREMENT
TECHNIQUES
ANDERSEN "OUTSTACK" IMPACTOR
atii
-------
LIST OF FIGURES
No. Title Page
1 Potential Dust Fabric Stresses Developed During Shake
Cleaning 15
2a Concepts of Cleaning Via Acceleration (a) High
Adhesive/Cohesion Ratio 16
2b Concepts of Cleaning Via Acceleration (b) Low
Adhesive/Cohesion Ratio 16
3 Internally Illuminated 10-Foot by 6-Inch Diameter
Cotton Bag, After Cleaning (Photo Approximately
4 Feet From Lower End) 20
4 Appearance of Shaking Sateen Weave Cotton Bag
10-Feet Long by 6-Inch Diameter (4 cps, 2-Inch
Amplitude, 7.4 Pound Shaking Tension) 23
5 Variation of Bag Tension With Position of Shaker Arm 26
6 Schematic Frawing, Bag Mounting and Shaking Assembly 39
7 Mountings Used to Support Bottom of Shaken Bags 40
8 Bag Dimensions and Loop Design 44
9 Tensile Properties for a 10-Foot by 6-Inch Sateen Bag 48
10 Inlet Fly Ash Size Distributions by Andersen Cascade
Impactor 56
11 Shaking Tension as a Function of Frequency for Cotton
Sateen Bag 65
12 Predicted Versus Observed Resonant Frequencies 66
13 Maximum and Minimum Envelope Curves and Average
Shaking Tension for Cotton Sateen Bag 67
14 Bag Displacement Versus Driven Frequency and Indicated
Dynamic Tension 71
15 Effect of Shaker Amplitude on Shaking Tension 75
16 Effect of Fabric Type on Shaking Tension With
10 Ft. x 6 In. Bags (1 In. Shaking Amplitude) 76
xiii
-------
LIST OF FIGURES (Continued)
No. Title
17 Effect of Length/Diameter Ration on Shaking Tension 77
&•
18 Effect of Initial Tensioning on Shaking Tension With
Unnapped, 10 Ft0 x 6 In, Cotton Sateen Bags 78
19 Comparison of Average Bag Amplitudes as Calculated by
Equation (2.8) or Equations (2.17) and (2.18). 81
20 Measured Power Inputs to Shaker Motor 82
21 Power Consumption, Phase Angle and Frequency Relation-
ships for a 10 Ft0 x 6 In« Cotton Sateen Bag (1 In.
Amplitude) 84
22 Force, and Velocity Versus Shaking Frequency and Period 85
23 Percent Fly Ash Removal Versus Number of Shakes and
Shaking Frequency at 1-Inch Amplitude 91
24 Effect of Shaking Frequency on Filter Capacity for
1-Inch Shaking Amplitude 93
25 Effect of Resonance on Dust Removal 95
26 Effect of Number of Shakes, 8 cps, and Shaking
Amplitude on Dust Removal From 10 Ft. by 6 In.
Cotton Bags 97
27 Effect of Shaking Amplitude and Shaking Frequency,
350 Shakes, on Dust Removal From 10 Ft. by 6 In.
Unnapped Cotton Bags 98
28 Effect of Shaking Frequency and Shaking Amplitude,
350 Shakes, on Dust Removal From 10 Ft0 by 6 In0
Cotton Bags 99
29 Dust Removal Versus Bag Acceleration for Cotton Bags,
350 Shakes 103
30 Dust Removal Versus Bag Acceleration for Cotton Bags,
40 Shakes 104
31 Filter Capacity Versus Average Bag Acceleration for
Sateen Weave Cotton Bags 106
xiv
-------
LIST OF FIGURES (Continued)
No. Title Page
32 Cloth Loading and Filter Drag Characteristics for
Typical Shaking Regimes - Composite Curve 108
33 Dust Removal Versus Cloth Loading for Fixed Shaking
Regime 109
34 Residual Fabric Loading Versus Average Bag Acceleration 111
35 Residual Fabric Loading Versus Total Number of Shakes
at Three Acceleration Levels for Unnapped Sateen
Weave Cotton 112
36 Residual Fabric Loadings for Various Fabrics With Fly
Ash Aerosol (8 cps, 1 In. Amplitude Shaking) 114
37 Residual Fabric Loadings for Various Fabrics With Talc
Aerosol (8 cps, 1-In. Amplitude Shaking) 116
38 Fabric Loading Versus Filter Drag for Woven Cotton and
Dacron Bags With a Talc Aerosol (360 Shakes at 8 cps
and 1-In. Amplitude) 117
39 Calculated Effluent Concentration Versus Time for Fly
Ash and Ambient Dust Based on B&L Measurements 125
40 Calculated Effluent Number Concentration Versus Time
and Particle Diameter for Fly Ash Filtration With
Sateen Weave Cotton 129
41 Changes in Effluent Size Properties With Filtering
Time for New (< 10^ Shakes) and Old (2 x 107 Shakes)
Bags (Sizing by Optical, B&L, Counter) 130
42 Calculated Concentration Versus Shaker Amplitude at
Constant Shaking Frequency
43 Calculated Effluent Concentration Versus Shaking
Frequency at Constant Amplitude 133
44 Inlet Concentration Versus Percent Weight Penetration,
Ambient Temperature 136
45 Comparative Effluent Concentrations for Fly Ash With
Different Fabrics 140
xv
-------
LIST OF FIGURES (Continued)
No. Title Pai
46 Total Dust Emitted Per Filter Cycle Versus Fabric Type
and Residual Dust Holding 141
47 Comparative Effluent Concentrations for Talc and Fly
Ash With Cotton and Dacron Fabrics 143
48 Resistance Changes During Fly Ash Filtration (see
Table 15) 15Q
49 Effect of Insufficient (Non-Equilibrium) Shaking on
"K", Filter Drag and Fabric Loading With Fly Ash at
3.5 grains/ft.3 and 3 ft./min. Filter Velocity
10 Ft. x 6 In. Cotton Sateen Bag 153
50 Residual Filter Resistance and Residual Fabric Loading
(Figure 34) Versus Bag Acceleration for Fly Ash/Sateen
Weave Cotton System 155
51 Bag Elongation Versus Total Number of Shakes for Used
and Clean, 10 Ft. x 6 In. Sateen Weave Cotton Bags 15^
52 Bag Elongation Versus Total Number of Shakes for
Various Used 10 Ft. x 6 In. Fabric Bags 154
53 Schematic of Pulse Jet Cleaning Assembly 18$
54 Standard Pulse Delivery System 18g
55 Average Outlet Concentrations for Fly Ash Versus
Reservoir Pressure With Direct and Damped Pulses;
(a) All Variables Averaged Except Pressure and
Pulse Type; (b) Variations in Velocity, Loading and
Pulse Type Shown Individually 198
56 Effect of Direct and Delayed Pulses on Bag Dif-
ferential Pressure for Valve Open Time of 0.15 Second 20Q
57 Effect of Reservoir Pressure and Pulse Damping on
Effluent Concentration for Fly Ash Filtration With
Dacron Felt (Note: Inlet Mass Concentration Converted
to Equivalent Number Concentration and Scaled by 10"^
for Comparison) 201
58 Effect of Frequency of Cleaning on Average Outlet Con-
centrations for Fly Ash Filtration With Dacron Felt 205
xvi
-------
LIST OF FIGURES (Continued)
No. Title Page
59 Effect of Frequency of Cleaning on Average Outlet Con-
centrations for Fly Ash Filtration With Dacron Felt 207
60 Effect of Pulse Duration and Pulse Interval on Effluent
Concentration for Fly Ash Filtration With Dacron Felt,
Direct Pulse (Note: Inlet Mass Concentration Converted
to Equivalent Number Concentration and Scaled by 10~4
for Comparison) 208
61 Effect of Pulse Duration and Pulse Interval on Effluent
Concentration for Fly Ash Filtration With Dacron Felt,
Damped Pulse (Note: Inlet Mass Concentration Converted
to Equivalent Number Concentration and Scaled by 10~4
for Comparison) 209
62 Comparative Performance Between Wool and Dacron Felts
With Fly Ash Filtration 211
63 Effect of Felt Type on Fly Ash Particulate Emissions 212
64 Comparative Filtration Characteristics for Talc and
Fly Ash With Dacron Felt 214
65 Particle Concentration Versus Time for Selected Sizes 226
66 Particle Concentration Versus Time for Selected Sizes 227
67 Particle Concentration Versus Time for Selected Sizes 228
68 Particle Concentration Versus Time for Selected Sizes 229
69 Particle Concentration Versus Time for Selected Sizes 230
70 Particle Concentration Versus Time for Selected Sizes 231
71 Relationship Between Residual Dust Deposit and
Residual Filter Resistance 234
72 Average Outlet Concentration Versus Average Filter
Resistance (Reservoir Pressure 40 to 100 psig, Pulse
Interval 1 min., and Pulse Duration 0.06 to 0.15 sec. 235
73 Dust Concentration Profile on Exit Side of Bag,
Bottom to Top 238
xvii
-------
LIST OF FIGURES (Continued)
No. Title Page
74 Resistance Characteristics for Fly Ash and Dacron Felt,
Inlet Cone. 12 grains/ft.3, Filtration Velocity
8.5 ft./min. 240
75 Resistance Characteristics for Talc and Dacron Felt,
Inlet Cone. 1.53 grains/ft.3, Filtration Velocity
8.5 fpm 241
76 Average Filter Resistance Versus Reservoir Pressure,
Direct and Damped Pulses, for Fly Ash Filtration With
Dacron Felt 243
77 Effect of Pulse Interval on Average Filter Resistance
for 70 psig Pulses 244
78 Comparative Resistance Properties of Wool and Dacron
With Fly Ash Filtration 247
79 Relationship Between Rate of Pressure Rise and
Residual Filtration Resistance at 8.5 ft./min.
Filtration Velocity 256
80 Residual Filter Resistance Versus (a) Pressure Rise
Rate and (b) Reservoir Pressure (see Figures 76 and
79, Respectively) 258
81 Postulated Removal Mechanism for Reverse Air Drag
Through a Loaded Felt 263
82 Schematic of Mechanical shaking System as Modified for
Reverse Flow Cleaning 286
83 Schematic View, Pulse Jet Assembly With Low Pressure
Reverse Air and High Pressure Nozzle Removed 287
84 Resistance Characteristics for Fly Ash/Unnapped Cotton
Filtration With Mechanical Shaking and/or Reverse Flow 293
85 Resistance Characteristics for Fly Ash/Dacron (Plain
Weave) Filtration With Mechanical Shaking and/or
Reverse Flow 294
86 Controlled Variations in Rate of Differential Pressure
Change Across Felt Bags (See Table 31) 300
xviii
-------
LIST OF FIGURES (Continued)
No.
87
88
89
A-l
B-l
B-2
B-3
B-4
C-l
G-l
G-2
H-l
H-2
J-l
K-l
Title
Air Volume Ejected Per Pulse With Commercial 1/4-in.
Nozzle and Solenoid Valve Assembly
Characteristic Pressure/Time Traces for Variations in
Reverse Pulse Duration (see Table 33)
Differential Pressures Across Dacron Felt Bag for In-
dicated Cleaning Conditions
Automatic Flow Control System
Load Cell Mounting for Shaken Bags, Model 1
Revised Design for Load Cell Mounting for Shaken Bags
Load Cell Mounting for Weighing the Pulsed Bag
Assembly
Circuit Diagram, Signal Output From Load Cell
Description of Bag Motion by Interception of Light
Beam
Position of Photographs Relative to Resonant
Frequencies
Bag Displacement Versus Shaking Frequency
Sample Data Sheet for Dust. Removal Tests
Typical Effluent Concentrations Versus Particle
Size Category and Filtration Time
Operating Variables in a Multicompartment Baghouse
Transport Effects on Particle Disposition: s: Small
Page
303
306
309
327
332
333
335
336
340
351
352
363
368
380
Particles, M: Medium Particles, 1: Large Particles 390
K-2 Particle (or Agglomerate) Sizes Which are Supported
by Upward Air Flow 394
L-l Static Displacement of Fabric Surface at Constant
Differential Pressure, Used Bags 399
M-l Comparative Mass Loading Measurements With All-Glass
Filters and Andersen Impactor (Foundry Dust, HMD =
~1 um) 408
xix
-------
LIST OF TABLES
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Title
Shaker Motion and Bag Suspension Used in Commercial
Equipment
Properties of Woven Fabrics Selected (Manufacturers '
Data)
Properties of Bags Tested in Mechanical Shaking
Systems
Summary of Test Dust Size Properties
Comparative Aerosol Concentrations as Determined by
Various Instrumental Methods
Envelopes of Motion for a Sateen Weave Cotton Bag
Calculated and Measured Average Bag Amplitudes
Sample Test Data Tabulation
Experimental Shaking Conditions
Dust Removal Versus Average Bag Amplitude Measured
and Calculated Test Parameters
Filtration Characteristics of Various Dust/Fabric
Combinations
Collection Efficiency and Effluent Concentrations
for Various Cleaning Regimes
Effluent Concentrations Versus Averaging Period
(Fly Ash; Unnapped Cotton Sateen Bags, 10 IN. X
6 FT.)
Fly Ash Emission Parameters for Different 10 FT. X
6 IN. Fabric Bags
Talc and Silica Emissions From New (< 10^ Shakes)
Cotton and Dacron Bags
Fly Ash Filtration Characteristics for New « 104
Page
q
37
42
43
51
61
72
74
90
100
102
118
135
137
139
142
Shakes) and Well-Used (2 X 107 Shakes) Bags
xxi
-------
LIST OF TABLES (Continued)
No.
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Title
Effect of Extended Shaking on Tensile Properties of
Clean and Dust Laden Cotton Bags, Fly Ash Aerosol
Tensile Measurements on New and Used Fabric Strips
With Instron Tester
Properties of Test Felts
Outlet Concentration and Fraction Penetration for
Various Reservoir Pressures and Direct and Damped
Pulses 2
Talc and Fly Ash Filtration With Dacron Felt at
8.5 ft./min.
Effect of Variations in Filtration Velocity and
Inlet Loading on Fly Ash Emissions With Dacron Felt
Summary of Test Parameters for Fly Ash and Talc
Filtration With Dacron and Wool Felts
Effect of Pulse Duration on Average Filter Resistance
for Various Dusts and Fabrics
Effect of Inlet Concentration and Filtration Velocity
on Average Filter Resistance for Fly Ash and Dacron
Fly Ash Effluents From Pulsed Dacron Bags as a
Function of Pulse Form
Typical Descriptors of Pulsed Bag Motion (Dacron
Felt)
Parameters for Use in Equation (3.2)
Fly Ash Filtration With Mechanical Shaking and/or
Low Pressure Reverse Flow Cleaning
Dust Removal Versus Reverse Flow Velocity, Volume,
Pressure and Duration
Effect of Varying Rate of Pressure Change on Filter
Resistance and Effluent Concentration.
Page
16O
166
189
203
216
217
219
246
249
25 X
259
260
292
297
301
xxii
-------
LIST OF TABLES (Continued)
Title _ ^ Page
32 Effect of Cleaning Frequency on Filter Resistance and
Effluent Concentration With Reverse Flow Cleaning 304
33 Effect of Low Pressure PUlse Duration on Filter
Resistance and Effluent Concentration 305
34 Effect of Pulse Jet and /or Reverse Flow Cleaning on
Filter Resistance and Effluent Concentrations With a
Fly Ash/Dacron Felt System 308
H-l Summary Listing of Data Sheets for Tension/Shaking
Frequency Studies 361
H-2 Summary Listing of Data Sheets for Typical Filtra-
tion and Mechanical Shaking Tests 354
H-3 Summary Listing of Data Sheets for Effluent Size and
Concentration Properties for Various Dusts and
Fabrics 370
1-1 Comparison of Measured and Predicted Lateral Forces 377
M-l Weight Losses for Loaded Lubricated, Petri Dish
Impactor Stages After Repeated Weightings 409
xxiii
-------
NOMENCLATURE
A Shaker arm amplitude
C. Inlet concentration
Df Characteristic collector diameter
E' Instantaneous energy in travelling wave per unit length
E Energy per shaking cycle
F Lateral force normal to shaker arm
F Maximum lateral force
max
G Flexibility modulus
K Specific resistance coefficient
L Length (bag)
M Elastic (tensile) modulus of bag
M" Shear modulus
N Number of individual shakes or particles
P Compressed air manifold pressure
P Average steady state power consumption
RA Shaker arm radius
A
S Effective (initial) filter drag
e
S Terminal filter drag
T Tension
T Initial midpoint bag tension (static) at maximum amplitude
A
T Dynamic tension produced by shaking
d
T Initial tension (static) at top of bag
T Initial average midpoint tension (static)
i,m
T Midpoint bag tension (dynamic)
m
T Pulse duration
p
TV Initial midpoint tension, (static) shaker arm vertical
V Velocity
V Average velocity
V Lateral velocity of shaker arm
ct
V Maximum lateral velocity of shaker arm
amax '
V Initial particle ejection velocity
V Average air velocity at resumption of air flow
K.
V_, Particle terminal velocity
xxiv
-------
NOMENCLATURE (continued)
W Filter bag weight
WR Residual fabric dust holding, weight/unit area
WT Terminal fabric dust holding, weight/unit area
Y Bag amplitude
Y Average bag amplitude
a Acceleration
a Maximum acceleration
m
c. Inlet concentration
cps Cycles/second
d Diameter or bag separation distance
e 2.71828
f Frequency
f Pulse frequency or cleaning frequency
f^ Resonant frequency
2
g Acceleration in gravity field (32.2 ft./sec. ) or gram
gr Grain
h Filter pressure drop or fabric thickness
h Residual filter pressure drop (effective)
H Terminal filter pressure drop
n Exponent
t Time
x Distance (along bag)
a Attenuation parameter (wave energy), dimensionless
6 Exponential damping rate, length
Ap Pressure lose, any filter bed
AT One-half tension excursion at resonance
E
AY Difference in bag amplitude for resonant and nonresonant states
£
C Dust transport effectiveness parameter
9 Phase angle relating force and velocity at shaker arm
A Wave length
X Average wave length
XXV
-------
NOMENCLATURE (continued)
U Viscosity
ym Micrometer
IT 3.14159
p Bag and/or dust mass per unit length
CT Geometric standard deviation
g
T Relaxation time or period
ij) Function symbol
xxvi
-------
CONVERSION FACTORS FOR BRITISH AND METRIC UNITS
To convert from
°F
ft. -
ft.2
ft.3
ft./min. (fpm)
3
ft. /min.
in.
. 2
in.
oz.
oz. /yd.
grains
2
grains/ft.
3
grains/ft.
]b. force
Ib. mass
lb./ft.2
in. H20/ft./min.
To
°C
meters
2
meters
3
meters
centimeters/sec .
3
centimeters /sec.
centimeters
2
centimeters
grams
2
grams/meter
grams
2
grams /meter
grams /meter
dynes
kilograms
2
grams /centimeter
cm. H-0/cm/sec.
Multiply by
| (°F-32)
0.305
0.0929
0.0283
0.508
471.9
2.54
6.45
28.34
33.89
0.0647
0.698
2.288
4.44 x 105
0.454
0.488
5.00
To
centimeters
centimeters^
centimeters-^
meters/sec .
3.,
meters /hr.
meters
meters
grains
grams/centimeter
Newtons
grams
2
grams/meter
2
Newtons/meter /cm/sec.
Multiply by
30.5
929.0
28,300.0
5.08 x 10~3
1.70
2.54 x 10~2
6.45 x 10~4
438.0
3.39 x 10"3
0.44
454.0
4880.0
490.0
x
-------
CHAPTER I
INTRODUCTION
ROLE AND CAPABILITIES OF FABRIC FILTER SYSTEM
Fabric filtration provides a highly effective means of dust control pro-
vided that the filter medium can be cleaned periodically at reasonable
cost without impairing collection efficiency or disturbing the system
gas flow. Although some electrostatic precipitators and wet scrubbers
also provide high efficiency collection, ~ 99 to 99.9 percent in some
applications, the greater collection capabilities of a fabric filter
system should be sought for very hazardous materials such as asbestos,
beryllium and plutonium. On the basis of equivalent inlet concentra-
tions, for example, weight collection efficiencies for fabric filters
range from 99.9 to 99.999+ percent and the effluent dust concentrations
are about 10 to 1000 times lower than those attainable with wet scrub-
bing or electrostatic precipitation.^ In terms of dollar investment,
fabric filter systems represent some 25 percent of the commercial market
and an estimated 105 fabric filter units are treating about 7.5 x 108
ft.3/min. of industrial effluents.1
PROBLEMS IN AEROSOL FILTRATION
Two specific problem areas have prevented more extensive usage of fabric
filter systems. The first relates strictly to the working environment
where high temperatures and/or corrosive materials lead to rapid degra-
dation or at least abbreviated service lives for the filter media. Ad-
ditionally, the inability to maintain highly humid gas streams at
-------
temperatures safely above the dewpoint can also prevent effective ap-
plication of fabric filters. The second problem area is that associated
directly or indirectly with the fabric cleaning process, the principal
reason for conducting the research described in this report.
PROBLEMS IN FABRIC FILTER CLEANING
Specific problem areas related to fabric cleaning, which must be carried
out on an intermittent or continuous basis to provide uninterrupted,
high efficiency gas filtration, are summarized below:
• Failure to remove sufficient dust in cleaning to
maintain acceptable filter resistance.
• Fabric damage and frequent replacement due to excess
or poorly distributed cleaning energy.
• Excessive cleaning energy requirement due to inef-
ficient conversion of energy into mechanical stresses
at the dust-fabric interface.
• Excessive time required for cleaning, requiring long
filter compartment down-times and extra system capacity.
• Improper selection of cleaning method for a given
application and improper selection of fabric for
that cleaning method.
• Improper matching of dust and/or fabric properties
and cleaning method.
Benefits to be derived from improved cleaning should: (1) make filtra-
tion technology feasible for new field applications; (2) reduce the
chance of filter tearing or perforation with the ensuing release of par-
ticulate; and, (3) provide for significant reductions in emissions by
avoiding over-cleaning in some applications.
Good fabric cleaning, with specific reference to the actual cleaning
process used, may be judged by the following yardsticks:
-------
• No significant reduction in collection efficiency
immediately following and during cleaning of filter
elements.
• Minimal average resistance to air flow over an extended
operating period.
• Minimal time required for cleaning.
• Minimal reduction of fabric service life due in any
way to cleaning including application of cleaning
energy, mechanical wear of fabric during cleaning,
or changes in the abrasive or plugging properties
of dust due to cleaning.
• Minimal cleaning energy requirement.
Ideally, the costs to meet the above criteria in conjunction with all
other costs related to the owning and operating of the filter system
should be at a minimum for treating the prescribed volume of gas at the
prescribed efficiency level.
The additional recovery of particulate accomplished by improved cleaning
for many existing installations is estimated to be of the order of 0.1
percent of the amount presently under control (or collected). Unless
the dust has a high intrinsic value, there might be little economic
advantage to the added recovery and probably some cost disadvantage for
increased handling or waste disposal. On the other hand, since most
fabric filters operate in the 99 to 99.9 percent efficiency range, very
significant reductions may be seen in terms of the discharge rate of
particulates to the atmosphere, ~ 5 to 10 times, depending upon the
specific installation.
PROGRAM OBJECTIVES
The objectives of the research program presented in this report were to:
• Determine how the effectiveness and utilization of
fabric filter systems can be increased through a better
understanding and application of filter cleaning methods.
-------
• Design and construct a laboratory pilot plant testing
facility and appropriate instrumental techniques that
provide maximum flexibility for the investigation of
existing and experimental fabric filtration and cleaning
systems.
• Investigate currently used or proposed cleaning methods
to identify and define the basic mechanisms responsible
for dust removal.
• Suggest design and operating parameters for maximum
utilization of existing cleaning methods and any new
or improved techniques that might arise from this study.
SELECTION AND SEQUENCE OF STUDY OF FABRIC CLEANING METHODS
A very brief discussion of the principal methods of filter cleaning and
the rationale for the selection and sequence of study of the fabric
cleaning methods investigated in this program are presented in the fol-
lowing text. More detailed background data, including appropriate back.*
ground and reference material, are provided in the individual chapters
of this report.
The major program effort was directed to a study of mechanical shaking
methods us^d with woven fabrics and reverse pulse-jet systems used with
felted fabrics. It was estimated that some 35 and 22 percent, respec-
tively, of existing filtration equipment are cleaned by the above
methods. Time restrictions dictated that a lesser effort be directed
to low pressure, reverse-flow cleaning approaches which, taken as a
broad class, are used in about 25 percent of existing filter systems.
Although high pressure, pulse-jet systems appear to be increasingly
popular filter cleaning methods, field surveys have indicated that
mechanically shaken systems, because of their basic design simplicity,
will continue to control a large fraction of the particulate emission
potential for some time in the future. More important, properly selecte
-------
greater particle retention than pulse-jet cleaned, felted media in most
applications.
Pulse-jet systems, on the other hand, afford significant advantages in
reduced space occupancy because of their high air-to-cloth ratios and
a high degree of stability in exhaust ventilation rates which is essen-
tial for good source control of dust generating processes. In contrast
to woven fabric filtration systems with mechanical shaking, pulse-jet
systems have also been reported as better adapted for the filtration of
very high dust loadings.
Despite the fact that filtration technology and its related fabric clean-
ing methods have been used for a long time, the design criteria for many
systems depend mainly upon prior field experience and empirical projec-
tions of very limited laboratory data. Although the qualitative aspects
of filter performance derive readily from theory and the cleaning action
of various filter surface restoration methods can be explained satisfac-
torily on a qualitative basis, the existing literature provides very few
quantitative guidelines to aid in the choice, method of application,
and the evaluation of filter cleaning methods.
In that the overall effectiveness of any fabric filter system hinges
upon the operator's ability to clean the media on some periodic basis
so that resistance levels are moderate, the system gas flow relatively
constant, and the particulate removal consistent either with process
needs or emission control regulations, it was believed that a detailed
investigation of fabric filter cleaning methods would represent a sig-
nificant contribution to the state-of-the-art.
The results of the studies on mechanical shaking, high pressure pulse-
jet cleaning and low pressure reverse-flow cleaning are described in the
order named along with a detailed outlining of specific conclusions and
recommendations in Chapters II, III, and IV, respectively. As pointed
out previously the above cleaning techniques have been estimated to
-------
represent about 82 percent of the market. Reverse jet (traveling blow
ring) designs, units cleaned by high frequency mechanical (or sonically
produced) vibrations and miscellaneous combinations of all cleaning
methods typify the remaining collector designs in current usage. Time
priorities did not allow for investigations of all possible cleaning
systems.
Several new instrumental methods were developed exclusively for these
studies that permitted measurements heretofore impossible either in the
the laboratory or in the field. Unfortunately, most devices were
restricted to laboratory use (or would at least be restricted to single
bag collectors) because of the need to integrate sensor and filter ays-
tern components. Details on the fabrication and the operation of special
instrumentation are provided in the appendices. The development of
theoretically and/or empirically derived relationships for describing
overall collector function; i.e., bag motion, dust dislodgment forces
and other aspects of particle and fabric kinetics are also presented in
the appendices to amplify their treatment in the main text.
SUMMARIZED CONCLUSIONS AND RECOMMENDATIONS
Detailed conclusions appear in Chapters II, III and IV and slightly ab-
breviated versions of the conclusions and recommendations evolving fro»
this study have been consolidated in Chapter V. Because of their large
number and diversity, only a brief summation of the principal conclusion^
and recommendations is given here.
Conclusions
Unless indicated to the contrary, the operating and per*
formance parameters cited for the specific dust/fabric
combinations investigated in this program cannot be ap-
plied safely to other dust/fabric systems without sup-
porting data.
Fabric filter performance is best judged by the con-
centration and size properties of the particulate
-------
effluent. There are no simple relationships for
predicting the above properties on the basis of the
concentration and size distribution of the inlet dust
and the clean fabric properties.
• Particle size distributions for filter effluents may
more often be controlled by the agglomerating and
dust release characteristics of the filter fabric
(and method of cleaning) than by the inlet dust
properties.
• Residual and average fabric resistance, dust holding
capacity and dust retention characteristics can be
predicted for specified dust/fabric combinations and
specific cleaning systems.
• Filter effluent concentrations from mechanically shaken
filters are, on the average, some 10 to 100 times lower
than those for high pressure, pulse-jet systems.
• Dust removal by mechanical shaking results mainly from
the acceleration/deceleration action of the bag pro-
duced by the shaker arm. Actual dust separation takes
place when tensile forces exceed the adhesive and
cohesive forces binding the dust to the fabric.
• Shaking frequency, shaking amplitude and duration of the
shaking interval define the performance of mechanically-
shaken, woven fabrics.
• Dust removal by high pressure pulse-jet action also
results from the acceleration/deceleration action
imparted to the fabric and the interaction of tensile
and adhesive/cohesive forces.
• The waveform, the rate of differential pressure change
across the fabric (a function of reservoir pressure, jet
location and the opening time for solenoid values), and
the average differential pressure over the duration of
the pulse determine the performance of felted media
cleaned by high pressure pulse-jet action.
• Aerodynamic forces per se play a very minor role in
dust removal both in high pressure, pulse-jet systems
where reverse cloth velocities seldom exceed 2 ft./sec.
and in low pressure, reverse air cleaning where reverse
velocities are seldom much greater than filtration
velocities. In most instances, bag collapse and/or
flexure caused by flow reversal are the major dust
dislodging forces.
-------
Recommendat ions
Fabric filter systems cleaned by mechanical shaking
should be used whenever highly toxic or harmful par-
ticulates are encountered. Insofar as possible, a
woven fabric such as sateen weave cotton or its
synthetic or mineral equivalent in terms of filtra-
tion properties should be selected.
Woven fabric filters should always be maintained at
sufficient tension to guarantee transmission of
shaking energy over the complete length of the bag.
No attempt should be made to use the shaking param-
eters suggested by laboratory studies in the field
until the structural integrity of the bag enclosure
and supporting members are determined.
The use of felted fabrics cleaned by high pressure,
pulse-jet air is recommended when the advantages of
reduced space requirements, minimal air flow varia-
tions at critical exhaust points, and adaptability
to high inlet dust loadings are important. The
above features must be weighted, however, against
higher outlet loadings and higher power needs.
Net power requirements for pulse-jet systems may be
lower at reduced compressed air pressure; e.g., 40
versus 100 psig despite greater fabric resistance
while significant reductions (~ 5 times) are attain-
able in effluent concentrations.
Regardless of reservoir pressure levels, a pulse-jet
unit must employ fast acting valves to produce rapid
changes in system pressure levels > 1,500 in.
water/sec.
It is recommended that pilot plant and field engi-
neering measurements be performed to determine the
appropriate "K" values for dust/fabric combinations
encountered in industrial practice.
Fundamental studies should be performed to determine
which particle and fiber (fabric) parameters, taken
singly or in combination, are most important in
determining characteristic "K" and residual resistance
values for a specified dust/fabric combination in con-
junction with a specified fabric cleaning technique.
8
-------
Carefully controlled laboratory pilot studies should
be performed to define accurately the role of filtra-
tion velocity, dust loading, method of jet pulse
admission, and bag dimensions (particularly length)
with pulse-jet systems.
Rigorous tests should be performed, initially on a
bench scale and later on a pilot scale, to determine
how the following factors control or may be utilized
to improve the filterability of various dusts: electro-
static charge, its presence or absence on particles
and/or fibers; particle size distribution with shape
factor constant; humidity control; particle size versus
fabric pore size; surface deposition versus interstitial
deposition; and the use of conditioning methods such as
induced agglomeration.
REFERENCES
1. Billings, C. E. and J. E. Wilder. Handbook of Fabric Filter Tech-
nology. GCA Corporation, Bedford, Massachusetts, Contract No.
CPA-22-69-38, Available From National Technical Information Service,
U.S. Department of Commerce, Springfield, Virginia 22151, Document
No. PB 200-648. December 1970.
-------
CHAPTER II
MECHANICAL SHAKING STUDY
OBJECTIVES AND APPROACH
Although shaking has been used for many years to remove accumulated dust
from fabric filters, no quantitative analyses of the physical mechanisms
controlling the cleaning process were published until about 1962. At
that time, Walsh and Spaite reported a series of experiments that
related bag cleaning to the peak acceleration of the shaken end of the
bag. Their tests showed distinct, nonlinear relationships between peak
acceleration and the dust removed by shaking and the resultant pressure
drop across the fabric. No data were presented on collection efficien-
cies, changes in fabric properties and expected bag service lives, nor
was the acceleration of the shaker arm related to the motion of other
parts of the bag. The optimization of the performance of a filter sys-
tem, however, may be presumed to require a beforehand knowledge of all
these factors.
In the present study attention was directed first to extending the rela-
tionship between shaker arm motion and cleaning and resistance charac-
teristics to the overall motion of fabric bags. Secondly, the effects
of shaking on dust collection efficiency, bag service life and the power
needed to provide effective cleaning were investigated. This was accom-
plished by first identifying a mechanism whereby bag acceleration might
reasonably promote dust removal, followed by a detailed study of the
motion of clean bags. The latter step was necessary to avoid the com-
plication of variable bag and system properties arising from
11
-------
progressive loss of dust. The same motion mechanics were then applied to
dust-laden bags with respect to the resulting dust removal and pressure
differentials. Other tests were also conducted to determine collection
efficiency and shaker power consumption as described in the following
sections.
In order to represent accurately the force and motion patterns of
typical shaken filter bags, apparatus was designed and constructed that
*
would simulate typical fabric filter equipment. The apparatus ulti-
lized cylindrical filter bags suspended from a horizontally-aligned
rocker shaft. For convenience in manipulation of instrumentation ex-
ternal to the bags, the filter was operated on a positive pressure basis
(fan upstream).
The ranges and magnitudes of the operating variables were selected so
that test results might be applied to field filtration systems with
minimal extrapolation. Tests were performed on a variety of woven
fabric bags, each of which depicted a commonly used size or fabric type.
A prime objective of this study was to investigate the underlying
physical processes that contribute to either good or poor field perfor-
mance in existing fabric filter systems. Although areas for improve-
ment attainable by redesign or operational changes in existing systems
were not overlooked, most tests were aimed at the need to define first
the cleaning processes now in common usage before initiating any
development work on new or novel designs.
The apparatus and various instruments and techniques for measuring bag
tension, bag displacement, dus|; concentration, and other variables,
which were developed during the program are described under APPARATUS,
TECHNIQUES AND MATERIALS, pages 36 to 63, and in more detail in Appen-
dices A, B and C.
12
-------
BACKGROUND AND THEORY
The qualitative aspects of dust removal by mechanical shaking are
fairly well recognized. The harmonic motion that is imparted to the
dust-laden bag accelerates and decelerates the fabric-dust system pro-
ducing tensile and shearing forces at dust-particle and dust-fabric
interfaces. When these forces exceed local cohesive and adhesive forces,
the dust is dislodged. The magnitude of the latter forces depends upon
several factors; e.g., physical properties of the dust and/cr fabric
system, gas velocity, dust aging and compaction, and temperature and
humidity of the gas.
The first constructive analyses of mechanical shaking processes were
reported by Walsh and Spaite in 1962. The tests performed on a single
bag system established practical limits for the number of shakes needed
to remove fly ash from sateen weave cotton fabrics. Measurements made
at various shaking frequencies, amplitudes and shaking times also
indicated that dust removal could be related to the peak acceleration
of the shaken end of the bag. No data were reported on collection
efficiency, projected bag service life, and the kinetics of the bag
motion. Thus, the above findings, although providing useful guide-
lines for the present study, did not furnish the necessary data inputs
to establish optimum shaking conditions.
Aside from the work of Walsh and Spaite little Additional understand-
ing of mechanical shaking processes has been achieved up to the present
time. Investigations have been made, however, on adhesive phenomena,
collection efficiency, fabric design factors, and various other filter
operating characteristics that contribute to an understanding of
cleaning by mechanical shaking. This work is summarized in the follow-
ing sections.
13
-------
Adhesion and Removal Mechanisms Including Acceleration
All fabric shaking involves some combination of:
• Stress in a direction normal to the dust-fabric
interface (tension).
• Stress directed parallel to the interface (shear).
• Stresses developed during warping, bending or flexing
of the fabric surfaces.
These stress conditions are illustrated in Figure 1. In addition to
shaking per se, there are two other general methods of cleaning, the
first relying on air pressure or viscous drag to remove dust and the
second relying on direct physical contact as with vibration, manual
rapping or sonic and shock cleaning. In the later instances, contact
stresses such as those produced by mechanical shaking appear respon-
sible for cleaning. While Figure 1 depicts the types of stresses en-
countered in all cleaning systems, the intensity and/or repetition
factors may differ between those of simple mechanical shaking and the
other cleaning techniques.
According to theory, adhesive or cohesive bonds will fail whenever or
wherever the local stresses exceed the bonding strength. Examination
of the stress models illustrated in Figure 1 shows that of the four,
only the first, (a), involves stresses directed away from the fabric.
Thus, although the other three mechanisms may loosen the deposited dust
and prepare it for separation, only when the tensile force produced by
acceleration directly opposes the adhesive and/or cohesive forces will
there be a true separation of the dust from the fabric.
Without a high-speed microscopic study of the dust separation process
(something beyond the scope of the present study) there is no way to
describe precisely the dust removal mechanisms. Two conceptual models
of separation by acceleration are indicated in Figures 2a and 2b. Bond
14
-------
FABRIC
OUST
DEPOSIT
(o) NORMAL (TENSILE) STRESS
DUE TO ACCELERATION
(b) PARALLEL (SHEAR) STRESS
DUE TO PLANAR ACCELERATION
COMPRESSION
TENSION
(C) STRESSES DUE TO FABRIC
FLEXING
(d) STRESS DUE TO PLANAR YAW,
OR YARN TO YARN SHEAR
Figure 1. Potential dust fabric stresses developed during
shake cleaning
15
-------
u
HYPOTHETICAL BOND TENSILE
STRENGTH
INTERNAL TENSILE
STRESS AT TIME t
OUST DEPOSIT THICKNESS
FABRIC SURFACE
AND
DUST DEPOSIT
INTERFACE
DUST
DEPOSIT
SURFACE
VI
w
HYPOTHETICAL BOND TENSILE
STRENGTH
INTERNAL TENSILE
--_ _/ STRESS AT TIME t*
"---
t
•DOST DEPOSIT THICKNESS-
FABRIC SURFACE
AND
DUST DEPOSIT
INTERFACE
DUST
DEPOSIT
SURFACE
Figure 2a. Concepts of cleaning via acceleration
(a) High adhesive/cohesive ratio
Figure 2b. Concepts of cleaning via acceleration
(b) Low adhesive/cohesive ratio
-------
strengths are characterized hypothetically by the solid curves. In the
first cage, the adhesion of the dust to the fabric surface is higher
than the cohesive strength within the dust deposit.
As one moves from the interface to the surface of the deposit, the
cohesive strength is assumed to diminish because the aerodynamic
stresses tending to compact the deposit also diminish. In order for
separation to occur, it is necessary that the tensile stress caused by
acceleration must equal or exceed the bond strength at some specified
depth in the deposit. When this occurs a flake or layer of dust is
removed down to that depth. Beginning at time zero, successive in-
stantaneous tensile stresses are represented by the dotted lines in
Figure 2a. As acceleration begins, the tensile stresses rise from a
zero value at the surface of the deposit to a value at a given depth
in the deposit that is proportional to the mass per unit area above
that depth. At successive times during each cycle, t., t«, t_, ... the
acceleration of the fabric rises until at time t. the tensile stress
4
finally equals the tensile strength at depth (hj). At this time, a dust
layer of depth (h,-h.) is detached, leaving the remainder of the deposit
unchanged. Immediately the tensile stress at the new surface drops to
zero and the stresses within the remaining cake become proportional to
the remaining mass (line t,' ). If the fabric acceleration continues
to increase, another layer of thickness (h..-h2) will dislodge at time
t_. Whatever depth remains at the end of the shaking period becomes
the new residual deposit along with the dust within the fabric structure.
After many consecutive cleanings, the residual dust deposit is com-
posed of those particles most difficult to remove. Thus, this material
may adhere more strongly than any fresh surface* deposits. This it? the
reason for showing relatively strong residual bonds in Figure 2a.
Although the bonding strength decreases as particle diameter decreases,
the particle mass (and hence the tensile stresses produced by acceler-
ation) decrease at an even more rapid rate. As the finer particles
are more difficult to dislodge than the larger ones, they probably
17
-------
tend to concentrate in the residual deposit. Their small size would
tend to increase the flow resistance of the filter or even blind it
in extreme cases. Filter system operators often report a gradual in-
crease of filter operating pressure over a period of several hours or
days operation with a specified cleaning cycle. They also report that
temporarily increasing the cleaning intensity usually lowers the pres-
sure drop within a short period such that the regular cleaning cycle
can be resumed. It is postulated that this increase is due to a well-
bonded layer of selectively retained fine particles. It appears that
the residual layer can be reduced by periodically resorting to a more
intense cleaning cycle.
In contrast to the strong interfacial adhesive forces shown in Fig-
ure 2a, the opposite situation is described in Figure 2b. In the
latter case, it is suggested that the cohesive, particle-to-particle
boncing exceeds the adhesive forces at the particle-fabric interface.
The dust detaches from the fabric more cleanly, leaving a negligible
residual deposit. Lower separation forces are required and larger
dust agglomerates are removed. Smooth-fibered, shallow-surface fab-
rics such as glass and certain synthetics, if suitably woven, are be-
lieved to perform in the above manner. The results of the present
study have shown a much lower residual dust weight on Dacron than on
cotton fabrics, probably because of the mechanism postulated in Fig-
ure 2b. Heavily napped fabrics, felts, or fabrics with deeply woven
surfaces may be expected to accumulate substantial deposits of dust
beneath the surface. The dust may be physically interlocked with the
fibers and consequently adhere strongly irrespective of particle size.
The overall effect should be similar to that indicated in Figure Za.
The size and range of pore sizes probably account for the differences
in behavior between deeply woven and smooth shallow fabric surfaces.
The factors that determine cohesion and adhesion have been shown in
studies of Interfacial phenomena to include the molecular character-
istics of the materials as well as the properties of adsorbed or
18
-------
absorbed surface films. In practice, the surfaces of most materials
are contaminated by gas molecules, liquids, or solid particles. For
2
example, Durham and Harrington found a small effect of humidity on the
effective residual pressure differential of a filter that might be ex-
plained by a change in the surface characteristics of either the fiber
or the dust particles. Other possible factors are the porosity of the
deposit, its degree of compaction, and electrostatic effects.
In those cases where high surface adhesion predominates, the depth
to which dust is removed may depend upon local variations in any one
of the several factors that appear to determine removal depth. In-
sufficient tensile stress to produce removal in one cleaning cycle
will result in a heavier than usual deposit at that location after
the next filtration cycle. Ihe added dust mass will tend to increase
tensile stress, thus increasing the probability of removal at that
location in subsequent cleaning. It appears probable, therefore, that
in successive cleaning cycles, the removal of dust at a given location
on the filter bag may vary over a considerable range from one cycle
to another. If so, one would predict a non-uniform deposit over the
filter surface at any given time. Just such a spotty effect was ob-
served in this study when a light bulb was placed inside a filter bag
and the light transmission viewed in a darkened room, as shown in Fig-
ore 3. Some areas of the fabric expected to have little normal accel-
eration, however, were also well cleaned, particularly the pleats around
the top of the bag. The latter removal probably was due to Type c and
Type d stresses as shown in Figure 1. All the stresses in Figure 1 con-
tribute something to loosening the dust, even though acceleration
predominates.
Walsh and Spaite in their studies of shaker arm acceleration found
2
that about 1 g (32.2 ft./sec. ) was required to cause discernible dust
removal and 2 g's was sufficient to attain significant cleaning. In
view of the fact that the dust is deposited in a gravitational field
of 1 S» ^ *-s understandable that accelerations should exceed 1 g to
19
-------
Figure 3. Internally illuminated 10-foot by 6-inch diameter
cotton bag, after cleaning (photo approximately
4 feet from lower end). Shake direction is normal
to the page.
20
-------
remove dust deposits. Walsh and Spatte found that accelerations be-
low 1 g lowered the residual drag, probably because the dust deposit
was made more permeable by fabric and dust flexing.
In this program, the residual fly ash dust deposits on cotton fabrics
2
were typically 300 grains/ft. following an extended shaking with a
10 g maximum acceleration. Based upon bag kinetics and observed dust
removal, average adhesive forces of the order of 200 dynes/cm.2 were
estimated. No directly comparable data could be found in the liter-
3
ature. However, Zimon has reported effective adhesive strengths of
the order of 100 to 300 dynes/cm, for layers of various particle
materials (under 50 microns diameter) bonded to various substrate
materials. Thus, the postulated mechanism of dust removal by accel-
eration appears acceptable as far as the magnitude of separation forces
are concerned.
.Bag Motion flieory
In the preceding section, fabric acceleration was established as one
important factor in dust removal by mechanical shaking. Although
Walsh and Spaite found a correlation between the acceleration of the
shaker arm, per se, and dust removal, their studies did not indicate
the actual acceleration undergone by the fabric. A theory has been
developed in this study, based on classical mechanics and laboratory
measurements, that provides reasonable definition of fabric motion.
Although this theory is actually a result of the study, it is summar-
ized here to establish a framework for the experimental results re-
ported later.
Basic Concepts - Periodic lateral motion applied to one end of a
flexible tubular body (or filter bag) produces lateral waves that
travel downward along the bag with a characteristic translational
velocity. Wave amplitude is progressively diminished, depending upon
the system damping properties. At the bottom of the bag, which is
21
-------
rigtdly fastened, there may be further attenuation of the wave while
the unattenuated portions of the incident waves are reflected and ex-
perience further damping as they travel back up the bag. The reflec-
ted wave adds to the oncoming downward wave, producing a standing wave
marked by a. characteristic nodal pattern along the bag. At the anti-
nodes, the bag undergoes a large, sometimes violent lateral oscilla-
tion. At the nodal positions, the bag undergoes a slight oscillation,
in contrast to remaining stationary as it should for the "ideal
vibrating string." Figure 4a shows the instantaneous sinusoidal
shape of the shaken bag. Figure 4b, a time exposure, indicates the
maximum lateral displacements of the same bag while generating a stand-
ing wave pattern. Although standing waves are present at all shaking
frequencies, the loop pattern is clearly defined only at certain reso-
nant frequencies when the reflected wave has sufficient strength to
return for constructive combination with its successor wave. The con-
structive combination of return and successor waves at resonant fre-
quencies depends on the phase relationship as determined by bag length
and wave velocity. At other frequencies, destructive interference of
these waves diminishes the standing wave.
Because the bag assumes a sinusoidal form while it is shaking, it must
be stretched by an amount necessary to produce the curvature. The
added stretching causes more tension in the bag, depending upon its
elastic modulus. This added tension, in conjunction with whatever
tension the bag displayed while stationary, may exceed by several times
that produced by its weight alone. Despite this tension increase and
the variations in stiffness and damping properties, the shaken bag be-
haves much like t:he classical ideal string. The main effect of dust
jr»
Ideal String: Uniform mass and tension at all points and all times;
zero stiffness; zero damping-.
22
-------
(a)
(b)
(•) Exposure time 1/120 ••<:.
(b) Exposure time 2 sec,
Fii ure 4. Appearance of shaking sateen weave cotton bag 10-feet long
by 6-inch diameter (4 cps, 2-inch amplitude, 7.4 pound
•baking tension)
23
-------
accumulation on the bag is to increase its weight per unit length, a
term used in the classical wave equations. The dust appears to exert
only minor effects on the other bag properties such as stretching,
damping and stiffness.
When a stationary bag is first shaken, energy is required to initiate
the shaking action. Once the motion is established, a fraction of the
energy is retained in the bag as kinetic and/or potential. All bags
undergo some degree of damping, however, thus requiring a continual
input of shaking energy to maintain a steady motion. The mechanical
power required to maintain steady shaking is typically a few watts.
The damping loss is mainly due to air drag and to friction between
fiber, yarn, and dust particles. If the dust projected from the bag
during shaking has appreciable kinetic energy, further make-up energy
is required.
Waves and Wave Velocity - In an ideal string (a uniform density cylin-
drical structure with its length much larger than its diameter) lateral
waves travel axially along the string with a velocity given by
JT
I P
(2.1)
4
where T is the tension and P is the mass per unit length. To a first
approximation, waves traveling in a shaking bag behave the same way,
allowing for the qualification set forth in Appendix D. If the wave
travels down the string or bag of length L and returns to its origin
while the shaker arm has made exactly N cycles, the returning wave wiU
coincide (and reinforce) the input wave to produce a resonant state.
In this case, the resonant frequency is represented approximately as
24
-------
A correction factor amounting to about plus 10 percent is developed in
Appendix D to correct empirically for a stiffness factor not encoun-
tered with the ideal string. Additionally, the weight of the bag pro-
duces a variation in tension from top to bottom that is ignored with
the ideal string. It is shown later that if the average tension half
way down the bag is used in Equation (2.2) the calculated frequency
usually compares well with the observed resonant frequency.
Tension Changes - The total tension in a shaking bag is the sum of the
initial static tension in the bag prior to shaking, T. , and an addi-
tional amount caused by the shaking called the dynamic tension, T , .
The initial tension, T., is the sum of the weights of bag and dust,
the clamping tension (if any) applied statically to the bottom of the
bag, plus any stretching due to deflection of the shaker arm. At a
given point along the length of the bag, the gross weight of material
below that point must be supported; thus T has its maximum value at
the top of the bag. For most purposes, the average initial tension
along the bag at its midpoint T^ m» may be used.
Bag tension also varies with the position of the shaker arm Figure 5.
The midpoint tension, TV> increases from the vertical position of the
shaker arm to a value T. when the arm is swung outward a distance A
A
from the vertical (or zero amplitude) position. At static conditions
or very slow shaking frequencies, the tension difference is given by
(2.3)
where R is the radius of the shaker arm, L is the bag length and M is
the elastic modulus of the bag (see Bag Modulus, pages 46 to 49). Con-
sequently, the time averaged initial tension is
when the bag is much longer than the shaker arm.
25
-------
h «A2/2RA whtn hR«R
SHAKER ARM
LENGTH
R*
5. Variation of bag tension with position of shaker arm
26
-------
Normal shaking takes place at 5 to 10 cycles per second, a speed at
which the wave shape of the bag requires that it be stretched more
than at very low shaking frequencies. This elongation produces an ad-
ditional tension, T,, that depends upon both the amplitude and the
length of the waves produced in the bag. In Appendix E, the formula
for estimating T, is developed
2
Td = ML(^) (2.5)
where X is the wavelength and Y is wave amplitude. Equation (2.5) in-
dicates that for a fixed bag amplitude, Y, the dynamic tension varies
inversely with the wavelength (and directly with shaking frequency).
The proper tension value to use in Equation (2.2) is the one that re-
flects the combined effects of bag weight, applied tension, shaker arm
position, and wave shape; i.e.,
T = T, + T. (2.6)
m i,m d v^.w;
where the right hand terms are defined as in Equations (2.4) and (2.5)
respectively.
The average wavelength of the shaking bag can be estimated from Equa-
tion (2.1) as
*-?-? *T (2'7)
By combining Equations (2.5), (2.6), and (2.7), the following expres-
sion can be derived for calculating average bag amplitude, Y.
In Equation (2.8), bag motion is related to three basic bag properties
(tensile modulus, length and mass per unit length) and to three param-
eters depicting the shaking conditions (shaking frequency, initial ten-
sion at the bag midpoint, and average shaking tension at the midpoint).
27
-------
All variables cited in Equation (2.8) are readily determined without
special measuring techniques except for shaking tension. During the
experimental phase of this study, however, methods for calculating
shaking tension in terms of readily measured system variables were
developed. Therefore, Equation (2.8) can be used to predict bag shak-
ing amplitude so that dust removal and related filter performance
criteria can be estimated.
Damping and Reflection - According to general theories of undamped
oscillations, bag amplitude at resonant frequencies should become ex-
tremely large in comparison to the driving motion. Since the actual
oscillations in the shaking bag seldom exceed the shaker arm amplitude,
there must be considerable damping. In fact, the appearance of the
wave when the shaker is suddenly stopped indicates that the wave is
only slightly reflected from the lower cuff fastening. This must be
due either to heavy damping of the downward wave, or to energy absorp-
tion in the reflection process, or both. When a rather slack bag is
shaken the waves tend to disappear before reaching the bottom of the
bag, particularly at the higher frequencies and lower amplitudes. It
is assumed, therefore, that the degree of damping and reflection in a
shaken bag can be responsible for the extremes of no motion at all in
parts of the fabric to violent motion in other cases. Thus, it
appears that next to shaker amplitude, damping and reflection are the
most important considerations in determining bag amplitude.
Closely associated with damping is bag stiffness that tends to increase
the bag's wave velocity above that of an ideal string. Damping,
reflection, and stiffness factors are very complex because they depend
on the cross-sectional contour of the bag. As the bag shakes, its
cross section changes from circular to elliptical, the amount of
flattening depending on tension and other factors. The flatter the
cross section, the Less the stiffness and probably the lower the
damping rate.
28
-------
In theory, one might develop a mathematical or computer model of a
shaking bag that would predict damping, reflection, and stiffness.
This model would have to include not only the wave mechanics cited
above, but also the flexural and shear stiffnesses of the fabric-dust
combination. The model would consist of a thin shell cylinder, with
significant adjustments to describe the pleated top and shape changes
found with shaken bags. Because of the lack of accuracy in defining
major variables and the expected complexity of mathematical treatment,
such a sophisticated approach was not considered worthwhile.
Instead, useful results were obtained by an alternative empirical
approach in which minor correction factors were added to the mechanics
outlined previously, Equations (2.1) through (2.3). Despite the acknowl-
edged simplicity of these relationships, they appear to provide a
strong predictive capability.
Bag Acceleration - Based upon measurements performed during this Study
the oscillation of a shaking bag was determined to be essentially a
simple harmonic motion. In such motion, a given point on the bag
attains a maximum acceleration given by
am
where f is the frequency of oscillation and Y is the local amplitude.
Amplitude is defined herein as one-half the peak-to-peak displacement.
By stroboscopic viewing, it was determined that all points on the bag
move with essentially the same frequency as that of the shaker arm
driving the bag. As a wave of amplitude Y moves along the bag,
successive points on the bag undergo a maximum acceleration given
approximately by the above formula. Consequently, the problem of
determining the local bag acceleration is reduced to one of determin-
ing the amplitude of motion along the bag. In order to determine the
29
-------
average acceleration, which is presumed to control average dust
removal, it is only necessary to establish the average bag amplitude.
One can neglect, therefore, the variation in amplitude along the bag
due to damping, changes in tension, and variations in dust deposit
weight. In a later section, results will be presented that show that
averaged bag acceleration, as computed from tension data, can be cor-
related with dust removal.
Shaking Energy and Power
Based upon many commercial shaking systems, it appears that a one
horsepower motor can shake simultaneously about 200 filter bags. Thus,
the power required for a single bag should be of the order of three to
five watts. The chief consideration with respect to the power require-
ment is not its cost, but the effect the dissipated energy may have on
the service life of the bag and on the removal of dust. Damping in
the bag is associated with dissipation of mechanical energy through
friction, air current generation, and extraneous random motion.
To maintain steady shaking, there must be constant and sufficient
energy replenishment at the driven end of the bag to transmit the
motion to the bottom of the bag without excessive damping. The
instantaneous energy in a lateral wave traveling in a flexible string
is defined by classical wave mechanics as
E' = 2it2 f2 Y2 p (2.10)
per unit length of string. The motion in a shaking bag, however, is
complicated by both damping and a reflection factor, such that it is
difficult to establish from a theoretical standpoint the precise
energy density at any point along the bag. Nevertheless, Equation (2.
is useful for estimating the energy flow into the bag under the tacit
assumption that most of the motion near the top of the bag is due to
30
-------
the downward wave. The above approach predicts a bag power consump-
tion that is directly proportional to the squares of bag amplitude and
shaking frequency and to the weight of the bag. Experimental results
presented later in this report appear to substantiate the above
relationships.
Power consumption can be measured at the top of the bag where the
instantaneous energy flow through the linkage point is determined by
the vector product of force and velocity. In the present case, the
energy transmission rate is the product of the lateral velocity, Vfl, of
the suspension point of the bag and the lateral force, F, at this
point taken at right angles to the shaker arm. The instantaneous
product of arm lateral force and arm velocity, averaged over a full
cycle, represents the energy input per cycle, from which steady state
power consumption can be computed:
P - (Ecyc) (f) = £ / F Va dt (2.ii)
Alternatively, since both lateral force and velocity are periodic and
have the same frequency, their maxima can be used:
P - (F V ) - (2.12)
max amax' 2 v '
where cos 0 expresses the phase relationship and is analogous to a
power factor in electric power transmission termission terminology.
The lateral forces discussed later in this report can be measured
directly or, alternatively, they can be estimated from shaking tension
data as given by the theory developed in Appendix I. By using the
latter formulas in conjunction with Equation (2.12) for computing power
input to the shaking bag, it is shown in a subsequent section of this
report, pages 80 to 86, that power consumption can be predicted with
fair reliability.
31
-------
Bag Life
Limited field data indicate that filter bags may last from a few weeks
or less to 13 years or longer, with one survey showing an average life
Q
of about 1 year. Bag failure may result from abrasion, excessive tem-
perature or burning particles, fiber degradation from acid or chemical
attack, mechanical problems such as seam failures or tears, and blinding
or plugging. Quatititation of the above failure conditions is very poor.
It is sometimes claimed that grit, having penetrated a fabric, can
abrade fibers during mechanical shaking to the point that the yarns are
severed. No good quantitative data are available, however, to evaluate
this effect. Glass fabrics are known to be seriously damaged by mechan-
ical shaking. It appears that nicks or scratches produced by abrasion
9 10
increase fiber breakage during flexure. ' Ihe breakdown of glass
fiber fabrics is a special case, however, not typical of filter fabrics
in general. In the present study, fabric life was measured by the
number of individual shakes given the bags during the testing period,
up to 2 x 10 .
At the outset of the study there was no background evidence or theory
that related the concurrent mechanical effects of tension, flexure,
shear, and dust to fabric deterioration in filter systems cleaned by
mechanical shaking. For this reason, a portion of this study was
devoted to the assessment of this problem.
Collection Efficiency
An information survey has shown that the relationship between mechan-
ical shaking and collection efficiency has not been examined in any
Q
great detail. Although numerous investigations have shown a percep-
tible correlation between efficiency and pressure differential across
the fabric, (and weight of dust deposited on the fabric) these inves-
tigations generally embraced a narrow experimental range, ttie results
have not been summarized in such a way that they can be applied to
32
-------
Q
other situations. As noted above, the relationship between mechan-
ical shaking and pressure differential has only been partially studied,
chiefly by Walsh and Spaite.
The fractional particle size efficiencies of various filter systems
are even less clearly defined. In 1961, Whitby and Lundgren studied
the relationship between cleaning intensity and efficiency of particle
collection for diameters between 0.06 and 3 micrometers. They loaded a
fabric filter with three standard test dusts, cleaned the fabric with
a varying number of shakes, and then measured the effectiveness of
the partially cleaned filter in collecting uranine dye particles of
known diameter. Their measurements showed, first, that collection
efficiency was reduced by additional cleaning but at a decreasing rate
and, second, that'collection efficiency was practically constant for
particles less than 0.3 micrometers, and increased steadily for all
larger sizes. Collection efficiencies from about 70 to 99.9 percent
were reported.
A substantial amount of data relating efficiency to various fabrics,
humidities, dust types, inlet concentrations and system capacity has
8 12
been reported. ' Draemel recently described an investigation of
several fabric design parameters and related collection efficiencies,
13
including analyses of filter pore geometry. While all these
investigations suggest that pore dimensions play an important role in
determining collection efficiency, the effects of cleaning methods on
efficiency are not described by these investigations. {Therefore it
appeared at the outset of the present study that a considerable effort
would be necessary to relate, for example, the shaker arm motion,
through several cause and effect steps, to particle collection.
Applied filtration technology is based upon two distinct particulate
collector designs. The first involves the use of woven fabrics with
relatively flat, shallow surfaces upon which a closely compacted or
sealing layer of dust accumulates. Subsequent filtration is primarily
33
-------
a sieving process that is usually very efficient as the aerosolized
and filter bed particles are the same size. On the other hand, napped
fabrics and felted materials present deep, open surfaces that may not
become completely sealed. The widely spaced fibers present a statis-
tical barrier to airborne particle penetration, thus gradually reducing
particle concentration as the bed depth increases. This deep-bed or
mat filtration concept may involve any of the classical collection
mechanisms - impaction, interception, or diffusion as well as sieving
in locations where the bed has become heavily loaded. Since few
commercial filter systems operate completely either as sieves or
deep-bed filters several collecting mechanisms must be considered in
explaining or predicting their performance.
Efficiency depends on several factors as shown in the following
Q
tabulation.
Dust Properties
Size - particles between 0.1 and 1.0 micrometers
diameter may be more difficult to capture;
Seepage characteristics - small, spherical
solid particles tend to escape;
Inlet dust concentration - the deposit
is likely to seal over sooner at high
concentrations.
Fabric Properties
Surface depth - shallow surfaces form a
sealant dust cake sooner than napped
surfaces;
Weave tightness - fabrics with high
permeabilities, when clean, show lower
efficiencies. Also monofilament yarns,
without fibrils protruding into the
yarn interstices, show lower efficiencies
than more "fuzzy" staple yarns having
similar interstitial spacing;
Electrostatics - known to affect effi-
ciency.
34
-------
Dust Cake Properties
Residual weight - the heavier the residual
loading the sooner the filter is apt to
seal over.
Residual particle size - the smaller the
base particles, the smaller (and fewer)
the particles likely to escape.
Air Properties
Humidity - with some dusts and fabrics,
60 percent RH is much more effective than
20 percent RH.
Operational Variables
Velocity - increased velocity usually gives
lower efficiency, but this can be reversed
depending on the collection mechanisms.
Pressure - probably not a factor except
that increase of pressure after part of
the dust cake has formed can fracture
it and greatly reduce efficiency until
the cake reseals.
Cleaning - relatively unstudied (see
following discussion).
A stabilized residual deposit often has the appearance of close-packed
tufts or aggregates. Between these aggregates lie pores or slits lead-
ing to much more permeable regions of the fabric beneath the residual
layer. As filtration begins, some particles escape through the pores
while others attach to and enlarge the aggregated structures (and re-
duce the pore dimensions). As the pores become blocked, further pas-
sage of air and particles through these regions is greatly reduced
because of the considerable loss of pressure. The flow resistance of
the filter continues to increase due to dust deposition on the surface
of the filter.
The usual process of cleaning consists of removing most of the fresh
(external) relatively loose deposit. The older, relatively firmly
bonded (internal) residual deposit often remains. Some areas of the
35
-------
fabric are cleaned more thoroughly than others and the degree of clean-
ing in any one location may vary from one operational cycle to
another.
Most efficiency data reported in the literature are based on lengthy
sampling periods (hours to days) and conventional gravimetric methods.
Although these measurements provide a practical estimate of the aver-
age emission rate (weight basis) they do not indicate the variations
in effluent concentration that usually take place over individual fil-
tering cycles. The deviations from average efficiency are highly
significant with respect to understanding the mechanics of the fabric
cleaning and filtration processes.
APPARATUS, TECHNIQUES AND MATERIALS
The requirement to test on a realistic dimensional scale led to the
design and construction of an experimental fabric filter system ac-
comodating bags 10 feet long by 6 inches diameter. This filtration
equipment and the testing facilities are described in Appendices A and
The lower end of each bag was clamped to a thimble through which dusty
air entered the bag during the filtration cycle. The upper end of each
bag was fastened to an oscillating shaker arm that imparted a vigorous
shaking motion to the bag during the cleaning interval. The motion of
the bag and the forces in the bag during filtration and cleaning were
determined as well as the dust concantrations entering and leaving the
system. Several types of fabrics and dusts were tested.
Shaker Apparatus
Prior to designing bag shaking apparatus, a survey was made of the dif-
ferent methods of shaking used in commercial equipment, the results of
which are summarized in Table 1. The travel of a bag(s) suspension
point may follow nearly horizontal or vertical paths as well as several
intermediate combinations with amplitudes up to 2.5 inches. The motion
36
-------
Table 1. SHAKER MOTION AND BAG SUSPENSION USED IN COMMERCIAL EQUIPMENT
Manufacturer
American Air Filter
Buffalo Forge
Cox
puller Dracco
Fuller Dracco
Duity Dust less
Johnson March
Pangborn
R««»
Research Cottrell
Ruenelln
Seversky
Tailor
Vheelabrator
Number of
collectors*
33
12
5
60
60
5
1
90
5
60
3
1
1
40
Shaker •otionb
Radius
(inches)
3
3
1
12
3
7
6
4
4
7
10
7
7
Estimated
mplitude
(inches)
1 V
1.5 V
0.4 H
H
1.5 V
0.75 H
0.5 H
2.5
0.75 V
1.3 H
2 H
--
H
2.5
Sketch
J— 1
M
~O~
~H
^->
*-"
{—5
*_*
^-^
Bag suspension
Cap
X
X
X
X
Grommet
X
X
X
Buckle
X
X
c
d
X
X
d
Other
c
d
d
*Based on 1969 sales data and type of shaking motion specified.
''Based upon manufacturers data; V = nominally vertical motion; H
nominally horizontal motion.
CCltp used with buckle and strap.
37
-------
is usually produced by a pivoting radial arm, up to 12 inches long, that
describes a progressively flatter arc as the arm length increases and
the amplitude decreases.
The 6-inch radius arm selected for the present study afforded a nearly
horizontal motion with but a slight upward displacement at the points
of maximum amplitude. The shaker arm was attached to a horizontal
shaft, Figure 6, that was given a rapid alternating rotation by an
eccentric drive. By varying the speed of a 1 hp dc motor, shaking
frequencies from 0 to 20 cps were readily attainable. Careful balancing
permited operation up to 80 cps. The amplitude of the harmonic motion
at the top of the bag was easily controlled by varying the arm length.
The top of each bag was tented and sewn into a loop, the latter con-
nected to a light but rigid wire loop attached to the end of the shaker
arm. The shaking motion reached full speed in 0.2 second or less and
the stop action was made very abrupt by the addition of a resistance
in series with the motor armature.
The top of the shaken bag was fixed in position by the location of the
thimble at the lower end of the bag. The shaft, however, was designed
to be moved to two other positions on a 6-inch arc, leaving the sus-
pension loop in the same rest location. The net result was that a ver-
tical or a 45° motion was obtainable as an alternative to a horizontal
motion. Additionally, the uniform hole spacing in the structural
frame of the baghouse allowed the entire shaker mechanism to be low-
ered in 6-inch increments to accomodate bags shorter than 10 feet.
Because the top of the bag was fixed in location, tension was usually
adjusted from the bottom of the bag by one of the various mechanisms
indicated in Figure 7. The simplest but least effective adjustment
method, "a", required trial and error shifting of the bag cuff until
the desired tension level was attained. In an improved design, "b",
the thimble was floated on a long radius arm that was locked in place
during filtration and shaking. When released, however, the weight of
38
-------
LOAD CELL
6 in. SHAKER
IARM RADIUS
BAG TOP SUSPENSION
LOOP
TENTED BAG TOP
FILTER BAG
CLAMP
CUFFED BOTTOM
ON THIMBLE
Figure 6. Schematic drawing, bag mounting and
shaking assembly
39
-------
(o) SIMPLE THIMBLE
(b) FLOATING THIMBLE
11 i i J i i I /'? !'rn
POSITION LOCK
FLEXIBLE COUPLING
FLEXIBLE COUPLING
CLAMPED SEAL
(c) HORIZONTAL FLOATING
THIMBLE V/ITH BAG
SEPARATION.
MANUAL TENSIONING
Figure 7. Mountings used to support bottom of shaken bags
40
-------
the thimble was counterbalanced by a supporting spring so that the bag
(and dust) weight could be determined in situ. In the final design,
"c", the thimble was permitted to float horizontally. The thimble was
readily positioned by a hand screw for adjustment of tension and it
could be completely detached from the bag so that very accurate weighings
could be made. These successive mountings evolved as the nonlinear and
time dependent properties of the bags became more evident. Simulta-
neously, modifications were also made in the load cell mounting at the
top of the bag.
Test Fabrics
While many commercial or specially woven filter fabrics might have been
selected in this program, practical considerations dictated that test-
ing be restricted to the most commonly used fabric media. Cotton
sateen is used for a wide variety of dusts at low to moderate temper-
atures because of its relatively low cost and high efficiency. Since
it is obtainable from several manufacturers, it appeared to be a good
choice as the primary test fabric. To extend the study, the same cot-
/jT\
ton fabric in napped form and several types of Dacrotv^fabrics were
also utilized as shown in Table 2. Although the above fabrics were not
necessarily designed for exclusive use in shaking filter equipment, they
did represent a broad range of mechanical properties that might be
encountered in shaking systems, as well as typical variations in per-
meability and surface properties.
•Rae Properties and Measurement Techniques
The selected fabrics were sewn into the specified bag configurations
shown in Table 3, by the firm supplying the fabric. The design of the
cuffs and seams for the bags, Figure 8, was left entirely to the sewing
DuPont Trademark
41
-------
Table 2. PROPERTIES OF WOVEN FABRICS SELECTED (MANUFACTURERS' DATA)
Fabric
1) Cotton
2) Cotton (Napped)
3) Dacron^
4) Dacron^
5) Dacron^
Weight
10
10
5
10
10
Weave
Sateen
Sateen
3/1 Twill
Plain
1/3 Crowfoot
Yarn count
95x58
95x58
66x56 (Filament)
30x28 (Staple)
71x51 (Filament)
Permeability
13
13
12
55
33
Mfgr.
no.
960
960C
C892B
862B
865B
Mfgr. 's
comment
For shaking
For reverse flow
For shaking
For both shaking
and reverse flow
10
Definitions
Weight: ounces per square yard
Yarn count: yarns per inch, warp x fill
3 2
Permeability: ft. /tnin. of air paflfcing through 1 ft. of clean, new fabric at 1/2 inch HO
pressure drop
Manufacturer: Albany International Corporation, Industrial Fabrics Division
DuPotit Trademark
-------
Table 3. PROPERTIES OF BAGS TESTED IN MECHANICAL SHAKING SYSTEMS
CO
Bag size, and state
Unnapped cotton, 10 ft. x 6 in.
clean
Unnapped cotton, 10 ft. x 6 in.
used, < 104 shakes
Unnapped cotton, 10 ft. x 6 in.
used, 2 x 107 shakes
Unnapped cotton, 10 ft. x 4 in.
clean
Unnapped cotton, 10 ft. x 4 in.
used, < 10^ shakas
Unnapped cotton, 5 ft. x 6 in.
clean
Plainweave Dacron, 10 ft. x
6 in. clean
Multifilament Dacron, 10 ft. x
6 in. clean
Weight3
(Ibs.)
1.11
1.33
1.78
0.77
1.26
0.60
1.11
0.56
Lineal density
(slugs/ft, x 103)
3.58
4.22
5.70
2.45
4.06
3.58
3.58
1.80
Modulus
(Ibs. /in.)
16.5
16.5
31 - 45
tension
dependent
9.66
9.66
33.0
13.9
73.0
Effective
filter area
(ft.2)
14.60
14.60
14.60
9.63
9.63
6.92
14.60
14.60
.Total weight of bag including cuffs and seams.
Mass per unit length of the uniform portion of the bag; i.e., excluding cuffs but Including
the lengthwise seam, Ib.-sec.^/ft.2.
cThe elastic stretch modulus for the overall bag, including the normal stretch characteristics
of the cuff, Ibs./inch of stretch.
-------
00
n
ro
oo
W
OQ
a
if
P
co
ID
I
n
CD
H-
"S
SEAM TO SEAM LENGTH
-9ft 5irv
-------
shop's normal procedures except for specifications as to the style of
cuff; e.g., "sewn loop at top" or "slip fit at bottom." The selected
bag dimensions included two diameters, 6 inches and 4 inches, and two
bag lengths, 5 feet and 10 feet, providing length/diameter ratios of
10, 20 and 30. The mechanical characteristics of these bags were
representative of standard commercial bags.
"Lineal Density - The weight of fabric filter media, often expressed as
ounces per square yard, Table 3, provides a rough index of anticipated
collection efficiency, bag life, and possibly pressure drop. Sewing
a fabric of a given weight into a cylinder with a fixed amount of
folding in the seam (see Figure 8) results in a bag having a fixed
mass per unit length or lineal density, p, defined by the unit slugs/ft.
The above factor has a direct bearing on wave velocity and also influ-
ences initial bag tension, damping properties and power consumption.
In normal filtering operations, the weight of dust on and in the fab-
ric may increase the lineal density of new, clean, fabric by a factor
of 2 or more. Since the dust is usually not distributed uniformly
along the bag, both before and after cleaning, the lineal density varies
along the bag. Thus, it is probably best to consider the average value
over the entire bag. As dust is shaken from the bag, lineal density
decreases with time thus changing the bag's tension, its resonant fre-
quencies, and probably its damping properties. Analysts may take these
variations into account in a step-by-step analysis of the cleaning pro-
cess, or they may opt to use a time-averaged value of lineal density
In computation processes.
pap; Weight and Bag Tension - A load cell was incorporated in the shaker
arm, Figure 6, to provide an indication of the tension at the top of
the bag. The original load cell assembly used for determination of
static or dynamic (shaking) tension is shown in Figure B-l, Appendix B.
Compression of the load cell attached to the movable slide rod frame
against the metal block clamped to the oscillating shaft, provided a
45
-------
direct measure of axial tension. Because of sliding friction between
the lubricated rods and the shaft block, however, the response to
higher shaking frequencies was attenuated to an unknown degree. An
improved design again utilizing load cell compression to define axial
tension, is shown in Figure B-2, Appendix B. By providing a pivoting
action between the shaker arm frame and the oscillating shaft, a rela-
tively frictionless (and fast response) linkage was achieved. Con-
necting the load cell to a pivoting member resulted in a varying me-
chanical advantage across the top of the bag, such that tension on one
side of the bag contributed more to the voltage signal than tension on
the other side. The fact that the load cell seldom compressed more
than 0.001 inch under the maximum safe load of 25 Ib. minimized the
above problem. The instrument provided a stable and linear response
(Appendix B) with fairly reliable time-averaged and instantaneous
tension data for static and shaking conditions.
During one series of tests, tension had to be maintained at a constant
level in a three bag system. Rather than design and fabricate addi-
tional load cells, a simple tension measuring device was fashioned
from a U-shaped loop of copper tubing. When slipped over a slack bag^
the loop was supported at an angle determined by the bag tension. An
adjustable side arm with a bubble gauge allowed for calibration against
any preset tension, see Appendix B. By installing turnbuckles in bag
shaker arms, any desired tension level could be attained.
Bag Modulus - The tension required to produce a unit change in bag
length, defined as the elastic modulus in this study, was shown to be
an important bag property. The elastic modulus of the fabric was not
included in any materials specifications furnished by bag suppliers.
It was measured on fabric samples in our laboratory by applying a
known tension to a strip, (about 4 inches wide and 18 inches long)
and measuring the resulting change in length. The above approach gave
only an approximation of the elastic modulus of an actual filter bag
46
-------
(~ + 30 percent) because the overall bag modulus depended not only on
the basic fabric properties but also on the style or tightness of the
seam, the method of cuff stitching, and the folds in the upper (tented)
end of the bag. It was preferred, therefore, to measure the modulus
of the bag directly by suspending a known weight from the bottom of the
bag and measuring the resulting bag elongation. Alternatively, the
modulus was measured in situ by lowering the clamped lower end of the
bag a known distance, and measuring the resulting tension increase
with a load cell. Typical results are shown in Figure 9. Bag modulus
values did not vary linearly with increasing tension, due mainly to the
uncrimping of yarns prior to the normal stretching of individual fibers.
Because of this nonlinearity, the reciprocal slope of the line (or
elastic modulus) was arbitrarily taken as the average value over the
range of typical shaking tensions.
Many composite materials show nonlinear rheology due to friction be-
tween their separate elements. Such a material usually displays a
tension hysteresis as it is cycled repetitively through a fixed linear
displacement. Since the degree of hysteresis may depend on the rate
of cycling, there was reason to suspect that the bag modulus might
depend on the shaker frequency. A limited attempt to measure this
property was made with a 10-ft. x 6-in. cotton bag by allowing the
lower bag cuff mounting to move up and down at variable frequency,
through a stroke of approximately 0.1 inch, while the top of the bag
remained rigidly fastened to the load cell. The average bag tension
that resulted was essentially independent of frequencies up to about
20 cps. The range of instantaneous tension values approximately
doubled, however, as frequency increased indicating some degree of
bag hysteresis. These results suggested that bag modulus values would
display similar variations. The hysteris factor was also demonstrated
when the tension-elongation measurements were made over extended time
periods. With respect to the calculations performed in this study,
average tension and average modulus values were used to estimate bag
displacements, energy requirements and dust removal.
47
-------
oo
TENSILE
MODULUS
NEW-16.5 Ibs. /in.
USED- 45 Ibs./ in.
USED (2 x I07thok«s)
345*78
APPLIED WEIGHT LOAD, Ibs.
Figure U. Tensile properties for a 10-foot by 6-inch sateen bag
-------
The elastic properties of a bag may be expected to vary not only in
the short term (frequency dependence) but also in the long run as many
materials are made less elastic or "work hardened" by repeated stretch-
ing or flexure. To determine the durability of cotton filter bags,
they were shaken about 20 x 10 times which is equivalent to several
(3 to 5) years of normal bag use. The net effect on modulus was to
increase it by a factor of 2 or more. One can infer that the static
modulus of a new filter bag might be controlled by preconditioning, a
process that would reduce its "break in" time when put into service.
During shaking, flexure of the sides of the bag parallel to the direc-
tion of motion probably causes some shearing action. Therefore, the
bag's resistance to shear, or its shear modulus, may influence its
motion. Although any detailed analysis of such motion was beyond the
scope of this study, shear modulus was examined superficially for pos-
sible future reference. A swatch of cotton filter fabric was mounted
between parallel jaws, one of which was free to move in a direction
parallel to itself. The force-displacement relationship as the jaw
was moved indicated an approximate modulus W ' h = 1.5 Ib./in. where M"
2
is the customary shear modulus (Ib./in. ) and h the fabric thickness.
Shaking Energy and Power - The power needed to operate a shaking bag
filter system depends on the velocity of the top of the bag and the
force at the top of the bag tangent to the arc of motion (or
roughly normal to the bag axis). The velocity of the top of the bag,
equal to that of the end of the shaker arm, was readily computed
since the motion was essentially a simple harmonic type. The tan-
gential or "side" force was measured by strain gauges mounted on the
shaker arm as shown in Appendix B. In addition, a magnet -and -coil
device mounted on the shaker shaft indicated the phase relation be-
tween velocity and side force. The latter device was also useful in
studying the timing of the bag tension excursions with respect to the
position of the shaker arm. Bag power consumption was also approxi-
mated by ammeter and voltmeter measurements on the field and armature
49
-------
circuits of the shaker motor. A significant part of the energy trans-
mitted to the bag at the shaker arm juncture was lost because of damp-
ing processes. The rate of energy loss along the bag depends on the
characteristic damping properties of the fabric and also its partic-
ular geometry. In this study, damping was inferred by observations
of tension changes and motion during shaking, there being no simple
way to measure it directly. The results are discussed in a later sec-
tion of this report.
Motion of the Shaking Bag - Bag motion was measured by photographic,
stroboscopic and photometric methods. The most common approach in-
volved time exposure photographs of the shaking bag that shoved the
maximum excursions of the fabric. Stroboscope observations were made
of the bounding wave patterns and local fabric motions. Still (short
exposure) photographs of instantaneous fabric positions enabled quan-
titative analyses of the wave patterns. A light beam and photocell
technique, described in Appendix C, was also used to monitor the max-
imum excursions of single points along the bag. The measurement tech-
niques and results are discussed further under RESULTS pages 63> to 157f
Test Dusts
Coal fly ash was selected as a primary test dust for several reasons.
With respect to density and size properties (see Table 4) it is repre-
sentative of many industrial dusts that are collected by fabric filtra*-
tion. It can be successfully redispersed by high pressure, > 90 psig,.
compressed air ejectors to produce a particulate suspension with size
characteristics approximating those of the parent material. Since most
fly ash particles from pulverized coal combustion are spheroids (ceno-
spheres), the analysis of aerodynamic behavior is simplified. On the
other hand, the sphericity leads to increased particle penetration,
relative to that for irregularly shaped particles having the same
settling velocity.
50
-------
Table 4. SUMMARY OF TEST DUST SIZE PROPERTIES
Coal Flv Ash
Source: Hopper of electrostatic precipitator, cyclone
boiler. Sold by concrete block manufacturer
($0.025/lb. in 50-lb. bags in 5-ton lots).
Size properties:
Light field microscopy
As received bulk dust
Dust shaken from bags
Andersen cascade impactora
Aerosolized dust
Coulter count
Aqueous dispersion
Sieve analysis
Talc
Source: Sierra Talc and Clay Co., So. Pasadena,
California. "EMTAL 599" ($0.0844/lb. in
50-lb. bags).
Size properties:
Andersen cascade impactor3
Aerosolized dust
Light field microscopy
As received bulk dust
jSilica
Source: Cabot Corp., Boston, Mass. "M-5 Cab-0-Sil"
($2.00/lb. in 10- Ib. bags).
Size properties:
Andersen cascade impactor3
Manufacturer
MMD
urn
5.0
2.4
8.0
14.2
27.0
3.2
3.6
0.012
°"g
2.13
1.77
2.0
2.4
3.7
2.9
2.0
^
aAndersen impactor data used as particle size standard for upstream
samples.
51
-------
Another reason for selecting fly ash was that it has been used by sev-
1 13
eral investigators in related studies. ' As far as can be deter-
mined, the aerosols used in past and present studies were sufficiently
similar to allow comparisons of experimental results. Additionally,
fly ash is available in large quantities at little expense (50-lb bags
at $1.25) in contrast with some specially prepared powders for micro-
scale testing costing as much as $25.00 per gram. The nature of the
production process for fly ash is such that there should be minimal
variation from bag to bag or from ton to ton. The present fly ash
was collected by electrostatic precipitator from a cyclone fired
boiler. Thus, its size was somewhat smaller than the ash from a con-
ventional pulverized coal fired system.
Dusts from other industrial processes, if available in large quanti-
ties, may prove to be useful test aerosols. There are apt to be prob-
lems, however, such as very broad size ranges from aggregate drying
kilns, hygroscopic properties with some calcined materials, and redis-
persability with metal fumes. Although dust recovered by fabric filter
systems appears desirable from the size point of view, the presence of
short fibers shed by the filter may lead to atypical collection
properties.
To extend the range of aerosols studied, two other materials, talc and
silica were selected to depict other size and shape properties encoun-
tered in the field (see Table 4). Both dusts are commercially available
in bag quantities at reasonable prices.
Fly ash and talc dusts were sized initially by conventional microscopy
wherein small amounts of the bulk dust were dispersed in an immersion
oil. The above approach provides only a rough estimate of size since
one can never be certain whether the final state of dispersion is re-
presentative of the aerosol phase. The same dispersibility problem
also applied to screen analyses and Coulter counter measurements. Gen-
erally the cascade impactor measurements were considered to provide
52
-------
better estimates of aerodynamic size although it was necessary here to
assume an average particle density. Because agglomerates were known to
be present in the aerosol, the assumption of a particle specific gravity
of 2.0 is at best an approximation.
It was observed that the aerosolized state of the talc dust based upon
Andersen impactor measurements suggested a finer suspension than the
fly ash. Higher resistance characteristics noted for the talc were
consistent with its smaller particle size. In that the talc and fly
ash aerosols did not differ radically in size, it was expected that
any apparent difference in filtration characteristics might be attribu-
table to particle shape factor (and packing density).
Re-precipitated silica is composed mainly of highly agglomerated sub-
micron particles. Its pneumatic redispersion was expected to produce
a cloud composed of agglomerates of the order of 1-micron diameter,
corresponding to an aged fume. The high resistance observed during
its filtration suggested that the aerosolized material was considerably
smaller than 1 micron. Extremely high efficiencies were obtained when
filtering silica particles, possibly due to electrostatic properties
of the particles being very different from fly ash and talc. No sat-
isfactory size measurements were obtained on the inlet aerosols with
the cascade impactor.
Dust Measurements
Aerosol Concentration and Particle Size - Inlet dust concentrations
for the various fabric filter systems investigated during this study
were determined directly by gravimetric methods. The following ap-
proaches were used:
53
-------
• Material balance by dust feed rate - The dust delivery
rate for the Acrison Dust Feeder was shown to be con-
trollable to within + 1 percent for feed rates in the
range of 0.032 Ib./rain. (14.4 grams/min.) and aver-
aging periods of 1 minute or greater. In conjunc-
tion with the recorded system airflow rate, the dust
concentration entering the hopper section was readily
measured. The actual dust concentration arriving at
the fabric surface was found to be about 30 percent
lower than the gross inlet loading because of settle-
ment in the hopper section.
• Filters^ all-glass or membrane types - Readily weigh-
able quantities of dust (~ 0.5 gram) were collected
on high efficiency filter media at typical inlet con-
centrations, ~ 1 to 10 grains/ft.•*, to the fabric
filter testing assembly. Sampling was performed iso-
kinetically at 7 to 15 liters/min. for 10 minute
periods. Since replication in filter circle weights
was of the general order of + 0.2 mg. for 100 mg.
filters, filter weight gains could be determined
easily within + 1 percent. Ordinarily, no change in
inlet concentration occurred during any one testing
sequence.
• Cascade impactor measurements - A standard Andersen
impactor (6 stage, 1 ft.3/min., outstack design) was
used whenever practicable to determine the mass con-
centrations and size distributions of upstream and
downstream aerosols. Duct dimensions were too small
in the test system to permit use of the newer instack
models. In the case of highly efficient filter sys-
tems, it was not possible to collect weighable amounts
of dust from the effluent air streams within reason-
able sampling periods, ~ 2 hours. Usually, outlet
concentration levels > 0.001 grain/ft.-' could be
determined by the Andersen impactor.
Comparison between parallel filter and Andersen irapactor measurements
indicated that the Andersen values could be considered accurate to
within + 10 percent at the 95 percent confidence level, Appendix M.
It is assumed here that because of far greater ease in sanple handling
Acrison Model 120 Volumetric Feeder, Acrison, Inc., Carlstadt, New
Jersey, 07072.
54
-------
the filter values are usually correct. Since the individual impactor
stages could be weighed to within + 0.3 rag., Appendix M, no appreciable
errors were introduced with the collection of 10 mg. or larger amounts
of dust on each stage.
Repetitive determinations of the size distribution of the fly ash aero-
sol showed fairly good agreement, Figure 10. Size distributions Nos. L
and 2 show mass median diameters, HMD, of 8.7 and 7.2 |im, respectively,
and similar geometric standard deviations. Why Distribution 3 should
indicate a slightly higher HMD, 9 pm, is not clear although the difference
might be attributed to the lower concentration level or to variations
in humidity. Generally, it is expected that the reproducibility in
MUD values should parallel that observed for total mass concentrations
determined by Andersen impactor.
Characterizing stage diameters for the Andersen device were based upon
the manufacturer's specifications for spherical particles of specific
2
gravity 2.0 at a 1 ft. /min. flow rate:
Particle diameter, urn
0.36
0.7
1.4
2.3
3.9
6.5
> 6.5
In view of variations in physical and chemical composition, shape
factor, and degree of agglomeration, the assumption of a density of
2.0 grams/cc for fly ash appeared acceptable.
Figure 10 also shows the size parameters for the fly ash aerosol as
determined by the light field microscope sizing of an oil dispersion
of the dust. Frequently, this approach produces a better redispersion
55
-------
Oi
CURVE: CONCENTRATION,
groins/ft.3
2.0
2.5
0.2
MICROSCOPE SIZING
OF OIL SUSPENSION
I I
\
L
\
I
L
_L
I
I
I 2 5 10 3O 50 70
PERCENT MASS ^ STATED SIZE
90 95 98 99
Figure 10. Inlet fly ash size distributions by Andersen cascade impactor
-------
of a dry powder than that obtainable by air ejector systems. For the
purposes of the present study, it was assumed that the Andersen impac-
tor measurements indicated the effective size of the aerosol.
Outlet dust concentrations for the various investigations conducted
during the program were measured by the following methods;
RDM, respirable dust monitor - Mass concentration meas-
urements in the range of 0.001 to 0.01 grains/ft, were
made with an RDM sampler. This device operates on the
principle of collecting a short term, ~ 1 min., dust
sample on a single impactor stage and determining the
mass of the dust spot by its beta attenuation charac-
teristics. The measurement accuracy for this instru-
ment ranges from about + 25 percent at the low end to
about + 6 percent at midscale. The principal use for
this device was with fairly permeable woven fabrics
cleaned by mechanical shaking and felt bags cleaned by
reverse air pulse. Except for B&L measurements to be
discussed later, this instrument provided the only way
to estimate low mass concentrations within a 1 min.
sampling period.
Single particle light scattering counter - The size
properties of the inlet dust loadings could be estab-
lished for periods as brief as a few minutes by cascade
impactor sampling. Such measurements, however, were
not ordinarily needed in this study because the inlet
aerosol was maintained at a nearly constant concentra-
tion level. On the other hand, the outlet du§t concen-
trations from the various filter systems studied changed
constantly over each filtration cycle. The fact that
the outlet concentrations often ranged from 10-* to 10°
times lower than the inlet levels precluded the use
of impactors except for the long term measurements
cited previously. Although it might have been possible
to collect very brief samples, ~ 10 to 30 sec., on spe-
cial membrane filter assemblies, the time needed for
visual assessment of sufficient samples to portray
accurately changes in both size and concentration pro-
perties was prohibitive.
*RDM-101 Respirable Dust Monitor, GCA/Technology Division, Bedford,
Mags. 01730.
57
-------
Therefore, the use of a fast response device, designed
for the detection and measurement of low particle con-
centrations was indicated. A Bausch and Lomb* (B&L)
counter was found in many ways to be singularly effect-
ive for the needs of this study. Even when used with a
single channel dial or digital display, changes in the
concentration of specified size ranges could be estab-
lished in 0.1 minute and within 0.5 minute five size
ranges could be scanned. It was possible, therefore,
to follow sequentially the changes in concentration
for a broad range of particle diameters as the filter
process continued. As expected, particle emissions
were greatest when filtration was resumed with a
"just cleaned" filter. However, use of the B&L in-
strument provided unusual insights into the effects
of type of fabric (and dust), and the method and In-
tensity of cleaning*
It was recognized at the outset of the study that size measurements, by
different systems could present correlation problems. Light scattering
equipment is highly sensitive to particle shape, surface smoothness,
and to refractive index. Furthermore, the working particle concentra-
tion range for these units is generally near ambient dust levels. Thug
unless special air dilution apparatus is used, electronic "choking" or
no display conditions may occur in the lower size channels at concen—
f\ *}
trations levels in excess of 10 particles/ft. . Because the gating
with the B&L device is designed to shew the number of particles
than the indicated size, choking in the lowest channel; e.g..* > Q..3;
does not completely invalidate measurements for the higher channels,
0.5, 1.0, 2.05 5.0 and 10 p.m. Some error is expected, however, possi-
bly in the factor of two range, when one computes mass concentrations
from the indicated number concentrations.
The B&L measurements provide a data output from which one can calculate
immediately a size distribution (with respect to particles. > 0.3 p.m)
Bausch & Lomb Dust Counter 40-1, Bausch & Lomb Inc., Rochester,
New York 14602.
58
-------
and the total number of concentration. If the distribution can be de-
scribed by some convenient mathematical function; e.g., logarithmic
normal distribution, and a constant density assumed over the size range,
the dust distribution with respect to mass, volume, or surface area
can also be readily calculated. In the absence of a rigorous mathe-
matical relationship, a conventional incremental method is employed
in which particles are grouped within size ranges characterized by
some mean diameter. This approach was used in the present study with
the B&L size groupings described as shown below:
* —
Size range, urn Average diameter, d
0.3 - 0.5 0.424
0.5 - 1.0 0.826
1.0 - 2.0 1.65
2.0 - 3.0 2.60
3.0 - 5.0 4.24
5.0 - 10.0 8.26
The accuracy of the conversion from number to mass concentration de-
pends upon the initial particle concentration and the validity of as-
sumptions relative to density of either discrete particles or agglom-
erates, optical properties, possible particle losses in sampling lines
and the stability of the aerosol concentration over the 0.5 to 0.75
minute period required to scan all size channels.
It was not intended that the B&L measurements be used for direct com-
parisons with other sampling methods; i.e., cascade impactor, ::ilter or
RDM unit. Its primary function was to determine the relative changes
in aerosol properties taking place over the filtration cycle. Cali-
bration procedures were carried out, however, in the form of parallel
V - (d.3+d93)/2
59
-------
sampling of the same aerosol streams with the B&L instrument and other
devices. The results of some of these tests, summarized in Table 5,
show the general order of the agreement between the B&L instrument
and the RDM unit. Although our reaction to differences of up to a fac-
tor of 5 may appear to be rather casual it should be kept in mind that
effluent concentration levels over a nominal filtration cycle with
sateen weave cotton bags sometimes varied by 4 to 5 orders of magnitude.
Dust Filtering Rate - Dust filtering rate was determined either by an
overall system material balance, or by weighing the filter bag before
and after dust loading. In the balance method, the dust from the screw
feeder was collected in a weighing pan for various time intervals and
the hopper fallout was collected and weighed after a fixed filtering
time at a constant air flow. The differencing of these two quantities
represented the amount of dust actually reaching the fabric. In the
weighing method, the bag was freed at the bottom whereupon the load
cell used in tension measurements was used as a conventional weighing
device. Readings before and after filtering indicated the weight of
dust deposit. For greater precision, however, the filter bag was
carefully removed and weighed on standard laboratory scales.
Dust Removal - Various techniques were used to determine the dust re-
moved during a single cleaning process. When cleaning by mechanical
shaking, the hopper was immediately opened following filtration and as
the bag was shaken, the falling dust was caught in a pan inserted under
the bag. Weighing the dislodged dust permitted precise determination
of the rate of removal after successive shaking intervals of varying
duration.
Residual Dust - The residual dust holding of a bag following cleaning
bore no relationship to the amount of dust deposited during the fil-
tration cycle. Although the residual accumulation was related
60
-------
Table 5. COMPARATIVE AEROSOL CONCENTRATIONS AS DETERMINED BY VARIOUS
INSTRUMENTAL METHODS*
Test dust
Fly ash
•
Talc
Concentration - grains/
ft. 3 x 10*
B&Lb
8.2
10.4
0.048
0.083
0.11
0.92
0.14
0.027
0.071
0.076
0.11
RDMC
11.8
7.0
0.13
0.34
0.23
2.0
0.23
0.10
0.23
0.23
0.25
Concentration
ratio
RDM/B&L
1.44
0.67
2.71
4.1
2.1
2.17
1.64
3.70
3.25
3.03
2.27
aBased upon outlet concentrations from wool and Dacron felt bags
cleaned by pulse jet air.
Bausch and Lomb Dust Counter, 40-1.
CRDM-101, Respirable Dust Monitor.
61
-------
approximately to the energy input of the cleaning cycle, the electrical
charge, relative humidity and the age and prior service of the bag
also had some effect on residual dust holding.
A beta radiation instrument was constructed according to a procedure
14
devised by Stephan, et al to measure the dust distribution on small
2
(~ 1 in. ) areas of the fabric. Some measurements of dust distribution
along the shaken bags with this instrument showed changes in cloth
loading after successive short periods of shaking. Variations in dust
loading during the actual shaking could not be measured, however,
because the instrument was sensitive to the position of the fabric.
The instrument was found to be about three times as sensitive to dust
on one side of the bag cylinder as the other side, due to lack of
radiation collimation in this type of instrument. The above problem,
coupled with the sensitivity of positioning, limited the usefulness
of the instrument in most tests.
A simple and very qualitative technique for observing the residual dust
deposit density in shaken bags was to run a light bulb up and down in-
2
side the filter bag. As little as 40 grains/ft, of additional deposit
was sufficient to completely block the transmission of a 50 watt light
when observed in a darkened room. Thus, small variations in residual
dust were clearly seen as variations in the bag brightness.
Pressure, Air Flow and Air Properties Measurements
The system for regulating airflow to the filter bag is described in
Appendix A. The pressure differential across a Stairmand disc, nortnallv
about 1 inch water, was automatically monitored by a pneumatic indicati
M5>
controlling, and recording system (Appendix A).
A Stairmand disc is a circular plate, symmetrically located within a
round duct, that diverts the gas flow through the outer annular region
62
-------
When the disc area is one half of the duct cross section, its resis-
tance properties are nearly the same as that of an orifice of similar
size. In addition to acting as a flowmeter, this device promotes good
mixing of suspended participates so that centerline sampling alone is
sufficient at downstream locations, ~ 4 to 5 pipe diameters. Accuracy
and precision of the flows indicated by this system are estimated to be
within about 5 percent.
The pressure differential across the filter bag was monitored by a
pneumatic indicator similar to that used in the flow control system
and also by a sensitive Bourdon-type Magnehelic gauge and a liquid
manometer. The Magnehelic device with.its fast response and high pre-
cision (better than 1 percent), was used when moderate accuracy was
acceptable (within 5 percent). The manometer was used as a standard
and for highly accurate measurements.
The air used for filtration was drawn from the room in which the filter
system was located, such that considerable recirculation was involved.
Nominal room conditions of about 70° + 3°F and 50 percent + 4 percent RH
were maintained during most of the tests. Based on measurements re-
2
ported by Durham and Harrington, the variations in relative humidity
during present testing were not expected to have any significant effect
on cotton fabric tests. Dacron multifilament weaves might hav«J shown
some changes in both collection efficiency and specific resistance
coefficient due to the + 4 percent RH range.
RESULTS
cleaning Forces and Bag Motion
•Rap Tension versus Shaking Frequency - Theoretical relationships, sum-
marized in a previous section, indicate that shaking tension should
increase with shaking frequency and show characteristic maximum values
at resonant frequencies. Confirming experiments, summarized in
63
-------
Figure 11, with a typical filter bag indicate distinct resonant fre-
quencies of about 2, 4.5 and 7.5 cps. The dynamic shaking tension, T.,
increases more or less steadily with increasing frequency, except for
the periodic resonant states noted above.
A comparison of predicted versus observed resonant frequency values for
the fundamental frequency and the first few harmonics is shown in Fig-
ure 12. The predicted values, computed by formulas developed in
Appendix D, confirmed the fact that observed tension maxima were as-
sociated with resonance points. The rather narrow scatter of data
points around the regression line suggested that the relationships de-
veloped in this study to describe bag motion were fundamentally sound.
The shaking tension at a specified frequency was not constant as implied
by Figure 11, because the display dial shows only the average or inte-
grated reading. It was actually the time average of a rapidly varying
force that was measured by a fast response load cell mounted at the
upper end of the bag in conjunction with an oscilloscope display. A
typical force envelope is shown in Figure 13. In each excursion of the
shaker arm from the vertical, the tension passed through a maximum and
a minimum (i.e., two maxima per shaking cycle). The average width of
the envelope appeared to be similar in magnitude to the variation in
static tension as the arm was fully displaced from the vertical,
Equation (2.3). On the other hand, the range in envelope width depended
upon proximity to the resonant state. The position of the arm at the
instant of maximum tension was found to vary from full displacement at:
zero frequency to the vertical position at very high frequencies (or at
resonance conditions at moderate frequencies). Although the instan-
taneous forces at the top of the bag were probably related to small per-
mutations in the bag's standing wave pattern, the analysis of this rela-
tionship in greater detail did not prove fruitful.
According to classical mechanics, the amplitude at which a body reso-
nates is highly dependent on the degree of damping in the system. A
64
-------
14
12
K>
UNNAPPEO COTTON SATEEN
BAG, I0iru6in., CLEAN.
INITIAL TENSION: 2.4 Ibs.
SHAKER AMPLITUDE: I in.
fe T
Q.
O &
I- 6
2 4
UJ
2 4 6 8 10
SHAKER FREQUENCY.eps
12
Figure 11. Shaking tension as a function of frequency
for cotton sateen bag
65
-------
Q.
U
••
O
UJ
s
QC
U.
LJ
OC
Q
UJ
§
O
_J
O
II
20
16
12
8
0
o COTTON, 10 ft. x 6 in., CLEAN
• COTTON, 10ft. x 6 in., USED
C-:> COTTON. 5 ft. x Gin., CLEAN
-------
14
IS
12
II
.0 10
e>
2 •
u.
O 8
S 6
O 3
(/)
Z
LU
UNMAPPED COTTON SATEEN
BAG, 10ft. x 6 in.
INITIAL TENSION: 4.3 Ibr
SHAKER AMPLITUDE: I In.
INSTANTANEOUS
MAXIMUM
TIME-AVERAGE
TENSION
x^ INSTANTANEOUS
^y MINIMUM
/
^\
I I /
\ I
\ I
U'
r\ i
I \ *
I ~
i
6
10
12
14
SHAKER FREQUENCY.eps
Figure 13. Maximum and minimum envelope curves and
average shaking tension for cotton
sateen bag
67
-------
high damping rate will greatly attenuate system amplitudes, while a low
damping rate may produce amplitudes large enough to destroy the vibra-
ting member. In the case of an oscillating filter bag, observations
indicated that damping and reflection were partially dependent on bag
tension, bag weight, and shaker frequency and amplitude. The main
causes of the energy dissipation, or damping process, were considered
to be friction between fibers, yarns, and dust particles, as well as
flexural heating within the fibers and air resistance to the moving
bag. Dust separated from the surface of the bag with appreciable ve-
locity also represents a significant kinetic energy loss.
Although a rigid quantitative theory of damping might be developed for
shaken bags, it appeared that too many parameters would be required to
yield practical damping coefficients. As an alternative, the classical
damped wave equation shown below
Y = A e"?X sin (2« x/X) (2.13)
was applied to a single shaken bag (Appendix F). It was. shown that the
average bag amplitude, Y, could be related to the amplitude of the
driven end, A, by an attenuation parameter, Q, for specific shaking
conditions ', i.e. ,
a = « (2pL)~ (2.14)
The ratio Y/A was usually less than one except near resonant frequencies
in low-damped bags where it sometimes exceeded unity. If the classical
exponential damping rate, 3, can be defined, the attenuation coefficient
for Equation (2.10) needed to obtain average bag amplitude can be easii-y
computed. At this time, shaking tension data for various bags and
shaking regimes have been analyzed to obtain empirical values of a. On
this basis, a can be estimated roughly in accordance with procedures
described later in this section.
68
-------
The wave damping properties of the bag presumably bear some relation to
those of the fabric. An approximate measure of damping tendency was
made by stretching a strip of cotton filter fabric between fixed supports
and tapping the center of the strip to initiate oscillation. The
ensuing oscillation and its decay were observed on an oscilloscope by
magnetically sensing a tiny bit of steel cemented to the fabric strip.
Each successive amplitude displacement was about 0.91 times its pre-
decessor, indicating a 20 percent energy loss per cycle. To obtain
meaningful data, damping in a fabric sample should be studied at ten-
sions, frequencies, and angles of flexure similar to those seen in a
real filter bag. Generally, it is preferred to make measurements on a
full scale bag under normal conditions of use as discussed earlier under
APPARATUS, TECHNIQUES, AND MATERIALS (pages 36 to 61).
Acceleration - As pointed out under BACKGROUND AND THEORY (pages 13
to 36) the maximum acceleration seen by a point on the filter bag can be
approximated by Equation (2.9). The specific acceleration assumed to
relate quantitatively to dust removal is the average value of the maxima
seen by all the points on the filter bag. The time-wise phasing of these
maxima from point to point is expected to exert second order effects
only on dust removal. Since all parts of the bag were observed to shake
at essentially the same frequency the determination of average bag
acceleration requires only a measure of average bag amplitude.
Bafi frequency - In almost every case, all points on a bag moved with
the same frequency as that of the shaker arm according to stroboscopic
measurements and a light beam interception method (see Appendix C).
Consequently, it was concluded that the average bag frequency needed
to compute average bag acceleration was simply the shaker arm frequency.
The light-beam interception technique also revealed that the motion was
essentially simple harmonic (i.e., not appreciably saw-toothed or
skewed). Limited measurements by the above method suggested that at
shaking frequencies greater than nine cps, the bag motion indicated the
69
-------
possible presence of beat frequencies (higher harmonics). Their inten-
sities, however, relative to the driving frequency of the shaker arm,
were sufficiently low to justify neglecting them.
Envelopes of motion - As a means of estimating the average bag ampli-
tude needed to compute average bag acceleration, time exposure photo-
graphs were made over ten or more bag oscillations. A series of photo-
graphs of a 10 foot by 6 inch used cotton bag with a residual dust
holding of about 1.2 pounds were made at selected frequencies to depict
resonant, antiresonant and immediate frequencies. Typical tracings,
Figure 14, show the largest lateral excursions made by the opposite
edges of the bag. Reference to the rest position of the bag, which is
normally a tented Vee, yields the approximate amplitude of bag motion
at any axial location. Action photographs of different bags at a
variety of tensions and frequencies were used to obtain an average
bag amplitude for comparison with dynamic tension data.
The motion envelopes deriving from the above photographs (Figure 14)
depict clearly the resonant and anti-resonant states. The latter
state, which occurs at a frequency slightly above a resonant level,
is characterized by minimum tension and amplitude values. Additional
envelope photographs are discussed in Appendix G.
Typical motion envelope measurements are summarized in Table 6. Over
the range of frequencies investigated, 0 to 11.4 cps, the maximum
lateral displacement at the nodal points was 1.69 inches from the rest
position at a shaking frequency of 3.95 cps. Midway between the nodes,
the fabric only moved about 0.74 inches from the rest position, indi-
cating about 44 percent as much acceleration in that region. These
data demonstrated that the bag amplitude became more uniform as fre-
quency was increased and that the average amplitude of the bag dimin-
ished as the frequency increased.
70
-------
PHOTO
2
3
4
5
6
7
8
TENSION, Ibs.
3.01
2.20
2.S3
3.66
3.16
3.94
4.56
FREQUENCY.eps
20
2.6O
3.70
4.4
4.75
615
8.0
TOP 5
50
Q
52 4.0
3.0
20
2 4 6 S
DRIVEN FREQUENCY, ep«
BOTTOM-5
5 4 ' 7 6 5 4" 7
BAG DISPLACEMENT, in
Figure 14. Bag Displacement versus driven frequency and indicated dynamic tension
-------
Table 6. ENVELOPES OF MOTION FOR A SATEEN WEAVE COTTON BAG
Photo
number
1
2
3
4
5
6
7
8
9
10
11
12
Frequency
(cps)
1.8
2.2
3.3
3.95
4.1
5.4
5.95
6.4
8.6
8.9
11.0
11.4
Resonance
state
1st Res.
Anti Res.
Midpoint
2nd Res.
Anti Res.
Midpoint
3rd Res.
Anti Res.
4th Res.
Anti Res.
5th Res.
Anti Res.
Average
shakingb
tension-Tm
(Ibs.)
3.23
2.19
3.38
4.27
3.46
4.18
4.75
4.59
6.03
5.39
7.13
6.19
Bag
amplitude0 -Y
(in.)
max.
1.62
1.35
1.40
1.69
1.60
1.22
1.30
1.10
1.20
1.14
1.20
1.16
mm.
-
1.06
0.90
0.74
0.72
0.88
0.84
0.70
0.88
0.90
0.94
0.93
avg,^
• — .
(1.81)
1.21
1.15
1.22
1.16
1.05
1.07
0.91
1.04
1.02
1.07
1.05
10 in. x 6 ft. bag with light residual dust holding, gross lineal
weight = 0.135 Ibs./ft. Initial tension (T. ) = 1.55 lb., shaker
amplitude = 1 in. *
Total time-averaged tension at bag midpoint.
Maximum at node points, minimum at anti-node points.
72
-------
Our major difficulty in interpreting these data is that the rest posi-
tion of the fabric probably changed during shaking partly due to flex-
ure along the bag. This flexure in conjunction with the increased
tension tended to straighten bags that were initially tented unevenly
or were creased from storage. Downward traveling waves produced a de-
tectable downward air draft outside the bag that also contributed to
the flattening process. Thus, the maximum displacements seen in en-
velope photographs represent only approximately the true amplitude at
a given point along the bag.
The tension data in Table 6 provide a means to compute the average
bag amplitude using Equations (2.5) through (2.7). The results are
compared in Table 7 with the apparent average bag amplitude es-
timated from the envelope photographs. The same trends are seen with
increasing frequency and/or resonance. However, the computed tension-
based amplitude is about 30 percent lower probably because the photo-
graph displays only the maximum displacement, while the calculated
tension depicts an average around the bag. The lower amplitude values
are more in keeping with the visual observation that bag amplitudes
are usually less than the shaker amplitude, especially in situations
involving heavy damping. We believe that the tension approach yields
a more accurate average bag amplitude with respect to correlations
with dust removal.
Amplitude from basic bag properties - Before undertaking dust removal
studies, three types of fabrics and three length to diameter ratios
were tested at different installed tensions, shaking frequencies, and
shaking amplitudes (see Table 3). The purpose of the tests was
to enlarge upon and validate the general concepts developed in the
previous section. Appendix H provides detailed summaries of all ten-
sion versus frequency measurements. Typical results for various ex-
perimental conditions are provided in Figures 15, 16, 17, and 18.
73
-------
Table 7. CALCULATED AND MEASURED AVERAGE BAG AMPLITUDES*
Photo
no.
1
2
3
4
5
6
7
8
9
10
11
12
Average dynamic
tension Tj
(Ibs.)
1.68
0.64
1.85
2.74
1.98
2.70
4.32
3.16
4.60
3.96
5.70
4.76
Bag
wavelength
(ft.)
15.4
10.4
8.6
8.1
7.0
5.9
5.6
5.2
4.4
4.0
3.8
3.4
Average bag amplitude - Y
(in.)
Calculated13
1.68
0.70
0.98
1.13
0.83
0.81
0.98
0.78
0.79
0.67
0.76
0.62
Photograph
(1.81)
1.21
1.15
1.22
1.16
1.05
1.07
0.91
1.04
1.02
1.06
1.05
See Table 6 for operating data.
Based upon tension data, shaking parameters and bag properties:
irf
r . T
m D
Equation (2.15)
74
-------
12
II
10
9
8
M
JO
00
u. 6
o
z
UJ
»- 3
I0ft. xGin. BAGS
UNNAPPED COTTON
r
1/2 in.AMPLITUDE
_L
4 6 8 10
SHAKER FREQUENCY, cps
12
Figure 15. Effect of shaker amplitude on shaking tension
75
-------
8
10ft. x6in. BAGS,
^ I in. AMPLITUDE
CD
Q.
O
Z
O
CO
Z
LiJ
5oz./yd.'
MULTIFILAMENT
DACRON
lOoz. /yd.'
PLAIN WEAVE DACRON
J_
J_
_L
J_
_L
J.
34 5 6 7 8 9
SHAKER FREQUENCY, cps
10 II 12
Figure 16. Effect of fabric type on shaking tension with
10 ft. x 6 in. bags (1 in. shaking amplitude)
76
-------
8
CD
U.
Q-
O
z
o
(?) 3
LJ
UNNAPPED COTTON BAGS
6 in. DIAMETER, lOoz. /yd.2
10ft. x6in., L/D=20
5ft. x 6in.,L/D =
10ft. x 4in., L/D = IO
_L
J.
2 4 6 8 K>
SHAKER FREQUENCY.eps
12
Figure 17. Effect of length/diameter ratio on shaking
tension
77
-------
CURVES I-4, USED BAGS, WEIGHT- 1.4 Ibf.
CURVES 5-7, UNUSED BAGS, WEIGHT « I.I Ibs
CURVE 6-7, NO INSTALLED TENSIONING
4 6 8 10
SHAKER FREQUENCY, cps
Figure 18. Effect of initial tensioning on shaking tension
with unnapped, 10 ft. x 6 in. cotton sateen bags
78
-------
Regardless of the fabric configuration or shaking conditions, all
tension- frequency curves exhibited several common features. Starting
from specified T. values that depended only upon initial tensioning
and bag weight, all curves described upwards paths with periodic ex-
cursions or oscillations from the mean associated with resonance
phenomena „
The mean curves of Figure 18 (excluding oscillations) can be described
approximately by a combination of Equations (2.16) and (2.17).
(2.16)
m
term a given in Equation (2.16) can be estimated readily from the
tension- frequency data given in Appendix H and the following empirical
relationship:
u =
1 - 7.:
/T. \ -I
f - 2.65 + I— - 0.75 )
VW '
f - 2
(2.17)
In Equation (2.17), the term T. refers to the initial average tension
at the top rather than at the middle of the bag and W is the total bag
weight.
The predicted average bag amplitude can be adjusted to take into
account the tension variations cited above by the following empirical
equation
AT
0.35
MA.
2R.
(2.18)
79
-------
where AT^ is one-half the peak-to-peak tension excursion that occurs
E *
near each resonant frequency. The magnitude of this tension excursion
is about one-third of the tension difference noted when a bag is
stretched from its vertical position (!„) to maximum displacement (T V
A
Equation (2.15).
Equation (2.17), in conjunction with Equation (2.18), provide a
means to estimate average bag amplitude without the need to perform
actual shaking tests as required by Equation (2.16). All terms in
Equation (2.17) can be evaluated prior to performing any shaking- meas-
urements. A comparison of average bag amplitudes as calculated by
both approaches, Figure 19, indicates fair agreement.
The corresponding deviation in average bag amplitude, AY.,, at the
resonance and nonresonance points may be approximated from the
ing expression
AYE =ATE I"
m
noting that the coefficient in Equation (2.18) decreases with increas-
ing frequency. The primary value of Equation (2.19) is that a better
estimate of bag amplitude can be made for the resonant state.
Bag Power Consumption - The most direct way to estimate bag power con~
sumption is to measure the electrical power needed to operate the shaker
with and without the bag. Figure 20 indicates that typical bag power
consumption, which is roughly a few watts, increases with frequency.
The coefficient 0.35 is an average that appears to depend on which
harmonic is excited. Review of data gives the following average co-
efficients for the 1st through 5th harmonic as 0.43, 0.39, 0.31,
0.26, and 0.21 respectively.
80
-------
c
'^f
00
CJ
cv
LJ
UJ
O
OL
li
O
<
00
LJ
O
<
cr
u
§
O I0in.x6in. UNMAPPED COTTON
X I0in.x6in. 5oz. DACRON
® I0in.x6in. 1002. DACRON, PLAIN WEAVE
CD 5in.x6in. UNNAPPED COTTON
© I0in.x4in. UNNAPPED COTTON
_L
I 2
AVERAGE BAG AMPLITUDE ( Eq. 2.8),in.
Figure 19. Comparison of average bag amplitudes as calculated
by Equation (2.8) or Equations (2.17) and (2.18)
81
-------
70
60
I
50
OH
LJ
45
40
35
(USED COTTON lOftxGinBAG, WEIGHT = 1.41 Ibs;,
I in. AMPIJTUDE; INITIAL TENSION 2.3«bs.)
(I)
(2)
(3)
Shaking one bog
Stocking without bog
Shaking without bag hanger
4 5 6 7 8 9 10 10 12
SHAKER FREQUENCY, cps
Figure 20. Measured power inputs to shaker motor
82
-------
During laboratory tests, the power requirements were so low that ac-
curate current/voltage measurements were difficult, particularly so
because of the special, solid-state switching circuit required to vary
shaking frequency. It was decided, therefore, that a better indication
of bag power consumption would be the measurement of the force/velocity
relationship at the juncture of the bag and the shaker arm. A concur-
rent oscilloscope display of the signal tracings denoting lateral force
and the position of the shaker arm, respectively, enabled determination
of the phase angle between the lateral (horizontal) force and the hor-
izontal component of the arm velocity. The latter quantity was deter-
mined readily by the classical wave equations and the shaking amplitude
and frequency.
Figure 21 summarizes typical results for tension, lateral force and
computed power levels based on tests with a 10 ft. x 6 in. cotton sateen
bag. Both tensile and lateral forces are seen to increase with shak-
ing frequency and to show peak values at resonance points. Ihe phase
angle between the lateral force and the horizontal component of the
shaker arm velocity decreases from 90 at very low, < 1 cps, frequen-
cies to approximately 0 at 8 cps or greater. Shaker power consump-
tion for a single bag increases approximately as the 1.5 power of the
shaking frequency because of the decreasing phase angle. Although
calculated power consumption is slightly higher at the lower frequen-
cies, the differences are not significant at those frequency levels
where the bag cleaning appears the most effective.
Since all shaken bags ordinarily assume the same phase relationship
between the instantaneous lateral force and the suspension point ve-
locity, the key variable to define is the lateral force. Figure 22
and Appendix I show that the instantaneous maximum lateral force is
roughly proportional to the average tension when plotted against fre-
quency • In Appendix I, the ratio of maximum lateral force to average
tension was postulated to be as follows:
83
-------
10 ft x 6in COTTON BAG (used)
I in. AMPLITUDE
TENSION AT TOP
LATERAL
FORCE H
2468
SHAKER FREQUENCY, cps
Figure 21. Power consumption, phase angle and frequency
relationships for a 10 ft. x 6 in. cotton
sateen bag (1 in. amplitude)
84
-------
o
-------
!max = A_ / MA/M ^ A_
T R I 4R T / R * '
avg A \ Am/ A
The actual ratios determined for a variety of bags, Appendix I, were
about 1.5 times greater than the theoretical ratio, Equation (2.20).
Noting that T in Equation (2.20) refers to the shaking tension at
the midpoint of the bag, T , Equation (2.12) may be revised to read
A T
1.5 -r-^ J (2nfA)
A
The second right hand term in Equation (2.21) expresses maximum arm
velocity as a function of the shaking conditions. As indicated pre-
viously, T can be estimated from the known physical properties of a
bag and the operating parameters for the shaking system.
Dust Removal
Dust removal was investigated over a broad range of fabric loadings
and cleaning methods by several controlled experiments. Generally, the
range of cleaning parameters studied was much broader than that expect-
ed in most field applications. To facilitate comparisons, inlet dust;
concentrations and fabric dust holdings were held as near constant as
possible at typical field levels. Except for limited tests at a cloth
velocity of 6 ft./min., most filtration was performed at 3 ft./min.
3
and inlet concentrations were 5 grains/ft. • As the result of elutri*-
tion losses in the dust hopper, dust concentrations arriving at the
fabric surfaces were reduced by 30 percent. At the same time the mags
median diameter (HMD) and the geometric standard deviation (a ) for
o
the inlet fly ash particles were reduced from 5.0 (j.m to 2.4 (j.m and 2.4
to 1.8, respectively.
Prior to starting a filtration cycle, the static tension (no shaking,
no airflow) as measured at the top of the bag was adjusted to some
86
-------
preselected level. During the succeeding filtering intervals, the
tension values at the same location and the pressure differential across
the bag were monitored periodically. Dynamic tension levels, which
always exceeded the initial static values, increased with dust
accumulation.
Stopping the airflow at the end of the filtration cycle dislodged but
a small amount of dust, < 60 grains, as the result of bag deflation.
The gradual flexure induced in the fabric may have promoted some sur-
face loosening that enhanced dust release during subsequent shaking.
About one minute after cessation of air flow, the shaking operation
was initiated.
The procedure used to establish the relationship between dust removal
and the number of shakes (the product of frequency and shaking time)
was to shake the loaded filter intermittently for several brief shak-
ing periods. The amount of dust dislodged and collected in a special
tray inserted beneath the bag was measured after each shaking interval.
Since the dust release rate decayed roughly exponentially, the length
of successive cleaning intervals was increased such that about 20
seconds of continuous shaking were used for the final shake period.
The sane time increments were used for those tests designed to estab-
lish the relative cleaning effectiveness of different cleaning regimes.
Although variations in shaking time increments affected intermediate
dust releases, the total amount of dust removed by either continuous
or interrupted shaking, was essentially constant, depending mainly on
the total number of shakes and the specific shaking conditions. The
total number of shakes was normally about 360, with some deliberate
variation.
•The dust release per increment of shaking was also determined by in-
dependent weighings of the bag before and after shaking by using the
tension measuring load cell as a simple weighing scale. As a sup-
plementary measurement, the total amount of dust collected in the
87
-------
hopper (fallout plus bag shakedown) was compared to that delivered by
the dust generating system. The latter technique was applicable only
for long term, steady state operations.
The dynamic tension (top of bag) was recorded during each successive
step of the incremental shaking process. During certain special tests,
the lateral forces exerted at the top of the bag were also measured.
Filter bags used for the evaluation of selected cleaning regimes were
alternately loaded and cleaned on a cyclical basis until no discernible
changes could be detected in the pressure loss (or drag) versus time
or cloth loading relationship. During the course of these measure-
ments, no significant behavior differences were observed for bags oper-
ated anywhere between 2 and 24 hours. Therefore, a reasonably constant
base line was established for comparing various shaking systems. it
is emphasized, however, that under long term service conditions, a
gradual increase in residual drag often takes place at a rate deter-
mined by the combined properties of the specific dust/fabric system
and the working environment.
Since the amount of dust removed was variable from test to test, the
residual drag and residual cloth holdings also differed from test to
test. Therefore, in order to establish a standard rating technique
all filters were given an extra cleaning (called supercleaning in this
study) to reduce all fabric residual loading and drag values to the
same starting level. The special shaking regime for the supercleanine
process consisted of a continuous 20 second shaking period at about
11 cps and a 1-inch shaker arm amplitude. This resulted in a residual
holding of approximately 0.9 Ib./bag (430 grains/ft. ) at the start of
the experimental filtration cycles. The weight gain per 30 minute
2
filtration cycle (0.7 lb. or 300 grains/ft. ) represented the differenc
between the dust entering the baghouse hopper and that settling imme-
diately to the dust bin. The total dust accumulation on each bag at
the end of the loading cycle was approximately 1.6 pounds.
88
-------
A typical experimental data sheet is shown in Table 8. Detailed in-
formation for all measurements in this series and related technical
nomenclature is presented in Appendix H. The terms "taut" and "loose"
refer respectively to the top tension in pounds when the bag is attached
for filtration, and to the total bag weight when suspended free for de-
termination of dust loading. Most tests were conducted with the bag
tension adjusted to a near "slack" condition; i.e., with the top ten-
sion ranging from about the weight of the free hanging bag to 1.0 pound
in excess. These tension levels conform to those cited for related ex-
periments performed by EPA personnel in past studies.
Dust Removal versus Number of Shakes - The dust removal curves of Fig-
ure 23 indicate that the dust is dislodged in decreasingly smaller
increments as the shaking process continues. The unloading process
can be approximated by an exponential decay. All curves indicate that
a practical upper limit in dust removal is approached, regardless of
shaking frequency, after 200 to 250 individual shakes. The leveling
off point is reached with fewer shakes at the higher shaking frequen-
cies. It appears that extending the shaking beyond 200 to 250 shakes
represents a needless power expenditure, a reduction in gas handling
capacity, and in the long term a possible reduction in service life
due to extra shaking motion.
It was noted that the rate of dust removal during the first phase of
the shaking operation was significantly less for the lower shaking
frequencies. This suggested that with less applied shaking energy,
the initial flexing motion served mainly to diminish the bonding
forces prior to the actual dust dislodgement by subsequent shaking.
Generally, the form of the removal curves at low frequencies indicates
that more shaking is needed to reach a limiting level of dust removal.
From a practical viewpoint, increased frequency and/or amplitude ap-
pear to provide the better approach to cleaning.
89
-------
Table 8. SAMPLE TEST DATA TABULATION
DUST REMOVAL TEST
Date: 4 May 71
Test No. 4
LOADING DATA: Bag No
Inlet 3
cone. : 5 gr/ft CFM: 43.8
Initial tension (Ibs.) Taut: _2
4 Bag type: 10'
Wei
x6"
cotton!
1 Used
;
7o RH
,54 Loose: 2.03 Re- taut: '
Filtering
Time(min) Ap(in) Tension (Ibs.)
Notes:
0
1
10
21
30
1.40
1.58
2.28
3.08
3.65
8.12
8.12
8.12
9.33
9.50
CLEANING DATA:
Amplitude: ± 1 inches Frequency: 10 Q,PS
Cum.
••condi
shaken
0
5
10
IS
25
45
cpi
(Increc't
ave.)
10.
10.
11.5
10.0
10.1
Cum.
No.
•hake*
0
50
100
150
250
452
Shake
tension
range
7.98
8.12
8.30
8.5-8.6
8.6-8.8
Duit
off
(grama)
0
233
107
30
16
11
Cum.
dust:
off
0
233
340
370
386
397
Cum. 1
duat
off
0
31.8
46.5
50.5
52.7
54.3
Raiult
-V
-------
OJ
UNMAPPED COTTON BAGS
iOft. x6ln.
5O |OO ISO 200 250 3OO 350
TOTAL NUMBER OF INDIVIDUAL SHAKES,N
.Figure 23. Percent fly ash removal versus number of shakes and shaking
frequency at 1-inch amplitude
-------
Since the total cloth loading for the tests described in Figure 23 was
2
essentially constant at 740 grains/ft. , the actual weight of dust
removed is directly proportional to the cumulative percent removed.
This weight is a measure of the fabric filtration or dust capacity per
filtration cycle- This parameter shows how much dust can be collected
with a specified filtration and cleaning system. If the maximum cloth
loading is held constant, any cleaning process that reduces the residual
dust holding such as increased frequency, shaking amplitude and, up to
a point, the number of shakes will also increase the filtration capacitv
Experimental data shown in Figure 23 are in qualitative agreement with
earlier tests reported by Walsh and Spaite, Figure 24. Although
fewer shakes were required to achieve optimum cleaning, 60 to 80
versus 100 to 200 in the present tests, filtration capacity and shak-
ing frequency followed the expected relationships. The higher capa-
cities indicated in Figure 24 are attributed to somewhat coarser
fly ash size properties that provide a more permeable dust deposit and
to a slightly different shaking motion. Because certain key measure-
ment defining static and dynamic tension conditions were not made, no
further attempt is made here to compare the data of Figures 23 and 24.
Dust Removal versus Shaking Frequency - Examination of Figure 23 shows
that for a fixed number of shakes at a constant driving amplitude,
the amount of dust removed is an increasing function of frequency. This
result, which confirms earlier tests, is consistent with the basic
theory relating the dust removal forces to the fabric's acceleration.
During successive shaking intervals, it was observed that the bag am-
plitude and axial tension increased after partial dust removal had
taken place. This behavior is explained by the bag motion studies de-
scribed earlier in the BACKGROUND AND THEORY section (pages 13 to 36).
According to theory, near-resonant frequencies can exist only over
limited ranges of dust holding. As the bag becomes lighter, it can
pass into a resonant state causing the amplitude of motion and shaking
92
-------
CO
1,000
CJ
^ 800
•^
c
"o
600
o 400
o:
LJ
H-
."*- 200
TEST CONDITIONS
FILTER - COTTON SATEEN FABRIC
DUST - ELUTRIATED FLY ASH
FILTER VELOCITY - 3ft/mia
TERMINAL DRAG - I IN. h^O/ft/mln.
8.2 cps
5.2 cpft
(DATA EXCERPTED FROM REFERENCE 1)
20 40 60 80 100 120
TOTAL NUMBER OF INDIVIDUAL SHAKES.N
140
Figure 24. Effect of shaking frequency on filter capacity for 1-inch shaking amplitude
-------
tension to increase. In addition, removal of dust probably diminishes
the damping characteristics of the bag, thus promoting greater amplitudes
of motion. A point-by-point computation of the varying resonant fre-
quencies during dust removal indicated that during a typical cleaning
cycle, a bag is likely to achieve a resonant state for at least one
frequency.
One might infer that more vigorous cleaning might be accomplished by
continually adjusting or "tuning" the applied frequency to compensate
for dust loss. Results of such an attempt are shown in Figure 25
where in the upper curve frequencies were varied to produce a maximum
tension. In a separate test, the frequency was shifted to minimiae
the tensicn as shown in the lower curve. Since several different max-
im/? and minima might have been selected over the complete range of
shaking frequencies, maximum and minimum tension values were sought
which were as close as possible to 9 cps. The results suggest that
10 percent greater dust removal might be obtainable by holding to a
resonant condition.
During prior motion studies on clean bags, the bag amplitude, when com-
puted over a range of frequencies, was found to increase only 10 per-
cent or so above the average level at resonant frequencies. Thus it
is concluded that while resonance does enhance dust removal, the im-
provement is a very modest one. In view of the postulated complexity
of the instruments and controls necessary to maintain resonance, the
"tuning" approach does not appear justified at this tine. On the other
hand, it does appear advisable to estimate and operate at the resonant
frequency of the combined bag and dust weight during the latter part
of its cleaning cycle.
Dust Removal versus Shaking Amplitude - In a separate series of experi-
ments, the shaker arm amplitude was varied from 0.5 to 1.5 inches dur-
ing sequential tests, while keeping the shaking frequency constant at
8 cps. Complete loading and cleaning cycles were carried out at
-------
I
LJ
cr
-------
selected amplitudes as shown in Figure 26. It appears that amplitudes
less than 0.75 to 0.50-inch were ineffective, at least over the first
few hundred shakes. This is because the wave motion is rapidly damped
at small amplitudes such that a large part of the bag sees no motion
at all. From 0.50 to 1-inch, amplitude increases were very effective
in promoting dust removal. The waves reached the bottom of the bag and
underwent reflection so that generally uniform bag motion was obtained
Although further removal was attained at higher shaking amplitudes,
the actual increases were relatively small in comparison to the total
amount of dust removal.
The relationship between dust removal and shaker amplitude is indicated
in Figure 27 by several constant frequency curves. The measurements
suggest that below certain threshold amplitudes, ~ 0.5 inches, poor
dust removal is attained regardless of shaking frequency. Over a lim-
ited range of amplitude increase, the dust removal is also increased
significantly with the greatest improvement shown at the higher shaking
frequencies.
Although shaking amplitudes greater than 2 inches were not studied,
the shapes of the curves of Figure 27, indicate that there are prac-
tical upper limits beyond which further dust removal is small.
A related graphing of dust removal characteristics is given in Fig-
ure 28 for shaking frequency variations at several amplitudes. Again
similar trends are suggested; i.e., at very low frequency energy trans-
mission throughout the entire bag is ineffective leading to poor dust
removal.
There is no apparent limit to the improvement over the frequency range
tested3 although above 50 percent removal, proportionately less dust
is removed. Frequency is, of course, limited in practice by other
considerations such as bag tension and equipment vibration.
96
-------
10
SO 100 150 200 250 300 350
TOTAL NUMBER OF INDIVIDUAL SHAKES
400
Figure 26. Effect of number of shakes, 8 cps, and
shaking amplitude on dust removal from
10 ft. by 6 in. cotton bags
97
-------
1/2 I I 1/2 2
SHAKER AMPLITUDE, in.
Figure 27. Effect of shaking amplitude and
spaking frequency, 350 shakes, on
dust removal from 10 ft. by 6 in.
unnapped cotton bags
98
-------
60
50
_J
-------
The effects of amplitude and frequency are combined in the concept of
fabric acceleration in the following section.
Dust Removal versus Acceleration - Previous EPA investigators have
indicated that the dust holding capacity of fabric filters could be
related to the maximum acceleration, a , of the shaker arm:
m
a
2_2
m
= 4* rA (2.22)
In Equation (2.22), f is the shaking frequency and A the shaking ampli-
tude. According to the EPA tests performed with sateen weave cotton
bags and a fly ash aerosol, there existed a strong correlation between
filtration capacity and acceleration for accelerations less than
2
46.2 ft./sec. (1.44 g's). Above this critical acceleration level, no
significant improvements in filtration capacity were obtained. As de-
fined in the present discussions, filtration capacity describes the
quantity of dust that can be placed upon a filter during its normal
2 2
filtration cycle (grains/ft, or grams/m ). Under steady state opera-
tions, it must also depict the quantity of dust removed by the specific
cleaning method. In conjunction with the reported drag values (effective
and terminal) the filtration capacity enables one to determine the
collector size and power requirements for a given dust/fabric system.
Table 9. EXPERIMENTAL SHAKING CONDITIONS
Shaker arm frequency
(cps)
4.7
6.0
8.0
11.0
Shaker arm amplitude (maximum)
(in.)
1.0, 1.5, 2.0
l.C, 1.5, 2.0
0.5, 0.75, 1.0, 1.15, 1.5
0.5, 0.75
Note: Inlet dust concentration, 3.5 grains/ft.
Filtration velocity, 3 ft./min.
Filtering interval, 30 min.
Number of shakes, 350.
100
-------
Several experiments were performed during the present study to provide
a broader range of amplitude-frequency combinations with which to test
the dust removal-acceleration relationship. Table 9 indicates the
various shaking systems investigated.
results of dust removal tests under the shaking conditions des-
cribed in Table 10 are presented graphically in Figure 29. Here the
ordinate refers to the average bag amplitude and the abscissa depicts
the bag shaking frequency which is the same as the driving frequency.
The parallel broken lines, which represent lines of constant input
acceleration (1,3, and 10 g's), are drawn with a slope of -2 on the
log-log plots.
The bag amplitude-frequency coordinate for each cleaning condition
cited in Table 10 has been identified by a number showing the related
percent dust removal. Dashed lines representing approximately lines
of constant dust removal appear to have a slope of -2. The encircled
number flagging each dashed line is the best estimate of dust removal
efficiency along that line. Despite the broad point scatter, the
experimental data suggest that the amount of dust removed is a function
— 2
of the product (Yf ) and hence the average bag acceleration, as shown
in the following Equation:
R = (I) = * (4«2f2Y) (2.23)
Figure 30 shows the results of similar measurements except that the
shaking action was restricted to 40 shakes. Although the dust removal
followed the same pattern shown in Figure 29, the absolute quantities
removed were considerably lower in accordance with the reduced number
of shakes.
The point scatter exhibited in Figures 29 and 30 was attributed to
several effects; the neglect of individual differences in bag tension-
ing with respect to resonant frequencies, statistical variations due to
101
-------
Table 10. DUST REMOVAL VERSUS AVERAGE BAG AMPLITUDE, MEASURED AND
CALCULATED TEST PARAMETERS
Test
number
SI
S2
S3
S4
S5
S6
S7
S8
S9
S10
Sll
S12
S13
S14
S15
S16
S17
S18
S19
320
321
322
Shaker
amplitude
(in.)
I
I
I
I
1
1
1
1
1
0.75
1.15
0.5
1
1.5
I
1,5
2
1
1.5
2
Q.5
0.75
Shaker
frequency
(CPS)
9.0
7.0
10.0
6.2
8.1
7.2
7.8
7.5
7.7
7.7
8,0
7.9
7,5
7.8
5,9
6.0
5,7
4,7
4.5
4,1
10.9
10,7
Bag3
amplitude
average
fin.)
0.72
0.75
0.73
0.85
0.77
0.77
0.73
0.83
0.69
0.46
0.76
0.23
0.75
1.4
0.74
1-1
2.2
0,58
1,8
1.6
0.18
0,37
a
Bag
acceleration
average
re's)
5.96
3.75
7.45
3.34
5.16
4.08
4.55
4.75
4.18
2.79
4.97
1.47
4.31
8.70
2.63
4.05
7.30
1.31
3.72
2.75
2.18
4.33
Tension
Shaking
(lb.)
7.9
7.1
8.3
6.8
7.4
6.8
7.1
9.5
7.5
6.0
8.1
4.5
7.2
11.9
6.4
8.4
14.3
5.1
10.4
9.4
3.9
5,4
Staticb
(lb.)
3.1
2.9
2.5
2.2
2.7
2.2
1.9
6.1
2.9
2.5
2
2.6
2,5
2.0
1.9
2.0
2.0
1,6
1.6
1,6
Dust
removal
(7.)
44
33
54
36
42
39
42
39
41
27
47
4.2
40
45
24
40
47
8.2
31
27
11
34
Filter0
capacity
(gr^/ft.^
326
246
414
264
312
289
314
289
304
202
350
31
296
336
179
300
348
61
233
197
83
255
Residual
fabric
loading
(sr./ft.2)
373
494
324
476
428
440
426
451
436
538
390
709
445
404
561
440
392
679
507
543
657
485
o
N>
values,
b
At 0 cps, shaker arm vertical,
-------
24 \ 424744
41 v
8'2\ \\ X X
\ \ v \ * \
4 6 8 10
FREQUENCY.eps
Figure 29. Dust removal versus bag acceleration for cotton bags, 350
shakes. Broken line indicates constant acceleration
contours. Dashed line and circled number shows constant
dust removal contour. Uncircled numbers show actual
dust removal at Y-cps coordinates.
103
-------
2.0 —\
LJ
O
l5
10
00
O
O
LJ
I-
15
Q.
^
O
O
l>"
06
o
*.
O.i
\
\
*\
\\ x \ \
\\ \ \ \
v\ \ H \
\\ \ \ \
v\\ \ N
\\ NN x
\ \% \ \
4 5 6 7 8 9 O
FREQUENCY, cps
Figure 30. Dust removal versus bag acceleration for cotton
bags, 40 shakes. Broken line indicates con-
stant acceleration contour. Dashed line and
circled numbers show constant dust removal
contour. Uncircled numbers show actual dust
removal at Y-cps coordinates.
104
-------
accuracy limitations in instrumental methods or control of operating
parameters, and the basic assumption that the acceleration force,
acting normally to the bag surface, was the predominant dust removal
factor. It was also noted that the dust removal rate did not always
follow an exponential decay pattern when the total number of shakes
was small (< 75). This is shown by the inflection points on some of
the curves of Figure 23 at the lower shaking frequencies. It was con-
cluded that the initial shaking, particularly at the lower frequencies,
was effective in loosening, but not necessarily dislodging, the dust
deposit until at some critical threshold the bonding forces were
exceeded. Practically speaking, the above variations are relatively
unimportant since the net dust removal after 200 shakes undergoes
little change. In addition, the lower frequencies at which the abnormal
unloading effects were noted, would not ordinarily be selected for
optimum cleaning. Although it has not been possible to distinguish
between loosening and separation forces in the cleaning process, it is
postulated that increased fabric stretching with increased shaker am-
plitude might enhance the interfacial shear between the fabric structure
and the dust deposit.
Above certain critical acceleration levels, the present measurements
and those of earlier investigators suggest that dust removal (and
filtration capacity) increases only slightly. Therefore, the pos-
tulated linear relation between dust removal and acceleration should
be expected to hold only over a limited range, roughly 3 to 10 g's.
In Figure 31, the relationship between average bag acceleration and
filter capacity is given for all data shown in Table 10. These meas-
urements indicate that the filter capacity for low acceleration sys-
tems varies (on an empirical basis) with the square of the average bag
acceleration. Beyond the 4 g level, however, subsequent filtration
capacities are much less dependent on acceleration, varying approxi-
mately as the square root of g. Over the central range of 2.5 to 6 g's
the hypothesized direct relationship between filter capacity and
105
-------
500
2 5
WE RAGE BAG ACCELERATION, g't
10
Figure 31. Filter capacity versus average bag acceleration for
sateen weave cotton bags
106
-------
acceleration appears to prevail as indicated by the dashed 45° line.
The relationships suggested above are intended as rough approximations
only of a process too complex for rigorous description at this time.
Dust Removal versus Initial Bag Tension - It was shown earlier that
after the complete bag is set in motion, good cleaning will result
provided that the proper combination of acceleration and number of
shakes is used. It has also been shown that a minimum initial tension
is necessary to establish good bag motion. Although the effects of bag
tension on dust removal were not investigated separately, it appeared
that a minimum initial bag tension was required to obtain good cleaning.
This threshold level was considered to be equal to the combined tension
at the shaker arm produced by the free hanging, dust-laden bag plus an
added tensioning of 25 percent of the loaded bag weight. Further ten-
sioning might hasten the development of bag motion as the shaking begins,
but unless it were important to minimize the shaking period, the added
tension would afford no advantages. In fact, fabric wear as well as
dust penetration properties might be increased.
ntist Removal versus Cloth Loading - The effects of given cleaning sys-
tems (frequency, amplitude, and duration) on dust removal, described
in the preceding sections, were determined with a constant terminal
' 2
cloth loading, W_, on the fabric, approximately 760 grains/ft. . Under
these circumstances, the residual cloth holding, WD, decreases and the
K
holding capacity, VL, - W , increases as the energy applied during the
cleaning cycle is increased. This is illustrated by the composite
curve in Figure 32 that shows the differences in WR values for selected
cleaning regimes as well as the corresponding filter drag and resistance
characteristics.
The data presented in Figure 33 show that the residual dust loading for
any specified cleaning system is independent of the cloth loading on
107
-------
o
GO
I
QJ
u
£
to
a:
5.0
4.0
3.0
? 1.0
I I I
10ft. x 6 in. COTTON SATEEN
SHAKING REGIME (350 SHAKES)
A-AMPL. I in., FREO. 10.1 cps
B-AMPL.3/4in.,FREQ. 10.7 cps
C-AMPL. 2in., FREQ. 4.0 cps
D-AMPL. I in., FREQ. 4.9 cps
1.6
-se J
-------
o 100
3
IT)
Q 50
u.
o
O
cc
UJ
Q-
20
10
% REMAINING
% REMOVED
i
10 ft. x 4 in. cotton sateen
Shaking Cycle ~ 8 cps,
I in. Amplitude,
350 Shakes
300 450 600 750 900
TOTAL CLOTH LOADING, groins/ft.2
Figure 33. Dust removal versus cloth loading for fixed
shaking regime
109
-------
the filter at the start of cleaning. Both the fraction and absolute
quantity of dust removed, however, increase with cloth loading, the
latter the result of increased inlet dust concentration, filtration
velocity, or filtration interval.
With respect to measurements on a 10 ft. by 4 in. cotton sateen bag
(using an 8 cps, 1-inch amplitude, 350 shake cleaning system) no dust
removal was observed for initial cloth loadings less than 330 grains/
2 P
ft. . When cloth loading was allowed to increase to 800 grains/ft. ,
the same cleaning system caused approximately 60 percent (470 grains/
2
ft. ) of the total cloth holding to be dislodged. For all final cloth
loadings, however, represented in Figure 33, the residual cloth loading
was the same. On a fractional basis, the quantity of dust removed for
a specified dust-fabric cleaning method is expressed by the ratio
(WT - WR)/WR.
The data provided in Figure 33 and in earlier figures, in which the
percent of dust removed was used as an experimental performance param-
eter, is meaningful when a specific fabric has the same terminal dust
loading. For field applications of laboratory findings, however, the
effect of the mechanical shaking operation is best described by the
relationship between residual fabric dust holding and the average accel-
eration imparted to the bag as shown in Figure 34. According to the
data presented in this figure the residual dust loading, W_, varies
c\
inversely with the square root of the average bag acceleration over the
range 1.5 to 7.5 g's. These tests were based upon 360 individual shakes
per bag. Examination of the effect of varying the number of shakes at
different acceleration levels, Figure 35, indicates again that there is
little to be gained by too extensive a shaking period. Beyond 100
shakes, the dependency between residual cloth holding, W_, and number
of shakes, N, is relatively weak; i.e., WR - * (N).0'09 Although the
linear curve fits do not apply for shaking levels less than 100, the
practical consideration is that the dust removal is generally inadequate
and unpredictable for N values less than 100.
110
-------
1000
0
S500
o»
i
O
3
o
£
0
£200
100
©
UNMAPPED SATEEN
WEAVE COTTON
FLYASH AT 3.5 GRAINS/ft3
FILTER VELOCITY, 3H/min.
360 SHAKES.
1
2 5
AVERAGE BAG ACCELERATION, g's
10
Figure 34. Residual fabric loading versus average bag
acceleration
111
-------
CM
£1000
c
'5
1 500
O
_J
o
on
CO
s
200
13
O
V)
UJ
cr
CURVE BAG ACCELERATION-g's
X A E.7
A 8 4,6
O C 7.6
1
1
10
50 100 200
TOTAL NUMBER OF SHAKES, N
500
Figure 35. Residual fabric loading versus total number of
shakes at three acceleration levels far
unnaf>pe4 sateen weave cotton
-------
By combining the empirical relationships shown in Figures 34 and 35.
the residual fabric loading can be expressed in the following manner
as a function of both average bag acceleration, a, and total number of
shakes, N,
WR = 1433 (a)'0'5 (N)-°'°9 (2.24)
Equation 2.24 applies specifically to the fly ash aerosol and the sateen
weave cotton bags used in this study. The amount of dust that can be
accommodated by the fabric; i.e. its filter capacity, is the difference
between the selected terminal and residual fabric loadings. For a given
cleaning system, the filter capacity is usually limited by the maximum
allowable resistance across the fabric or the available fan capacity.
As discussed previously, knowledge of the K value for the specific dust-
fabric combination, in conjunction with the residual cloth loading and
related resistance, permit ready estimation of either the terminal fil-
ter drag and the terminal fabric holding, Figure 32.
Dust Removal versus Fabric Type - Further tests were performed with the
previously described fly ash aerosol to permit comparison of filtration
behavior of different fabrics. Again, inlet loadings were approximately
o
3.5 grains/ft, and the filtration velocity 3 ft./min. The bag dimen-
sions were also the same, 10 feet long by 6 inches diameter, as well as
the mechanical shaking system, 8 cps at 1-in. amplitude for 360 indi-
vidual shakes. Test results are shown in Figure 36 for two Dacron
weaves, plain and crowfoot, and a napped sateen weave cotton.
The curve shapes follow closely those for the unnapped cottons shown in
Figure 35. In fact, there does not appear to be any significant differ-
ence between the two cotton fabrics based upon the limited experimental
data. After extended use as simulated by the accelerated life testing
procedures, the residual dust holding capacity of all bags decreased
after 2 x 10 shakes. The observation that outlet dust concentrations
were generally higher after 2 x 10 shakes suggested that irreversible
stretching in the media had enlarged the pores.
113
-------
1000
- N
CM
V)
c
'5
o»
§
Q
UJ
cc
CURVE FABRIC
|- 1,2 NAPPED SATEEN WEAVE COTTON
3,4 PLAIN WEAVE DACRON
5,6 CROWFOOT DACRON
N = NEW, < I04 SHAKES,O=OLD,2*IO
i i . SHAKES.
50 K)0 200 500
TOTAL NUMBER OF SHAKES,N
Figure 36. Residual fabric loadings for various fabrics with
fly ash aerosol (8 cps, 1 in. amplitude shaking)
114
-------
In contrast to the cotton fabrics, the steeper slopes of the curves for
Dacron media, Figure 36, indicate that the dust release process is more
gradual. Approximately 20 percent of the total dislodged dust was
removed between the 100th and 360th shake of the Dacron bags in contrast
to only 10 percent for the cotton fabrics over the same shaking interval,
Dust Removal versus Dust Type - Limited tests were performed to inves-
tigate the filtration characteristics of other dusts with the previous-
ly described cotton and Dacron fabrics. Measurements with a resuspended
talc dust having approximately the same size parameters determined for
the aerosolized fly ash (see Table 4) are shown in Figure 37. The
residual fabric loadings varied approximately as the -0.2 power of the
number of shakes. The approach to equilibrium cleaning for both cotton
and Dacron media appeared to follow the same pattern in contrast to
the distinctly different slopes shown in Figure 36 for fly ash
filtration.
The difference in release characteristics for fly ash and talc may
be associated with the relative strengths of adhesive and cohesive
bonds. One possible explanation is that a higher adhesive force be-
tween dust and fibers may lead to a gradual spalling-off of dust rather
than a scaling-off at the dust/fabric interface.
The relationship between filter drag and fabric loading given in
Figure 38 shows that residual fabric loadings and residual (effective)
drags were lower for talc dust than for fly ash. This comparison
is also indicated in Table 11 along with the filter capacities for
both talc and fly ash with cotton and Dacron fabrics. Filter capac-
ities were based upon a filtration velocity of 3 ft./min. a 360 shake
cleaning cycle at 8 cps and 1-in. amplitude, and a terminal drag, S ,
of 1-in. water/ft./min.
According to Figure 38, the K values for talc are generally in the
range of 2.5 to 3 times greater than those for fly ash. The fact
115
-------
500
CM
«#-
V.
CO
c
'6
o 200
z
o
o 100
o:
00
o
CO
UJ
(T
50
o I
A 2
x 3
NAPPED SATEEN
WEAVE COTTON
UNNAPPEO
WEAVE COTTON
PLAIN WEAVE DACRON
NEW BAGS,
-------
1.2
c 1.0
00,8
o 0.6
<
oc
o
£ 04
0.2
OOI
002 OO3
FABRIC LOADING, lb*./ft.s
0.04
OX)5
Figure 38. Fabr-lc loading versus filter drag for woven cotton and Dacron bags with a
talc aerosol (360 shakes at 8 cps and 1-in. amplitude)
-------
Table 11. FILTRATION CHARACTERISTICS OF VARIOUS DUST/FABRIC
COMBINATIONS3
Fabric
Unnapped cotton
Napped cotton
Plain Dacron
Crowfoot Dacron
Effective drag
(in. H20/ft./min.)
Fly ashb
0.67
0.23
0.35
0.43
Talcb
0.3
0.13
0.08
-
Filter capacity*1
WT - wR
(lbs./ft.2)
Fly ash
0.024
0.058
0.059
0.035
Talc
0.019
0.027
0,032
m
Mechanical shaking, 360 shakes at 8 cps and 1-inch amplitude.
]j
See Table 4 for aerosol size properties*
£
Filter capacity, WT - WR, at a terminal drag of 1 inch H O/ft./min,
that the filter capacities appear in the ratio of 1.7:1 is attributed
to the greater ease with which the talc is dislodged from the fabric.
As a result, filtration is begun at a lower resistance or drag. One
explanation for the lower residual fabric loadings may be that a lare
fraction of the talc reaching the filter face remains as a superficial
deposit. If, in fact, the estimated pore deposits for talc are com-
posed of loosely agglomerated particles as suggested previously, $.£
follows logically that the superficial layer may predominate.
The limited data obtained for talc and fly ash filtr*!tt
-------
that the dust unload properties versus number of shakes does follow dif-
ferent patterns for talc and fly ash. Generally, the results indicate
that more studies are required to pinpoint critical relationships where
dust/fabric interactions are involved as shown by differences in K values.
Attempts were made to establish filtration parameters for a resuspended
amorphous silica with sateen weave cotton fabrics. Generally, no ap-
proach to steady-state filtration could be reached until inlet concen-
3 3
trations were reduced to 0.075 grains/ft, (versus 3.5 grains/ft, for
talc and fly ash tests) and filter velocities to 1.5 ft./min. Finally,
at 11.6 cps shaking frequency and 1-in. amplitude, 360 shakes appeared
sufficient to stabilize the filtration process. It was noted that
the dust shaken from the bags was composed of chip-like agglomerates
such as those often found when filtering zinc oxide fumes. Examina-
tion of the bag suggested that some areas were poorly cleaned, leading
to an appreciable dust layer that displayed surface cracks when the
bag was flexed. The results of the silica tests are too limited to
warrant any but the obvious conclusions that certain fine particles
are very difficult to filter and that existing shake cleaning methods
require considerable improvement.
Summary of Dust Removal Studies - Test results given in this se-tion
have indicated the following relationships and/or interactions among
dust removal characteristics, dust type, fabric type and the intensity
and duration of mechanical cleaning:
• After 200 shakes, practically all of the dust
removable by a specified combination of shaker
frequency and amplitude was dislodged. Tests
indicated, however, that the fraction of dust
removed at 200 shakes ranged from 80 to 95
percent of the dust dislodgable by extending
the cleaning period to 360 shakes (~ 1 min.
at 6 cps). The variations cited above depended
upon the single or combined effects of dust and
fabric properties.
119
-------
• Although dust removal was shown to increase
with an increase in either shaker amplitude or
frequency, the most useful parameter describing
dust removal was the average bag acceleration.
The above term is readily defined in terras of.
the average bag amplitude, Y, and the shaking
frequency, f.
•• The residual fabric loading for a specified*
dust/fabric combination correlated well wttft
the reciprocal of the square root of the aver-
age bag acceleration (expressed in g's) over
the test range of 1.5 to 7 g's.
• The residual fabric loading and residual filter
drag also appeared as unique functions of the
dust/fabric combination under study and the
specified cleaning approach. Regardless"of
the initial fabric loading, application of
the same mechanical shaking routine (fre-
quency, amplitude, and total number of shakes)
led to the same residual cloth holding.
• Filter capacity, defined as the amount of dust
that can be collected upon the fabric surface
during each successive filtration cycle, is a,
parameter that depends upon the acceptable
terminal filter drag, available fan capacity,
and the ability of the fabric to withstand and
operate effectively at a specified pressure
drop across the media.
» Because of the limited tests~ per farmed with
dusts other than fly ash, it has not been
clearly determined whether the shaking rela-
tionships and the effect of bag acceleration-
on dust removal can be treated independently,
irrespective of dust and/or fabric ty-pea.
• The measurements presented in this and: ewriier
sections enable one to estimate average resis-
tance characteristics (power costs) and fabric
area (collector size) for dust and fabrics very
similar to those used in the experimental program.
In the following section, important performance parameters relatimr
to the filter effluent properties are presented.
120
-------
Collection Efficiency and Emissions
The following filtration and cleaning experiments were designed to
determine effluent dust concentrations and particle size properties
under the same test conditions described previously for inlet dust
loadings, bag types, fabric loadings and methods of cleaning.
One test procedure involved the measurement of particulate emissions
immediately upon resumption of filtration following the use of the
superclean cycle described previously. In a second testing procedure,
particulate properties were determined after steady state filtration
conditions had been established for the selected cleaning conditions.
No supercleaning was employed in the latter case.
To permit maximum accessibility for instrument placement, mechanical
shaking systems were operated under positive pressure, (inside to
outside) filtration. Thus, it was necessary to surround the lower
8 ft. of a 10-ft. bag by a grounded metal enclosure with a small
exit port for venting a representative fraction of the effluent aero-
sol. Failure to use this design would have resulted in dilution of
the bag effluent with the ambient air prior to sampling. In those
—8
cases where the particulate concentrations were very low, 10 to
o 3
10 grains/ft. , the B&L sampler was the only effective measuring
instrument. On some occasions, the RDM sampler provided an index of
mass concentrations when exit concentrations were in the 10 to
-2 3
10 grains/ft. range or larger.
The use of these instruments and the quality of the resulting data
have been discussed under Aerosol Concentration and Particle Size,
pages 53 to 60. Again, because of the rapid response and high sen-
sitivity of the B&L unit, it was possible to categorize the effluent
aerosol in five size fractions within a 0.5 min. sampling interval.
By repeating this process as long as necessary throughout the filtra-
tion cycle, the initial and subsequent particle concentrations could be
121
-------
measured until they decreased to levels below the detection limit
—8 3 -•••••..>
~ 10 grains/ft. . This lower limit of detectability was usually
reached after 4 to 5 min. of filtration.
As described previously, it was possible to estimate .rpughly the rtprtal
mass concentrations at various times during the filtration cycle bv
summing up the computed masses within each size fract^^pp. ,Qn a rre|la,-
tive basis, the changes in number concentration -And .si^e
were probably in agreement within at least a factor p/f -two.
sons of B&L and RDM measurements on an absolute basis ,indica-ted
the ratio of mass concentrations, RDM/B&L, ranged from 2 rtp 5 on -t
average. With respect to this study, where concentration qhan$as
five orders of magnitude were not uncommon, the discrepancies ^
B&L and standard measuring devices , such as the RDM, .were considered
unimportant.
General Observations - The experimental results described ,in this
section suggest that the particulate emissions from a filter can :be
attributed to some combination of the following sources:
1. Inlet dust that because of its small size
passes directly through the filter, usually
in progressively smaller amounts as the
filter pore structure becomes plugged.
2. Dust that migrates through the filter iby
successive deposition and re.-entrainment
under the combined effects of aerodynamic
and mechanical (vibrational) -forces. -Such
dust penetration is often referred ;tp ,f$
"seepage" in commercial parlance, it jaay
be more pronounced in the case of JmrlfeL.-
filament yarns, spherical or smooth sur_-
faced regular particles, and in the absence
of electrostatic or other forces enb-anc-ing
adhesion or cohesion.
3. Dust dislodged from the shaken fabric during
cleaning that has penetrated to the clean
air region. Resumption of air flow fljj,she.s
out the clean air side of the system, often
producing a visible puff of dust.
122
-------
Dust loosened during the cleaning process
whose bonding to the fibers or interstitial
dust structure is not sufficiently strong
to resist the combined dislodging forces
(aerodynamic and mechanical flexure) when
system air flow is resumed.
Although it appears difficult to weigh the relative importance of the
above sources, it is suspected that items (3) and (4) may account for
a large fraction of the total mass emission, probably in the form of
a few large particles, whereas items (1) and (2) are responsible for
the discharge of most of the submicrometer material.
One can infer from the above that effluent concentrations based upon
long-term measurements represent the combined, unresolvable effects
of these four emission sources. On the other hand, short-period
measurements, of the order of seconds to minutes, provide an oppor-
tunity to identify the dust source and to assign reasonable values to
particle sizes and number concentrations.
The four suggested sources of dust emission contribute in some degree
to the total emission from all types of fabric filters. The predomi-
nant source with respect to particle number is probably emission during
g
filtration. According to theory, this emission should relate inversely
to the pressure drop across the filter, since both are dependent on
the degree to which the pores in the fabric and dust deposit have
become filled and sealed. In other words, the total emission of dust
per unit area of filter over a given period of operation should relate
to the weight of dust on the filter.
Based upon the relationship between adhesive and/or cohesive forces
and particle size, it appears that particles collected singly on the
filter surfaces will most likely be dislodged in the form of agglom-
erates. One could conceive of the extreme case where a freshly
123
-------
generated fume composed of particles less than 0.5 (jm could very readily
produce a filter effluent composed of much coarser particles despite a
high overall mass collection efficiency.
Effluent Concentration versus Filtration Time - Comparison of the
concentration properties of filter effluents when the inlet aerosols
j *5
consisted of ambient dust at < 10 grains/ft, and fly ash at 3,5
3
grains/ft. , respectively, is given in Figure 39. The time scale in
Figure 39 and many similar figures has been terminated at 5 rain, because
subsequent emissions over the remainder of the 30 min. filtration cycle
used in most tests were below the limit of detection, ~ 10 grains/ft. .
In both cases, steady state conditions had been established for fly ash
filtration in conjunction with a 360 shake cleaning cycle at 7,2 cps and
1-in. amplitude. The results indicated that effluent concentrations at
the resumption of filtration were roughly equivalent to the prevailing
atmosphere levels, regardless of the inlet aerosol. Furthermore,
both effluents decayed by more than three orders of magnitude within
5 minutes time and both decay curves suggested that the total mass
emissions over the filter cycle were similar and hence practically
independent of the inlet concentration to the filter.
Since very little dust accumulation was expected during atmospheric
dust filtration, the initial concentration level and the subsequent
decrease in effluent concentration were attributed to two factors.
The first was the dislodgement of interstitial and surface deposits
loosened by prior shaking, and the second was the transient inlet
loading produced at the resumption of air flow by material dislodged
from the inlet duct system.
The fact that resumption of full dust load, Curve B, Figure 39, led
to emission levels of the same order as the "clean air" tests points
out that effluent concentrations are not related in any simple fashion
to inlet loadings as one finds with many centrifugal collectors.
124
-------
1000
A«ROOM AIR FILTRATION
-------
Actually, the effluent concentration appears to be almost independent
of inlet dust loading. In addition, the dust that penetrates the fil-
ter probably represents a composite of inlet particles that pass un-
changed through the filter plus particle agglomerates and perhaps
discrete particles that are swept from their former deposition sites
by the entraining air stream.
It is also important to note that approximately 90 percent of the total
quantity of dust emitted during the complete 30 min. filtration cycle
was released during the first minute of filtration. During this
period the filtration surface underwent partial restoration. Similar
2
results were shown by Durham and Harrington at 50 and 60 percent
relative humidities.
The dust penetration characteristics shown in Figure 39 typify nearly
all tests with fly ash and cotton sateen fabrics in which the effluent
concentrations fell to levels below the limit of detection by the B&L
instrument in about 5 minutes. On the other hand, similar tests with
Dacron fabrics showed a continuing emission throughout the entire
filtration period.
Figure 39 shows an initial fly ash weight penetration of about 0.006 p«r.
cent, which means that initially only one out of every 33,000 enterii^
particles passed through the filter. The above statement is based
upon the simplifying assumption that the aerosol is monodisperse.
The emission then decreased about one order of magnitude for every
2
6 grains/ft, deposited. If distributed evenly among the number of
2
yarn interstices (about 5500 pores/in. ) in this cotton fabric, a
deposit of about 30,000 average-sized particles per interstitial
pore would result. Thus, it appears that the probability of a single
particle passing through a pore is reduced one order of magnitude
following each successive filtration interval of about 0.55 minutes.
This assumes that all filtration and emission occurs at the inter-yarn
pores, which may not necessarily be true in all cases.
126
-------
2
It was estimated that a fabric dust holding of 6 grains/ft, on a per-
pore basis would occupy a volume characterized by an 85 tarn cube. In
comparison, the inter-yarn pores of the fabric with their residual
loading were estimated to be of the order of 75 jam diameter, based on
the residual drag.
Despite the broad assumptions made in the above analysis, it is easy
to visualize why the dust depositing within the first few minutes of
filtration is so effective in reducing particulate emission levels,
Figure 40.
Effluent Concentration versus Particle Size - Upon restoration of the
filter surface, the larger particles are the first to be effectively
retained as shown in Figure 40. Because of difficulties in performing
these measurements, the data points may appear quite scattered as shown
in Appendix H. The larger particles, > 2 ^m, are emitted for such brief
periods as to make accurate measurement difficult. Whether they come
from the surface of the fabric as seepage dust or are actually pene-
trating particles is difficult to ascertain. Based upon the ambient
dust filtration results described in Figure 39, it is believed that
dislodgement of particles loosened by prior shaking may constitute a
significant fraction of the early, < 1 minute, emissions. Discharge of
the large particles is usually diminished at a greater rate than for
the small particles and with cotton sateen filtration not even the
smallest particles (> 0.3 |_im class) could be found after 3 to 5 minutes.
Other fabrics, mainly Dacron weaves, apparently sealed incompletely
such that particles greater than 0.3 \m continued to penetrate through-
out the 30 minute filter cycle, following a moderately rapid initial
flhe lowest equivalent mass concentration detectable by the B&L device
was estimated to be 10"8 grains/ft. , roughly 103 times lower than
typical clean air concentrations of approximately 10"5 grains /ft.-3 or
127
-------
decrease in penetration. Hie larger particles, 0.5 to 1.0 \.m, however,
usually ceased to be emitted after a few minutes. Therefore, the
poorer the filtration, the greater the divergency of the particle size
curves in Figure 40.
A review of a number of tests with cotton fabrics (Appendix H) showed
that the size distributions of the emitted dust were fairly similar in
all cases. The size distributions were truncated at the high ends by
some limiting characteristics of the filter in its just-cleaned state,
probably the size of the largest pore in the media. Smaller particles
appeared to be emitted in concentrations controlled by the parent
aerosol size distribution. This applied not only immediately after
cleaning, but throughout the filtration process, with the concentrations
of all but the extreme particle sizes diminishing at about the same
rate. As time progressed, first the largest size and then the next
2
smaller passed below the detectable range (150 particles/ft. ). The
rate of decrease of individual particle sizes was about the same as that
for the total mass emission; i.e., roughly one order of magnitude for
2
each additional deposit of 6 grains/ft. .
Hie decrease in effluent size properties with filtration time is in-
dicated in Figure 41 for times ranging from 0.7 to 2.8 minutes. After
2.8 minutes, the indicated size distribution parameters for the effluent,
CMD * 0.75 urn, ag =2.1, appear to be approaching those for the typical
room air, CMD = 0.5, ag = 1.5, as determined by membrane filter sampling
and oil immersion, light field microscopy.
Based upon an average efficiency in the range of 99.9999 percent for a
typical 30-minute filtration period (Curve B, Figure 39) the following
rough estimates of fractional size efficiencies can be made:
128
-------
I 2 3
FILTRATION TIME.mln.
Figure 40. Calculated effluent number concentration versus
time and particle diameter for fly ash
filtration with sateen weave cotton. Filtering
and cleaning conditions of Figure 39,
129
-------
4.0
u>
o
NOTE:
2.0
I in AMPLITUDE.
BAG TIME AFTER SHAKING
min.
2.1 mm
2.8mm
0.7 mm.
2.1 mm
J L
J.
J I L
_L
J.
0.5 I
10 30 50 70 90
PERCENT NUMBER! STATED SIZE
95 96 99 99.5
Figure U. Changes In effluent sire properties with filtering time for
nev (<\^ shakes) and old (2 x 107 shaVes) baRB. Sizing
b-y optical, Ht«l>, counter
-------
Size range, urn . Efficiency, percent
0.3 - 0.5 99.9970
0.5 - 1.0 99.9965
1.0 - 2.0 99.9992
2.0 - 5.0 99.9999+
The above calculations derive from cascade impactor measurements of
mass distribution for the inlet aerosol and effluent size properties
estimated to be similar to atmospheric dust, MMD = 1 \im, ag = 1.5.
Although the larger sizes were collected at the higher efficiencies as
theory would predict, highly efficient collection was also observed for
particles as small as 0.3 \m.
Effluent Concentration versus Shaking Amplitude - Effluent concentra-
tions were found to be strongly dependent on shaking amplitude based
on a series of tests with 10 ft. by 6 in. unnapped cotton sateen bags.
This is most evident in Figure 42a which shows differences in ef-
fluent concentration when filtering room air after shaking the bags
at amplitudes of 1/2, 1 and 2 inches. A similar but less pronounced
trend was indicated when filtration was resumed with fly ash at inlet
loadings of 3.5 grains/ft.3, Figure 42b. The best explanation seems
to be that the residual dust cake experiences more breakup with greater
stretching of the fabric, thus creating larger gaps between the resid-
ual deposits and prolonging the sealing process. It was noted that a
factor of two increase in amplitude caused a factor of ten increase in
average effluent concentration. This difference could be of consid-
erable importance in marginal applications.
Effluent Concentration versus Shaker Frequency - In a series of related
tests, shaker frequency was found to have little effect on effluent con-
centration. Results of tests at three different frequencies and a con-
stant of 1-inch amplitude are shown in Figure 43. The higher shaking
tensions and greater rates of fabric deformation associated with higher
131
-------
to
I04
10'
o
x
CM
«*-
to
o (O2
o>
O
cc.
H
z
W
o
o
o
10
(0
Id 10*'
_J
u.
u_
Id
.?••* •
UNNAPPED COTTON SATEEN
|0ft.x6in. BAG
CLEANING CYCLE-360
SHAK£S,7-5cps
in. AMPLITUDE
.L
- 10
3-5 groins/ft3
See (a)
- 10"
- 10'
,-3
- 10
,-5
-6
- IO
-lio-7
-------
OJ
VI
c
O
z
UJ
O
O
O
UJ
IOJ
•IOZ
10'
10°
10
a K>'a
10
qroin»/ft.s
UNNAPPEO COTTON SATEEN
10 ft. xGin. BAG
CLEANING CYCLE -226 TO 387
SHAKES AT I ift. AMPLITUDE
1387
SHAKES}
ll.3cp*
7.l5cps
(357 SHAKES)
I
3-5
Se« (a)
5cpt
113 cp*
I
43ept
1
10'
10
'* 1
a.
io-
IT
UJ
o
10'
10
-7
2 3
TIME, min.
2 3
TlME.min.
(a) ROOM AIR FILTRATION
(b) FLY ASH FILTRATION
Figure 43. Calculated effluent concentration versus shaking frequency at constant amplitude
-------
shaking frequencies did not appear to influence dust penetration even
though more dust was removed from the fabric at the higher frequencies.
The finding that the degree of stretching and not the rate of fabric
flexure determined filter emission levels was consistent with the view
that the effective pore dimensions after cleaning were controlled by
the cracks in the dust cake.
A summary of test measurements showing the effects of amplitude and fre-
quency variations in given in Table 12. On the average, effluent
concentrations during the first minute of filtration were 30 times
greater than those for the complete 30-minute filtering cycle, irre-
spective of the inlet concentration. The highest tension level was
associated with the largest shaking amplitude and also the largest ef-
fluent concentration.
The average emissions for the fly ash/sateen weave cotton system were
at first suspected of being too low, possibly because of the limita-
tions of the B&L instruments. This impression was based partly on
several tests reported by Draemel in which outlet concentrations f°r
several fly ash/Dacron systems were found to be in the 10~ grains/ft*
range. Examination of measurements on several different sateen weave
cotton, industrial filter systems, Figure 44, showed that CCA fly «*h
data were in excellent agreement with the field data. Since the
size properties of the foundry dust, Figure 44, were similar to those
for the GCA fly ash, it appeared that the mass concentration data devel
oped from B&L number count measurements were well within an order of
magnitude of the true values.
Effluent Concentration versus Bag Life and tag Stretch - As part of
the accelerated life testing procedure for cotton sateen filter bags*
particulate measurements were made periodically in an attempt to de-
tect potential signs of bag failure. After 2 x 10 shaking cycles,
equivalent to 3 to 5 years of normal bag usage, no physical damage
could be discerned visually. Effluent concentrations for fly ash
134
-------
Table 12. COLLECTION EFFICIENCY AND EFFLUENT CONCENTRATIONS FOR VARIOUS CLEANING REGIMES
a
Shaking
system
cps
7.5
7.5
7.5
4.3
7.15
11.3
Ampl.
(in.)
2
1
1/2
1
1
1
<*
Average effluent concentration - grains /ft.
Fly ash filtrationb
First minute
3 x 10'4
3 x 1
-------
COTTON; SAT-GEN
MECH:ANICALL.Y' SHAKEN;
Q FOUNDR.Y. DUST;S; (REE.
; C 0 Ali. FLY/ AS.H] (THIS STTUDTK
O W.O;OU. FELrT.ltBESrl>R'EV'ERSE
J.ET(:BLOW;RING:) A.BRASI.V.E
- SiC'2.AI2Q5.,B-BC tRSr.
0.01
INLET
l.0>
ZJ
i,,, groins^ fft,
Figure 44..
Ihletr concentration, versus percent, weight penetration,
ambient temperatures*
L36
-------
increased gradually throughout the shaking process, however, as indicated
in Table 13. The factor of two or three increase is almost negligible
when compared with some of the other changes encountered in this study.
During the extended shaking process, the bags underwent progressive
stretching. This must have opened the inter-yarn pores in the longi-
tudinal direction, possibly accounting for the modest concentration in-
crease. Tests with the more highly tensioned bag indicated slightly
higher outlet concentrations.
Table 13. EFFLUENT CONCENTRATIONS VERSUS AVERAGING PERIOD (FLY ASH;
UNNAPPED COTTON SATEEN BAGS, 10 IN. X 6 FT.)
Number of
cycles"
6 x 106
10 x 106
15 x 106
20 x 106
Average effluent concentration - grains /ft.^ x 10b a
Taut bagc
First minute
750
750
500
900
30 minutes
25
25
17
30
Loose bag°
First minute
250
350
450
30 minutes
8.7
8.3
12
15
Measurements made after loading filter to ~ 700 grains/ft. , and
then cleaning.
Shaking cycle 8 cps, 1-inch ampl. 360 shakes.
CStatic tension = 3.1 Iba., shaking tension 6.5 Ibs.
Static tension - 1.3 Ibs., shaking tension 4.5 Iba.
Efficient Concentration versus Fabric Type - In conjunction with life
tests on four types of fabrics, comparative emission measurements were
4
made on equilibrated nearly new bags (< 10 shakes) and on equilibrated
bags shaken 2 x 10 times. Generally, the two cotton fabrics had very
high efficiencies compared with the Dacron media. All four fabrics,
however, showed efficiencies in excess of 99 percent, and the Dacron
fabric performance would be completely acceptable in many applications
137
-------
despite the higher effluent concentrations. Test conditions and results
are summarized in Table 14 and Figure 45.
The napping of the cotton sateen fabric appeared to improve: its perform-
ance .. The net emission was; slightly lower,, although it took slightly
longer for the napped surface to, seal... The effect of the napped sur-
face was almost insignificant when compared with the emissions differ-
ences noted for the Dacron fabrics. Although there is no* reason why the
weight of the dust on a fabric should uniquely determine the penetrating
properties of a dust, the general correlation shown in; Figure 46 was ex-
pected. Deviations from the mean curves- are readily explained: by differ-
ences in deposit density, fabric weave, pore structure: and other physical
factors.
Efficient Concentration versus Dust Type - Summary measurements in
Table 13 indicate that the effect of extended cleaning, up to 2 x 10
shakes,, on various fabrics was to increase emission levels by a factor
of 2 to 4 over the' outlet concentrations for new filter media. In some
eases, the greater penetration may have resulted from lower residual
deposits. Additionally, the stretching of all fabrics: over the life
testing period undoubtedly produced a more open pore structure that
contributed to increased dust emissions.
Limited tests were performed with talc and silica to study the effect
of particle^ properties on dust penetration., Test results shown in
Table 15 and Figure 47 indicate that higher efficiencies were obtained
with talc than for fly ash. The greatest improvement was noted for
the plain weave Dacron bags although the talc/unnapped cotton emissions
were also slightly less than thos;e: for the fly ash/unnapped cotton
combination tested previouslyv On the other hand, a slightly higher
talc emission was noted with a napped cotton bag than that determined
for fly ash. Estimated resddual drag values for talc1 and fly ash
tests, 0»27 and 0.13 in./ft./min., respectively,, appear consistent with
the higher emission level for fly ash although the difference may not
138
-------
Table 14. FLY ASH EMISSION PARAMETERS FOR DIFFERENT 10 FT. X 6 IN. FABRIC BAGS*
Nearly new fabrics < 101* shakes
Residual drag
Residual dust
Peak exit concentration
Avg. exit concentration for 1st min.
Avg. exit concentration for 30 min.
Total emission per 30 min. cycle
Avg. efficiency, 30 min.
Well used fabrics 2 x 107 shakes
Residual drag
Rcsixhuil (hint
Stretch, 20 x 10 period
Peak exit concentration
Avg. exit concentration for 1st min.
Avg. exit concentration for 30 min.
Total emission per 30 nin. cycle
Avg. efficiency, 30 min.
Unnapped
cotton
0.47
437
136
50
1.9
180
99.99994
(0.60)
397
1.40C
710
170
5.7
510
99.9998
Napped
cotton
(0.20)b
475
26
20
1.2 •
110
99.99997
(0.53)
355
1.20
80
(60)
2.5
222
99.99993
Plain-weave
Dacron
(0.17)
219
3,600
1,900
1,635
147,000
99.95
(0.00)
120
0.79
17,000
(12,000)
8,200
738,000
99.77
Crowfoot
Dacron
(0.37)
97
10,400
( 6,000)
( 2,330)
(210,000)
99.93
(0.02)
77
0.69
20,000
(15,000)
10,400
939,000
99.71
in. /ft. /min.
gr./ft.2
gr./ft.3 x 106
gr./ft.3 x 106
gr./ft.3 x 106
gr./ft.2 x 106
percent
in. /ft. /min.
St. /ft.2
inches
gr./ft.3 x 106
gr./ft.3 x 106
gr./ft.3 x 106
gr./ft.2 x 106
percent
o
Measurements made after equilibration at 3 £t./min. velocity and 3.5 gr./ft. inlet loading. Clean-
ing cycle of 360 shakes of 1-inch amplitude at 8 cps.
^Parenthesis indicates questionable data.
CBag tension adjusted periodically to 3 Ibs.
-------
10
<0
o
s
o»
z
o
cc
UJ
o
z
o
I-
z
UJ
u.
u.
UJ
PLAIN WEAVE DACRON
CROWFOOT DACRON
NAPPED COTTON
UNNAPPED
COTTON
8 12
FILTRATION
16
2O
2B
Figure 45. Comparative effluent concentrations for fly a»h with
different fabrics
140
-------
D IU
o
X
•s.
(A
O>
uF
° IO4
-------
5?
ro
Table 15. TALC AND SILICA EMISSIONS FRCM NEW « 104 SHAKES) COTTON AND DACRON BAGS
Residual drag
Residual dust
Peak exit concentration
Avg, exit concentration for 1st roin.
Avg, exit concentration for 30 min.
Total emission per 30 min. cycle
Avg, efficiency, 30 min.
Talc3
Unnapped
cotton
0.27
97
32
12
0.38
34
99.99999
Napped
cotton
0.13
152
92
40
1.6
140
99,99996
Plain -weave
Dacron
0.047
96
713
358
14
1,300
99.99996
Silicab
urmapped
cotton
1.3
,.b
0,012
~0.001
-0,0001
-0.01
> 99,99999C
in, /ft, /min.
gr./ft, 2
gr./ft,3 x 106
gr./ft.3 x 106
gr./ft.3 x 106
gr./ft.2 x 106
percent
Inlft loading ~ 3,3 grains/ft,3, filter velocity 3 ft, /min., 20 wia, filter cycle, gleaning 8 CDS,
1-in, amplitude and 360 shakes,
^Equilibrium not «tutn«4,
w r#duc»4 tgl«t loading of Qtl grain/ft.
-------
10'
io2
10'
UJ
o
o
s
10
Curve Dust
Fly Ash
Talc
Fly Ash
Talc
Fly Ash
Talc
Fabric
Plain Weave Docron
ii H n
Napped Cotton
Unnapped Cotton
II M
\
\
\
Inlet Looding, ~ 3.5 gr./ft.3
Filter Velocity, 3ft7min.
Cleaning, 360 Shakes
(3)8 cps, I-in Amplitude
\
8
12
16
20
24
FILTRATION TIME, min.
\'
\
\
28
Figure 47. Comparative effluent concentrations for talc and fly ash
with cotton and Dacron fabrics
143
-------
be significant. The same situation seems to apply to the talc emissions
for napped and unnapped cotton, Table 15. Again, the outlet concentra-
tions and total emissions varied inversely with the residual drag levels
as expected.
As indicated previously, it has been postulated that the characteris-
tically fluffier deposit formed by the talc (roughly one quarter the
bulk density of fly ash) probably leads to less interstitial and greater
superficial deposition. Therefore, a larger fraction of the dust is
released during cleaning. Because of its irregular form, however, there
is assumed to be less talc migration or "seepage" during the actual
filtration cycle.
Exceptionally high efficiencies were noted during the si liea/unnapped
cotton tests although the minimal holding capacity, high resistance
and cleaning difficulties did not make the system attractive from the
power cost or capacity perspectives. Actual outlet concentrations
were estimated to be several orders of magnitude lower than those
determined for the other test dusts. One can infer that the high resis-
tance and low dust holding capacity demonstrated by the silica/unnapped
cotton system is related largely to the extremely low bulk density of
the silica pigment, ~ 0.04 g/cc.
Based upon the limited tests performed with talc and silica and the
numerous measurements made with fly ash, it appears that filter resis-
tance and holding capacity is more closely related to the volume than
to the mass of dust residing on the filter surface.
Discussion of Particle Emission Studies - Dust collection for the high-
ly efficient filter bags tested in this study can best be viewed in
terms of the particles penetrating the bag as a function of time during
the filtration cycle. The particulate emission may be related to the
following factors:
144
-------
• Weight and/or volume of dust residing on filter - The quantity
of dust discharging to the atmosphere over any time interval
depends upon the degree of fill of the Interstitial or pore
volume of the fabric. Present tests suggest that deposition
of sufficient dust to fill the estimated pore volume reduces
particulate emissions by a factor of 10. The time to achieve
this condition depends upon the inlet dust concentration, gas
flow rate and, as noted in the previous section, the bulk
density of the collected dust. The accumulation of successive
increments of dust (equal to that of the initial deposit) leads
to an exponential decay in outlet concentrations that may
embrace several orders of magnitude during a single filtration
cycle.
• Particle size - The larger particles are the first to be blocked
from further penetration followed by progressively smaller
particles as the filtration cycle progresses. The smallest size
blocked appears to be inversely proportional to the mass or
volume of dust deposited on the filter.
13
According to Draemel, it appears that complete filling or bridging-
over of the pore openings may be expected if the mass median diameter
of the dust is greater than 0.1 the average pore diameter. This estimate
must, of course, be tempered by the nominal packing density assumed by
the dust as determined by particle shape and electrical charge proper-
ties and the uniformity of the pore structure. Failure to attain com-
plete bridging was evidenced by pinhole leaks through which dust pene-
13
trated in the form of high velocity jets. Such dust penetration was
observed during this study with Dacron media when filtering fly ash.
Shaking amplitude and rate of flexure - Increasing shaking
amplitude from 1 to 2 inches led to a nearly tenfold in-
crease in mass emission rates whereas variations in shak-
ing frequency from about 4 to 12 cps had little impact on
total dust discharge. This leads to the conclusion that
it is the stretching or enlarging of the pore openings
rather than the rate of flexure that causes increased
particle emissions.
Bag life - After an extended shaking period, equivalent to
3 to 5 years of service, the average fly ash emission rates
for cotton fabrics increased by 2 to 3 times. Because of
the very low effluent concentrations determined for fly
ash, Table 14, this increase was not considered important.
145
-------
After a similar shaking period, a 4- to 5-fold increase
in fly ash emissions was observed for the Dacron fabrics.
In the latter case, the fact that the absolute quantities
of dust released to the atmosphere were 1000 to 2000 times
greater may be very important with toxic materials. Irre-
versible stretching and related pore enlargement were
assumed to be responsible for the increased emissions.
Dust type - Aside from differences In size, shape, charge
or hygroscopic properties, the single factor that appears
to influence most the resistance and dust holding capacity
of a specified dust/fabric combination is the bulk density
of the dust as it actually deposits on and within the
filter structure. Therefore, in searching for character-
izing parameters with which to predict filter performance
one might use bulk density as one means of defining the
interaction of several independent variables.
Filter Resistance
On the resumption of filtration, the pressure differential across the
filter bag was recorded as a function of tine. Due to the initial
rapid rate of pressure increase during the brief period of accel-
erating flow, it was difficult to determine accurately the true
residual pressure differential. However, since pressure data were
reasonably accurate after about 10 seconds of filtration, it was
possible to determine the effective residual pressure differential in
most cases by extrapolation of the nominally linear portion of the
resistance-time curve.
A single filter element (or compartment) may be operated in two ways.
First, it can be loaded with dust until a predetermined terminal
resistance is reached at which point the flow i* stopped and the filter
cleaned. Alternatively, it can be cleaned at predetermined time
intervals. In the first instance, the average operating pressure loss
across the element, h, is given by
h -
-------
where hr is the effective residual resistance immediately after cleaning
and ti£ the terminal resistance. The term "effective residual resis-
tance" refers to the time zero intercept value of the linear extra-
polation of the resistance-time curve that appears as a nearly re-
producible quantity during repetitive filtration and cleaning cycles.
In contrast, the instantaneous values of resistance when filtration is
resumed, and those during the first few minutes of filtration, are dif-
ficult to measure and may vary from cycle to cycle because of rapidly
changing cake structure. In practice, use of the more readily deter-
mined "effective residual resistance", results in a slightly con-
servative (higher) estimate of average operating pressure from
Equation (2.25). Thus, the cost of fan power, a major consideration
in filtration, is dependent on cleaning only insofar as the effective
residual pressure resistance is determined by the specific method of
cleaning.
lt: should also be recognized that the use of Equation (2.25) to esti-
"^te the average filter resistance over a filter cycle applies only
to the time interval over which filter resistance increases from its
nr to ht levels. If only a negligible change (decrease) in air flow
r
-------
if the filtration velocity, V, the inlet concentration. C., and pre-
sumably the specific resistance coefficient, K, for the specified
dust/fabric combination remain invariant over the period of filtra-
tion, the estimation of h depends only upon the filtering interval,
t. Variation in the other parameters merely requires that Equation
(2.26) be modified to show their time dependencies.
It is not readily apparent that when several filter bags or comparts
merits are cleaned sequentially, the overall average system resistance
may be expressed in terms of the same variables describing a single
element or chamber. Because the mathematical procedures necessary to
illustrate the above approach are cumbersome although actually quite
simple, the details have been relegated to Appendix J. It Is
shown that if a linear relationship exists between resistance increase
and time, and the specific resistance coefficient is independent of
the method of shaking, the average system resistance can be calculated
on the basis of single bag (or compartment) parameters.
As shown in Appendix J, the average system (multi-compartment) filter
drag may be expressed in the alternate forms given by Equations (2,27}
and (2.28).
+ KV C.t
If it is assumed that the average air flow (and filtration velocity)
does not change significantly with time, Equations (2.27) and (2.28)
reduce to the same form and involve the same parameters appearing in
Equations (2.25) and (2.26).
148
-------
Linearity of Resistance-Time Curves - Resistance versus time curves
based upon constant filtration velocity and constant inlet concentra-
tion are shown in Figure 48 for various fabrics and a fly ash
aerosol. Despite the variations between the intercepts for the actual
and effective residual resistance values, a good approximation to
linearity was usually achieved within 5 minutes. Generally, these
findings are substantiated by performance data for woven fabrics ap-
Q
pearing in the literature. In the case of heavily napped media
or the typical felts discussed in Chapter III, non-linear resistance
versus time curves are often encountered. As pointed out in the litera-
ture, high seepage rates or cake compression may also lead to non-
linearity. Inspection of Figure 48 indicates clearly that the use
of Equations (2.25) through (2.28) to estimate average resistance and
power requirements will provide slightly conservative results seldom
more than 5 percent higher than the true values.
Specific Resistance Coefficient, K - While comparing the results of fly
Jf
ash filtration with four different fabrics, K values were determined foi
relatively new and well-used fabrics shaken ~2.0 x 10 times,
Table 16. In each instance, the bag was repetitively loaded and
cleaned many times to equilibrate it before making measurements.
Althouth the residual dust holdings of all fabrics were substantially
reduced, ~ 25 percent, by the extended shaking, no consistent nor
large changes were noted for K values. Considerably more variation
was found among different fabrics, suggesting that the surface struc-
ture of a fabric affects the average pressure drop during filtration
more than differences in cleaning method. For example, Figure 32
indicates that K values were substantially unaltered despite a broad
range in shaking frequencies and shaking amplitudes.
... . Resistance Change
*K = Specific Resistance Coefficient = mter Velocity . cloth Loading
149
-------
Ul
o
Unnopped Cotton
Plain Weave Dacron
Figure 48, Resistance changes during fly ash filtration
for well used fabrics (see Table 16)
-------
Table 16. FLY ASH FILTRATION CHARACTERISTICS FOR NEW « 10 SHAKES) AND WELL-USED
(2 X 107 SHAKES) BAGS
Fabric type*
r
Residual drag
in H20/ft./min.
Effective
residual drag
in H2p/ft./rain.
Terminal drag
in H20/ft./min.
Dust collected**
per cycle
grains/ft.2
"K" value in
H20 ft./min./
lbs./ft.2
Residual dust
grains /ft. ^
Tm dust retnovedc
by shaking
Plain weave
Dacron
N
0.17
( 0.35)
( 0.81)
278
11.6
207
57
U
—
0.30
0.73
255
11.0
113
69
i
Crowfoot Dacron
N
( 0.37)
0.43
1.12
288
16.8
92
76
U
( 0.02)
0.47
1.11
275
16.3
73
79
Napped
Cotton Sateen
N
( 0.20)
0.23
0.82
295
14.0
449
48
U
( 0.53)
0.67
1.17
312
11.2
336
48
Unnapped
Cotton Sateen
N
0.47
0.67
1.24
284
14.1
413
41
U
( 0.60)
( 0.73)
1.41
290
16.4
375
41
10 ft, long x 6 in. diaai. bags, N «= New, U = Well used.
Inlet loading ~ 3.5 grains/ft.3, Filter velocity - 3 ft./min., 30 min. filter cycle.
Cleaning cycle - 360 shakes, 1 in. amplitude, 8 cps.
-------
In another series of tests, the fabric was deliberately undercleaned
with respect to the number of shakes to determine the impact: upon K.
Figure 49 shows the relationship between filter drag and total
fabric dust holding for four successive filtration and cleaning cycles.
Tests were initiated with a bag that had been shaken, at 11 cpa for lu
minutes. In real practice, no such cleaning cycle- would be used- he-
cause of loss of service (filtration) time and fabric orcrstressing.
The first cleaning, however, provided a base line with which to com-
pare the effect of shaking duration on performance paaramKtera for a.
fixed shaking technique (1 in. amplitude and 6.8 cp*).
During the first filtration cycle the average value for IS was 12.4 as
the fabric dust holding rose to 450 grains-/ft.^. Us* otf the statsdard
cleaning regime, ~ 360 shakes, reduced residual filter dtrajf tor roughly
the starting level although the residual dust hold .ing was considerably
greater, ~ 210 grains/ft.2. A slight increase in "K" was observed
for the second filtration cycle. Of more significance, however, was
the fact that use of only 86 shakes led to a still higher residual
drag and a nearly doubled residual fabric loading, roughly 400 grains;/
ft.2. A third filtration interval indicated that "K"^ values were
continuing to increase, 14.3 to 16.5, a* the number- of shakes were
reduced. A minimal shaking process, 30 shakes, led to a large* de-
crease in filter drag but was quite ineffective in reaovlag dua*.
f
Along with a large residual dust holding of 620 grains/ft. , the K.
value was almost doubled during the last filtration' cycle.
These tests, which confirm the earlier mcasureraexrits of Walsh- and. Spalte,1
indicate that the number of shakes required to maxln£ce? filtra>-
tion capacity exceeds those needed to reach a minima* dfeag, value.
According to data presented earlier, Figure 32, the E valas? for a fixed
dust/fabric system undergoes little change provided: that the- number of
shakes is in excess of 200. It is postulated that with leva shaking.,
the dust residing upon the filter undergoes compaction and thereby
152
-------
e
I
>
£ 2.0
O
'" 1.2
3
gO.8
£ 0.4
K «I2.4
I
0 200 400 COO tOO IOOO
TOTAL FABRIC LOADING, fraifis/ft*
Initial Fabric State
Extended (Excess) Cleaning
356 Chakes, in arapl., 6.8 cps
83 shakes, in ampl., 6.8 cps
30 shakes, in ampl., 6.8 cps
Subsequent "K" Value
12.4
14.3
16.5
26.9
*in H20/ft./inin. x lb/ft.2
Figure 49.
Effect of insufficient (non-equilibrium) shaking on "K",
filter drag and fabric loading with fly ash at 3.5 grains/
ft.3 and 3 ft./min. filter velocity 10 ft. x 6 in. cotton
sateen bag
153
-------
presents a more resistant path for air flow. As indicated previously
K values depend not only upon dust properties but also upon the
surface properties of the filter media as shown in Table 16.
Effective Residual Resistance - The filter resistance immediately fol-
lowing cleaning was examined with respect to the energy imparted to
the shaken bag at various acceleration levels. According to Figure
50, an inverse relationship is suggested despite the scatter of
points. Exclusion of the flagged point, however, indicates a pro-
gressive reduction in residual resistance as the shaking intensity
(acceleration) is increased. The solid line, Curve C, depicts the
normalized linear regression line having a correlation coefficient
r = -0.83, and a slope of -0.26. The dashed line, Curve A, represent,
an empirical fit to a slope of -0.5 to simplify rough calculations
Curve A', which shows the approximate relationship between residual
fabric loading and bag acceleration, also appears in Figure 34.
In that the actual slope of this curve, -0.39, as shown by the linear
regression line Curve B (r = -0.93) was roughly -0.5, the force fit
curve with a slope of -0.5 was presented in Figure 34 for calculation
purposes when estimated bag accelerations are in the 3 to 6 g range.
It was not unexpected that a broad spread of data points would obtain
for the residual resistance measurements because fabric flexure upon
resumption of air flow, and re-entrainment and re-deposition of ayataai
dust deposits upon the filter surface could well cause wide variation.
in the character of the residual dust/fabric structure.
From a practical viewpoint, the actual magnitude of the resistance
changes do not appear to be very large relative to the range of ac-
celerations over which most fabric filters would be expected to
operate. In the limiting case depicted by extended shaking periods
the resultant dust holdings and resistance, respectively, would ulti-
mately be expected to converge, regardless of the intensity of shakir
154
-------
10
3-
JS
5.0
5
-I 2.0
1.5
1.0
1.0
I
I
I
23 5
BAG ACCELERATION, g's
1000
**
500
2
Q
O
O
ft:
CD
200
CO
u
oc
Figure 50. Residual filter resistance and residual fabric loading
(Figure 34) versus bag acceleration for fly ash/sateen
weave cotton system
155
-------
Past studies by Walsh and Spaite appear to confirm the latter
observation.
As a matter of interest, the relationship between fabric dust holding
and residual resistance may be inferred by comparing the curves of
Figure 50. First, in the strictly hypothetical case where dust
removal is assumed to result from the spallation of an essentially
superficial dust layer, a linear relationship should prevail between
resistance and the amount of dust remaining on the filter. In the
above situation, it is assumed that no significant change in pore
structure takes place and that a decrease in dust holding is reflected
by a proportionate decrease in dust layer thickness. Were this con-
cept valid, the slopes of Curve* B and C would be identical. The fact
that they differ, however, is understandable if one oMunes that in
addition to increased acceleration removing more dust, the increased
intensity of shaking also compacts and thereby decreases the porosity
of the interstitial dust deposits. In accordance with a simplified7
expression of the factors determining resistance to air flow in
beds,
gas velocity, (V), gas viscosity Qa) and characteristic collector
diameter (Df) are assumed to be constants. Despite the fact th*t
dislodgement of dust will lead to a decrease in effective bed deptn
(L), the compaction process will increase the overall pecking densi
(p) of the residual dust/fabric structure. Since the exponent for
is about 1.5, the residual filter resistance should be proportional
to the residual fabric holding raised to the same power, n < 1.0.
156
-------
Insofar as fly ash was concerned, the range of effective residual
resistances was small, approximately 1.5 to 2.4 in, water and equiva-
lent to residual drag levels of 0.5 to 0.8 in. water/ft./min. During
4
the early use period for the bags, up to 10 shakes, a gradual in-
crease in residual resistance was noted, roughly 25 to 35 percent,
indicating a plugging of fabric interstices.
After extended usage as simulated by accelerated shaking (up to
2 X 107 shakes), cotton fabrics appeared to develop slightly higher
residual and terminal drag levels. Thus, power requirements for
treating equivalent gas volumes would increase in roughly the same
proportion. According to tests with the Dacron fabrics, Table 16,
extensive shaking did not appear to change s,ignificantly the resistance
characteristics nor the dust holding capacity. It should be noted,
however, that the Dacron media were also less efficient than either
the napped or unnapped cotton bags.
Discussion of Operating Parameters for Various Filter Systems - Tests
results summarized in Tables 11 and 16 and Figure 36 provide
basic operating parameters that enable estimation of power require-
ments, cloth area for arbitrary limits set for terminal drag, and
frequency of cleaning. It was pointed out previously that the holding
capacity of a given fabric for certain dusts appeared to be a combined
function of both dust and fabric properties. Thus, K values for fly
ash ranged from 11.0 to 16.8 depending upon whether plain weave Dacron
or Crowfoot Dacron were the filter media, with cotton fabric generally
lying in the 11.2 to 14 range. On the other hand, talc filtration
with various fabrics, Figure 38, showed much smaller variations
for K.
As far as fly ash was concerned, extended shaking did not appear to
have any significant bearing on K values, regardless of the fabric
used. Therefore, power requirements would increase in direct pro-
portion to the average filter resistance over the cleaning cycle,
157
-------
Equation (2.25), provided that a rise in terminal drag were permis-
sible. If a definite upper limit were set for terminal drag, it
would be necessary to either increase the cloth area or increase the
cleaning frequency to handle the inlet dust loading.
Napping of cotton sateen appeared to reduce resistance character-
istics appreciably for relatively new media but the advantage
diminished after an extended shaking period. Nevertheless, according
to Table 16, the dust holding capacity for the napped cotton ap-
pears to be about twice that for the unnapped material when a con-
straint of 1 in. water/ft. /min. is set for terminal drag-
In that residual resistance (and drag) values were consistently
for talc, irrespective of filter media, the average resistance and
power level would also be consistently lower with a 1 in. H20/ft.
set for the terminal drag. On the other hand, the dust handling
capacity for equivalent filter cycles would be about half as much for
talc as shown in Table 11 and implied by the higher K values for
talc, Figure 38.
Bag Life
It is reasonable to assume that bag motion will always contribute to
fabric deterioration whether by fatigue or embritt lament, by abrasion
between dust and/or fibers or by some heasetoior unidentified munha
nisras. Degradation due specifically to fabric sotion is difficult to
isolate in commercial practice because heat, abrxxion and corroaiv,
atmospheres may also augment the damage done by flexure alone.
Because the cost of filter media represents a significant fraction of
overall system operating costs, a portion of this study was devoted
to determining the effect of lengthy mechanical flexure on filter per-
formance. As described previously, life tests, based upon a con-
ventional shaking procedure, exposed various full-size filter bags to
158
-------
some 2 x 10' individual shakes, representing a nominal 3 to 5 year's
continuous field service. Periodic loading of the filters with fly
ash to their normal dust holding capacities permitted the investigation
of the combined effects of mechanical flexure and abrasion on filter
performance.
A summary of test parameters and the observed changes in bag length
and tensile modulus after 2 x 10? shakes is given in Table 17 for
parallel accelerated life tests on three unnapped, cotton bags. Turn-
buckles were installed in the bag suspending arms as a means of ad-
justing the tension to the initial values.as stretching progressed.
After the first few days, during which time frequent tension adjust-
ments were necessary, tension was adjusted once per day over the
remainder of the test period.
In Figure 51, the elongation properties for the test media described
in Table 17 are shown as a function of the cumulative number of
shakes. As a rough estimate, the total stretch varied as the 1/4 power
of the cumulative number of shakes given the bag.
Over the test period of about 4 weeks, the unnapped cotton bags stretched
approximately 1-in. while the elastic moduli of the bags more than
doubled. In these tests, the new bags stretched about 1/8 of an inch
in the first 2 minutes of shaking, the rate of stretching decreasing
significantly thereafter. Note that this amount of stretching without
re-tensioning would produce a tension decrease of 2 to 4 pounds,
depending on the individual bag modulus. These tests clearly show the
need for frequent adjustment of tension, particularly during the early
life of a bag, if excessive slack conditions are to be avoided. They
also suggest that bags should also be well-broken in, prestretched or
deerimped before installation, to minimize early maintenance problems.
Following the above tests, these bags were hung in storage (top sup-
port only) for about 2 months. At that time, the modulus of bag
159
-------
Table 17* EFFECT OF EXTENDED SHAKING ON TENSILE PROPERTIES OF CLEAN AND DUST LADEN COTTON BAGS,
FLY ASH AEROSOL
Bag number*
Ttonaoftfed cotton
Reference6
8
*
9
io
Bafrnttfl cc&£dft
ii
^t,(Htt*JwiaYtt ftf "%'*fl*t
in
13
Initial
tension
Top 3*1
Avg4 2.2
top Oi9
Avg, 0*4
top 1*3
Atg. 0*4
Button 0*4
BfettOft 0,5
Shaking0
tension
(top)
6.5
3*0
4*3
dust
loading
(grains /ft.2)
420
None
(clean)
420
420
420
420
Cttnulative
stretch
(in.)
0
1.4
0.61
1.03
1.2
0.19
0.69
Elastic*1
nodulus
(Ibs./in.)
16.5
44.5
31.4
39.6
Initial*
26.5
10.1
33*0
final
37.2
10.1
39.1
*ld ft* * 6 ihi bag! i
Shaking fiyst&at 8 fepa, 1 ih» attpiit«tdek 2 it IO7 shakes*
C0fae 30 AintiU f iU»*tieh (tytlt pet day*
Special ttoodului applying duly to a 10 ft* x 6 in. bag of specified fabric.
*ft«t 30 mlnuttt
-------
2.0
1.0
0.5
x
o
fc;
Of.
w 0.2
o
0.10
0.05
Bag No. Condition
• 8
o 9
A 10
Taut, Used
Slock, Clean
Slack, Used
I03
10"
10'
10'
TOTAL NUMBER OF SHAKER CYCLES, N
Figure 51. Bag elongation versus total number of shakes for used
and clean, 10 ft. x 6 in. sateen weave cotton bags
161
-------
No. 10 was found to have recovered to approximately 20 lb»./in. Since
the moduli values in Table 17 were measured a few hours following the
end of the extended shaking period, the bags mey have already under-
gone partial recovery. Because of this hysteresis factor, it is
possible that the bag moduli were actually higher during the shaking
period.
Practically no sign of bag wear was observed. Duat collection ef-
ficiency tests conducted once per week during the life testing process
indicated a gradual increase of dust penetration to about 10 times
the initial level, Table 13. Despite the tenfold increase in ef-
fluent concentration, the cotton fabric efficiencies for fly ash were
«o high that outlet concentrations were still about 1O3 times lower
than normal atmospheric levels.
There was absolutely nothing that could be termed a bag failure with
respect to filtration capability. The outer surfaces of the bags
•bowed no fabric deterioration except for some raising of a light nap
on the lower portions of all three bags. The outer (clean) surfaces
of the loaded bags were stained by low level seepege of dust through
the fabric, particularly in the lower portions of the baft* n*»r the
nodal areas and in folds of the bags near the tented tops (probably
the areas of greatest flexure). The clean filter beg being shaken
adjacent to the two loaded bags also acquired a staining in the
vicinity of the nodes due to external contamination. However, no
changes in the mechanical properties nor the filtration capabilities
of the bags were observed that could be attribute** to the fly ash
particles b«yond the effect of deposit weight on shaking tensions.
Since a large fraction of the fly ash particle* is spherical
(cenospheres), it ig expected that they would cause less damage than
hard, jagged particles.
prtcttonaL wear of the bag suspending loops was observed at their
Contact pointt between the loop and the supporting l/8th-inch steel
162
-------
rods. At some points the loops were perforated by either a rod end
or by rough spots on the rod surface. Presumably, the same type of
inconsequential wear occurs in field installations (assuming that the
above damage is external to the enclosed filtering volume).
The degradation of glass filter fabrics due to shaking has been ex-
tensively studied by several laboratories. This deterioration
is of a unique type, however, in which abrasive scratching or corrosive
pitting of the surface of individual glass fibers produces stress
points that lead to complete fracture under subsequent flexure. Be-
cause most filtration fibers are not as brittle as glass, their service
8
lives are usually much longer. Most fibers and fibrous materials
are tested for their chemical, thermal, and mechanical resistance
properties by the manufacturers. Unfortunately, manufacturers are often
not cognizant of the ultimate use of their materials in filter bags nor
to the types of stresses and abuse encountered in the field. As a
result, the effects of filter cleaning, including mechanical shaking,
on filter life are generally poorly defined.
Life tests were also performed on three other bag types; napped cotton,
plain-weave Dacron, and crowfoot Dacron, Table 17. The total elon-
gation features for these bags, Figure 52, were approximately the
same as those of the unnapped, Figure 51. Napping of the cotton
bag apparently raised its initial modulus relative to that of the un-
napped media, perhaps because the bag was mechanically stretched in
the process of napping. Note that even the crowfoot Dacron with multi-
filament yarns was stretched moderately by the extended shaking. It is
not clear why the plain-weave fabric had the same modulus after shaking
in contrast to the other fabrics tested. Again, the three fabrics
continued to filter effectively after lengthy shaking and no mechan-
ical wear of the fabric surfaces was visible.
The results of special tensile tests performed by an independent
laboratory with an Instron Tester on several 16 in. x 2 in. wide samples
163
-------
2.O
1.0
0.5O
E
DC
fc
23 a
o
O.K) -
0.05
Bog No. Typ«
/
o II
• 12
A 13
Mopped Cotton
Ploin-W«nve Docron
Crowfoot Docron
J_
I03 K>4 10s K>*
TOTAL NUMBER OF SHAKER CYCLES
Figure 52. Bag elongation versus total number of shakes for variotic
used 10 ft. x 6 in. fabric bag*
164
-------
of new (unshaken) and used (2 x 10? shakes) fabrics are summarized in
Table 18. The purpose of these measurements was to determine by
conventional textile rating apparatus whether any important changes in
fabric mechanical properties had resulted from extensive shaking.
Fabric specimens were clamped between jaws 6 in. apart and stretched
at a rate of 2 in./sec.
As far as the unnapped cotton bags are concerned, it can be seen that
there are no significant differences between descriptive parameters
for new and well-used bags. Furthermore, no specific locations on
the surface of a bag appeared to show signs of excess stress and/or
strain. The ultimate load (failure) levels for cotton were approxi-
mately 200 times greater than the dynamic tensions seen by the bags
during shaking. Generally, the negligible impact upon filter per-
formance both efficiency- and resistance-wise and the lack of visual
signs of damage or changes is supported by the tests described in
Table 18.
Initial "ultimate loads" for both Dacron fabrics were considerably higher
than those determined for cotton bags. Extended shaking, however,
appeared to have a weakening effect on the Dacron bags, although not at
a level to suggest any impairment of filter function.
The results of the test series indicated that mechanical shaking, per se,
of woven cotton and nonmineral synthetic fabrics does not damage the
media over relatively long simulated working periods, 3 to 5 years.
Additionally, nonabrading type dusts having smooth surfaces do not
appear to detract from effective field service over the previously stated
time period. With respect to the media described in Table 18, it is
suggested that thermal effects and chemical corrosion singly or in
combination with mechanical flexure are major causes for bag failure.
One might also include such factors as installation at excessive tension
165
-------
Table 18. TENSILE MEASUREMENTS ON NEW
INSTRGN TESTER
USED FABRIC STRIPS Will!
Bag number
Unnapped cotton
Reference bag6
8TJ£
9U
10U
Happed cotton
Reference bag
11U
Plain-weave
Dacron
Reference bag
12U
Crowfoot Daeron
Reference bag
13U
Unnapped cotton -
ba« 88
Top, left
Top, rear
Center, front
Bottom, rear
Bottom, right
Ultimate*
load
(Ibs./ih.)
100
106
103
103
107
112
173
152
421
366
101
107
106
107
104
Standard1'
deviatldh
7.7
6.7
5.6
8.7
7.*
6VO
18.9
4.0 '
. 5.5
19.'5
\
I
- *;s . j
i* 'i '
7.6
•en '
-5.8 (
,.7*_. J
Ultttfwtec
attain
0.24
fc»44
0.25
'0.24
0.2A
O.SS
0.43
0.40 '
0^40 i
0A6
10^4
!'
_^ I
Brmk point?1
modulus
416
444
413
481
44ft
SOB
402
390
-1060
32*8
^.
^
«.
•*.
^Maximum stress per inch of strip width it br«ki1»o-tri6.
Standard deviation of break point stress.
CStrain immediately before breaking, clamps *et*at 6'in.
*Soint function only.
^Reference bag, no prior shaking.
fU indicates 2 x 107 shakes,Js«e'fable 17.
strips from various sections of single^^f.
166
-------
or slackness and the failure to prevent the formation of cementlike
deposits by hydration reactions.
CONCLUSIONS TO MECHANICAL SHAKING STUDY
In the following section we summarize the major conclusions that can be
drawn from the mechanical shaking studies discussed in Chapter II.
With respect to the order in which these conclusions are presented, two
key factors have been considered. First, one of the most pressing
concerns is the effectiveness of various particulate control systems
in the removal of fine particles, < 1 nm. Our current efforts are
now directed not only to overall mass collection efficiencies but also
to the actual concentration and size composition of collector effluents.
Second, given the specification that particulate emissions from a given
source must not exceed certain boundaries, the use of mechanical shaking
has been examined with respect to optimizing methods to achieve these
criteria while minimizing those factors contributing to capital and
operating costs.
General Conclusions
To avoid misinterpretation and misapplication of the results of this
study we present the following general conclusions.
1. Limits of data application - Unless it is clearly indicated
that a conclusion may be given broad interpretation, the
reader should assume that descriptive and operating param-
eters cited for a specific dust and fabric combination may
apply only to that specified dust/fabric combination.
Measurements reported in this study and also in the lit-
erature indicate that test results for a single dust or
fabric are not sufficient to permit a generalized extra-
polation to other dusts or fabrics. This problem results
from the large number of independent variables needed to
describe completely a dust and/or fabric and their res-
pective interactions.
2. Need for further research - A major conclusion based upon
the previous paragraph is that much more research on a
167
-------
bench and pilot scale is required if one is to be able to
predict the field performance characteristics of a fabric
filter system without preliminary trials with the dust/
fabric system involved.
3. Outlet versus inlet concentrations - It is strongly emphasized
that no simple relationship exists between typical outlet
and inlet concentrations for most fabric filters. More
often, for a specified dust/fabric combination and a
fixed operating mode for the collector, the average mass
emission and its related size properties may be nearly
independent of the concentration and size of the inlet
dust. Furthermore, the fact that the dust penetrating a
filter consists partly of agglomerates (which are fortned
within the filter from single particles but detached as
clusters) prevents any accurate analysis of collector
fractional particle size efficiencies.
It is also emphasized that the particle concentrations departing, the
fabric filter may vary by several orders of magnitude over a complete
.filter cycle. Thus, for all practical purposes, the only useful size
parameters that can be generated for field application are those des-
cribing average performance over a complete and representative filter
cycle.
Particulate Emissions
Those conclusions relating to specific tests on various' dust/fabric; com>
binafcions are presented first with respect to the impact of the mechan-
ical shaking process upon particulate effluent characteristics. When-
ever possible, an attempt has been made to indicate toe Mmitf t& wfr&tfi
test results may be extrapolated to other-dust* and' fabrics*
1. Fly asb and talc aerosols or other particulates- having,
similar size properties can be filtered at very high
efficiencies: with, sateen weave cotton bags,, ~99).9^9^;
percent for inlet ^concentrations, of 3.5 graina/ft^..
Filtration efficiencies- for Dacron media are high,
99.7 to 99'.9* range, but based upon an- inlet concen-
tration of 3.5 grains/ft. , outlet, concentrations
much greater as pointed out below.
168
-------
2. Average outlet concentrations for the cotton fabrics
range from 10~^ to 10~^ grains/ft.-* depending upon the
number of shaking cycles, the degree of tension
in the bag, the precise mode of shaking, and the
residual dust holding. Average outlet concentrations
for the Dacron media with fly ash and talc aerosols
are about 10~3 to 10~2 grains/ft.3, about 1000 times
greater than those for the cotton fabrics. One concludes
that where dust toxicity is important the fabric perform-
ance should always be examined on the basis of the out-
let concentrations and not efficiency alone.
3. For a specified set of operating conditions, the magnitude
of the aver age outlet concentration is nearly independent
of the inlet dust concentration. Thus, an order of magni-
tude increase in the inlet loading may result in a corres-
ponding rise in collection efficiency but a negligible
change in the mass and size distribution of the discharged
particulates.
4. Effluent particle size distributions for fly ash and talc
aerosols, as measured by instruments having 0.3 ^m as the
smallest resolvable size, were about the same as for
atmospheric dust, GMD =0.5 urn, og = 1.5.
5. Instantaneous values for outlet concentrations measured
from the beginning to the end of a normal filtration cycle
may decrease by less than one to several orders of magni-
tude. The greatest changes take place with the fabrics
having the highest efficiencies.
6. The fact that as much as 90 percent of the total dust
emission from a filter may take place during the first
minute of use following cleaning suggests strongly that
a recycle of the early effluent would greatly improve
the system performance.
7. The key factor determining how rapidly the restoration
of an efficient filtering layer is accomplished is the
rate at which the inlet dust can completely fill the
pores within the fabric weave.
8. The filling or bridging-over of the fabric pore structure
appears to depend upon the characteristic specific
volume of the dust within the filter interstices as well
as the pore diameter, and the ratio of particle to pore
diameter.
9. The slightly lower outlet concentrations observed for
talc (relative to fly ash) result from a low bulk density
that facilitates closure of the pore deposits and the
irregularity in shape that retards seepage or slippage
such as found for spherical particles; e.g., fly ash
cenospheres.
169
-------
10. Variations in mechanical shaking parameters have the
following impact upon fabric filter effluent
c oncent r a t ions.
• An increase in shaking frequency from 4 to 11 cps does
not produce significant increases in effluent concen-
trations. Thus, the rate of fabric flexure is act con
sidered an important factor in determining participate
emissions.
• An approximate 10 times increase iti outlet concentra-
tion occurs when shaking amplitude is increased from
1 to 2 inches. It is concluded Chat it is not tfre
amplitude per se but the tension increase on the fil-
ter bags and the resultant increase in pore dimension
that leads to greater emissions.
• Particulate emissions relate inversely to toe mater
of shakes given a filter op to about 2OO shalees,.
Beyond this point, further dust removal is negligible
and filter emissions are essentially constant.
11. Accelerated life testing of cotton and Dacron bags by
shaking each approximately 2 by 107 times (equivalent
to about 3 to 5 years' field service) leads to incon-
sequential increases in the outlet particle concentra-
tions, 2.5 times for the napped and unnapped cotton
and 4.3 times for the Dacron media. It is concluded
that increased dust penetration results from the
lower residual dust holdings observed for both media
after extended shaking.
12. Average particle outlet concentrations
-------
Dust Removal by Mechanical Shaking
The following conclusions are concerned with the relationships between
the mechanical shaking process and the system performance parameters
exclusive of particle effluent description.
1. Dust removal by mechanical shaking is best defined by relating
the energy supplied by the shaking process to the adhesive
and/or cohesive forces binding the dust to the fabric.
2. An increase in shaking frequency and/or shaking amplitude
has a direct effect on shaking power and dust removal.
3. For a specified dust/fabric combination there is a cha-
racteristic residual dust loading (grains/ft.3) and an
effective residual resistance for each specific mode
of shaking that are nearly independent of the fabric
loading at the end of a filtration cycle.
4. The total number of shakes at a specified shaking fre-
quency and amplitude should not be carried much beyond
200 to 300 since further dust removal and reduction of
filter resistance is negligible. Continued shaking will
only decrease the effective filter on-line time, in-
crease the probability of shortened service life by
unnecessary stress and abrasion damage to the fabric,
and as stated previously increase the effluent
concentration levels.
5. With respect to a given dust/fabric combination, the
residual dust holding is best related to the average
acceleration imparted to the bags during the shaking
process. Based upon readily measurable (or predictable)
physical properties of the fabric bags; the assigned
shaking parameters (frequency, amplitude and duration);
and the installed and dynamic tension, average bag
accelerations are readily computed.
6. Residual fabric dust loadings correlate well with
average bag accelerations for fly ash/sateen weave
cotton systems. Over the acceleration range, 1.5 to
10 g's, residual dust holding varies approximately
with the inverse square root of average bag
acceleration.
7. Increase of accelerations beyond 6 to 8 g's does not
lead to any significant decrease in residual dust holding.
8. For all practical purposes, there are upper limits to
both the intensity and duration of the shaking process.
Given sufficient shaking under strictly experimental
171
-------
conditions, the residual duat holdings of many filters
cleaned at different acceleration level* would ultimately
converge. Thus, a mathematical description of the fabric
cleaning process is necessarily constrained to operate
within pre-set limits.
9. Installed bag tens toning does not appear to exert any
significant effect upon the average bag acceleration pro-
vided that the initial values are la the 0.5 to 5 Ib. rang*-
It is the difference between shaking tension and averaged
installed tension that determines average bag shaking am-
plitude and acceleration.
10. Installed bag tension levels should be as close as
possible to a true slack condition where the top tension
is that of the bag weight alone. If too slack, however;
i.e., no actual stretching of the bag occurs even when
the shaker arm sees its maxima* displacement frost
center, transmission of shaking energy over the entire
bag length Is incomplete and cleaning is poor and
nonuniforra. Over-tensionlng of the fabric bags
affords no advantages with respect to dust removal
but will expand the pore structure (greater du«t
penetration) and possibly shorten fabric life through
excessive stretching.
II, Although slightly greater dust removal, ^ 10 percent,
was obtained by adjusting shaking frequencies to
resonance levels, it was concluded that the coat and
complexity of constant "tuning" to compensate for dust
loss was not justified.
Flltar Resistance and Power Requirements
1. The actual energy expenditure for mechanical shaking
is very small, approximately a tew watta per bag, »"
that the cleaning power requirement is negligible
compared to that needed to overcome the re«i»t*»c«
to air flow through the filter media.
2. The residual (effective) filter resistance (and/or
filter drag) can be related to the residual do»t
holding on a weight or volume basis for • •pacific*
dust/fabric system. Additionally, the residual
resistance for different dusts may be related
approximately on tha basis of the voloms- of dost
residing on the filter after cleaning.
3. Dusts with low bulk densities such as talc are col-
lected more as superficial rather than as Inter-
stitial deposits. Therefore, dust is dislodged
172
-------
more readily such that lower residual resistance
levels obtain.
4. Specific resistance coefficients appear to be a
combined function of both dust and fabric pro-
perties. In the absence of field data it is risky
to attempt to predict "K" values without some
trial data.
5. Undercleaning of a filter; i.e., less than the recom-
mended 200 shakes, has an adverse effect on its K values,
leading to decreased system filtration capacity
for a specified working resistance range.
6. Filtration capacity, per se, is an indeterminate quan-
tity unless one defines the upper allowable working
resistance for the filter.
7. Average filter resistance with respect to estimating
fan power requirements is the arithmetic average of
the residual (effective) and terminal resistance values.
This will apply for single bags or sequentially cleaned
multicompartment units just so long as the system gas
flow is reasonably constant.
REFERENCES
1. Walsh, G. W. and P. W. Spaite. An Analysis of Mechanical Shaking
in Air Filtration. J. Air Poll. Control Assoc., 12:57, 1962.
2. Durham, J. R. and R. E. Harrington. Influence of Relative Humidity
on Filtration Resistance and Efficiency. NAPCA, PHS, U.S. DREW,
AICHE 63rd Annual Meeting paper, Chicago, 111., November 1970.
3. Zimon, A. D. Adhesion of Dust and Powder. Plenum Press, New York,
1969.
4. Lindsay, R. B. Mechanical Radiation. McGraw-Hill, New York (1960).
5. Shortley, G. and D. Williams. Elements of Physics. Vol. II, 4th
Ed. Prentice-Ha11, Inc., Englewood Cliffs, New Jersey (1965).
6. Sears, F. W. and M. W. Zemansky. University Physics. 3rd Ed.,
Part 1. Addison-Wesley Publishing Company, Inc., Reading, Massa-
chusetts (1963).
7. Fink, D. and J. Carroll. Standard Handbook for Electrical Engineers,
10th Edition. McGraw-Hill, New York (1968).
173
-------
8. Billings, C. E. and J. E. Wilder., Handbook, of: Fabric, Filter Tech-
nology, GCA Corporation.,. Bedford,. Massachusetts,,; Contract Nov
CPA-22-69-38, Prepared for National Technical. Information. Service,
U.S.. Department, of .Commerce,, Springfield,, Virginia! 22151.,, Document
No. PB 200-648, December 1970 .,
9. Spaite:, P'. W-. , J. E.. Hagan and; W. F.> Tbddi, A Protective Einisli for
Glass-Fiber Fabrics. Ghent.. Engg. Prog;.,,, 59::54,, April. 196'3.
10v Hicks, R. E. and W. G'.. B., Mandershoofcv Flexing; Fatigue: o£ aiass;-
Fiber Filter Cloth.. Textile; Res;., "Jv,, 9421,, SepJrember 19-!6.8).,
11.. Whttby,. K., T. and tt.. A. Lund'gren., Fractiona'i. ESl'iaiency- Character-
istics of a. 'Tor.lt. Type Cloth Collector.,, Tbr.lt Manufaceuring Company,
St.. Paul, Minnesota:, August 1961.
12. Spaite,. P. W. and7 G. W. Walsh. Effect, of ' Fabric: StcuctuEei om Filter
Performance:. Am-. Ind. Hyg. Assoc. J.,,. 24's357;,.
13:.. Draemel, D. C.. Relationship Between Fabric Structure and Filtra-
tion Performance in Dust Filtration. Control Systems; laboratory,
U.S. Environmental Protection Agency, Research Triangle; Park
Report No. EPA-R2-73-288, July 1973.
14. Stephan, D, G. et. al.. A- New Technique; for. Fabric Filter' Evaluation
AIHA J.,, 28:276, 1958..
15-.. Dennis^ R..,. G. A.. Johnson:,, M!., W..» Firstr and' L.. Silvermaa'.: How Dust
Collectors; Perform., Chemv Eng.., 5^:.196^,; 1952v
16-. Robinson, J. W. „ R. E. Harrington and! Pv Wv Spaite;^ AL New Method
of Analysis) for Multicompartmented FaBric: Eiltration^ Atmos^
Envir., 1:495', 1967.
17. Spaite,, F. W. and R. E.. Harrington,. Endurance, of: Fiberglas, Filter
Fabrics. JAPCA, 17:310/, 1967.,
174
-------
CHAPTER III
PULSE CLEANING STUDIES
OBJECTIVES AND APPROACH
Within the last 15 years, reverse pulse cleaning has seen increased use
for many types of dust and fume control. Nearly every large manufac-
turer of fabric filter equipment off erf. at least one model cleaned by
this method. Because of the growing utilization of filter systems
cleaned by high pressure reverse air pulses, it is important that the
mechanics of this cleaning process be fully understood. Accordingly,
a study of reverse pulse cleaning, as reported in this chapter, was in-
stituted to achieve the following objectives:
• To determine the effect of design and operating parameters
on the performance of pulse-cleaned equipment with special
emphasis on the particle size and concentration of the ef-
fluent dust.
• To determine the power requirements for various operating
modes.
• To establish an improved base for future studies, field
testing, design modifications, and preparation of opera-
ting guidelines.
In order to represent commercial equipment as it operates in the field,
a full-scale, pulse-cleaned fabric filter system was constructed. Con-
siderable flexibility was incorporated in the design to permit simula-
tion of several common configurations and operating conditions. Instru-
mentation was provided for measuring bag motion, bag weight, and the
175
-------
instantaneous pressure differential across the bag. These variables,
which are seldom monitored in the field, were considered fundamental
to describing the pulse cleaning procesi.
TWO types of filter bags (Dacron and wool felt) and two test dusts (fly
ash and talc) were used in the test program. The various cleaning cycles
and filtration conditions evaluated in the study encompassed those con-
ditions commonly encountered in industry. Operating pressure differen-
tials, cleaning power requirements, and dust concentrations emitted by
the filter were measured for all test runs, since these factors represent
the main criteria for assessing cost and performance. Additional details
concerning the equipment and test procedures used are discussed under
APPARATUS, MATERIALS AND TECHNIQUES.
The approach taken in the study was to determine first the variations in
operating pressure differential and outlet dust concentration with se-
lected operating and design parameters. These variations were then ex-
amined in terms of more fundamental parameters such as bag motion and
particle adhesion versus removal mechanisms, in an attempt to explain the
observed results.
BACKGROUND
Applications and Advantages
Pulse cleaning is accomplished by admitting a short-duration pulse of
reverse flow, high pressure air to the clean air side of the filter bag.
This method of cleaning was introduced in the mid-fifties by Pulverizing
2
Machinery Co. It has proven to be a utilitarian means of filter bag
cleaning that is now offered by all large manufacturers of bag filter
equipment. While there are no published statistics to indicate the
proportion of the market represented by pulse-cleaning equipment, it is
estimated that about one-third of current sales or about 3,000 installa-
tions per year are of the pulse-cleaned type. This equipment is employed
176
-------
for many dust control applications as well as for chemical process
operations. Its application has been less common in areas where the
gas temperatures are high, the dust loadings low (infrequent cleaning)
or the dust especially fine (freshly generated fume).
Several characteristics of reverse pulse cleaning make this system very
attractive for a large number of applications. Because the pulse of
high pressure air is brief, typically 0.1 second in duration, the pri-
mary air flow does not have to be shut off during the cleaning. Instead,
the pulse temporarily reverses the flow through the fabric due to the
transient reversal of pressure gradient. The particle agglomerates dis-
lodged by the pulse and reversed flow are sufficiently large so that
much of the dust can settle to the hopper during the brief interruption
of filtration. Thus, the equipment operates essentially on a continuous
basis and with almost no moving parts except for solenoid valve elements.
The fabric receives little mechanical wear when properly installed and
thus has a long service life, unless damaged by some other environmental
factor. Felts rather than woven fabrics are generally used in this
equipment, because woven fabrics tend to be over-cleaned by pulsing,
resulting in excessive leakage following cleaning. In addition, the
frequent fabric cleaning rate permits filtering velocities typically
three to four times greater than those used in shaken or reverse flow
equipment without an accompanying pressure increase. Because the bags
move less during cleaning, they may be packed together more closely.
Additionally, little or no extra capacity need be installed to replace
the filter surface temporarily removed from service during cleaning.
The latter factor is especially advantageous with high dust loadings.
In combination, the above features contribute to a reduced space (and
volume) requirement for cleaning a specified gas volume.
Problem Areas
Despite the attractive features of the pulse jet system, there are still
insufficient data to properly assess its cost and performance in many
177
-------
field applications. Because of the need to use compressed air, it is
recognized that the power costs associated with cleaning may often be
higher than those for mechanically shaken systems of equivalent capac-
ity. At the same time, however, the increased air-to-cloth ratios may
provide compensating cost advantages with respect to reduced collector
size and space occupancy requirements. If no alternate method is
available to perform the required filtration process, however, any
added cost must be accepted as a prerequisite to meeting the control
needs.
An important disadvantage in certain cases is the generally lower effi-
ciency provided by a pulse-jet system. For example, typical effluent
concentrations from pulse-cleaned filters are about 100 times greater
than those attainable with mechanically shaken fabrics. Additionally,
because the energy of the high pressure air-pulse tends to dissipate
rapidly as it travels down a bag, the length of pulse-cleaned bags is
limited to 10 feet or less in most models. If the felt does plug
it is usually the result of interstitial rather than surface deposition.
This may require the removal and external cleaning of the bags (an
expensive procedure) or a media replacement, both of which add to
operating cost.
Collector Performance
Resistance and Particle Removal - The filter operating pressure dif-
ferential and its relation to the cleaning regime has been studied by
3
Dennis, et al. Based upon tests with fly ash on one of the first
types of reverse pulse equipment, the authors found the average opera-
ting pressure differential to be defined approximately by the following
relation:
178
-------
3.8 (c/'25
P (In. H,0) = 1.1 + si r-r (3.1)
!^
100
o
where dust loading (c^) varied from 1 to 14 grains/ft. , the manifold
compressed air pressure (P ) varied from 60 to 120 psig, and the clean-
ing frequency (f ) varied from 1 to 6 pulses per min. per bag.
Equation (3.1) applies to a nine-bag, single compartment, sequentially
pulsed system using wool felt (6 ft. x 4.5 in.) bags at a filtering
velocity of 8 ft./min. Pulse duration, nominally about 0.16 seconds,
had no apparent effect on resistance or efficiency, at least with a
nine-bag unit.
The same quantitative relationship was shown for a fine silica dust
except that the resistance was considerably higher. Freshly generated
iron oxide could not be filtered successfully with the fabrics available
at the time of the earlier studies. Transient pressure levels were also
estimated within filter bags during the pulse interval, but only during
3
nonflow conditions. Therefore, although these tests indicated the
general wave form and magnitude of pressure pulses, they did not re-
3
fleet true operating conditions.
Although considerable laboratory and field data have been reported»with
r\ I f-
respect to operating pressures, * ' data on collection efficiency and
effluent dust properties for pulse cleaned filters are very sparse.
Dennis, et al. reported weight penetrations ranging from 0.1 to 0.01
percent for wool felt bags with fly ash and vaporized silica, respec-
tively. No apparent correlation of the efficiency with any of the op-
erating variables was noted. Berg reported 0.1 percent particle
penetration for Nomex felt in asphalt-concrete plant operations at gas
temperatures of 400 F and inlet dust loadings (sand and limestone) of
3
about 25 grains/ft. . There are several reasons for the scarcity of in-
formation in the above areas:
179
-------
• In a very practical sense, the lack of any visible
emissions coupled with the accepted generalization
that all fabric filters operate at the 99 percent
efficiency level or greater is often the basis for
selection and favorable acceptance of fabric filter
systems. Many users of filter equipment are disin-
terested in any details of particulate emissions so
long as no complaints arise and no product loss is
involved.
• Collection efficiency may be so high, or possibly
so difficult to measure because of accessibility
problems, that the sampling process is undesirably
long.
• Because efficiency depends upon several factors;
e.g., particle size, humidity, charge character-
istics, state of fabric surface (just cleaned to
fully loaded), all of which may vary considerably
at a given plant site, one cannot rely on a few
measurements to give an accurate portrayal of col-
lector performance. Since a lengthy sampling period
may not be acceptable, often for economic reasons,
the net result is that few, if any, efficiency mea-
surements are performed.
Ihere are reasons to suspect that many factors affect cleaning effec-
tiveness and dust collection efficiency, including dust particle size
and physical properties; air temperature, velocity, and humidity; in-
dividual fiber properties; fabric weave, surface depth, porosity, and
mechanical characteristics; the dimensions and physical properties of
the filter bags; the arrangement of bags in the dust collector; the
cleaning pulse - its intensity, duration, and frequency, and the ac-
companying reverse flow of air through the filter.
The experiments performed during this study were designed to accomplish
the following objectives:
• Show the relative importance of several of the design
and operating parameters cited above.
• Determine how these parameters can be controlled to
attain optimum collector performance for specified
dust removal criteria.
180
-------
Factors Involved in Dust Removal
During one complete filtration cycle for a single bag at equilibrium
conditions, fresh dust is deposited upon and within the fabric struc-
ture while an equal amount of dust consisting partly of the newly ar-
rived material and partly of previously deposited dust is dislodged by
the cleaning action. Of the total amount of dust dislodged by the
pulse, a large fraction immediately redeposits when normal air flow
is resumed. Some particles may remain within the fabric structure
indefinitely, becoming part of a residual deposit such as described
under Shake Cleaning.
Ajhesloo . ». residual and fresh dust is seldom deposited as a uniform
layer as postulated for smooth-surfaced fabrics, tte dust is usuaUy
concentrated in the form of tufts because the mediu* is thicker, much
more open, and the pore structure .ore irregular. Uus, aUhough the
on wrasse internal strength profile analogous
deposited dust may have an average ine ,_ f ,
to that discussed under a postulated spaUation theory for shake clean-
Ing, the average profile concept may be misUading in the case of
felted media. Dividual fiber-to-particle and part de-to-particle
jiaec af cleaning method. However, be-
bonds should be similar regardless of cleaning
cause the deposit on a felt is for^d and dislodged at freouent inter-
vals vith pulse cleaned system the adhesive strengths •-•£*•
deaning stresses can vary considerably fron, iocation to location.
that the fabric moves or is deformed during
n.a-t re^a! .chanis. apply as dUcussed
puise cleaning, t»«= acceleration, sheer, flexure-
in Chapter II for shaking cleaning; i.e., ««!« •
*.v,a morion of a pulsed bag is less com
stretch, and warping. However, the motion ot P
, , KHO since it consists mostly of radial
Plex than that of e shaken bag, since .n-tM, flexure
a.r>*leration and one-dimensional tiexure
dispiacements. B-erefore, -"le"^ ^chanlsms. In ad-
appear to be the «.t Probable «ch - ^ ^^
dition, a substantial reverse flow of Btresses as
cuaning interval. *is contributes additlona! remova! stresses
181
-------
well as transporting the dislodged dust from the fabric interstices.
One major goal of the pulse cleaning experiments was to determine the
relative importance of the four mechanisms contributing to cleaning;
i.e., acceleration and flexure coupled with reverse drag and
re-entrainment.
Reverse Air Transport - The reverse flow of air through the fabric
during the pulse interval flushes loosened particles from the fabric
pores. Although not considered a primary dislodgement mechanism, the
effect of the reverse flow on system operation may be considerable.
After the bag is pulsed, there is little time before the flow recovers
and filtering is resumed. During this interval, the particles must
either fall into the hopper or else be re-deposited on the fabric.
Only the largest agglomerates may have sufficient weight to continue
falling after normal filtration flow has been resumed. Many agglom-
erates, however, will be drawn toward the fabric as they descend and
be re-deposited at some lower station. The major transport processes
associated with pulse cleaning are discussed in detail in Appendix L.
It is shown on theoretical grounds that the main effect of extending
the pulse duration is to provide added time for the agglomerates to
fall. The benefit gained depends on the fall velocity of the agglom-
erates and hence their size and density. Thus one suggestion is that
the performance of a filter might be improved by controlling the prop-
erties of the agglomerates removed in pulse cleaning.
APPARATUS, MATERIALS AND TECHNIQUES
A description of the apparatus, materials, and experimental techniques
used in the pulse cleaning study is presented in this section. When
appropriate, reference is made to Chapter II for previously detailed
descriptions of test procedures, apparatus and test aerosols.
182
-------
Filter Assembly
The basic filter assembly used in the pulse cleaning study is shown in
Figure 53, which indicates schematically the major operating and de-
sign features. Dusty gas enters the hopper section at the base of the
bag enclosure losing some of its particulate loading by gravity and in-
ertial separation to the collection drum attached to the hopper. Fil-
tration of the gas then follows via outside to inside flow through the
felt bag, the latter supported by a wire cage to prevent its collapse.
Cleaned gas departs through the upper exit plenum from which reverse
pulse air is ejected into the bag exit in accordance with the programmed
cleaning schedule.
The equipment was designed with sufficient flexibility to represent in
principle, if not exactly in dimensional scaling, the more common types
of commercial equipment. The 4-1/2 inch diameter bags were enclosed in
an 8-inch by 8-inch chamber to provide the normal upward dust trans-
port velocities found in the field. Bags of any length up to 8 feet
could be accommodated in the housing by adding extension sections.
The volume of the housing surrounding the filter bag, including the
dust collection drum indicated in Figure 53 was about 7 cu. ft.
To facilitate observation of the bag during filtration and cleaning, a
3/8 inch thick Plexiglas window was installed in one wall of the
housing. The other three walls of the housing were fabricated as a
single movable section to allow rapid access to the filter bag and the
instruments attached to it. It was constructed from a sheet of 10
gauge steel bent into channel shape and suspended on rollers to form
a sliding enclosure. This enclosure was held in place by quick dis-
connect clamps and 1/8 to 1/4 inch neoprene sponge gasketing was used
to prevent gas leakage.
The supporting wire cage for the filter consisted of 10, 1/8 inch steel
rods spaced at equal intervals around the circumference of the bag, and
183
-------
120 PSI AIR SUPPLY
PRESSURE CONTROL
TIMER
UNIT
SOLENIOD VALVE
OUTLET
PLENUM
GASKETS
INTER-
CHANGEABLE
^- NOZZLE
WINDOW
DIRTY
AIR
INLET
Pm
1
W
RESERVOIR
L
r
CLEAN AIR
/ OUTLET
BAG AND CAGE
•ASSEMBLY
4 FT. LONG
4-1/2 IN. DIAM.
SLIDING
ENCLOSURE
OPEN
POSITION
HOPPER
COLLECTION
DRUM
GATE
7
MEASURING POINTS:
E-EXIT LOADING
F-FALLOUT COLLECTION
Pd - INSIDE BAG PRESSURE
P0 Pm- UPSTREAM PRESSURE
W- LOAD CELL BAG
WEIGHT
Figure 53. Schematic of pulse jet cleaning assembly
184
-------
five stiffening rings spaced at equal intervals along the length of
the bag. The cage was fastened to a steel thimble by a 4-1/2 inch diam-
eter steel strap clamp that also served to prevent any by-pass leakage.
A Venturi nozzle was attached to the steel thimble usually in the con-
figuration shown in Figure 54. The Venturi section, wire cage, com-
pressed air valves and also the pulse control apparatus discussed below
were assembled from commercially available components.
Pulse Jet Cleaning Equipment
The system used to deliver compressed air to the bag, which was de-
signed to adapt to a number of pulse cleaning configurations, is shown
in Figure 54. All components were standard shelf items, black iron
pipe, hose, and sheet metal, with the exception of the diaphragm and
solenoid valve system, which was a commercial system purchased from a
major manufacturer of pulse cleaning equipment. The damping tank and
valve shown in Figure 54 were introduced in some tests to determine
how the form of the pressure wave affected dust removal and dust pene-
tration properties. The nozzle was a 1/4 inch diameter section of
black iron pipe, adjustable in the vertical direction with respect to
the location of the Venturi nozzle. The jet of compressed air through
the Venturi induced a secondary clean air flow from the top plenum in
addition to the primary jet air.
In operation, the compressed air reservoir was maintained at the de-
sired cleaning pressure via the regulator valve. Release of the elec-
trically controlled solenoid valve tripped the diaphragm valve, dumping
air from the reservoir through the nozzle for as long as the valve re-
mained open, usually about 0.1 second. During this time, the reservoir
pressure usually decreased to about two-thirds of its initial level.
This pressure difference, in conjunction with any necessary temperature
2
adjustment, a reservoir volume of 0-1/2 ft. , and an assumed adiabatic
expansion provided one means of estimating the compressed air
requirement.
185
-------
FROM COMPRESSOR
I
"""^•^J PRESSURE REGULATOR
J* AND GAGE
COMPRESSED AIR
RESERVOIR (0.5 FT.3)
DAMPING TANK
AND VALVE _
(0.06 FT3)
* STANDARD COMPONENTS
MIKRO PUL DIVISION
U.S. FILTER CORPORATION
SUMMIT, NEW JERSEY
DIAPHRAGM
I SOLENOID VALVE
SYSTEM*
PLENUM
OUTLET
Figure 54. Standard pulse delivery system
186
-------
For repetitive operation of the pulse system, two solid-state timers
were used, one controlling the interval between pulses; i.e., the length
of the operating cycle, and the other controlling the length of time
the solenoid valve was open. These timers had capabilities of 0.01 to
99.99 min., and 0.01 and 99.99 sec., respectively, with precisions
of about 0.008 sec. according to the manufacturer.
The source of compressed air was a compressor operating at 120 psig that
delivered air at room temperature and essentially free of water droplets.
The pulse properties as delivered to the bag depend on the pressure-flow
characteristics of the compressed air supply, and the nozzle shape and
location relative to the bag inlet. The effective pulse volume depends
on the induction action of the Venturi, if a Venturi is used. The total
air volume entering the bag is controlled by the nozzle induction ca-
pacity that depends, in turn, upon plenum dimensions and the absolute
pressure within the plenum. The reverse pressure differential across
the bag caused by the pulse depends strongly on the pressure that
rapidly builds up on the dirty side of the bag. This build-up depends
not only on the volume of reverse flow air, but also on the pressure-
flow characteristics of the main fan when it is momentarily blocked.
Additionally, the pressure build-up also depends on the volume of the
housing surrounding the bag including the hopper volume and perhaps on
the length/width (aspect) ratio of the housing. Thus, in practice, the
presence of other filter bags in the same housing may affect the clean-
ing whether they are cleaned simultaneously or not.
It should be noted that the pressures measured inside the filter bag
during the pulse interval were of the order of 1.0 psig or less above
atmospheric, much lower than that of the reservoir pressures. This
follows from the fact that the air pressure beyond the nozzle discharge
depends only upon the reservoir venting rate, the external pressure,
and the external system pressure release capability. For the purposes
of analyzing dust penetration, dust removal, and operating pressure
drop, one should consider the pressure within the bag and not that of
187
-------
the compressed air reservoir. However, tank pressure, which is much
more easily measured, is commonly used as a descriptor for the cleaning
process. For this reason, compressed air pressure will be reported in
this study as an operating parameter, while bag differential pressure
will be used to explain filter performance.
The apparatus shown in Figure 54 provided most of the pulse control
capability needed in the study. For other special tests, the gate valve
above the collection drum was sometimes closed, a pressure release valve
was installed in the inlet duct ahead of the hopper, and a supplementary
fan system was used to inject cleaning air into the top plenum outlet.
This secondary system included a damper, blower, and timers, as de-
scribed in Chapter IV of this report.
Test Fabrics
Felted fabrics are commonly used in filtration equipment cleaned by high
pressure reverse pulsing. The Dacron felt used in the majority of tests
discussed in this section was representative of fabric media used in
commercial practice. However, an abbreviated series of comparative
tests were also run with a woolen felt that also sees frequent com-
mercial application. Wool felts were used for many years in equipment
cleaned by other than pulsing methods, notably by the traveling blow
ring method. They are still being used to some extent in various kinds
of equipment including the pulse type. Table 19 provides a descrip-
tion of the properties of the two fabrics used in the tests. The sim-
ilarity of the two fabric structures is further reflected in their es-
sentially similar filtration performance as shown later in this
Chapter.
The Dacron and wool felts supplied by the manufacturers were very sim-
ilar in fiber diameter, fabric thickness and density. The felt density
or volume fraction of fiber was computed from the felt weight and thick-
ness using discrete fiber densities of 1.40 and 1.30 gram/cc for the
188
-------
Table 19. PROPERTIES OF TEST FELTS
Property
Type
Style
2
Weight (oz./yd. )
Permeability (air flow, ft./min.
1/2 in. pressure drop)
Fiber diameter (micrometer), Df
Felt thickness (cm)
Fiber volume fraction, a
Pore size (micrometer), Dp
Dacron
felted, needled
Al36Ba
18
35
20
0.17
0.26
34
Wool
felted, without
scrim, HCE
silicone treatment
M1778b
16
30 to 40
20
0,20
0.21
39
aAlbany International Corp., Globe Albany Division, North Monmouth,
Maine.
Menardi and Co., Augusta, Georgia.
/, \2
D = D.C
P f
a
Dacron and wool, respectively. Calculated pore sizes and air flow perme
abilities were also very similar.
Felt structures differ from woven fabrics in many important respects.
They are depth type media and generally contain many more pores and a
larger free area than woven filter fabrics. The performance of felted
6 7 R
media has been the subject of numerous investigations. * ' Although
the performance of clean (previously unused) media can be adequately
predicted by classical filtration theory, the behavior of media con-
taining substantial amounts of residual dust cannot be quantitatively
depicted because of the complexity of the geometrical array of fibers
189
-------
and collected dust. There is great need for a systematic study of the
effect of dust size/pore size on felted fabric performance similar to
9
that conducted by Draemel on a multiplicity of woven fabric structures
using dust loadings representative of those encountered industrially.
The average pore size can be expected to play an important role in the
penetration of test dusts. An extensive investigation of pore size/
dust size correlations with filter efficiency has been conducted by
Q
Draemel for several woven fabrics. Extensive penetration was noted
when the average pore size was more than 10 times the mass median par-
ticle diameter. Earlier workers have also noted the same relationship
between woven fabric structures and filter performance.
Bag Properties and Measurement Techniques
Bag weights before and after cleaning were measured to determine the
gross bag weight and/or the amount of dust collected upon the filter.
The mounting plate was clamped against the bottom of the top plenum
using a neoprene sponge seal to facilitate easy removal. Because this
method was time consuming, a much quicker but less accurate method was
devised to weigh the bag in place. The mounting plate was undamped,
and a piezoelectric load cell located beneath the bottom of the bag
cage was elevated until it came in contact with the bag. Pressing the
load cell against the bag produced a signal proportional to the weight
of the bag, cage, Venturi, and mounting plate. The method selected
depended upon the accuracy requirement.
The elastic modulus of the bag was determined by inflating a bladder
inserted inside the bag and noting the differential changes in circum-
ference and length with increasing pressure. The static position of
the bag as it draped around the supporting cage was measured by calip«ra
Small increases of pressure inside the bag reduced the dimpling and
ultimately caused it to balloon out into a smooth surfaced cylinder.
The relationship between displacement and pressure, which afforded a
190
-------
measure of the bag's flexibility, was used in computing the motion of
the fabric during a pulse.
Motion
Motion of the bag during pulses was also measured in three different
ways.
Accelerometer - A piezoelectric accelerometer weighing only 0.15 grams
was attached to the bag and the output signal displayed on an oscil-
loscope screen during the bag pulse. Some electronic diffulties were
encountered; i.e., any motion of the two wires carrying the extremely
small signal apparently produced electromagnetic noise even when the
wires and accelerometer were shielded. Additionally, high-frequency
vibrations, apparently present either in the pulsed air or in the bag
cage, sometimes overwhelmed the desired lower- frequency signal. Suf-
ficient information was obtained, however, to confirm other instru-
mental measurements despite the limitations of this technique.
strain Gage - A special strain gauge sensor, similar in principle to
a phonograph pick-up cartridge, was designed for measuring small fabric
displacements (up to 1/2 inch) . The device consisted of a light-
weight finger in light but positive contact with the bag and with the
strain gauges bonded to opposite sides. Bag motion flexed the finger
causing imbalance in a resistance bridge. Photographs of the resulting
signals displayed as distance/time coordinates on an oscilloscope screen
permitted estimation of fabric velocity and acceleration as well as
physical displacement of the bag during pulsing.
*Accelerometer: Model 91 piezoelectric accelerometer from Wilcoxon
Research, P.O. Box 5798, 9 Ericsson Road, Bethesda, Md. 20014.
"^Pressure transducer: Model PT-H2 Pitran Pressure Transistor, nominal
linear pressure range, 015 psid; from Stow Lab., Inc., Kane Industrial
Drive, Hudson, Mass. 01749.
191
-------
High Speed Photography - Cinematography at approximately 3500 frames
per second produced a viewing time magnification of about 220X. The
movies were also used to make estimates of the displacement, velocity,
and acceleration of the fabric and also to depict the dust cloud leav-
ing the fabric.
By and large, pressure differential measurements provided the best in-
dication of bag motion, using mechanical properties of the bags and
computing the motion. The above techniques were used during the study
to verify the computational approach.
Test Dusts
The fly ash and talc dusts described earlier in Chapter II of this re-
port were also used in pulse cleaning tests. They were reaerosolized
in the same way; i.e., using an aspirator with a jet of 90 psig com-
pressed air.
Dust Measurements
Generally, the same procedures were used in the pulse jet studies as
were used in the Mechanical Shaking Studies. Dust was metered and mon-
itored in the same ways. A portion of the dust introduced to the system
was lost to the hopper as fallout without reaching the filter bag. fov
fly ash, this was of the order of 50 percent, depending on air flow rate,
whereas practically all the talc reached the bag. The most accurate
method of determining rate of dust filtering was by bag weighing tech-
niques described above. Weighing techniques were accurate to within
about 8 grams, or 2 percent.
Dust removed by mechanical shaking, Chapter II, could be determined
readily by weighing the dust shaken into a special container placed in
the dust hopper. Because there was no air flow during and for 1 minute
after shaking, it was estimated that the dust reaching the hopper
192
-------
represented most, — 99 percent, of the dust dislodged by shaking. Once
steady state conditions were established, the dust removed by a spe-
ific mode of shaking also depicted the filtration or holding capacity
of the fabric for the indicated increase in resistance to airflow. In
cleaning by pulsing, however, a large portion of the dislodged dust
deposits on the walls of the enclosure around the bag, where it mixes
with dust previously deposited during the filtering state. Furthermore,
there was no direct method of measuring the amount of dust dislodged
from the bag that would not interfere either with the inlet flow or
the flow reversal during the pulse interval. More important, the amount
of dust falling to the hopper per pulse does not depict the actual quan-
tity of dust removed by the pulse, but only that fraction of the dis-
lodged material that reaches the hopper. The remaining dust mixes with
the oncoming air stream and then redeposits, usually at some lower posi-
tion along the filter tube. From a practical standpoint, however, the
true filtration capacity of each bag at steady state is the quantity of
dust conveyed to the filtration surface between pulse cleaning intervals,
In turn, this quantity represents the difference between the amount of
dust dislodged by each pulse and the amount of dust that redeposits on
the filtering surface before reaching the hopper.
Emission of dust from the filter bag was determined by the B&L
counter and the RDM monitor described in Chapter II. The B&L in-
strument indicated particle size avid concentration based upon light
scattering phenomena, while the RDM instrument provided mass concentra-
tion only. The B&L instrument was used principally to observe var-
iations in outlet concentration with time over the operating cycle
with a time resolution of about 0.1 minute.
Mass concentration and particle count measurements were determined for
the effluent air stream at the exit from the clean air plenum (see Fig-
ure 54). The size of the device required that an 18-in. tygon sam-
pling tube (dwell time, 4 sec.) be used between the plenum outlet
and the instrument. For measurements along the length of the bag a
193
-------
longer Tygon tube was inserted via the plenum into the bag to the de-
sired sampling height. While some particle diffusional losses were un-
avoidable, the B and L test results are comparable on a relative basis
and also show moderately good agreement with concurrent RDM measurements
Pressure and Flow Measurements
The instrument systems for measuring operating pressure differential
across the filter bag and controlling the system air flow to the bag
were described earlier in Chapter II and are described in detail in
Appendix A. These devices, which relied on pneumatic sensing of pres-
sures at various points in the system, provided accuracies and pre-
cisions of the order of 5 percent.
Operating pressure differentials were also monitored using manometers
&
as reference standards and Bourdon-type dial indicators for fast re-
sponse direct readout. Neither instrument, however, was suited for
measuring the rapid pressure changes during pulse intervals. For the
latter application, two good quality diaphragm-type pressure trans-
ducers were used, one inside the bag and the other outside and with
one side of each transducer sealed. Each was calibrated to have iden-
tical linear responses and zero points so that when their separate sig-
nals were subtracted on an oscilloscope, the pressure differential
trace was displayed. Alternatively, the individual pressure signals
also could be viewed. Although response times of the transducers and
circuits were excellent, there was a problem of a drifting zero point
that occasionally allowed one transducer or the other to shift into a
*Magnehelic gauges, 0 to 10-inch water range.
+Pitran silicon transistor, 10 microsecond rise time, 2 microinch dia-
phragm displacement, 1/3 gram weight, 14-inch water nominal linear
range. Model PT-H2, Stow Laboratories, Inc., Kane Industrial Drive,
Hudson, Mass.
194
-------
region of nonlinear response. Infrequently, dust caked over the dia-
phragm of the dirty-side transducer, causing an erratic response and
making it difficult to calibrate until cleaned. Photographs of the
pulse differentials recorded on the oscilloscope are included with the
^f
data in Appendix M. These two pressure transducers were mounted on
opposite sides of the fabric, at a point about half way down the bag.
Compared to the typical pulse duration of about 0.1 sec., air shock
fronts traversed the 4-ft. bag in about 0.004 sec. Thus, the pres-
sures measured by the transducers were very nearly equal to the time
averaged pressures throughout the entire bag.
The regulating circuits for the establishment of pulse intervals and
pulse durations were standard commercial hardware items. It should be
noted, however, that the dial settings of 0.01 and 0.1 second for con-
trolling valve open times actually produced 0.06 and 0.15 second pulses,
respectively. The above values were derived by examination of oscil-
loscope traces depicting the transient pressure changes within the fil-
ter bags as discussed later in this Chapter.
•Jhe volume of air delivered during a single pulse was determined by at-
taching a 2 ft. diameter meteorological balloon to the nozzle and by
3
accumulating sufficient volume, ~ 2 ft. , by sequential pulsing so
that the balloon contents could be determined with a laboratory spirom-
eter. The primary air volume per pulse was also calculated on the basis
of critical flow through the 1/4 in. nozzle and the absolute pressure
and temperature of the reservoir air.
In retrospect, a single transducer element across the bag to provide a
direct differential signal would probably have been a better approach.
It was rejected in this program because of the anticipated difficulties
of grommeting a pneumatic tube through the bag without picking up me-
chanical vibrations to which these transducers were sensitive; the pos-
sibility of dust plugging the pneumatic tube, and the inability to view
separately the inside and outside pressures.
195
-------
Although the automatic flow control and recording system was adequate
for normal flow regulation, it :could not respond to sudden surges in
flow. Therefore, the reduced resistance of the "just-cleaned" filter
bag caused a flow increase which took the flow control system about
10 seconds to return to the preset level. The residual filter pressure
drops reported in the following chapters have been adjusted to reflect
normal filtering velocities.
RESULTS
The results of extensive tests with a single bag, full-scale pulse jet
system in which system performance was studied with wool and Dacron
felt bags are presented next. Test aerosols, sampling equipment and
the design and operating features of the filter system have been
described previously in this chapter and/or in Chapter II.
Although the laboratory system was designed to simulate full scale in-
dustrial systems to enable maximum use of experimental data, it was
heavily instrumented to provide information that could not be obtained
practically under normal field conditions. The primary functions of the
special measurement systems were to relate the particle collection and
resistance characteristics to the key operating variables of the com-
pressed air pulse jet system.
Particulate Emission Characteristics
Effluent dust concentrations from pulse cleaned bags are appreciably
higher than those for the mechanically shaken bags described in Chap-
ter II. Penetrations of the order of 0.01 percent, however, can be
obtained with suitable choice of felt and operating conditions. As
with shaken bags, filter emissions are highest immediately after clean-
ing. The effect of the cleaning cycle and other system operating and
equipment design parameters on dust penetration are reviewed in the fol-
lowing paragraphs.
196
-------
Compressed Air Pressure - The pulse of air used to clean the bag was re-
leased from a reservoir tank by a quick-acting solenoid valve. The air
pressure within the reservoir, 30 to 100 psig, was found to be a major
factor in controlling filter efficiency and fabric resistance. Because
of the ease with which pressure adjustments could be made, reservoir
pressure was treated as a primary variable in this study.
Figures 55a and b indicate particulate emissions measured by the RDM for
a series of experimental runs in which fly ash was filtered with Dacron
felt. The reservoir pressure was varied between 40 and 100 psig and the
mode of introduction of the cleaning pulse to the bag was either direct
or damped. Analyses of the data indicate that the emission increases
with reservoir pressure (statistically significant at < 1 percent level
of confidence). A pressure of 40 psig was determined to be the approx-
imate minimum level necessary to obtain any appreciable cleaning of the
bag. These results indicate that, at least with fly ash, one should be
able to control collection efficiency by the proper choice of reservoir
pressure.
The data points shown in Figure 55a represent the combined effects
of variations in inlet concentration, filtration velocity, pulse inter-
val and pulse duration for each reservoir pressure level. It was in-
tended that the smoothing and averaging process would provide an improved
picture of the emission/reservoir pressure relationship. The right hand
graph, Figure 55b, shows a family of curves in which the operating
parameters were separated for each curve. The latter data have been used
to assess concentration and velocity effects in later discussions.
The difference between the effluent concentrations resulting from the ap-
plication of the cleaning pulse to the bag by the direct and damped modes
is significant at a < 1 percent level of confidence. The decrease in ef-
fluent concentration resulting from damped or delayed pulses is attrib-
uted to two factors. When the valve is opened to the damping tank shewn
197
-------
00
rO
O
X
10
V)
§5
CD
Ul
Z
O
O
Ul
UJ
O
-------
in Figure 54, part of the compressed air vented from the main pressure
reservoir (about 10 percent) is diverted to the damping tank. There it
remains until closure of the solenoid valve in the vent line from the
main pressure reservoir. Because part of the total air volume released
from the main reservoir is temporarily stored in the damping tank, the
transient reverse pressure gradient across the bag during the open time
of the valve and the rate of pressure change in the system are slightly
reduced. Both factors contribute to less effective cleaning such that
the greater residual dust holding leads to improved particle collection.
Another distinguishing feature of the delayed pulse is that upon closure
of the solenoid valve to the reservoir tank, the contents of the damping
tank vent to the clean air side of the bag thus forestalling a too rapid
resumption of filtration. In the case of the direct pulse, there is no
back pressure to oppose the transient flow and pressure drop increases
that often occur at the end of the pulse. The higher air flow coupled
with a transient stretching of the felt augments the higher effluent
concentration caused by the greater rate of pressure rise and larger
reverse pressure gradient associated with the direct pulse. Figure 56
illustrates simplified tracings and actual photographs of typical pres-
sure/time relationships shown by oscilloscope display of the differential
pressures across direct and delayed pulse systems.
Ihe interaction between reservoir pressure and the mode of presentation
to the bag (i.e., direct or damped) was found to be significant (< 1 per-
cent level of confidence) indicating that the cleaning effectiveness of
a pulse delivered from a reservoir operating at a given pressure is also
a function of the pulse form. Thus, the collection efficiency and the
average pressure drop can be optimized within the system by properly
selecting the reservoir pressure and the mode of admission to the bag.
Overall particulate emissions presented in Figure 55 have been re-
solved into specific size fractions in Figure 57 to show the separate
effects of reservoir pressure and pulse damping on the effluent
199
-------
PHOTOGRAPH, DIRECT
PHOTOGRAPH,DELAYED
to
O
O
s
S
c
uT
in
oc
a.
UJ
K
UJ
TRACE, DIRECT
TRACE.DELAYED
Figure 56. Effect of direct and delayed pulses on bag differential pressure for valve open time of
0.15 sec.
-------
NJ
O
PARTICLE NUMBER CONCENTRATION, N/ft?
S So So, S* S« S« °-
II III
DIRECT PULSES
i
— g-O-Q w ...r^ — A — i . .
CODE DESCRIPTION
INLET CONC. X 10"*
ANDERSEN IMPACT
-A — A- OUTLET CONC. , B a
M w OUTl ET CONC B fl
O— O- OUTLET CONC., B8
ii ii
I I 1 I
M"
^ BY
OR
L , 100 PSIG
L, 70 PSIG
kL, 40 PSIG
. i i i i
i i i i i i i i i i
DAMPED PULSES
V ,
f\ *-\ i
1
A A
PULSE INTERVALS, >-O O—
0.4 TO I.O MIN.
PULSE DURATION,
0.06 TO 0.15 SEC.
INLET CONC., II. 6 GRAINS/ FT.3
CLOTH VELOCITY, 8.5 FT./ MIN.
CONVENTIONAL 1/4 IN. NOZZLE /!
VENTURI SECTION
1 i ii i
_
w w
O <*"»
\J \J
kND
10
PARTICLE DIAMETER,
10
Figure 57. Effect of reservoir pressure and pulse damping on effluent concentration for fly ash fil-
tration with Dacron felt (note: inlet mass concentration converted to equivalent number
concentration and scaled by 10~^ for comparison)
-------
characteristics. To allow comparisons between inlet and outlet concen-
trations with the fly ash/Dacron felt system, the inlet concentrations
estimated by Andersen impactor have been converted to their equivalent:
number concentrations and classified according to the B&L size intervals.
The inlet values appearing on the graph have also been reduced by four
orders of magnitude to facilitate the comparison. When the dashed im-
pactor lines and those depicting the effluent are exactly superimposed
on Figure 57, the collection efficiency for the indicated si?e cate-
-4
gory is 99.99 percent (or the fractional pentration 1 x 10 ).
Examination of the histogram bars of Figure 57, aside from a con-
forming qualitatively the results in Figure 55, shows that a higher
proportion of coarse particles appear in the collector effluent during
direct pulse cleaning. This effect, which is most pronounced at the
higher pressure levels, is more clearly shown in Table 20. Inspec-
tion of the calculated fractional particle size penetration values
indicates that the apparent collection efficiencies decrease with in-
creasing particle diameter for direct pulse cleaning at 70 and 100 psig.
What actually takes place is that previously deposited dust is Ipo.sened
and flushed out in the form of agglomerates when the normal filter flow
.resumes. Therefore, the effluent particles usually represent two popu-
lations, those upstream particles that would ordinarily penetrate the
felted media with its residual dust holding and the agglomerate fraction
dislodged as a result of the cleaning process.
The damping procedure appears to reduce effluent concentrations signif-
icantly, particularly so at the lower reservoir pressures. .The Improved
effluent conditions are attributed in large part to the less .rapid change
in bag pressure differential, Figure 56, when the main solenoid valve
is closed. It appears tha± the "snap" .of the bag :when filtration is re-
moved coupled with the higher transient filtration velocity /are the
principal causes of the higher average dust penetration. It should be
noted that the experimental points in .Figure 57 and Table 20 in-
clude variations in pulse duration and p.ulse frequency. Therefore,,
202
-------
Table 20. OUTLET CONCENTRATION AND FRACTION PENETRATION FOR VARIOUS
RESERVOIR PRESSURES AND DIRECT AND DAMPED PULSES. FLY
ASH FILTRATION WITH DACRON FELT (SEE FIGURE 57)
Size
range
(inn)
Dust concentration particles/ft.
Inleta
Outlet
100
70
40
Apparent fraction penetration
Outlet0
100
70
40
Direct pulse
I- 2
2- 3
3- 5
5-10
2.2 x 1010
1.0 x 1010
1.5 x 109
1.2 x 108
._
6.2 x 105
5.1 x 105
2.2 x 105
5.7 x 105
2.9 x 105
1.8 x 105
2.3 x 104
6.3 x 105
2.2 x 105
1.1 x 105
7.0 x 103
..
6.2 x 10"5
3.4 x 10"4
1.8 x 10"3
2.6 x 10~5
2.9 x 10"5
1.2 x 10'4
1.9 x 10'4
2.9 x 10"5
2.2 x 10"5
7.3 x 10"5
5.8 x 10'5
Dampe4 pulse
•111 1 ' •
1- 2
2- 3
3- 5
5-10
2.2 x 1010
1.0 x 1010
1.5 x 109
1.2 x 108
--
1.7 x 105
5.1 x 104
9.0 x 102
--
7.1 x 104
1.7 x 104
5.0 x 102
2.7 x 105
4.2 x 103
1.0 x 103
5.0 x 10
-.
1.7 x 10"5
3.4 x 10"5
7.5 x 10' 6
--
7.1 x 10'6
1.1 x 10'5
4.2 x 10'6
--
4.2 x 10"7
6.7 x 10"7
4.2 x W7
*Basod on Andersen impactor data converted to equivalent number concentration.
Based on B&L measurements. Outlet concentration levels partly attributable
to agglomerates re-entrained immediately after pulsing.
°Many coarse particles dislodged from fabric as agglomerates. Not due to
direct penetration.
203
-------
although the data appear to align logically on a qualitative basis,
they should be used only as guidelines until more precise measurements
can be made.
Cleaning Frequency (Operating Cycle) - Several tests were performed to
determine the effect of cleaning frequency on particulate effluent prop-
erties. If one assumes that two filtering operations are identical in
every respect except for a difference in cleaning frequency it would
appear that the more rapid pulsing process might cause increased dust
emissions. This follows from the fact that the filter surface is pre-
sented more frequently in the "just cleaned" state in which greater par-
ticle penetration occurs. Increased emissions during the very early
filtrations phase are attributed to two effects, first the expected in-
creased penetration due to the more open filter structure and second
the re-entrainment of particles (usually agglomerates) whose bonds have
been weakened by the preceding cleaning process.
The above mechanisms may be counterbalanced by two additional effects.
The amount of dust associated with the transient puff is probably di-
rectly associated with the prior dust holding of the filter. Thus a
greater dust release should occur when the cleaning frequency is re-
duced; i.e., more dust on the filter. Figure 55b, for example, indi-
cates that the filter effluent concentration is strongly dependent on
filter loading for a fixed cleaning system. Although part of the total
dust penetration is presumed to be due to a less complete sealing of the
fabric surface, it is suspected that most of the effluent is associated
with the initial puff. In turn, the "puff" concentration depends upon
the transient concentration produced on the dirty air side of the filter
immediately after pulsing. On the other hand, the filter with the
greater dust loading should also show decreased penetration in accord-
ance with all filtration logic. The net result of these opposing
effects is that cleaning or pulse frequency may not exert a strong in-
fluence on effluent concentrations, as shown in Figure 58 with respect
to cleaning frequencies greater than 0.4 minutes.
204
-------
O
Cn
0.5 1.0
CLEANING FREQUENCY, min.
2.0
Figure 58.
Effect of frequency of cleaning on average outlet concentrations for fly ash filtration
with Dacron felt. Data reflect variations in reservoir pressure, pulse duration and
pulse form for commercial 1/4 in. jet and Venturi system.
-------
It should be noted, however, that very frequent pulses appear to in-
crease significantly the effluent concentrations. Note that revised
values for mass concentrations given by B&L samples should be at least
as high as the values read from the right-hand ordinate. Actual cali-
brations under experimental conditions applying specifically to pulse
jet tests suggest that the ratio of RDM to B&L concentrations is
roughly 5.0.
Practically, the use of pulse intervals less than 0.5 minute would pre-
sent severe compressed air requirements as well as increasing the par-
ticle emissions. The above high frequency measurements are confirmed
by the plenum pulse tests shown in Figure 59. Here we see a very
consistent structure to the concentration/pulse frequency relationship
for time periods of up to 0.2 minute. The data align logically with
respect to both reservoir pressure and pulse duration. Despite the fact
that filter resistance levels were quite low, it is emphasized that the
low pressure plenum pulses actually represented prohibitive compressed
air demands (about six times greater than those for typical commercial
1/4-in. diameter jets). The main purpose of the plenum pulse tests
was to permit a better understanding and definition of the pulse clean-
ing process.
Figures 60 and 61 examine the effects of cleaning frequency on
the particle size distribution and number concentration of the filter
effluent. The histograms also indicate the relative fractional con-
centrations of the inlet dust (fly ash). The term relative has been
used because the actual number concentration levels have been reduced
by 10 to permit rapid visual estimation of apparent efficiencies in
the indicated size ranges. The effects of direct and damped pulses
as well as pulse duration, 0.06 and 0.15 second, are also shown in
Figures °0 an(* **1.
With respect to the size ranges resolvable by the particle measurement
instruments (Andersen itnpactor and B&L counter) there appears to be
206
-------
10
o
PRESSURE
PSIG
50
40
30
PULSE
0.06 SEC
O
A
DURATION
0.10 SEC 0.15 SEC.
NOTE: SPECIAL TEST SERIES. PRESSURE PULSE
APPLIED TO PLENUM SECTION. NOZZLE
AND VENTURI SECTION REMOVED.
50 PSIG
50 PSIG
40 PSIG
40 PSIG
30 PSIG
3O PSIG
L
0-1
0-2 0-3
CLEANING FREQUENCY, min.
0-4
0-5
Figure 59. Effect of frequency of cleaning on average outlet concentrations for fly ash filtration
with Dacron felt. Data reflect variations in reservoir pressure and pulse duration.
-------
10
O
00
10
!*
o
*
Of,
PARTICLE NUMBER CONC
5
1 1 1 1 I 1 1 1 1
DIRECT PULSE ,0.15 SEC. DURATION
N»
M«*
—A A
CODE DESCRIPTION PULSE*
•" — INLET CONC. X I0~4 , BY
ANDERSEN IMPACTOR
Nw f\i ITI FT cvkM/* n A Kin i n A
MIN.
A A OUTI ET CONC B AMD L 1 O
MIN. -
* PULSE INTERVAL, MIN.
i i i i i i t i i
2 5 1
M 1 1 1 1 1 1 l 1 1
DIRECT PULSE , 0.06 SEC. DURATION
A A
u Or-"
A, A^ x W
~
* *
RESERVOIR PRESSURE , 70 PS 16
INLET CONC. , 11.6 GRAINS /FT.3
CLOTH VELOCITY, 8.5 FT./ MIN.
CONVENTIONAL 1/4 IN NOZZLE AND
VENTURI SECTION
S\ l i f i i i i i i
0 2 5 11
PARTICLE DIAMETER,
Figure 60. Effect of pulse duration and pulse interval on effluent concentration for fly ash filtra-
tion with Dacron felt, direct pulse (note: inlet mass concentration converted to
equivalent nu»b«r concentration and scaled by 10~t for comparison)
-------
to
O
vo
w
|io6
0
PARTICLE NUMBER CONCENTRATI
o ° o o o
W M Cri A *•«
1 1 1
DAMPED PULSE , 0.15
- A A
CODE 0
INL
ANI
Xw
ESCRIPTI
\ \ \ \ \ \ r
SEC. DURATION
ON
ET CONC. X 10-4
3ERSEN IMPACTOR
X M OUTLET CONC. ,
B AND L
•A— A- OUTLET CONC.,
B AND L
* PULSE INTERVAL, WIN.
ii iii
-
PULSE *
0.4
A A
1.0
*4 1 I 1 I 1 1 1 I 1
DAMPED PULSE , 0.06 SEC. DURATION
RESERVOIR F
INLET CONC.
CLOTH VELO
CONVENTION
AND VENTL
A i i
Ky
>RESSURE,70PSIG
,11.6 GRAINS/ FT.3
CITY, 8.5 FT./MIN.
AL 1/4 IN. NOZZLE
IRI SECTION
-
X**
A A
I 1 1 i
5 10
PARTICLE DIAMETER,
10
Figure 61.
Effect of pulse duration and pulse interval on effluent concentration for fly ash filtra-
tion with Dacron felt, damped pulses (note: inlet mass concentration converted to
equivalent number concentration and scaled by 10~^ for comparison)
-------
no discernible evidence that the effluent particle size distribution is
altered significantly with direct or damped pulses. Note that super-
position of an effluent histogram bar with the corresponding dotted
influent bar depicts a collection efficiency of 99.99 percent. As
stated previously, the apparent "coarseness" of the effluent dust re-
sults from a large contribution from agglomerates released immediately
after the pulse. Examination of particle size concentration versus
filtration time curves given later in this section confirm this hypoth-
esis. It should be noted that reservoir pressure variations depicted
in Figure 57 exerted a significant effect on particle emission. At
100 psig, for example, the outlet aerosol was much coarser than the
3
inlet aerosol despite the large, > 10 times, reduction in mass
concentration.
Cleaning Pulse Duration - Preliminary tests during this study and
3
earlier results reported by Dennis et al. indicated that pulse duration
had no significant effect on filter resistance. It is recognized that
in comparing tests upon single bag and multi-bag collectors, any effect
noted with a single bag may be obscured in the larger system. Fig-
ures 60 and 61 show that pulse duration does have a measurable
effect upon effluent concentration. The most pronounced change, howev
occurs with damped pulses where the effluent concentration appears to
increase by about 3 times for the longer pulses in the 2 to 5 ym sise
range. The fact that there is less of a change for direct pulse sys-
tems is attributed to the fact that the initial emissions immediately
following a pulse constitutes the major part of the total emission.
Felt Type - Comparative fly ash emissions for wool and Dacron felts ar
given in Figure 62 for direct and damped pulses over the reservoir
pressure range 40 to 100 psig. Although wool felt emissions were
slightly lower than those for felt, the differences in fabric density
permeability, residual dust holding and filtration resistance provide
no clear cut explanation for the slightly improved performance. A
second perspective on comparative performance is given in Figure 63
210
-------
en
2 6
'6 5
Z
LU
U
o
U
O
LU
O
UJ
NOTE: A) DACRON DATA FROM
FIGURE 55o
8) WOOL FELT TESTS
Cj = 9.2 TO 11.6 GRAINS/FT3
PULSE INTERVAL,
0.4 TO 2 MIN.
PULSE DURATION,
0.06 TO 0.15 MIN.
DACRON
DIRECT
WOOL
DIRECT
DACRON
DAMPED
_J
1
20 40 60 80 100
INITIAL RESERVOIR PRESSURE, psig
Figure 62. Comparative performance between wool and Dacron felts with
fly ash filtration
211
-------
10'
"L
2 in6
10
z
o
£ io5
z
1U
z
J2 0 4
E u io4
QC
UJ
CO
5
z 10
11 1
u
< »o2
Q.
10
i
1
1 I
i i i 1
DIRECT PULSES
-
—
CODE
,,M V . 1
Ux>^
-°— 9=1
~
-*-O O-K-
DESCRIPTION
INLET CONC. X I0~4
ANDERSEN IMPACTOR
WOOL FELT
-O— O- DACRON FELT
i
2
i i i
i iii
sj 1 1
DAMPED
1 f
1 1
1 1 1 1
PULSES
_
0 0-
—
RESERVOIR PRESSURE, 70 PSIG
INLET CONC., 11. 6 GRAINS/FT.3
CLOTH VELOCITY, 8.5 FT./MIN.
PULSE INTERVAL, 0.04 TO
2 WIN.
-
—
«/ w
l—o — o— :
PULSE DURATION ,0.06 TO
XI 1
0.15 SEC.
i i 1 i i i i i
5 10 2 5 1(
PARTICLE DIAMETER,
Figure 63. Effect of felt type on fly a*h particulate emissions
-------
in which the fractional emissions are shown for direct and damped pulse
systems. According to the latter data, little distinction can be made
between the Dacron and wool felts. It should be noted that wool emis-
sions are also coarser for the direct pulse conditions.
Dust Type
Limited tests were performed with a resuspended talc dust which was
slightly finer than the fly ash based upon cascade impactor measure-
ments, Table 4. Microscope examination indicated that the particles
were irregularly shaped with few spherical forms (cenospheres) as found
in fly ash. Although the dispersed dust was finer than the fly ash, it
appeared that it formed more loosely packed agglomerate structures ac-
cording to filtration studies with mechanical shaking systems,
Chapter II.
A comparison of filter effluent concentrations for talc and fly ash
given in Figure 64 shows no extreme differences in performance.
Generally, the apparent collection efficiencies for the indicated size
ranges average close to 99.9 percent or greater for both talc and fly
ash aerosols. The reason that talc concentrations were held at the
2
1.6 grains/ft, level was due to a handling problem with the dust feed
system and not to limitations in pulse jet cleaning capacity. For ex-
ample, the average filter resistance for the 40 and 70 psig talc tests
shown in Figure 64 was 2.4 and 1.9 inch water, respectively, with
the lower resistance associated with the higher pressure pulses as ex-
pected. On the other hand, it is also indicated that the pulse damping
process exerted a greater effect on effluent concentration than does a
lowering of pulse pressure from 70 to 40 psig.
The limited measurements performed with the talc/Dacron system showed
that increased reservoir pressure and extended pulse duration led to
increased particulate emissions, thus conforming to the pattern exhibited
by the fly ash/Dacron tests.
213
-------
N>
IU
"J
>s^
z B
Z
o
PARTICLE NUMBER CONCENTRATI
5 5 o o o
0 N W * W
i i i i 1 1 1 1
TALC DUST
INLET CONC., 1.6 GRAINS/FT.3
- V = 8.5 FT./MIN.
A ^
£» <_>
CODE
„ IN
AN
-O-O- 40
_ -£-£• 70
NOTE: PU
PU
1
:gigq^
..J\ yv,.
DESCRIPTION " °"
FT CONC X I0~4 -£ ^ ,
DERSEN IMPACTOR
PSIG , DIRECT PULSE
PSIG, DAMPED PULSE
LSE INTERVAL , 1 MIN.
LSE DURATION , 0.06 SEC.
1 l 1 I 1 1 I
SI 1
FLY ASh
A \
1
1
INLE
O — D-
i i 1
T CONC. , 12
L-A A-
i i
i i i i
\ GRAINS/FT3
A A
1 1 1 1 1
2 5 10 2 5 1<
PARTICLE DIAMETER, pun
Figure 64. Comparative filtration characteristics for talc and fly ash with Dacron felt
-------
A comparison of talc and fly ash filtration with Dacron felt tubes,
Table 21 suggests that there are no dramatic differences between
the capture characteristics of these two dusts. It should be noted,
however, that the system capacity for talc is considerably lower than
that for fly ash. For example, in tests to be discussed later, the
effective Rvalues for fly ash and talc, respectively, appear as 14.5
2
and 80 inch water/ft./min./lb./ft. . The fact that, on the average,
2 2
the residual talc weight is 290 grains/ft, versus 583 grains/ft, for
2
fly ash/Dacron tests and 492 grains/ft, for fly ash/wool tests indi-
cates that proportionally more talc is retained on the basis of inlet
3 3
loading (12 grains/ft, for fly ash and only 1.6 grains/ft, for talc).
Filtration Parameters - Limited tests were performed in which the inlet
dust loadings were reduced by roughly 12 times. In contrast to prior
shaken bag tests where outlet concentration correlated very weakly if
at-all with inlet loading, (Table 12) the pulse jet systems are seen
to be highly responsive to loading changes. The outlet concentrations
are 3 to 5 times greater at the higher load levels over the range of
reservoir pressures studied. Based upon approximate measurements, the
residual dust holdings immediately after cleaning are nearly the same
for all operating conditions shown in the matrix of Table 22. After
a 1-minute filtering interval, one should find that about 12 times as
much dust has deposited under standard filtering conditions (11.6
3
grains/ft, and 8.5 ft./min. filtration velocity). It follows that dust
penetrations should be appreciably lower if the majority of the dust
particles escape the filter during the filtration period. The fact
that the reverse is true can only be explained by assuming that most
of the dust release takes place immediately upon cessation of pulsing.
When normal air flow is resumed, the dust concentration on the dirty
air side of the fabric momentarily exceeds the normal level due to en-
richment by particles dislodged from the filter (but too small to have
settled completely to the hopper). Several measurements by B&L
counter presented later in this section suggest strongly that most of
the dust escapes the bag during the early phase of filtration.
215
-------
Table 21. TALC AND FLY ASH FILTRATION WITH DACRON FELT AT 8.5 ft./min.
Cleaning
parameters
Direct pulse, 40 psig
Pulse frequency, 1 min.
Pulse duration, 0.06 sec.
Damped pulse, 70 psig
Pulse frequency, 1 min.
Pulse duration, 0.06 sec.
Damped pulse, 40 psig
Pulse frequency, 1 min.
Pulse duration, 0.06 sec-
Pulse duration, 0.30 sec.
Damped pulse > 70 psig
Pulse frequency, 4 min.
Pulse interval, 0.06 sec.
Talc emissions
c
Concencra tion
grains/ft? x 106
251
218
100
20
Percent penetration
0.016
0.014
0.006
0.001
b
Fly ash emissions
c
Concentration
grains/ft.3 x 106
1960
613
330
Percent penetration
0.016
0.051
0.028
3 t
Inlet concentration = 1.6 grains/ft. .
Inlet concentration = 12 grains/ft.-^.
£
Concentration estimated gravimetrically by RDM mass monitor.
-------
Table 22. EFFECT OF VARIATIONS IN FILTRATION VELOCITY AND INLET LOADING ON FLY ASH EMISSIONS
WITH DACRON FELT
Outlet concentration grains/ft.3 x 10^ a
Reservoir
pressure
(psig)
40
70
100
Ci = 0.87 gr/ft.3
V = 8.5 ft. /rain.
(low load)
65 (177)b
102 (160)
852 (163)
Ci = 9.9 gr/ft.3
V = 6.2 ft./min.
(low flow)
44 (220)
98 (185)
607 (192)
Ci = 11.6 gr/ft.3
V « 8.5 ft./min.
(standard)
250 (240)
500 (170)
2500 (167)
Ratio
Standard
Low load
3.8
4.9
2.9
Standard
Low flow
5.7
5.1
4.1
ro
a
Concentration measurements by RDM monitor
lumbers in parenthesis are residual dust holdings in grains.
Note: Pulse interval, 1 min.; pulse duration 0.06 sec. Damped pulses.
-------
A second sequence of tests in which the filtration velocity was reduced
from 8.5 to 6.2 ft./min. (and the inlet concentration reduced slightly)
also shoved a significant; i.e., a 5-fold reduction in emissions.
Part of the decrease in effluent concentration is charged to the fact
that about 40 percent less dust is deposited on the filter over a 1-
minute filtration cycle. It appears, however, that the magnitude of
the filtration velocity may also be important during the first part of
the filtering cycle. On the premise that the jet pulse loosens con-
siderable dust without actually dislodging it, the lower (6.2 versus 8.5
ft./min.) filtration velocity should entrain fewer particles and there-
fore lead to lower effluent loadings.
Design Modifications - It has been pointed out previously that the exper-
imental pulse jet system was constructed so that many operating and de-
sign features could be varied. During the present study, it was possible
to examine in at least cursory fashion the following factors that were
suspected to influence overall system performance:
• Pulse damping - Clean air side - In prior descriptions of the
impact of reservoir pressure on particulate emissions, con-
siderable data were presented showing the effect of pulse
damping. Data reported in Figures 55 and 62 for fly ash
and Dacron and wool felt bags indicated that use of the
damped pulse technique resulted in about a 5-fold de-
crease in effluent concentrations. Although a secondary
damping reservoir (Figure 54) was the means used to
accomplish a more gradual return to the normal (filtra-
tion) pressure gradient, it appears that the same results
could be attained more simply by changing the closing
time of the solenoid valve.
• Pulse generation - Clean air side - In the normal arrange-
ment for the jet pulse system, the nozzle tip is located
about 2 inches above the inlet to the Venturi element
attached to the exit end of the bag. The nozzle itself
consists of a 2-inch length of 1/4-inch pipe with a true
inside diameter of 0.364 inch. When the complete nozzle
assembly is detached, however the air vents directly into
the plenum through a 0.84 inch diameter opening located
12 inches above the bag exit. Limited measurements pre-
sented in Table 23, Tests 1 through 15 indicate that the
218
-------
Table 23. SUMMARY OF TEST PARAMETERS FOR FLY ASH AND TALC FILTRATION WITH DACRQN AND
WOOL FELTS
Test
number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Dust/4
fabric
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
Air flow
(£t.3/min)
40
45
40
40
40
40
40
40
40
40
40
40
40
40 .
40
40
40
40
40
40
36
40
40
40
Inlet
cone.
gr./ft.3
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12,0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
Outlet**
cone.
gr./ft 3
x 103
B 1.17
B 0.53
B 0.97
B 1.13
B 1.55
B 0.56
B 1.47
B 2.12
B 1.64
B 2.60
B 1.59
B 2.30
B i.29
B 1.66
B 1.84
B 1.27
B 1.85
B 1.62
B 1.31
B 0.21
B 0.32
B 2.66
B 0.57
B 1.00
Filter resistance
in. water
Average
3.1
4.3
4.5
5.0
4.7
5.1
4.7
2.1
3.0
2.3
2.9
1.5
1.4
3.0
4.7
5.3
4.2
5.6
4.6
6.2
7.3
2.4
4.6
6.2
Residual
2.6
4.0
4.4
4.8
4.6
5.0
4.2
2.0
2.9
1.7
1.8
1.3
1.3
2.7
4.3
4.8
4.0
5.4
4.2
6.1
7.1
2.2
4.2
5.9
Reservoir
pressure0
(psig)
40
40
30
30
30
30
40
40
40
50
50
50
50
40
40
40
40
40
60
D 60
40
90
D 90
40
Pulse
freq.
miti.~L
0.2
0.1
0.05
0.06
0.03
0.1
0.5
0.03
0.06
0.1
0.2
0.06
0.04
0.1
0.1
0.2
0.06
0.06
0.1
0.1
0.1
0.1
0.1
0.1
Pulse
duration
(sec.)
0.15
0.15
0.10
0.13
0.06
0.15
0.06
0.06
0.06
0.15
0.15
0.06
0.06
0.15
0.15
0.15
0.06
0.06
0.06
0.06
0.15
0.06
0.06
0.15
Comment
Plenum pulse, no
jet or Venturi;
Tests 1 through 14
5/8-inch nozzle
and Venturi;
Tests 15 through 18.
No collection drum
Commercial con-
figuration, 1/4-
inch jet with
Venturi; Tests
19 through 33
-------
Table 23 (Continued).
N>
to
o
SUMMARY OF TEST PARAMETERS FOR FLY ASH AND TALC FILTRATION WITH
DACRON AND WOOL FELTS
Test
number
25
26
27
28
29
30
31
32 .
33
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
Dust/3
fabric
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/P
F/D
F/D
F/D
F/D
F/D
Air flow
ft.3/niin.)
30
40
32
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
- 40
40
40
40
40
40
Inlet
cone.
^r./£t.3
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
11.6
11.6
11.6
11.6
11. 6
11.6
11.6
11.6
11.6
11.6
11.6
9.3
0.87
0.87
0.87
0.87
Outletb
cone.
&r./ft,J
x 103
B 0.62
B 1.08
B 0.75
B 0.59
B 2.05
B 1.Q5
B 0.67
B 0.95
B 0.78
R 1.46
R 3.32
R 0.80
R 3.54
R 2.07
R 0.09
R 1.09
R 0.15
R 2.06
R 3.48
R 2.24
R 1.85
R 0.85
R 0.12
R 0.07
R 0.10
Filter resistance
in. water
Average
7.2
5.3
6.9
6.0
3.2
4.6
5.5
5.5
5.5
4.4
4.1
3.5
2.9
3.7
5,1
5.6
7.0
5.5
3.8
5.5
5.6
3.7
5.0
6.5
5.0
Residual
7.1
4.9
6.7
5.7
2.8
4.3
5.0
5.0
5,0
3.4
3.0
2.9
2.0
2.8
4.8
5.3
6.8
4.8
3.2
4.4
4.8
3,3
4.6
6.0
4.5
(.eservoir
>ressurec
(psig)
40
70
D 70
D 100
100
90
90
90
90
D 70
70
D 70
70
70
D 70
40
D 40
D 70
70
70
D 100
D 100
D 70
D 40
D 70
Pulse
freq:!
min. -1
0.1
0.1
0.1
0.1
0.1
0.1
0.2
0.2
o.i
1.0
1.0
0.4
0.4
0.4
0.4
0.4
0.4
1.0
1.0
1.0
i.o
1.0
1.0
1.0
1.0
Pulse
uration
(sec.)
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.10
0.20
0.15
0.15
0.15
0.15
0.06
0.06
0.06
0.06
0.15
0.06
0.06
0.06
0.06
0.06
0.06
0.06
Comment
ommercial con^
figuration, 1/4^-
inch jet with
Venturi; Tests 19
through 33
Low loading tests,
Tests 47 through 50
-------
Table 23 (Continued).
SUMMARY OF TEST PARAMETERS
DACRON AND WOOL FELTS
FOR FLY ASH AND TALC FILTRATION WITH
Test
number
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
Dusty3
fabric
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/D
F/W
F/W
F/W
F/W
F/W
F/W
F/W
F/W
F/W
F/W
Air flow
(ft.3/nin.:
29
29
29
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
Inlet
cone.
gr./ft.3
9.9
9.9
9.9
13,1
13.1
13.1
13.1
11.0
11.6
11.6
11.6
11.6
11.6
11.6
9.5
9.2
11.6
11,6
11.6
11.6
11.6
11.6
11.6
11.6
i i
Outlet1*
cone.
gr./ft 3
x 103
R 0.10
R O.£l
R 0.04
R 2.00
R 0.26
R 1.96
R 0.33
R 3.16
R 6.87
R 3.20
R 1.88
R 2.29
R 0.23
R 0.61
R 2.40
B 0.58
B 0.15
R 1. 00
R 0.20
R 0.66
R 0.13
R 1.96
R 1.51
B 0.38
Filter resistance
in. water
Average
4.4
3.8
4.7
4.3
5.0
4.6
6.4
4.2
3.3
6.7
7.4
4.8
5.3
5.2
4.2
3.4
3.6
4.0
4.9
7.2
4.3
6.4
4.2
4.2
Residual
4.0
3.5
4.4
3.3
3.8
3.7
5.5
3.2
2.4
5.8
6.5
3.8
4.3
4.2
3.4
2.5
2.8
3.1
3.9
6.7
4.6
3.4
3.4
3.9
Reservoir
pressurec
(psig)
D 70
D 100
D 40
D 70
D 50
40
D 40
D 100
100
70
70
70
D 70
D 70
70
100
D 100
D 90
B 70
D 40
D 70
D 70
70
70
Pulse
freq.
min.~l
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
2.0
2.0
2.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
0.4
2.0
1.0
1.0
Pulse
duration
{see. )
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0,06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.15
0.06
Comment
Low flow tests;
Tests 51 through 53
Nozzle raised
4 inches
to
-------
Ni
to
N>
Table 23 (Continued).
SUMMARY OF TEST PARAMETERS FOR FLY ASH AND TALC FILTRATION WITH
DACRON AND WOOL FELTS
Test
number
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
Dust/3
fabric
F/W
F/W
F/W
F/W
F/W
F/'W
F/W
T/D
T/D
T/D
T/D
T/D
T/D
T/D
T/D
Air flow
(ft.3/min.
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
Inlet
cone.
gr./ft.3
11.6
11.6
11.6
11.6
11.6
11.6
11.6
1.53
1.53
1.53
1.53
1.53
1.53
1.53
1.53
Outletb
cone.
gr./ft.3
x 103
B 0.38
R 1.33
R 2.62
R 4.85
R 0.09
R 0.29
R 0.30
R 0.22
R 0.02
R 0.05
R 0.10
R 0.22
R 0.44
R 0.25
R 0.22
Filter resistance
in. water
Average
4.9
5.3
5.0
4.1
4.6
4.7
5.2
1.9
6.0
4.7
2.9
2.4
2.4
2.5
2.1
Residual
3.9
—
--
--
3.7
3.6
4.3
1.4
2.7
2.4
2.1
1.4
1.5
1.7
1.6
Reservoir
pressure0
(psig)
D 70
70
70
70
D 70
D 70
D 70
D 70
D 70
D 70
D 40
40
40
40
40
Pulse
freq.
min , " *•
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
4.0
4.0
1.0
1.0
1.0
1.0
1.0
Pulse
duration
(sec.)
-• i • *• ™~
0.15
0.06
0.15
0.10
0.06
0.06
0.06
0.06
0.06
0.30
0.30
0.30
0.15
0.06
0.15
Comment
Supplementary
plenum pulse, 100
100 psig, 0.35 sec.
Loose bag
Taut bag
*F/D - fly ash/Dacron; F/W - fly ash/wool; T/D - talc/Dacron
bB Indicates B and L measurement; R indicates RDM jneasurement.
CD indicates damped pulse, otherwise direct.
-------
pulse technique (no Venturi used) did not cause any significant
difference in particle emissions relative to the conventional
1/4-inch nozzle system cited previously. The compressed air
requirement is increased by about 10 percent, and the average
filter resistance is somewhat lowered.
When a 0.622-inch nozzle was substituted for the 1/4-inch
model, the outlet concentration was higher and, consistent
with this higher emission, the average filter resistance
was lower. Again the Venturi section was not used during
the above tests.
Nozzle location - Clean air side - Raising the 1/4-inch
nozzle tip to a position 6 inches above the Venturi inlet
rather than the noraml 4 inches, permitted improved bag
cleaning during special tests in which the cleaning interval
was extended to 2 minutes, Tests 62 and 63 in Table 23.
It appeared that with an approximate doubling of the dust
deposit upon the bag very poor dust removal was obtained
over the upper 10 percent of the bag. Allowing the air
jet to entrain more air from the plenum reduces the
possibility of air entrainment from within the bag. The
latter process can induce air from the dirty air side
which reverses the flow direction sought during the clean-
ing interval.
P rirhninr -o-i.mw* - Dirtv air side - During typical field
operation, it is highly unlikely that more than 10 per-
cent of the bags (or compartments) will be undergoing
cleaning at the same time. Hence, during the transient
cleaning period when a small fraction of the system gas
handling capacity is lost, the remaining on-line com-
ponents accommodate the diverted flow with but a small
increase in chamber pressure.
unfortunately, in the single bag test unit, the pulse
Unfortunately, i filtration flow. Therefore,
?he main fan continued to pump air into the dirty air
?5 of fhTsvstem up to its static pressure capacity.
Because the chamber containing the bag was designed
Because en e u h velocltieB on the dirty air
SdTS t « F* tfose found „ jarse fieW assemblies,
side cyPlcai , b ut 3 ft.J Therefore, even
its volume was only about ^^ &
ienbak surea of 5 inches of water might
sient fcacK pr « yalve was open between
*
ienbak su
sient fcacK pr « yalve was open betwee
be expected. fl™eVf*ollection barrel beneath the
the baghouae and the co lie 5 ^ ^ (,dir
thTsystel Thus, a 0.06 second pulse
pressure increase, approximately
223
-------
2-1/2 inch of water, on the dirty air side. The effect
of varying the air volumes on the dirty air side of the
system is shown by Tests 17 and 18, Table 23. Closing
the slide damper led to a decreased pressure differential
across the filter during the pulse interval. As a result,
the cleaning was less effective, the effluent dust con-
centration slightly lower, and the operating resistance
slightly higher. Although these data are far too limited
to allow any firm conclusions, we do not believe that the
volume factor introduced by the special test system had
any serious effect on the simulation of field performance
conditions.
Bag tension - Dirty air side - The circumferential snug-
ness of fit of the bag on its supporting cage was lessened
by reducing the cage diameter by 3/8 inch. In a second
test, a wedge was inserted between the bag and the frame
to produce a tighter fit. The results of these varia-
tions are reflected by Tests 47 and 28 (Table 23) upon
wool felt bags with fly ash. Surprisingly, the data
showed negligible differences both for outlet concen-
tration and average bag resistance. It had been expected
that a slack bag might undergo a more vigorous motion
(greater acceleration) and hence be cleaned more effectively.
Particle Size and Concentration Changes During Filtration Cycle
Data have been presented previously in this Chapter on those factors
that exerted (or were expected to exert) measurable effects on par-
ticulate emissions from pulse cleaned fabric filters. We have shown
that the reservoir pressure level, the pulse form as controlled by
damping tank, the inlet loading level, the filtration velocity, and
the duration of the compressed air pulse exert significant effects on
effluent dust properties. On the other hand the behavior of Dacron and
wool felts with respect to the collection of both fly ash and talc
showed only minor differences. Likewise a reduction in the frequency
of cleaning and minor permutations in nozzle dimensions and location
did not produce any important changes in system performance, suggesting
that other areas be examined for possible improvements in system
operation.
224
-------
In the following paragraphs, the observed changes in concentration for
various particle diameters over the period of the filtration cycle are
examined. Again, a single particle light scattering counter such as
used for the mechanical shaking systems proved to be the only measure-
ment technique capable of providing the desired time resolution.
It has been shown earlier in Figures 58 and 59 that when the fabric is
cleaned very frequently (filtration interval < 0.2 minute) the emission
increases with frequency, but with less frequent cleaning it is rela-
tively independent of frequency. This suggests that emission is largely
associated with the instant of cleaning and the transient period imme-
diately thereafter. This hypothesis appears to be confirmed by Fig-
ures 65 through 70 which show particle concentrations by size and
total mass concentration as estimated by light scattering measurements
over typical operating cycles. In fact, after correcting for the delay
in signal output (sampling time and particle counting time combined
were approximately 0.1 minute) the emission peak appears to be almost
coincident with the termination of the pulse.
Based upon the clean air volume (1.44 ft. ) and the filtration flow
O
(40 ft. /min.) any residual dust in suspension in the clean air side of
the system would be rapidly flushed out. On a very conservative basis,
a concentration reduction of at least 16 times at the end of 1 minute
is predicted by an ideal logarithmic dilution process. This appears to
be the apparent decay rate for the larger particle sizes during some
tests although with high pressure pulses, large particles are found in
the effluent over the complete filtration cycle. The data would tend to
suggest, therefore, that the observed emissions can be only partially
explained by a simple flushing out process. The remaining particles
must represent a combination of those that penetrate continuously over
the filtration cycle or those that are re-entrained from the filter over
the same time period.
225
-------
0.0
0.2
> 0.3pm
>0.5jim
FLY/ASH DACRON ^
70 PSIG , DAMPED PULSE
PULSE INTERVAL, 1 WIN.
PULSE DURATION , 0.06 SEC.
DASHED LINE, AVERAGE
MASS CONCENTRATION
_L
0.4 0.6
TIME , min.
1.0
to1
Figure 65. Particle concentration versus time for selected sizes
226
-------
10'
10"
o
Of
8
10
5
Z
10"
FLY ASH/DACRON
70 PSIG, DIRECT PULSE
PULSE INTERVAL , 1 MIN.
PULSE DURATION , 0.06 SEC
DASHED LINE , AVERAGE
MASS CONCENTRATION
> 3pm
I
1
0.0
Figure 66.
0.2
0.4 0.6
TIME, min.
OJ
-o
o
Z
O
u
o
ID1*
10
Particle concentration versus time for selected sizes
227
-------
10
ro
U.
V
z
a:
H
z
IU
O
z
O
O
n:
UJ
CD
tu
o
oc
g
TALC/DACRON '
70 PSIG, DAMPED PULSE
PULSE INTERVAL ,4 WIN. '
PULSE DURATION , 0.25 SEC
DASHED LINE , AVERAGE
MASS CONCENTRATION
O O
X
10
£
V)
s
U
U
o
I-
Ul
_]
H
o
Ul
eg
Ul
2.0 3.0
TIME.min.
4.0
Figure 67. Particle concentration versus time for selected sizes
228
-------
10*
10
Ul
U
O
O
DC
UJ
CD
UJ
_l
O
P
tr
T
T
T
FLY ASH, WOOL
70 PSIG, DAMPED PULSE
PULSE INTERVAL , I MIN.
PULSE DURATION 0.06 SEC.
DASHED LINE , AVERAGE
MASS CONCENTRATION
> 2 jim
I03
(O
o
x
10
K*i
(£.
10'
UJ
UJ
Ul
^
a:
0.0
0.2
0.4
0.6
TIME, win.
0.8
Figure 68. Particle concentration versus time for selected sizes
229
-------
10
ro
H
Z
o
fe
111
o
z
o
o
o:
UJ
CO
id
_l
o
I-
o:
10=
10* —
T I I
FLY ASH/DACRON
100 PSIG, DAMPED PULSE
PULSE INTERVAL, I MIN.
PULSE DURATION 0.06 SEC.
DASHED LINE , AVERAGE
MASS CONCENTRATION
I
> 5pm
I
I
u>
103 O
,
u.
v.'
(O
10'
a:
o
a:
i-
ui
o
o
o
lil
U
o
I01 y
0.0
0.2
0.8
1.0
0.4 0.6
TIME , min.
Figure 69. ParticleJcoricentration versus time for selected sizes
230
-------
FLY ASH, DACRON
40 PSIG, DAMPED PULSE
PULSE INTERVAL, I MIN.
PULSE DURATION, 0.06 SEC.
DASHED LINE.AVERAGE
MASS CONCENTRATION
•o
o
o*.
10
c
"5
o»
•%
Z
O
TIME . win.
Figure 70. Particle concentration versus time for selected sizes
231
-------
Generally speaking, dust penetration decreases progressively over the
filtration cycle with the greatest changes observed for the larger par-
ticles. The estimated mass concentrations (calculated on the basis of
the indicated number concentrations and an assumed particle specific
gravity of 2.0) appear as the right hand ordinates of Figures 65
through 70. The decay paths for the overall mass concentrations,
which are shown as dotted lines, tend to follow those of the predominant
particle sizes. Prior filtration tests with mechanically shaken, cotton
bags indicated that the overall weight collection efficiencies for fly
ash decreased about one order of magnitude for each addition of 5 to 6
grains of dust per square foot of filter area. Based upon a cloth load-
2
ing rate of approximately 100 grains/ft, /min. for pulse jet filtration
one might predict that efficiency would decrease one order of magnitude
in 0.055 minute. The family of curves appearing in Figures 65 through
70, however, not only indicate much less rapid decay rates but even
a tendency to level off in effluent concentrations after the first
1/2 minute of filtration.
It appears, therefore, that the filtration process is quite different
for the felt media. Whereas experimental measurements and theoretical
calculations indicated that the interstitial pore structure filled
quite rapidly with mechanically shaken, woven fabrics, the predominant
filtration process with pulse-cleaned felt media appears to be mainly
a statistical or deep bed process in which particle and fiber size
fiber packing density, bed depth, filtration velocity and inlet concen-
trations may play very important roles.
Regardless of the mechanisms involved in the filtration process using
damped pulses, the dust concentrations at the end of each 1 minute fil-
ter cycle ranged from 20 to 100 times less than that, estimated after
0.1 minute of filtration for systems. High early emission rates are be-
lieved to be associated mainly with the pulse action upon already
loosened particles and a transient dilation of the pore structure.
232
-------
Residual Filter Resistance
The fact that about 95 percent of the operating pressure differential
is due to imbedded dust rather than to the fiber matrix (clean media
resistance at 8.5 ft./min. is 0.12-inch water) suggests that most of
the particle collection is associated with the dust rather than with
the fiber matrix. Unfortunately, the in-situ measuring technique used
to estimate gravimetrically the residual or final filter dust holdings
was insufficiently accurate to establish useful correlations between
filter dust holding and average particulate emissions or residual
resistance. It is believed, however, that the amount of dust initially
present on the filter should relate closely to the average, initial, or
final filter resistance for a fixed set of filtration and cleaning para-
meters. Figure 71 indicates a fairly high correlation (~= 0.66) bet-
ween residual dust deposit and residual filtration resistance despite
the fact that the data were collected under wide range of dust, fabric
and pulsing conditions.
In Figure 72, the average filter emission (assumed to be proportional
to the peak emission after cleaning) is plotted against the average fil-
ter resistance. A consistent relationship is indicated in which the
outlet concentration decreases approximately an order of magnitude for
each additional 2-1/2 inch water of pressure differential. The data
points depicted on Figure 72 should be considered in three general
categories. The first represents a sample population based upon fly ash/
Dacron measurements in which the pulse interval was constant at 1 minute
and the inlet dust concentration was 12 grains/ft.3. Otherwise, reser-
voir pressures ranged from 40 to 100 psig, pulse durations were 0.06 and
0.15 second and the pulses were eigher direct or damped. Despite the
indicated variations, the effluent concentration seems to depend mainly
on the filter resistance as shown by the regression line labeled A. A
second data set depicts a limited test sequence in which the inlet fly
ash loading was reduced to 0.87 grains/ft. . Previous analyses of the
measurements shown on Curve B (also shown in Figure 55), suggested that
233
-------
6 -
1 1 1 1 1 1
O FLY ASH/DACRON
A FLY ASH/ WOOL °
f I
A
£
O
-------
«£ 10.0
1 5.0
g
< 2.0
z
ui
-------
the emission related not only to the filter average dust holding but also
to the inlet concentration. In view of the limited information available,
there appears to be no reason to assume a different slope for Curves A
and B.
The third data set, Curve C, is based upon talc filtration with a
•i
Dacron felt. Except for the fact that the loading was 1.6 grains/ft. ,
the variables were the same as cited for Curves A and B. It was not
expected that the talc data would fall on Curve A, because inlet loading
had already been shown to have a significant effect on outlet loading.
The fact that filter resistance is appreciably lower for talc filtration
than for fly ash based upon equal outlet concentrations, is attributed
to the lower bulk density of the talc as it deposits within the fiber
matrix. It is assumed that the volume and not the mass of dust within
the filter pores is the key factor in determining efficiency. This con-
cept has been discussed earlier in Chapter II in which filter efficiency
for shaken bags has been related to the volume of dust upon the filter.
Although the few data points based upon fly ash filtration with wool
felt are scattered broadly, they appear to fall within the realm of the
fly ash/Dacron results.
Concentration Profiles, Dust Emissions
In the preceding discussions, the reported emissions are those at the
outlet of the top plenum which depict the weighted average concentra-
tions over the entire filter bag. Based on the premise that emission is
related to the degree of cleaning, one expects that effluent concentra-
tions may vary along the filter bag. By lowering a flexible sampling
line (attached to the inlet of the B&L counter) into a fourth long bag
it was possible to extract samples at selected locations. A vertical
(upward) traverse of the bag interior provided a series of data points,
each describing the average concentration of particles in the effluent
air up to the point of sampling. Some particle loss took place in the
236 -
-------
sampling lines and the shock of the cleaning pulses probably caused inter-
mittent sloughing off of particles from the walls of the sampling tube.
Despite the above problems, the measurements were considered useful as
an indication of concentration variations along the bag.
Figure 73 shows that the local dust concentration levels increase
as one traverses from the bottom to the top of the bag. The greatest
increases takes place from the mid position to the top end of the bag
in which region the cleaning appears to be the most vigorous. Based
upon inspection of the bag, the densest deposits appear to be at the
bottom with a steady decrease noted up to approximately the end of the
Venturi section (roughly 6-inches from the top of the bag). From the
Venturi element upwards, the cleaning is poor according to the apparent
thickness of the residual deposit after pulsing.
Since average filter emission correlates with average pressure differ-
ential (Figure 72), it is probable that the filter bag was poorly
cleaned at the lower end and over-cleaned in the region just below the
Venturi. Since a concentration gradient appears with a 4 foot bag it is
proposed that the potential for non-uniform cleaning might be much
greater for longer bags.
Operating Filter Resistance
One of the major costs in operating a gas cleaning system is that for
moving the dusty gas through the collection device. In the case of
filter systems, the resistance to gas flow through the fabric medium
may represent a large part of the fan total static pressure requirement.
Therefore, a full understanding as to how such filtration resistance
may be minimized while meeting a specific effluent criterion, is very
important. In the case of mechanically shaken bags, the resistance
increased linearly with time (and cloth loading) over more than 80 per-
cent of a 30 minute filtration cycle, (see Figure 45, Chapter II).
Therefore, the average operating resistance was readily determined. On
237
-------
(O
I
O
J£ 6.0
Z
a 5.0
^
Z
O
< 4.0
oe.
-------
the other hand, it was noted that only during the last 60 percent of a
normal, 1 minute filtering cycle was the resistance/time curve essen-
tially linear for a pulse jet system, Figure 74. If the resistance
curves are examined in terms of the quantity of dust deposited on the
2
filter (expressed as grains/ft. ) then 0.4 minute of pulse jet operation
o
at 8.5 ft./min. and 12 grains/ft, should be equivalent to approximately
4 minutes of filtration with the mechanically shaken system described
o
in Chapter II at 3 ft./min. and 3.5 grains/ft. . Therefore, it appears
that the amount of fly ash required to seal over of the fabric surface
is roughly the same for both the mechanically shaken cotton fabric and
the pulse jet cleaned Dacron felt. If one compares the relative efflu-
ent concentrations for the mechanically shaken and the pulse jet sys-
tems, it is also seen that the greatest reduction in effluent concen-
tration takes place during the initial 4 minute and 0.4 minute filtra-
tion periods, respectively.
Because of the rapid changes in the system when normal air flow resumes,
it is impossible to define accurately the fabric resistance immediately
after cleaning. It was found, however, that consistent relationships
were obtained for all resistance values measured 0.1 minute after the
resumption of filtration, Figure 74. Furthermore, when the average re-
sistance of the filter for the complete (1 minute) filtration cycle was
estimated by taking the actual curve path into account, it was found
that it differed from the 0.5 minute value by no more than a few per-
cent. Therefore, it was considered acceptable to use the resistance at
the midpoint of the filtration cycle as the average value.
Filter resistance characteristics for talc dust and Dacron felt are
o
given in Figure 75 for an inlet loading of 1.53 grains/ft, and a
filtration velocity of 8.5 ft./min. The slopes of several curves were
essentially constant irrespective of cleaning frequency, 1 to 4 tninutes,
and the apparent "K" value was 80 as compared to 14.5 for fly ash. In
contrast to fly ash measurements, the linearity of the curves extends
over a much broader range even though the surface dust loading is about
239
-------
o
to
LU
Of.
Of.
LU
5 2
D
SYMBOL
O
A
x
D
SPECIFIC RESISTANCE
TEST COEFFICIENT
NUMBER IN. WATER/FT./MIN./LBS./FT.2
63 17.1
55 13.8
64 15.6
56 13.9
54 13.6
58 14.0
0-2 0-4 0-6
FILTRATION TIME, min.
Figure 74. Resistance characteristics for fly ash and Dacron felt»
inlet cone. 12 grains/ft.^, filtration velocity
8.5 ft./min.
240
-------
U
z
LU
Of.
OL
LU
SPECIFIC RESISTANCE
TEST COEFFICIENT
SYMBOL NUMBER IN. WATER/FT./MIN./LB./FT.2
O T-4 83.0
A T-5 80.5
D T-7 83.5
X T-8 69.7
J-
JL
J_
0-2 0-4 0-6 0-8
FILTRATION TIME, min.
1-0
Figure 75. Resistance characteristics for talc and Dacron felt, inlet
cone. 1.53 grains/ft.3, filtration velocity 8.5 fpm
241
-------
8 times lower. It is believed that because of the lower bulk density
for talc the pore filling phase of the dust collection process is com-
pleted more rapidly. Under the latter circumstances, subsequent dust de-
position is mainly superficial, which leads to an approximately linear
relation between resistance rise and subsequent dust deposition.
In the following chapter, those operating factors that were expected to
influence filter resistance are examined.
Compressed Air Pressure - Figure 76 indicates average filtration
resistance for several measurements performed with a fly ash/Dacron sys-
tem as a function of reservoir pressure. Operating filter resistance
is seen to be roughly inversely proportional to reservoir pressure, in
view of the limited amount of experimental data available, several
variations in operating and/or cleaning parameters have been grouped for
each reservoir pressure level for purposes of data smoothing. A com-
parison of direct and damped pulse measurements indicates an average
resistance lowering of about 1-inch water when direct pulses are used.
The corresponding impact on effluent concentration was much greater*
i.e., a five fold increase in effluent concentration when using direct
pulses according to Figure 55.
Pulse Interval - Selected data shown in Figure 77 indicate that
extending the interval between pulses causes an essentially linear in-
crease in average filter resistance. A few trials were performed in
which the pulse intervals were reduced to very brief periods, ~ 0.06 to
0.1 minute. Operating resistances appeared to be higher during these
tests than those for the 0.5 minute pulse intervals given in Figure 75.
Presently, it is believed that some undefined instrument error led to
these results. From a practical viewpoint, the reduction of pulse in-
tervals to less than 0.5 minute for any one bag, would lead to pro-
hibitive compressed air requirements, increased effluent concentrations
as shown in Figures 58 and 59, and certainly increased air handling
costs if in fact, the filter resistance actually increases for the very
brief, ~ 0.06 to 0.5 minute pulse intervals.
242
-------
. 5
UJ
U
Z
4
UJ 4
Of.
Z
o
uJ 2
IO) DAMPED
(x) DIRECT
NOTE:
DATA POINTS FOR EACH PRESSURE REPRESENT
AVERAGE OF THE FOLLOWING VARIABLES'-
INLET FLY ASH LOADING,0.87 TO 13.1 GRAINS/FT.*
FILTRATION VELOCITY, 6.2 TO 8.5 FT./MIN.
PULSE INTERVAL, 0.4 TO 2.0 MIN.
PULSE DURATION , 0.06 TO 0.15 SEC.
1
i
Figure 76.
20 40 60 80 100
INITIAL RESERVOIR PRESSURE, psig
Average filter resistance versus reservoir pressure, direct
and damped pulses, for fly ash filtration with Dacron felt
243
-------
NJ
O 10
c
x
uj 6
QC
1^ J
Ul
O
oc
IU
CODE DESCRIPTION
O FLY ASH/DACRON, DIRECT PULSE, DURATION 0.06 SEC.
A FLY ASH/DACRON, DIRECT PULSE, DURATION 0.15 SEC.
A FLY ASH/DACRON, DAMPED PULSE, DURATION 0.15 SEC.
• FLY ASH/WOOL, DAMPED PULSE, DURATION 0.06 SEC.
O TALC/DACRON, DAMPED PULSE, DURATION 0.06 SEC.
NOTE' FLY ASH CONC. 12 grains/ft.3
TALC CONC. 1.6 grains/ft.5
JL
JL
JL
2-0
PULSE INTERVAL, min.
3-0
4-0
Figure 77. Effect of pulse interval on average filter resistance for 70 psig pulses
-------
Pulse Duration - Although prior studies had indicated that pulse dura-
tion did not exert any strong effect on filter resistance,-* it was be-
lieved that single bag tests might reveal some differences. However,
according to data shown in Table 24 extended pulses have no signif-
icant effect on filter resistance, whereas they do lead to slightly
higher particulate emissions. Only two tests in Table 24 suggest
that there might be some advantage resistance-wise in extending the
pulse duration. More tests will be required, however, to determine
whether the resistance decreases are significant. From a practical
viewpoint, it should be noted that the extended pulse duration repre-
sents a nearly threefold increase in compressed air demand. Addition-
ally, prior measurements have also shown an adverse effect on emissions
( > 60 percent increase in outlet concentration).
Felt Type - When wool felt was substituted for the Dacron media, the
average operating resistance for fly ash was basically unchanged,
Table 24. Since felts were similar in weight and permeability no
large difference in filtration resistance was expected. In Figure 78
average resistance values for several fly ash/wool tests are shown for
direct and damped pulses as a function of reservoir pressure. Inlet
dust concentration, pulse interval and pulse duration were constant
o
at 1.53 grains/ft. , 1 min., and 0.15 sec., respectively. There does
not appear to be any significant difference in the resistance proper-
ties of wool and Dacron felts for damped pulse systems. Curves 2 and 4
represent the data for fly ash/Dacron systems given in Figure 76.
The apparent difference indicated for direct pulses suggests a greater
response to reservoir pressure for the fly ash/wool combination.
Dust Type - Although the reported resistance values for talc were lower
during most tests than those observed for fly ash filtration it should
be noted that the inlet loadings were also very much less, approximately
8 times. In order to place the resistance changes of fly ash and talc
on a comparable basis they must be compared with the incremental dust
deposit rates. This has been accomplished by calculating the specific
245
-------
Table 24. EFFECT OF PULSE DURATION ON AVERAGE FILTER RESISTANCE FOR VARIOUS DUSTS AND FABRICS
10
Operating parameters
Wool felt, flv ash
70 psig, pulse interval
1.0 min.
70 psig, pulse interval
1.0 min. reservoir
and rrRulator inter-
changed
Dacron felt, fly ash
70 psig, pulse interval
l.-O min.
Dacron felt, talc
40 psig, pulse interval
1.0 min.
70 psig, pulse interval
4.0 min.
Average filter resistance, in. water
Direct pulses
Pulse duration
0.06 sec.
4.20
5.31
5.5
2.54
-
0.15 sec.
4.19
5.0
4.3
2.40
-
Ratio
1.00
1.06
1.28
1.06
-
Damped pulses
Pulse duration
0.06 sec.
4.90
4.9
—
6.0
0.15 sec.
4.90
4.9
—
0.30 sec.
4.7
Ratio
1.0
1.0
_
1.28
Mote: Inlet concentrations, fly ash 12 grains/ft. , talc 1.53 grains/ft.3.
-------
o
.. 5
01
U
i
UJ A
ex *
Z
O
5 3
u.
UJ
o
§ 2
\
SOLID LtNES
CURVE 2, DAMPED PULSE, DACRON
CURVE 4, DIRECT PULSE, DACRON
DATA FROM FIGURE 76
DASHED LINES
CURVE I, DAMPED PULSE, WOOL
CURVE 3, DIRECT PULSE , WOOL
20 40 60 80
INITIAL RESERVOIR PRESSURE, psig
100
Figure 78, Comparative resistance properties of wool and Dacron
fly ash filtration
247
-------
resistance coefficients, K values, for the data presented in Figures
74 and 75. The average K values for £ly ash actually were
quite similar, ~ 13 to 16, to those found for fly ash filtration with
woven cotton and Dacron fabrics in mechanical shaking systems. On the
other hand, talc K values were generally about 2.5 times greater (80
versus 32) suggesting that the nature of the dust deposition must be
different for the felt media. As stated previously, the increase in
cloth loading over the first 0.4 minute of fly ash/felt filtration was
the same as that for prior fly ash/cotton sateen at the end of 4 minutes
In both cases the slope of the resistance versus loading curves were
linear beyond these points. However, the linear portions of the talc/
Dacron felt curves appear to start at cloth loading levels greater than
2 f\
10 times lower; i.e., < 4 grains/ft, rather than 40 grains/ft. . This
behavior, which suggests that less talc dust penetrates the felt inter-
stices, is confirmed by the lower residual dust holding for talc as
2
compared to fly ash (330 versus 660 grains/ft. ).
The fact that the K values are appreciably greater for talc is consis-
tent with the effective size of the dusts as measured by cascade im-
pactor; e.g., HMD = 8 urn for fly ash and 3.2 um for talc. Since K
values, in theory, are inversely related to the diameter squared for
monodisperse particles, one should expect to find the K values for talc
of the order of 6 times greater than those for fly ash.
Filtration Parameters - Although both decreased filtration velocity and
inlet dust loading were shown to exert a significant effect on effluent
concentration, Figure 55, their effects on average filtration resis-
tance appear to be minimal, Table 25. This is somewhat surprising in
that one might expect those filters with the lower dust holdings to
68 ^
show lower resistance. ' Earlier tests by Dennis et al indicated
that average filter resistance rose approximately as the fourth root of
the inlet concentration. The above measurements, however, were per-
formed on a sequentially cleaned, 9 bag unit, such that considerable
masking of any variable change should be expected. In the case of
248
-------
Table 25. EFFECT OF INLET CONCENTRATION AND FILTRATION VELOCITY ON
AVERAGE FILTER RESISTANCE FOR FLY ASH AND DACRON
Reservoir
pressure
40
70
100
Average filter resistance, in. water
Normala
filter cycle
6.4 (281)e
5.3 (177)
4.9 (168)
Reduced**
dust loading
6.5 (177)
5.0 (160)
3.01 (163)
Reduced6
filtration velocity
4.7 (220)
4.4 (185)
3.8 (192)
Adjusted
resistance
6.5
6.1
5.2
a!2 grains/ft, inlet concentration. Cloth loading 51 grains/ft.2 at
0.5 min.
0.87 grains/ft. ^ inlet concentration. Cloth loading 3.7 grains/ft, at
0.5 min.
C9.9 grains/ft.3 inlet concentration. Cloth loading 30.5 grains/ft.2 at
0.5 min.
Filter resistance adjusted to 8.5 ft. /min.
Dumber in parenthesis is residual dust loading in grams.
Note: All cleaning based on 1 minute pulse intervals, and 0.06 second
pulses. Normal filtration velocity 8.5 ft. /min. Reduced filtra-
tion velocity 6.2 ft. /min.
reduced velocity, the observed resistances were significantly lower, but
primarily because of the lowered filtration rate. Pro-rating these data
to a velocity of 8,5 ft. /min. actually indicated equal or slightly higher
resistances despite the fact that the average filter weight gain was
about 40 percent lower than for normal conditions. Because the inlet
velocity was reduced, there is the possibility that a somewhat finer
aerosol approached the filter due to increased hopper losses. Hence,
some increases in filter resistance might be postulated. * It was
observed that the filter resistances recorded for the low loading tests
rose almost immediately to the average (or 0.5 minute) value. There-
fore, it appears reasonable to assume that system operation at reduced
load levels may lead to more interstitial penetration of the dust. In
249
-------
contrast, a high loading probably exerts a choking action that confines
a larger fraction of the dust to the superficial region. The net re-
sult should be a higher operating filter resistance. Unfortunately,
there are not sufficient data at this time to warrant any firm conclu-
sions. The results are sufficiently interesting, however, to suggest
strongly that further experiments be performed.
Design Modifications - The effect of equipment designs outlined in
Table 23 upon particulate emissions has been discussed previously.
In the following paragraphs, the modifications are treated in the same
sequence starting first with those on the clean air side of the bag.
Pul»e damping - clean air side - As pointed out previously, pulse damping
produced on the average a 4- to 5-fold reduction in effluent concentration
over the reservoir pressure range, 40 to 100 psig. If one equates re-
duced dust penetration with increased filter resistance, one should expect
to see the higher resistance to air flow with damped pulse systems. This
proved to be the case according to the data given in Figure 76, that
shows an approximate 20 percent resistance rise when the damping approach
i a employed.
Despite the fact that test data are limited and subject to errors be-
cause of complex instrumentation problems, the pulse damping studies
indicate that control of pulse form by mechanical or aerodynamic means
is one avenue where power requirements might be optimized with respect
to some pre-selected effluent standards.
Pulse generation - clean air side - By admitting the compressed air di-
rectly into the exit plenum chamber (1 ft.3 volume) and removing the
Venturi section from the exit end of the bag, the filter resistance was
lowered by about 30 percent, Table 26. Although no significant in-
crease in effluent concentration could be established, the compressed air
requirement was increased by about 10 percent. In that the elimination
250
-------
Table 26. FLY ASH EFFLUENTS FROM PULSED DACRON
BAGS AS A FUNCTION OF PULSE FORM
Pulse type
40/24b
Plenum pulse,
no nozzle.
Vent opening
0.84 in.
40/24b
Large nozzle,
I.D. = 0.622 in.
40/24b
Small nozzle
I.D. = 0.364 in.
Pulse
interval
(min.)
0.1
0.1
0.1
Pulse
duration
(sec . )
0.1
0.1
0.1
' • r— _________
Average9
outlet
concentration
(grains/ft.3 x 106)
1660
1840
1000
Average
filter
resistance
in. water
3.0
4.6
6.2
Particle concentrations estimated from B&L counter data.
Reservoir pressure at beginning and end of pulse.
of Venturi elements and individual bag nozzles represents a simplifica-
tion in collector design it appears important to explore further the
possibilities of the plenum pulse approach.
Nozzle position - clean air side - Limited measurements, Table 23,
Tests 62 and 63, indicated that elevating the 1/4-inch pulse jet nozzle
permitted more effective bag cleaning, at least for those tests where
the cleaning interval was extended to 2 minutes. Until the nozzle was
raised 4 inches, the pulse failed to dislodge the dust from the upper
section of the bag (6 to 10 in. from the top). In addition to decreas-
ing the filtration resistance and permitting equilibrium operation, the
effluent concentration increased as expected. These tests suggest that
it is very important to determine at precisely what nozzle locations the
best cleaning conditions can be obtained.
251
-------
Baghouse volume - dirty air side - Eliminating about 57 percent of the
housing volume on the dirty side of the bag (by closing a slide gate
above the dust collection drum) resulted in a greater than 30 percent
increase in operating resistance. The dirty side volume is apparently
necessary to prevent excessive back-pressure on the dirty side of the
bag during the pulse interval. Higher pressures on the dirty air side
appear to attenuate the rate of reverse pressure rise. For example^ the
pressure rises for Tests 17 and 18, Table 23 were 1540-inch water/sec.
and 1910-inch water/sec., respectively. Qualitatively, these values are
in agreement with pressure rise/filter resistance relationships discussed
later in this section. The same phenomenon) is presumed to take place in
a sequentially cleaned, multi-bag system, although the effect would be
much less noticeable because of the far greater expansion volume.
Bag tension - dirty air side - An attempt was made to simulate what mi ah
be typical "uniformity of fit" conditions for the felt bags on the wire
cage frame. Although the experimental data were confined to two mea-
surements, the results appear to be in general agreement with what one
would expect to find; i.e., the slack bag showing a lower average filter
resistance, 4.6-inch water, compared to the tight bag, 5.2-inch water
The rational for the above is that a lowered resistance to motion in the
slack bag should permit a greater acceleration rate. In turn, greater
dust dislodgement and hence a lower residual filtration resistance should
result.
Control of Operating Resistance
It has been shown that the average filter operating resistance depends OK
a variety of design and operating factors. In the following section, an
attempt has been made to analyze several specific relationships; e»g
that between filter operating resistance and reservoir pressure so
252
-------
that the results of the test program may be extended beyond the bound-
aries of the limited dust/fabric combinations and operating parameters
examined in the present study.
The following discussion relates the two components of average operating
pressure, namely the residual and the transition pressures, each to
their determining parameters. It is shown that residual pressure depends
on the intensity of the cleaning pulse and on certain fabric and dust
properties and that the transition pressure depends on the initial or
just-cleaned state of the filter, on the K value of the dust, and on
the amount of dust added between cleanings. With these simple rela-
tionships, the average operating pressure can easily be estimated.
Residual Resistance - As shown by Figure 74, the pressure differential
immediately after cleaning changes so rapidly that it is difficult
to measure accurately and, therefore, has little practical value. Most
investigators prefer to use an "effective" residual pressure differential;
i.e., the value obtained by extrapolating the linear portion of the
p(t) curve back to time zero. This process was carried out for the
pressure-time curves given in Figures 74 and 75. As stated pre-
viously, these curves achieved practical linearity after about 40 per-
cent of the filter cycle had been completed. Pressure values from 0.1
to 0.3 minutes showed the characteristic rapid rise during the early
phase of filtration. The average of 0.1 and 1 minute resistance values,
however, was within a few percent of the 0.5 minute resistance. This
indicated that except for highly unusual situations, the 0.5 minute
"average" value was adequate for estimating power needs.
The filter pressure differential is due mainly to dust on or within the
felt since the latter differential is about 25 times greater than that
for a clean fabric at the same filter velocity. Residual pressure dif-
ferential, therefore, must depend on the amount of dust on the filter
after cleaning, and also on its distribution on and within the fabric
structure. Several other factors, however, such as fabric properties,
253
-------
dust properties, and system design characteristics must also be con-
sidered. An extensive discussion of filtration performance and its
relationship to cloth-fabric structure has been presented recently by
Draemel.
In the following sections, experimental and theoretical approaches are
discussed for estimating the amount of dust remaining on the fabric
after cleaning.
Residual dust - Measurement of the dust remaining on the filter after
cleaning were made by weighing the whole filter assembly (bag, cage,
and Venturi). Because the dust weight represented only 5 to 10 percent
of the assembly weight, accurate determinations of the dust weight
could not be made. Consequently, caution should be exercized in apply-
ing the relationship between residual dust holding and residual resist-
ance. On the basis of the data prescaled in Figure 71, it appeared
that no valid distinctions could be made between the behavior of fly
ash/Dacron, fly ash/wool, or talc/Dacron filter systems.
Talc filtration with Dacron felt gave results similar to those for fly
ash, except that both resistance and dust holdings were about 1/3 lower.
Although the relationship between residual dust weight and pressure is
a fundamental one, it has no practical use because residual dust weight
is not under the direct control of the filter operator. In order to
characterize residual resistance, one must look beyond the residual
dust weight to one or more characteristics of the cleaning process.
Pulse rise rate - It has already been shown that average operating re-
sistance does not depend on the duration of the cleaning pulse, based on
results obtained both in this study and in previous studies. Thus, the
residual resistance fraction of the operating pressure is presumably
independent of pulse duration. However, residual resistance does depetid
on pulse intensity, as will be demonstrated here and in the subsequent
treatment of low pressure, reverse flow cleaning.
254
-------
In the case of shaken bags, the amount of dust remaining on the bags
after cleaning was found to correlate with the acceleration imparted to
the fabric. This suggests that in pulse cleaning too, acceleration may
determine the residual state of the filter. The maximum acceleration
experienced by the bag can be estimated by Equation (3.2), whose develop-
ment is given in Appendix L.
\
(3
dt
The terms G and M are the flexibility and stretch characteristics of the
bag, respectively, and p is the gross mass per unit area of the filter
(felt plus dust). The pressure expression is the rate at which the pulse
differential pressure changes across the bag at the beginning of the
cleaning pulse. One can infer from Equation (3.2) that the residual resis-
tance of the filter depends on the rise rate of the pulse differential.
The rise rate, when plotted against residual filter resistance, Figure
79, indicates that the residual filter resistance AP is described
approximately by the following relation:
104 ,3 3s
(3 }
(d
where both pressures are in inch water and time is in seconds. In its
present form, Equation (3.3) applies only to the the specific dust/fabric
combinations and operating parameters cited in Figure 79. The main
value of this relationship lies in the fact that the rise rate of the
pulse differential is determined by the bag dimensions and design of the
pulse generating equipment, both controllable variables. By using con-
ventional engineering calculations, one may predict the rise rate for a
given system and, from equations similar to the above, estimate the
residual operating filter resistance.
255
-------
FLY ASH/DACRON
5 2
UJ
at
A
A
D
DIRECT PULSE
DAMPED PULSE
0.87 GRAINS/FT.3
DAMPED PULSE
FLY ASH/WOOL
12 GRAINS/FT.3
DIRECT PULSE
DAMPED PULSE
I I
I
I
1000 2000
PRESSURE RISE RATE
Figure 79.
3000 4000
in. water/sec.
Relationship between rate of pressure rise and residual
tration resistance at 8.5 ft./min. filtration velocity
256
-------
During fly ash tests, the rise rate ranged from 1500 tp 3000-inch water/
sec. for damped pulses and from 3000 to 4000-inch water/sec, for direct
pulses. (Note that these values are a function of the design of the
equipment and do not apply to longer bags nor to other pulse delivery
configurations.)
If one refers to Figure 80, it can be seen that residual resistance
is roughly inversely proportional to both pressure rise rate and reser-
voir pressure. This means that the pressure rise rate varies linearly
with reservoir pressure for the systems studied, and, according to
Equation 3.2, the maximum acceleration imparted to the bag varies linearly
with reservoir pressure. Additionally, the residual filtration resis-
tance is seen to vary inversely with bag acceleration for the pulse jet
system. In contrast, the impact of acceleration on residual resistance
in mechanical shaking systems appeared to be less, Figure 50 wherein
residual resistance varied roughly as the inverse square root of the
acceleration.
The data scatter in all these test results was greater than expected
considering the care taken during the investigation to control all ex-
traneous variables. One possible reason for the dispersion should be
mentioned. If the flexibility of the felt, G, rises with improved
cleaning, then the resulting greater fabric acceleration should further
improve cleaning. Conversely, a heavier dust deposit would probably
lower G, thus decreasing the acceleration and causing a still heavier
deposit. In other words, the residual operating point may be unstable
and dependent to some degree on the effect that dust within the filter
matrix has on the filter properties.
Acceleration - In an effort to test the validity of Equation (3.2) that
gives the acceleration of the fabric, several alternative measurement
techniques were used (see APPARATUS, MATERIALS AND TECHNIQUES). The re-
sults in Table 27 suggest that a typical level of acceleration is
about 250 g's, which is 25 to 50 times greater than that encountered in
257
-------
ju 10
TT— 1 1 1 1 1 1 I I I.
_ N
I '-
to
oc
UJ
9
uj 1.5
' V''.
. i
i
(a)
J
\
\
J I I I I I
1.0 2.0
PRESSURE RISE RATE
5.0 10.0
in. WATER/SEC. XK>-»
10
ct
UJ
I
c
i
(/>
in
oc
9
UJ
oc
T
I
i i
DACRON
- O DAMPED
X DIRECT
' WOOL
A DAMPED
P DIRECT
(b)
I
I
I I 1 I I
20 40 60 80 100
RESERVOIR PRESSURE, psig
150
Figure 80. Residual filter resistance versus (a) pressure rise rate
and (b) reservoir pressure (see Figures 76, 78 and 79,
respectively)
258
-------
Table 27. TYPICAL DESCRIPTORS OF PULSED BAG MOTION (DACRON FELT)
1. Static measurements
Displacement versus pressure
2. High speed cameraa»b
Fabric
Dust
3. Strain gauge3
4. Accelerometer3
5. Computations
Expansion modulus and
pressure-time traces
6. Dust trajectory
Maximum
displacement
(inch)
0.26
0.33
0.24
0.26
Maximum
velocity
(ft./sec.)
11.4
6.3
6.55
4.35
2.0
Maximum
acceleration
g's
296
272
225
252
'cleaning regime - 70 psi, 0.06 sec. pulse, 6 sec. intervals.
Observed frequency - 133 cps.
shake cleaning. In the latter case, however, the bag was put through
the acceleration process some 200 to 300 times. With repetitive shaking
and particle loosening in conjunction with a statistical spread in par-
ticle adhesive and cohesive bonds, significant dust removal was accom-
plished despite the much lower acceleration. Substituting the charac-
teristic parameters for Dacron and wool felts from Table 28 into
Equation (3.2) in conjunction with the pressure rise term, dAp/dt, calcu-
lated from oscilloscope pressure/time traces gave acceleration levels of
290 and 1870 "g's," respectively for Dacron and wool felts. The Dacron
Value checked reasonably well with the other measurements given in
Table 27. There is some doubt as to the high acceleration level
259
-------
estimated for wool because both resistance and dust penetration proper-
ties did not differ radically from those of the Dacron felt* Possible
errors in the G or M parameter are suspected.
Table 28. PARAMETERS FOR USE IN EQUATION (3.2)
Fabric
Dacron
Wool
G
ft./(lb./ft.2)
0.00087
0.0071
P
Ib. sec.
ft.3
0.0070
0.0070
M
ft./(ib./ft.2)
0.000032
0.000053
dAp/df
in. water/sec.
1000
1000
amax
g'a
290
1870
Fore size and dust removal - If one assumes a nominal bulk density of
2 gro/cc for the collected fly ash, an acceleration of 250 g's should be
capable of dislodging an 8 um thick planar deposit from the fabric sur-
face. At the same acceleration level, a hemispherical deposit should be
dislodgable from a hemispherical surface pore provided that the pore
diameter is 48 (am or larger. In both cases, it is assumed that the
2
adhesive strength is roughly 200 dynes/cm as reported by previous in-
g
vestigators and confirmed in the mechanical shaking tests discussed in
Chapter II. Thus, it is expected that a large fraction of the dust
removed during the pulse cleaning process should consist of agglomerates
50 ym in diameter or greater.
During a typical cycle of 1 minute at inlet concentrations of 12 grains/
3
ft. , a dust layer with an average depth of 68 ym should be deposited.
On the premise that surface pores smaller than 48 ym in diameter
are permanently plugged, most dust probably deposits not uniformly but
in select locations centered over previously cleaned locations. There-
fore, in these regions, the deposit depth is probably nearer to 100 um
than the previously cited average value of 68 um. Since, according to
dust trajectory measurements most of the dislodged material fell within
260
-------
the agglomerate size range of 100 to 120 ym average diameter, the
concept of deposition cited above appears acceptable. It has also been
indicated in earlier studies that pore diameters less than lOx greater
than the characteristic particle diameter are readily bridged such that
1 9 12
high collection efficiencies are attainable. ' ' Thus with respect to
nominal 5 urn HMD particles, it appears reasonable that most of the fil-
tration should take place within those pore structures in excess of
50 jim diameter.
Microscopic examination of the filter in its residual state showed that
much of the surface was still loaded with dust. Tunnel-like cavities
of order 100 ym diameter were visible, however, from which dust had
evidently been removed during the previous cleaning. It is probable
that these cavities are responsible not only for most of the dust collec-
tion but also for most of the filter emission. Knowledge of the size
distribution of all cavities above the theoretical pore size (an estim-
ated 48 ym in this case) would permit computation of the residual
pressure differential and improved estimates of dust penetration
properties.
If a certain fraction of the fabric pores remain plugged for a specific
filtration application, one can conclude that these areas are wasted.
If it were possible to maintain a more uniform pore structure it would
be possible to provide more active filtration channels for a selected
gas handling capacity. Thus, although the superficial velocity would
remain unchanged, the provision of more "active" pores will lead to a
reduced interstitial velocity and a corresponding reduction in resis-
tance. It is recognized, however, that the control of pore structure
in fabric systems where the desired pore diameters, 50 to 100 jim, might
not be much larger than nominal fiber sizes 20 to 30 (am, may be difficult.
Reverse drag cleaning - Another possible cause of dust removal, aside
from acceleration, is the drag exerted on the dust deposit when the air
261
-------
flows backwards through the fabric during the pulse. These drag stresses
are indicated schematically in Figure 81. The dust deposit estab-
lishes a resistance to flow in proportion to the permeability of the
deposit, the air velocity, and air viscosity. Integrating the deposit**
resistance from the dirty side of the felt gives the approximate local
stress in the dust deposit tending to detach it from the fiber matrix.
This stress is offset throughout a good part of the felt layer by adhe-
sive forces that depend on fiber, fabric and dust properties. When the
removal stress exceeds the fiber holding strength, usually near the
dirty filter surface, the dust will be removed.
In normal pulse cleaning, a reverse flow accompanies the mechanical
(acceleration) stresses in every pulse. Thus, it is impossible to tell
which mechanism is causing removal without highly sophisticated measure-
ments. To isolate the separate effects of these mechanisms, two brief
series of tests were made. In one series, the slackness of the bag was
varied in order to change the acceleration of the fabric during the pulse
without appreciably altering the pulse itself or the associated backflow.
The cage diameter of the bag was reduced in one test, and in another a
rod was wedged between the cage and the bag to tighten it. The results
as shown below indicated that reducing the acceleration by constraining
the bag ,and without changing the backflow did tend to raise the residual
pressure differential, although not as much as anticipated. Unfortunate-
ly, the small differences and the limited data preclude drawing any firm
conclusions.
Residual resistance
Bag fit in. water
Loose bag 3.53
Normal bag 3.70
Tight bag 3.78
Several tests discussed previously and presented in Table 23 and
Figure 61 indicate that it is the intensity and not the length of
262
-------
(a.) LOCAL RESISTANCE
TO FLOW (PRESSURE
GRADIENT AND
STRESS)
TOTAL
DUST
CLEAN
SIDE
DIRTY
SIDE
FELT THICKNESS
(b.) STRESS BALANCE
FIBER
HOLDING
STRENGTH /
REMOVAL
STRESS
CLEAN
SIDE
DIRTY
SIDE
(C.)RESULT
REMOVED
DUST
REMAINING
DUST
CLEAN
SIDE
DIRTY
SIDE
Figure 81. Postulated removal mechanism for reverse air drag through
a loaded felt
263
-------
the cleaning pulse that determines the amount of fabric cleaning. This
conclusion is based on the fact that no resistance increase was observed
when pulse durations were extended from 0.06 to 0.15 sec. and as much as
4 sec. in the case of talc, Table 23. The inspections of pressure/
time traces describing the transient pressures over the duration of £he
pulse show that a pressure spike appears within the first 50 to 100
milliseconds followed by a rapid decay within the next 50 milliseconds
to a nearly constant pressure differential across the bag for the re-
mainder of the solenoid valve open time. The initial rapid decay in
pressure is associated with the removal of the dust from the filter and
the consequent decrease in filter resistance. The relative stability of
the pressure differential during the remainder of the pulse interval
suggests at best only minimal dust removal by the transient reverse flow
phase when it is preceeded by a high energy pulse.
It is pointed out in Chapter IV, however, that introduction of a low
pressure, reverse flow air at cloth velocities of the order of 40 to
60 ft./min. led to significant dust separation. The latter tests in-
dicated, however, that the initial bowing or bonding of the bag by
pressure differentials of the order of 15 inches was responsible for the
dust dislodgement. This conclusion is based on the fact that dust
removal determined over a threefold range of pressures were essentially
the same. As stated previously it is believed that the primary role of
the reverse flow air is to transport the dust dislodged by the pressure
pulse away from the filter so that the larger agglomerates have an op-
portunity to fall towards the dust hopper.
Although fabric resistance to air flow does not appear to depend upon
pulse duration, the particulate emissions show an apparent dependence-
i.e., a 60 percent increase in effluent concentration when pulse dura.-
tion was increased from 0.06 to 0.15 seconds. In the latter instance
it appears that the extended pulse duration permits more time for
settlement of the coarser particulate agglomerates. Accordingly, with
264
-------
resumption of air flow a somewhat finer size spectrum with the poten-
tial for increased penetration appears at the dirty face of the filter.
Transport of agglomerates - The reverse flow of air through the filter
may affect the trajectory and ultimate disposition of the agglomerates
that are loosened from the filter by other means. The velocity and net
displacement of the air reversing through the filter were estimated in
various ways:
(a) Representative reverse pressure differentials of
15 inches water, and apparent residual filter
resistances of 0.35 in.water/ft./min. indicate
average reverse flow velocities of about 40 ft./
min. Over a pulse duration of 0.06 seconds, this
would produce a reverse displacement of air of less
than half an inch through the filter. The true
residual filter resistance may be much lower,
however.
(b) Rough measurements of the reverse flow produced
during the pulse showed that the aerosol was
being forced backwards a typical distance of
2 feet in the ducting. Allowing for compression
of the air inside the housing, this indicated a
reverse displacement of air of about 1 inch
through the filter during a typical pulse.
(c) Cinephotographic observations of the velocity of
the dust cloud (Table 26) indicated particle
velocities of about 380 ft./min. at the fabric
surface.
From the above approaches, it appears that the air back flows through
the filter a distance of the order of 1 inch during the pulse, with a
maximum velocity of about 100 ft./min. With a single enclosed bag,
as in the present test, the result is that the reverse air velocity
in the vicinity of the bottom of the bag may increase to 1200 ft./min.
The possible consequences of this air movement are considered in Ap-
pendix K. Because the air movement reverses at the end of the pulse,
assuming approximately the same flowlines as when it advances, the net
265
-------
transport is essentially zero. However, the dust agglomerates may be
projected across flowlines, resulting in an effective cleaning will
occur under the following conditions.
• The flowlines are steep; i.e., nearly parallel to the
bag. Use of baffles between bags that are not cleaned
simultaneously and close bag packing should favor the
above.
• The recovery of forward (normal) flow immediately fol-
lowing the pulse is gradual, resulting in a relatively
long period in which the particles are airborne; e.g.,
damped pulses.
• The initial particle ejection velocity is high compared
to the velocity of the air initially flowing backwards
through the fabric.
• Short bags are used.
• Particles are dislodged from the fabric in the form of
large, dense, agglomerates.
In Appendix K, a transport effectiveness parameter is developed based
on the above concepts. This expression, presented below, is intended
only for qualitative guidance until more definitive measurements on
many dust/fabric combinations can be performed
V,
T
where VT = terminal falling velocity (Stokes velocity)
L = bag length
T = length of pulse (time)
VQ « initial particle ejection velocity
V = average pulse velocity
d = one-half the bag separation distance
g = gravity constant
266
(3-4)
-------
V = average air velocity during resumption of flow
K.
V = terminal (free-fall) velocity
The importance of having a large terminal velocity is evident. In the
event that the V /V ratio is not very large, the pulse duration should
1 K
be increased. This probably explains why extending pulse duration had
little effect on fly ash filtration but was effective with talc; i.e.,
the talc agglomerates were lighter and fell more slowly such that
settlement was enhanced by longer pulses.
Thetterminal falling velocity of agglomerated dust cannot always be given
by simple Stokesian mechanics. If the agglomerates fall as a dense
cloud rather than as isolated particles, the falling velocity may exceed
13
the Stokesian velocity. For example, dust loosened from the plates of
an electrostatic precipitator often falls rapidly in the form of a
single sheet of dust. Therefore, pulse cleaning a fabric in such a way
as to remove the dust in a thin, dense layer should enhance dust trans-
fer to the dust hopper.
In the following paragraphs, the controlling factors affecting residual
filter resistance have been summarized.
• The residual filter resistance is determined by the
amount of dust retained on the filter. Although a
good quantitative correlation might be obtained, the
difficulty in performing these measurements in any
simple manner precludes practical use of this
relationship.
• The residual pressure differential and the residual
deposit weight are determined by the acceleration
seen by the filter rather than by reverse (aero-
dynamic) drag.
• Filter acceleration is determined by the rise rate
of the pressure pulse, and by flexibility, tension,
and other properties of the fabric. Unfortunately,
the amount of acceleration obtained may be un-
predictable, depending on how the fabric properties
change with the residual dust.
267
-------
• Fabric acceleration Levels are approximately 25 to
50 times those used in shake cleaning. In contrast
to mechanical shaking, in which low, ~ 4 to 8 "g"
acceleration levels are imparted to a filter some
200 times, only one high energy "shake" is given
the bag cleaned by pulse jet air.
• Dust is removed as agglomerates having diameters
of the order of 100 ^m. Agglomerate size (and
density) are important factors in determining how
much of the dust falls to the hopper without being
redepoaited on the bag.
• In the absence of adequate falling velocity of the
detached agglomerates, a longer pulse duration is
necessary for effective cleaning. Other parameters;
i.e., bag geometry, agglomerate ejection velocity,
and filtering velocity are involved in this
relationship.
• Pores in the surface of the felt with diameters less
than approximately 48 um appear to be permanently
plugged for fly ash filtration. Because plugging
reduces effective use of the fabric media with res-
pect to filtration resistance, a better control of
pore size is recommended.
• Most of the dust penetrating the filter escapes through
the larger pores which are emptied and filled in
almost every cycle of the filter.
Transition Pressure - Figure 70 shows that, beginning at a low residual
pressure differential, the filter differential «t first rapidly, In-
creases as the filter loads. At the end of the loading cycle, the dif-
ferential has reached the terminal pressure. Evidently the terminal
pressure depends on the residual level and on the amount and type of
dust added during filtration.
It will be recalled from Chapter II that the rate «f -presaure -differen-
tial increase given by
K C, V2 - IT g (3.5)
268
-------
is a constant for a fixed specific filter resistance, K, a constant
filtering velocity V and constant inlet concentration C. . The main dif-
ference in pulse cleaning is that a constant rate of increase is not
reached until, about 40 percent of a 1-minute filtration cycle is com-
pleted (or about 20 percent of a. 2 -minute, cycle). However, the resis-
tance curves steadily approach linearity with less frequent cleaning
such that they resemble those for mechanically shaken systems. Extrap-
olation of the linear portion of these curves to time zero permits
estimation of the effective residual resistance.
The difference between the final and effective residual resistance val-
ues, which may be defined as the transitional pressure, is expressed by
the following relation
= KVW = KC.V2t (3.6)
in which t is the filtration time. As pointed out previously, Equa-
tion (3.6) permitted convenient estimation of "K" values for the various
dust/fabric combinations studied. Despite the deviations from linearity
in resistance/time curves, use of the midpoint resistance value (that
at 0.5 minutes for a 1-minute filtering cycle) actually provides a very
good estimate of the average resistance with respect to determining fan
power requirements. With few exceptions, selection of the midpoint
pressure leads to a slightly conservative estimate of power requirements
no more than 5 percent in excess of the time value.
Operating Resistance - The factors making up the filter operating
resistance, the residual and transition pressures, respectively, have
been discussed separately to permit a better understanding of the
cleaning function and dust/fabric properties. From an operational view-
point it is their summation, as depicted by a characteristic average
value of the filtration cycle, that is essential to determining the con-
tribution to total air handling costs chargeable to filter resistance.
With respect to the dust and fabric combinations tested in this program,
269
-------
average resistances were in the nominal 4 to 6 inch water range for
pulse jet filtration at 8.5 ft./min. as compared to 1.5 to 4 inch water
for mechanically shaken systems filtering at 3 ft./min. Based upon
filter drag criterion, the average values of 0.6 and 0.9 inch water/ft./
min., respectively, for pulse jet and mechanical shake systems indicate
reduced air handling power for the former system.
Factors in the Selection of Optimum Operating Conditions
It has been stated previously in the analysis of mechanical shaking
systems that, before any optimization of a filtration process can be
undertaken, one must decide what constitutes an acceptable outlet con-
centration (or mass emission rate). Practically speaking, the target
emission levels will ordinarily be dictated by Federal or State agen-
cies and should include some built-in safety factor.
With respect to acceptable operating resistances, two basic constraints
will probably apply. First, maximum resistance losses across the filter
media should not exceed 10 inch water in any case. Secondly, unless
special blowers of the turbo-compressor type are to be used for gas
moving, the practical limit in operating static pressures will be in
the range of 10 to 12 inch water. With some allowance made for resis-
tance losses in hoods and ducting, it will probably be preferable to
think in terms of 4 to 8 inch water for the typical filter resistance
in a pulse Jet system.
According to the results of this study, effluent concentrations vary
approximately as the third power of the reservoir pressure, with all
other filtration and cleaning parameters held constant. Conversely*
the residual filter resistance varies inversely as the absolute re««'~
volr pressure. When the pulse duration la increased, no observable
reduction in filtration resistance is obtained. Therefore, except tor
the unusual caae, it appears that it is advantageous to minimize pul*«
duration so as to minimize compressed air requirementa.
270
-------
On the other hand, if the pulse intervals are extended, for example,
from 1 to 2 minutes, the average operating filter resistance is increased
by the transitional resistance of the 1-tninute pulse frequency system.
Therefore, there appears to be an opportunity here to reduce the com-
pressed air requirement by a factor of 2 while increasing air handling
power needs by roughly 30 percent. If one is given an average allowable
outlet concentration, reference to curves such as shown in Figure 62
indicate that low pressure direct or high pressure damped pulses will
meet the effluent criterion. In almost every case, however, the added
power requirement associated with the higher reservoir pressure used
with a damped pulse will be much greater than any reduction attained
in resistance to air flow.
It may be possible, however, to decrease the reservoir pressure with
either direct or damped pulses so that the compressed air power is sig-
nificantly reduced provided that the pressures are not less than 40 psig.
To afford any advantage it is necessary that the increase in fan power
due to higher filter resistance be more than offset by the decrease in
compressed air demand. If the added filtration resistance does not
tax the fan capacity, the above optimization path offers the added ad-
vantage of yielding lower effluent concentrations. Usually, the cor-
responding increase in fan power is less than that saved by reducing
the compressed air volumes.
According to limited measurements, a 12 fold decrease in inlet loading
had no effect on average filtration resistance suggesting that large
variations in inlet dust loadings might be accommodated. In contrast
to mechanically shaken systems, however, effluent concentrations cor-
relate strongly with inlet concentrations. Thus, one must consider
the fact that a large gain in dust handling capacity without significant
cost penalties may be accompanied by higher effluent concentrations.
271
-------
CONCLUSIONS TO PULSE JET CLEANING STUDIES
The conclusions that can be drawn from the pulse jet cleaning tests have
been presented in essentially the same format as that used for the me-
chenical shaking studies. General conclusions are given first followed by
more detailed conclusions on (1) particulate emission characteristics,
(2) dust removal and filter resistance; and, (3) system design and operating
factors. Insofar as possible, the behavior of experimental pulse jet sys-
tems has also been investigated with respect to system effectiveness in the
collection of 1 urn particles or smaller. Consideration has also been di-
rected to those factors affecting cost optimization for system operation
once an emission criterion has been established.
General Conclusions
To avoid misinterpretation and unwarranted extrapolation of test re-
sults we present the following conclusions.
• Limits of data application - Unless clearly specified,
test results should not be extrapolated beyond the
bounds of the specific dust/fabric combinations stud-
ied and the actual filtration and cleaning parameters
used. This caution is more important for the pulse jet
studies than for the mechanical shaking studies because
of the many probing tests performed in which the measur-
ing techniques were deliberately of an approximate nature
and where time did not always permit experiment replication.
• Need for further research - Aside from confirming the
earlier observations that a dust/fabric interaction is
often unique, it appears that the role of many variables
identified and/or investigated that relate to collector
geometry and cleaning system parameters require further
study.
• Outlet versus inlet concentrations - It is necessary to
make a very important distinction between the behavior
of mechanically shaken and pulse jet cleaned fabrics.
Whereas only minimal changes in effluent concentration
and size properties are found in the former case for a
huge range, 105, in inlet concentration, the emissions
from pulse jet systems are strongly dependent upon the
inlet concentration. Limited data suggest a linear fch
relationship between effluent concentration and the n
272
-------
root of the inlet loading where n may vary from 0.5 to 1.
Roughly speaking, this states that it is the collection
efficiency and not the effluent concentration of pulse
jet systems that is more apt to be nearly constant.
Effluent particles from pulse jet systems contain many
large particles such that there may not appear to be any
significant size reduction during filtration even though
the collection efficiencies are in the 99.9 to 99.99 per-
cent range. The relative coarseness of effluent particles
is attributed to the large fraction of agglomerates dis-
lodged by the higher energy cleaning pulses. Upon re-
sumption of filtration, re-entrained agglomerates from the
fabric interstices as well as the transient penetration of
those still in suspension on the dirty air side of the
filter constitute a major part of the dust loading. It
might be argued that the coarse nature of the pulse jet
effluent may represent an advantage with respect to
restricting its atmospheric transport.
It is again emphasized that the concept of fractional par-
ticle size efficiency has no practical significance with
respect to the pulse jet system. To begin with, the size
and number concentrations change radically over a filtra-
tion cycle so that, at best, the only practical measure-
ment ia one depicting some long term average value.
Secondly, the presence of agglomerates in the effluent
(generated by the cleaning process) can lead to fictitious
relationships between particle size and fabric collection
efficiency.
Particulate Emissions
The conclusions presented in the following text describe the impact of
system operating parameters, cleaning parameters and fabric type upon
fly ash and talc dust emission.
• Fly ash and talc aerosols having roughly the same size
properties can be filtered at efficiencies in the 99.9
to 99.99 percent range. Compared to the behavior of
mechanically shaken cotton bags the average outlet con-
centrations were at least 2 to 3 orders of magnitude
higher for pulse jet systems.
• The average outlet concentrations for the Dacron and
wool felts, which were in the range of 0.001 to 0.0001
grains/ft.3 for both fly ash and talc, increased with
increasing inlet dust loading as cited previously* Al-
though the data were insufficient to establish a
273
-------
precise quantitative relationship, the effluent loading
appeared to depend on the inlet loading raised to some
power between 0.5 and 1.0.
The average outlet concentration showed a strong response,
five fold decrease, to a 25 percent reduction in filtra-
tion velocity (8.5 to 6.2 ft./min.). Test data were too
limited to justify any quantifying expression for the
above results. It is believed, however, that the reduced
reentrainment of agglomerates at the resumption of
filtration may partially account for the lower loadings.
The average outlet concentration varied directly with
pulse intensity showing an approximate five-fold increase
over the reservoir (pulse) pressure range of 40 to 100 psig.
The average outlet concentration was affected by the wave-
form (pressure-time trace) as well as the intensity of the
pressure pulse. By using a pulse damping technique in
which the rate of pressure release from the bag was re-
tarded, outlet concentrations were reduced approximately
five-fold, irrespective of pressure intensity.
The average outlet concentration was moderately dependent
upon the duration of a pressure pulse increasing roughly
as the square root of the pulse interval. One concludes
that there is little merit to extending the pulse duration
because of the adverse effects of increased particle dis-
charge and increased compressed air usage. At the same
time, no significant improvement in filter cleaning (and
reduction in operating resistance) is attained.
The average outlet concentration did not vary appreciably
with the pulse frequency over the time intervals 0.4 to
4 min. These results indicate that the major dust emis-
sion occurs immediately following cessation of the pressure
pulse.
The apparent fractional particle size penetration appeared
to be essentially independent of size for particles less
than 5 |im in diameter for reservoir pressures up to 70 psig
when direct pulses were used.
The apparent fractional size efficiencies for particles 2
to 10 ym in diameter actually varied inversely with diameter
with high energy, 100 psig pulses. This phenomenon was at-
tributed to the large population of agglomerates dislodged
from and penetrating the filter during the initial phase of
filtration.
The apparent fractional size efficiencies for damped pulse
systems indicated the expected; i.e., a direct relationship
between efficiency and particle diameter for the size
range 1 to 10 ym. It is believed that the more gradual
274
-------
bag deflation process reduces the "snap" or projection
capacity of the bag when returned to service as well as
eliminating the transient increased flow velocity. Both
factors, the former by mechanical dislodgment and the
latter by increased re-entraining power, favor increased
emissions.
Fractional particle size measurements for fly ash/wool
systems under typical pulse cleaning conditions were
essentially the same as those for the fly ash/Dacron
combinations.
No significant differences could be detected between the
behavior of talc and fly ash with respect to fractional
particle size measurements.
Outlet particle concentrations (number basis) decreased
rapidly during the filtration interval such that cal-
culated mass concentrations varied by as much as 100 times,
Although a similar pattern was established during mechan-
ical shaking tests, the differences in the former case
were often as much as 105 times. These results point out
that the pulse jet filtration process has more deep bed
collection characteristics compared to the "cake" type
filtration found in many mechanical shaking systems.
A relationships between effluent concentrations and
average filter resistance was established. Similarly,
the residual filter resistance was directly related,
as expected, to the residual dust deposit.
Dust Removal and Filter Resistance
The following conclusions concern those measurements that relate to the
key factors responsible for dust removal.
• The residual dust holding for a filter was inversely related
to the intensity of the pressure pulse as determined by the
reservoir pressure.
• The residual dust holding did not appear to vary significantly
for direct and damped pulses and with the duration of the
pulse.
• The residual dust holdings were approximately the same for
the fly ash/Dacron and fly ash/wool systems. This was not
surprising in that both felts had very similar properties
and the collection efficiencies were about the same.
• The residual dust holdings for the talc/Dacron systems were
about half those for the fly ash/Dacron (or wool) systems.
275
-------
The difference Is presently attributed to the much lower
bulk densities observed for talc.
The residual dust holding and residual resistance depend
upon the acceleration imparted to the fabric. This suggests
that the principal mechanism of dust removal is that of
simple tensile forces exceeding those of adhesion and cohesion
in essentially the same manner as predicted for mechanical
shake cleaning.
It is estimated that for many bags cleaned by pulse jet air
the maximum fabric acceleration is directly proportional to
the rate of rise of the pressure differential across the
bag, M.OOO to 4000 in. water/sec.
The residual resistance varies inversely with the rate of
change of bag differential pressure and also with reservoir
pressure. Thus, the rate of rise in bag pressure differential
can be equated to reservoir pressure.
Residual resistance values depend not only upon the rate of
change in pressure differential across the filter but also
upon the dust/fabric combination and the inlet loading to the
filter. More tests are required to establish quantitative
relationships that include the above factors.
The data presented in this report represent preliminary and, in many
cases, exploratory measurements. Therefore, despite the many qualita-
tive insights gained, we do not think the results should be extra-
polated for design purposes unless the dust and fabric properties con-
form very closely to those tested in this study. On the other hand,
the many relationships developed in this program, coupled with reliable
field data from industrial sources, should provide useful guidelines
for improvements in existing systems and for more extensive investiga-
tions of the variable influencing pulse jet system performance.
System Design and Operating Factors
The following conclusions relate to the probable impact of design and/or
operating parameters.
• The location of a pulse jet nozzle may exert significant
effects on system effluent and resistance properties by
aiding or detracting from the uniformity of bag cleaning.
Limited tests showed that secondary air flow induced by
276
-------
the pulse jet should be entrained from the plenum region
so that absolute pressure within the bag (clean air side)
always exceeds that on the dirty side.
Based upon observed effluent dust profiles, tests suggest
that long filter bags may see inadequate cleaning in the
lower region. The practical design limit should be
established by controlled laboratory study because in-
creased bag length can be advantageous from the perspective
of equipment space requirements.
Bag material should be flexible, light weight, and inelastic
to obtain maximum acceleration. It should have sufficient
weight (i.e., number of fibers per unit area) to present
many targets for entering dust particles and the pore struc-
ture should be as uniform as possible.
A large housing and hopper volume on the dirty side of the
filter bag will minimize the pressure build-up in this
region during the pulse, and thus enlarge the magnitude of
the pulse differential. Likewise, any relief of the build-
up pressure in this region during the pulse, including
temporary stoppage of the incoming dirty air, will result
in a larger .pulse differential for the same compressed air
energy expenditure. The larger pulse differential should
be associated with a larger rise rate, and acceleration.
The pulse delivered to the bag should begin as abruptly as
possible with sufficient inflating flow to subject the
entire bag length to a sudden pressure differential. It
might be better if several concurrent pulses could be
released at points long the bag, each with sufficient sharp-
ness and volume to give the felt a maximum velocity per unit
of pulse energy used.
The back flow of air through the filter accompanying the
pulse assists cleaning in several ways. It flushes ag-
glomerates loosened by the acceleration from the pore struc-
ture. It can itself loosen agglomerates if the shock itself
has been insufficient, although this appears to be"*a very
inefficient use of compressed air. It also accelerates
agglomerates that have already left the felt surface helping
to convey them to the hopper.
Pulse intensity should be as low as can be tolerated to
save on compressed air (and reduce power needs) but suf-
ficiently high to maintain an equilibrium cleaning process.
Pulse duration should ordinarily be as short as possible.
This will probably be limited by the care taken in designing
the pulse generating and delivery system, particularly in
the pressure release valving. In some operations, slightly
longer pulses may be necessary to (1) inject sufficient
air volume into the bag to enable the rise to develop
277
-------
maximum acceleration along the entire bag; or, (2) to
prevent redeposition and/or to promote transport of the
agglomerates to the hopper.
Optimum cleaning frequency is determined by the cost of fan
power on one hand, and the cost of compressed air or the
average emission requirement on the other hand.
REFERENCES
1. Billings, C. E. and J. E. Wilder. Handbook of Fabric Filter
Technology. Volume 1: Fabric Filter Systems Study. GCA/Technology
Division. Department A, Clearinghouse, U.S. Department of Commerce,
Springfield, Va. 22151. Report Number GCA-TR-70-17-G, APTD-0690,
Contract No. CPA-22-69-38, PB-200-648, December 1970.
2. Pulverizing Machinery Company, Summit, New Jersey. No Moving Parts
Inside Hot Dust Collector. Chem. Eng., 64:188, August 1957.
3. Dennis, R. and L. Silverman. Fabric Filter Cleaning by Intermittent
Reverse Air Pulse. ASHRAE J., 43:76, 1962.
4. Phillips, N. D. and R. J. Wright. A New Technique to Achieve Surface
Renewal in Fabric Filtration. Presented at the 1969 Meeting, Air
Pollution Control Association, Paper No. 69-201.
5. Berg, D. B. Dust Filter Reclaims 10,000 Pounds Per Hour of Hot
Asphalt Plan Aggregate. Chemical Processing, 77, April 1968.
6. Spaite, P. and G. Walsh. Effect of Fabric Structure on Filter Per-
formance. AIHA J., 24:357, July 1963.
7. Borgwardt, R. and J. Durham. Factors Affecting the Performance of
Fabric Filters. 60th Annual Meeting AIHE, Paper No. 296, November
1967.
•
8. Herrick, R. A. Theory and Application of Filter Drag to Baghouse
Evaluation. Air Engineering, 18, May 1968.
9. Draemal,.D. Relationship Between Fabric Structure and Filtration
Performance in Dust Filtration. Control Systems Laboratory, U.S.
Environmental Protection Agency, Research Triangle Park, North
Carolina, Report No. EPA-R2-73-288, July 1973.
10. Stephan, D., G. Walsh and R. Herrick. Concepts in Fabric Air
Filtration, AIHA J., 21:1, 1960.
278
-------
11. Williams, C. E., T. Hatch and L. Greenburg. Determination of Cloth
Area for Industrial Air Filters. Heat Pipe and Air Cond., 12:259,
1940.
12. Winchester, S. C. and J. C. Whitwell, Multivariable Studies of Non-
woven Fabrics. J. Eng, for Ind., 89:1, 1967.
13. Fuchs, N. A. The Mechanics of Aerosols. The MacMillian Company,
New York (1964).
279
-------
CHAPTER IV
REVERSE FLOW STUDIES
OBJECTIVES AND APPROACHES
Based upon data provided in the Handbook of Fabric Filter Technology,
it is estimated that about 57 percent of existing commercial collector
models are cleaned either by mechanical shaking (35 percent) or high
pressure pulse jet air (22 percent). Another important approach, re-
verse flow cleaning, is used in about 25 percent of existing collector
types based upon the statistics cited previously. The latter category
does not include those units cleaned by high pressure pulse jet air,
as discussed in Chapter III, and those cleaned by continuous application
of a high velocity air jet (traveling blow ring).
Data for low pressure-reverse air flow cleaning systems are quite limit-
ed with respect to their dust removal capabilities. In many cases, it
appears that the reverse air flow functions mainly as an adjunct to the
primary cleaning process; e.g., mechanical shaking as described pre-
viously or bag collapse as applied to highly tensioned, heavily laden
glass bags. It should also be noted that as pressure levels increase
and the rates of pressure and flow change become more rapid, it becomes
progressively harder to differentiate between reverse flow and pulse
jet cleaning processes.
Because the reverse flow concept is used in some 25 percent of available
commercial equipment, and also because some operating features of re-
verse flow systems appear similar to those of the high pressure pulse
281
-------
jet units, several exploratory measurements were performed in this area.
In contrast to the more detailed studies of the better defined shaking
and pulse jet cleaning systems, most of the experiments described in
this section were simplified. The principal test objectives were as
follows:
• Identify key variables that influence collection efficiency
and operating resistance and postulate their relationships
at least on a qualitative basis.
• Determine the apparent practical boundary conditions for
the variables determining system performance.
• Develop a framework for the more effective evaluation
and utilization of the reverse flow concept.
BACKGROUND
Postulated Cleaning Actions
Despite the different techniques used to create the dust separating
forces for mechanically shaken and pulse jet systems, the data pre-
sented in Chapters II and III, respectively, suggest that tensile forces
generated by rapid acceleration of the dust/fabric layer constitute the
primary cleaning mechanism. When tensile forces exceed the adhesive
and/or cohesive forces, dust is detached mostly in the form of large
agglomerates at those locations where the binding forces are weakest or
the acceleration forces the greatest. Visual observations indicate that
the separation frequently occurs at the interface between the dust par-
ticles and the main fiber-structure.
The data presented in Chapter III indicated that a reverse air flow fol-
lowing pulse application serves to flush out loosened particles from
fabric interstices. More important, however, is the transient arrest-
ing of system flow so that dislodged agglomerates can settle towards the
collecting hopper. Based upon gravimetric measurements and filter resis-
tance characteristics, there is little evidence to suggest any
282
-------
significant removal of dust particles by aerodynamic action alone. The
2
above findings are in agreement with those of Larson who states that air
velocities of the order of 200 ft./min. are required to remove a single
3
20 ym particle from a fiber and Zimon who indicates that air velocities
sweeping over a layer of dust (tangentially) must be in the 400 ft./min.
range before any appreciable dust removal is attained. The latter ob-
servation explains why the reverse jet (traveling blow ring) system with
constant slot velocities of 4000 to 6000 ft./min. constitutes an effect-
ive fabric cleaning method for felted media.
Since significant dust removal is attained in some reverse flow systems
one must conclude that separating forces other than aerodynamic drag
are involved. According to the drag approach, the separating force for
a single particle produced by viscous drag should be essentially the
same at any point within the filter if one assumes a constant intersti-
tial velocity. At the same time, however, the pressure gradient pro-
duced by the reverse flow causes the filter to compress. Therefore, the
normal adhesive forces between adjacent particles are actually increased
as one moves radially from the dirty to the clean air side of the fil-
ter. The net effect is that any dust dislodgement is more likely to
follow a simple spallation process with the adhesive bond failure occur-
ing close to the dirty face of the filter. It should be noted that the
above process is the opposite of that for cleaning by shaking or high
energy pulse in which the dust/fabric experiences tension rather than
compression when it is decelerated. From the practical viewpoint,
reverse flow alone does not appear conducive to high dust removal unless
air velocities are much higher than those used in typical reverse flow
systems.
It should be noted, however, that a high reverse pressure gradient across
the filter may also increase the lateral or hoop tension in the dust/
fabric system so that many of the adhesive and cohesive bonds are dis-
turbed. Given this situation, aerodynamic drag could be expected to
flush out more loosened particles. Another adjunct to dust separation
283
-------
is the flexure produced in the fabric when the flow is reversed. In
many systems, sufficient bending or warping of the fabric surface occurs
to cause a significant spallation at the dust/fabric interface. The
effect is most pronounced for woven fabrics in which a larger fraction
of the dust can appear as a superficial layer.
Basic Design Concepts
A comprehensive review of reverse flow cleaning processes has been pre-
sented by Billings and Wilder. In contrast to cleaning methods dis-
cussed previously, it is more difficult to characterize rigorously many
of the reverse flow cleaning systems. Generally, the reverse flow is
produced by a gradual venting of low pressure air from a secondary fan
rather than a compressed air source. Thus, there is usually no rapid
rise in pressure, ~ 2000 in. water/sec., such as seen with a high pres-
sure pulse jet system.
If the dust releases readily from the fabric, a reversal of flow in it-
self may suffice for adequate cleaning. In many cases, however, the
reverse flow is used in combination with shaking, pulsing or bag col-
lapse to facilitate dust removal. Because of its structural depth,
and hence greater dust retentivity, a felted fabric is not usually
cleaned by reverse flow.
When reverse flow is used as the sole method of cleaning, bag attrition
is low provided that proper support structures are used to prevent ex-
cessive bag flexure. It is emphasized, however, that the rate of flex-
ure is probably the controlling factor with respect to fabric (or fiber)
failures. Thus, a gradual inflation or deflation process is unlikely
to cause any serious fabric damage. Increased bag tension and reduced
reverse flow rates also minimize the degree of flexure as well as pre-
venting a complete flattening of the bag. Under the latter circum-
stances, there is no opportunity for loosened dust to fall to the hopper
284
-------
nor is there a pathway for the reverse flow air. The insertion of re-
straining rings or a supporting cage eliminates complete bag collapse
but introduces a potential problem of fabric chafing. Sewing the rings
to the bag minimizes the chafing or attrition problem.
APPARATUS, MATERIALS AND TECHNIQUES
Experimentation with reverse flow cleaning was carried out with the
shaking bag and pulse jet systems described in Chapters II and III after
modifying the above units as shown schematically in Figures 82 and 83.
Modified Mechanical Shaking System
Figure 82 shows the arrangement by which a reverse air flow was pro-
duced with the mechanically shaken equipment. Since dusty air was de-
livered to the filter bag under positive pressure, flow reversal was
accomplished by exhausting air from the dirty air side of the filter.
As stated earlier in Chapter II, positive pressure operation facilitated
the installation and use of special testing instrumentation.
During these experiments, Valve A was closed when the main fan was
stopped. Concurrently, Valve B was opened and the reverse flow ex-
hauster operated when the shaking cycle was in progress. In most cases,
the Valve B solenoid drive was connected to the electrical circuit for
automatic regulation of the shaking cycle.
Tests were performed with and without the installation of a flexible
coil spring, diameter ~ 3 in., inside the bag to prevent its complete
collapse when the flow was reversed. The static pressures cited during
the actual reverse flow periods, which were measured by conventional
manometers (liquid), reflect the negative pressure within the hopper (or
dirty air chamber).
285
-------
PARTIALLY
ENCLOSED
BAGHOUSE
FILTER
INLET
L<— SHAKER
• —
REVERSE FLOW
EXHAUSTER
COTTON
SATEEN
OR
PLAIN
WEAVE
DACRON
BAG
DUST
\ / HOPPER
FILTRATION MODE
VALVE A OPEN
VALVE B CLOSED
CLEANING MODE
VALVE A CLOSED
VALVE B OPEN
Figure 82. Schematic of mechanical shaking system as modified for
reverse flow cleaning
286
-------
NORMAL
COMPRESSED AIR
INLET
(BLOCKED)
REVERSE FLOW
AIR INLET
CONICAL
REDUCER
EXIT
PLENUM
BAG
FAST ACTION
BUTTERFLY
VALVE
(OPEN POSITION)
FILTER
HOUSE
TOP VIEW
SIDE VIEW
APPROXIMATE SCALE
lV2" =T
Figure 83. Schematic view, pulse jet assembly with low pressure
reverse air and high pressure nozzle removed. No bag
Venturi.
287
-------
Modified Pulse Jet System
The evaluation of the effect of reverse flow cleaning for the wool and
Dacron felts normally cleaned by pulsing was conducted using the basic
equipment described in Chapter III. The system was modified as shown
in Figure 83 by the addition of the same 600 ft. /min. blower used with
the shaken bag unit. The ducting was arranged with the fan reversed to
introduce a flow of air into the outlet plenum above the bag whenever
a fast-acting damper closed. The damper was connected with the pulse
solenoid, with a controlled time lag between them when desired. Alter-
natively the damper was used alone instead of the pulse solenoid, and
controlled by the regular pulse timing system.
Because a commercial damper with sufficient size and response time could
not be located, one was constructed on the butterfly valve principle.
The shaft was turned by a powerful solenoid acting against a return
spring. The rate of opening of the damper was regulated by attaching
weights to the shaft, enabling a study of the effect of the rate of
change of reverse flow.
When the damper closed, the primary system flow was temporarily blocked
and, with sufficient pressure from the cleaning fan, the flow was re-
versed through the filter. Although the schematic arrangement of
Figure 83 shows the pulse jet compressed air source as blocked off,
several measurements were also made where the high pressure pulse was
admitted directly to the plenum (plenum pulse as described in Chapter
III) or by the conventional 1/4-inch nozzle with the Venturi element
located in the bag exit.
The magnitude of the reverse pressure or reverse flow used during clean-
ing was regulated by damper control of the reverse flow fan. In most
cases, the reverse flow fan was allowed to run continuously during opera-
tion, to minimize start-up delays. The fact that considerable power was
wasted by this technique was immaterial from the experimental viewpoint.
288
-------
Commercial equipment designed on these principles would, of course,
utilize the reverse flow fan more efficiently because several compart-
ments would be operated in parallel.
Test Fabrics
The filter fabrics used in this phase of the study included the woven
and felted fabrics described in Chapter II, Tables 2, 3 and 17 and
Chapter III, Table 19. The woven bags, well-used unnapped cotton sateen
and plain weave Dacron, are normally cleaned by shaking. The term
"well-used" refers to those bags that had been given about 2 x 10 in-
dividual shakes as part of their accelerated life testing. The wool
and Dacron felt bags studied are normally cleaned by reverse pulse.
The felted media were included, however, even though rarely used for
reverse flow cleaning, to provide some data that is currently lacking
in the filtration literature.
Measurements and Instrumentation
No new control and measurement techniques or concepts were introduced
during the reverse flow tests. Primary system air flows and the record-
ing of filter resistance as well as the timing of cleaning and filtering
cycles were carried out with the same apparatus described in Chapters
II and III. The initiation and regulation of the reverse flow processes
were accomplished by making parallel or delayed time connections to the
basic control circuits. Transient pressure differentials across the
bag during the reverse flow intervals were again detected by paired
Pitran transistors (see Chapter III) and displayed on an oscilloscope
screen. This approach was particularly valuable for defining the pres-
sure wave form and the rate of pressure rise for high and low pressure
pulses.
Most dust concentration estimates for the effluent gas stream were per-
formed with the RDM monitor discussed previously. Less emphasis was
289
-------
placed on effluent characteristics during the reverse flow test» be-
cause the primary objective was to determine whether successful, steady
state operation could be attained prior to any further evaluations.
Test Dusts
Fly ash (5 micrometer MMD), Table 4, was used exclusively as the test
dust in this series of experiments. It was chosen because of the ex-
tensive amount of work done with fly ash in other phases of the program,
because of its moderately good release properties and because reverse
flow cleaning has also been used for industrial filtration of fly ash
from oil and coal combustion sources.
RESULTS
Low Pressure, Reverse Flow with Mechanical Shaking
The use of reverse flow air in conjunction with the mechanically shaken
apparatus shown in Figure 82 represents an augmentation approach in
which the reverse air function is: (a) to flush out interstitially
loosened dust; and (b) to hasten the removal of the finer aerosol frac-
tion from the dirty air side of the bag.
Aside from the reverse flow modification, the basic filtration parameters
for fly ash with unnapped cotton sateen and plain weave Dacron bags
were essentially the same as described in Chapter II; i.e., 3 ft./min.
filtration velocity and an inlet loading of 3.5 grains/ft. The shak-
ing conditions were those indicated in Chapter II as a standard cleaning
cycle; i.e., 8 cps, 1 in. amplitude and 360 shakes. The filtration
period was approximately 20 minutes, as compared to the 30-minute
period used previously.
Although the reverse flow fan had the capacity to exhaust air at 600
3
ft. /min., it was ordinarily throttled so that the typical range of
290
-------
3
reverse flows was 20 to 220 ft. /min. Unless otherwise specified, the
reverse flow fan and the shaker motor were operated simultaneously.
Less than 5 seconds were required for the reverse flow fan to reach
full speed.
According to the data summarized in Table 29, Tests 1 and 6, the base
line (mechanical shaking only) tests were reasonable replications of
prior measurements discussed in Chapter II. The resistance/time data
shown in Figures 84 and 85 indicate that the base K values for fly ash/
cotton and fly ash/Dacron, respectively, are 15 and 12 in. water/ft./
2
min./lb./ft. The first reverse flow augmentation experiments in this
series were conducted with bags that were not supported on the inner
side. This led to varying degrees of collapse depending upon the
vacuum (or differential pressure) across the bags during the reverse
flow interval.
In the absence of internal support, increased reverse flow made it pro-
gressively more difficult to dislodge the dust and condition the filter-
ing surface by the reverse flow. The result was that the K values also
increased gradually, Table 29, Tests 1 to 3 and 6 to 8, and a similar
increase in the amount of dust retained by the filters was also observed.
Upon the installation of an internal supporting structure (highly flex-
ible helical coil) some reduction in filtration resistance was obtained
as shown by Test 4. One is forced to conclude, however, that supple-
menting mechanical shaking with a low level reverse flow (plus support
structure) afforded no advantage. Limited tests with reverse flow only
and including an internal support showed much higher operating resis-
tance and greater K values.
Although the weighings were considered to be rough, it was noted that
the amount of dust dislodged (~ 200 grams) during Tests 1, 2 and 3, as
shown in Table 29, was about equal to that deposited during the filtra-
tion cycle, one criterion for true steady state operation.
291
-------
Table 29. FLY ASH FILTRATION WITH MECHANICAL SHAKING AND/OR LOW PRESSURE REVERSE FLOW CLEANING
Test
nucber
Test conditions3
Filter resistance in. water
Initial
Effective
residual
(estimated
value)
Terminal
Reverse flow
Applied
vacuum
in. water
Estimated
velocity
ft./min.
Dust parameters
Weight
dislodged
grams
Residual
weight
grams
Effluent
concentration
gr./ft.3 x 103
Urmapped cotton sateen
1
2
3
4
5
Shake only*"
Shake plus reverse flow
Shake plus reverse flow
Shake plus reverse flow
with support
Reverse flow only with
support0
1.5
1.6
1.4
1.4
1.3
2.0
2.1
2.1
2.0
2.5
3.4
3.8
4.0
3.4
6.2
-
0.7-0.6
2.2-1.8
1.0
5.2
-
1.2
4.3
2.1
12.0
193
212
203
-
-
372
520
535
-
290
\
Estimated
from
Table 13
2 x 10'2
to
V0
'to
Plain weave Dacron
6
7
8d
9fe
10
Shake only
Shaken plus reverse flow
Reverse flow only
Reverse flow only with
support
Reverse flow only with
support
0.2
0.2
-
3.1
1.0
1.0
0.9
-
3.5
1.4
2.1
2.0
6.7
7.0
4.5
-
0.5-0.25
4.5-4.3
0.75
5.4
-
5.6
-
0.7
16.2
208
225
50
80
•
414
386
1520
1490
720
6.6
11.4
-
14.0
"•
a!0 ft. x 6 in. diameter bags. Loaded to equilibrium «t 3.5 grains/ft.3, 3 ft./min. filter velocity.
Shaking, 3 cps, 1 in. amplitude, 360 shakes.
Reverse flow period, 45 seconds.
No equilibrium attained.
Si ear equilibrium.
-------
_ MECHANICAL SHAKING
R»REVERSE FLOW
I«INTERNAL SUPPORT
1
10
FILTRATION TIME, min.
Figure 84. .Resistance characteristics for fly ash/utmapped cotton
filtration with mechanical shaking and/or reverse flow
(shaking cycle, 8 cps, 1 in. amplitude, 360 shakes)
293
-------
I I I I I I
ESTIMATED K
* IN. WATER/FT./MIN./LB./FT.2
S
S,R
cc
111
I
UJ
o
I
(O
.55
UJ
-------
The fact that the K values increased, Figure 84, suggested that the col-
lapse of the bag led to a nonuniform cleaning process. The latter ef-
fect, which was reported in this study (Chapter II), was also observed
4
by Walsh and Spaite in their mechanical cleaning tests when insuffi-
cient shakes, <150, were used. Although the residual resistance levels
indicate good cleaning, the subsequent pressure rise shows a decreased
holding capacity (or larger K value). Despite the fact that the tests
reported in Table 29 show no benefits for reverse flow augmentation, it
is emphasized that in other applications reverse flow has been used ef-
fectively. The present tests point out that once vigorous mechanical
shaking has taken place there is little to be gained by the secondary
treatment. In many operational systems, however, the compartment isola-
tion valves may not completely stop filtration flow. Thus, during shak-
ing there is a tendency not only for dust redeposition such as encoun-
tered with pulse jet cleaning but also for dust transport to the clean
air side of the system. A gentle reverse flow in the latter case should
prevent this dust penetration.
When the dust deposit is mainly superficial and the bag tension is high
enough to minimize the collapse and internal blockage problems, the
reverse flow approach has merit in low velocity filtration applications.
In the present test series, it was not clearly defined what role the
flexing of the fabric exerted relative to the dust removal by flexing.
Some exploratory measurements are described in the next section that
shed some light on the amount of dust removed by aerodynamic action.
Low Pressure, Reverse Flow With Pulse Jet Equipment
Ordinarily, felt bags and similar filter media are not used in simple
reverse flow processes because pf reportedly poor cleaning, usually
attributed to complex dust/felt interactions. It was believed, however,
that with the experimental equipment on hand it would be advantageous
to carry out more detailed investigations of reverse flow processes to
further the understanding of the physical mechanisms involved. Several
295
-------
combinations of high and low level reverse flow pulses were investigated
to establish possible relationships between shock and aerodynamic effects.
In a series of tests where reverse flow alone was used, the following
factors were investigated with respect to filter cleanability.
• Slow inflation rate
• Rate of pressure rise (bag differential)
• Frequency of cleaning
• Flow duration
Separate test sequences were performed in which reverse pulse and reverse
flow cleaning were combined.
Slow Inflation Rate, Reverse Flow Only - The exploratory tests summarized
in Table 30 were designed to examine the effects of reverse flow
velocity on dust removal from another perspective. As well as could be
determined by the combination of reverse flow and mechanical shaking,
Table 29, the aerodynamic removal of collected dust was probably of
minor importance.
Several degrees of reverse pressuring of the clean air side of the bag
were investigated with the pulse jet system. The test procedure was to
load a Dacron or wool bag with fly ash under the routine filtering
3
conditions used in many tests; i.e., inlet loading 11.6 grains/ft. ,
filter velocity 8.5 ft./min. and a 1-minute filtering interval. At the
end of the filtering period, the main air flow was shut off completely.
Reverse air was then admitted to the exit plenum chamber by the gradual
opening of a valve installed in the compressed air delivery system such
that the average rate of pressure increase on the clean air side of the
bag was about 0.6 in./sec. The estimated acceleration in reverse flow
2
velocity was 0.014 ft./sec. Once the indicated pressure levels were
attained, Column 2, Table 30, the flow was held constant for 2 seconds
and then gradually reduced to zero.
296
-------
Table 30. DUST REMOVAL VERSUS REVERSE FLOW VELOCITY, VOLUME,
PRESSURE AND DURATION. FLY ASH AND WOOL AND
DACRON BAGS
Testa
number
11 W
12 W
13 W
14 W
15 W
16 W
17 W
18 W
22 D
23 D
24 D
19 Wc
20 Wc
21 Wc
Reverse
pressure
differential
in. water
3
4
6
8
3
4
6
7
5
10
14.6
7
7
7
Net dust
removed
grains
4.1
4.6
4.5
5.4
4.5
5.4
6.3
2.3
9.9
10.4
8.4
1.5
0.7
0.3
Reverse air&
Estimated
maximum
velocity
ft./min.
4.3
5.7
8.5
11.4
5.6
5.7
8.5
9.9
7.1
14.2
20.7
9.9
9.9
9.9
Estimated
total volume
ft.3
0.50
0.83
1.70
2.90
0.80
0.83
1.70
2.3
1.23
4.36
9.08
2.3
2.3
2.3
Estimated
cleaning
time
sec.
12
15
22
29
28
15
22
25
19
35
51
25
25
25
^ = wool; D - Dacron.
bRate of change in cloth velocity during valve opening and closing,
0.014 ft./sec.2.
cThree successive cleanings without reloading.
Note: Dust load applied at 11.6 gr./ft.3 and 8.5 ft./min., filter
cleaned first by reverse flow as indicated. Following
measurements, filter cleaned by pulse jet, 70 psig damped,
0.06 sec. duration and 1 min. frequency, cycle then repeated.
297
-------
Approximate values for maximum reverse cloth velocities and total re-
verse air volumes were based upon the terminal resistance value for the
loaded filter, ~ 6 in. water. The above estimating process was used
because only a small fraction, ~20 percent, of the 31 grams of dust
added to the filter during the loading period was removed during each
test. Thus, no large increase in overall permeability was expected.
Additionally, the velocity estimates were also conservative.
It appears very clear from Table 30 that regardless of the pressure
differential, the amount of reverse flow, and the flow velocity as
determined by the pressure, the dust loss is very small and for all
practical purposes a constant quantity. The only common factor that
appears to afford a reasonable explanation for the constant removal
is that the fabric surface was flexed once (dilate and contract) during
each test. At the indicated range of pressures, the curvature assumed
by the bag was probably the same and the change rate of curvature was
dictated by a uniform expansion rate in each test. Thus, the adhesive
and cohesive forces were diminished by the extent of the flexure so
that even a minimal air motion sufficed to detach those particles not
separated by gravity fall.
It had been observed that closing down the main fan in routine mechanical
shaking tests, Chapter II, resulted in a 5 to 10 gram loss of dust as
the woven cotton or Dacron bags collapsed. This loss was inconsequential
relative to the ~300 grams dislodged during the shaking process, about
2.5 percent. The approximate time interval for the bag relaxation was
5 to 10 seconds such that the rate of flex corresponded to that of the
tests summarized in Table 30.
The tests discussed in this section confirm the fact that even a very
gentle bending of a fabric is sufficient to dislodge dust surface layers.
Note that the rate of pressure increase for the Table 30 measurements,
~ 0.6 in./sec., was very small compared to that usually found for con-
ventional pulse jet systems, ~ 2000 in./sec. They also appear to bear
298
-------
2 3
out prior theories on adhesion ' that indicate that air velocities
must be in the 200 to 400 ft./min. range before aerodynamic forces
become significant.
Pressure Rise Rate, Reverse Flow Only - The shape of the pressure pulse
delivered to an equilibrated Dacron felt bag was varied by attaching
weights to the damper used to divert the flow from the positive side of
the high volume blower to the inner or clean side of the filter unit,
Figure 83. This, in effect, retarded the opening and closing of the
damper and reduced the rate of pressure change at both ends of the clean-
ing cycle. The wave forms of the pressure pulses across the bag are
shown in the oscilloscope traces of Fig.ure 86.
Each test shown in Table 31 was conducted at typical pulse jet filtra-
tion conditions; i.e., an inlet fly ash concentration of 11.6 grains/ft. ,
a filtration velocity of 8.5 ft./min. and a 0.5 minute cleaning fre-
quency. As far as the general effect was concerned, the pressure changes
produced in the filter system by the opening of the damper from the
high volume blower were very similar to those resulting from pulse jet
action. The major difference was that the rate of rise of differential
pressure, dAp/dt at its maximum value of 750 in./sec. was 3 to 4 times
less than that produced by a compressed air source. The actual maximum
pressure differentials at steady state were roughly the same as those
found for pulse jet systems, Chapter III.
Key test parameters for the measurements shown graphically in Figure 86
are given in Table 31. As expected, an inverse relationship between
rate of pressure rise and average filter resistance was indicated.
Similarly, a direct relation between average effluent concentration and
299
-------
40 i-
20
20
ui
S
0 0.2 O.4 0.6
A. dAP/dt =750 in./sec.
0.8
40 r
20
0.2 O.4 0.6
RELATIVE TIME SCALE, SEC.
B. dAP/dt = 650 in./sec.
40 r
20
0.8
0.2 0.4 0.6
C. dAP/dt = 335 in./sec.
0.2 0.4 0.6
RELATIVE TIME SCALE, SEC.
D. dAP/dt = 220 in./sec.
0.8
I I I I
0.8
Figure 86. Controlled variations in rate of differential pressure change across felt bags (see Table 31)
-------
OJ
o
Table 31. EFFECT OF VARYING RATE OF PRESSURE CHANGE ON FILTER RESISTANCE AND EFFLUENT
CONCENTRATION. FLY ASH/DACRON SYSTEM, 11.6 GRAINS/FT.3 INLET LOADING,
0.5 MINUTE CLEANING INTERVALS.
Test
number
25
26
27
28
Pressure
change
in. /sec.
Rise
750
650
335
220
Fall
520
545
140
140
Bag resistance in. water
Residual
7.5
7. A
9.5
10.1
Average
7.8
7.6
10.6
10.9
Terminal
8.1
7.8
11.6
11.6
Effluent*
concentrations
gr/ft.3 x 103
0.85
0.35
0.17
0.04
Average
reverse
pressure
in. water
17.0
18.0
14.0
20.0
Residual
dust
deposit
grams
145
213
263
268
Air volume
per
0.2 sec. pulse
ft.3
0.30
0.32
0.20
0.16
RDM measurement.
Based on cloth velocity of 8.5 ft./min. at indicated residual resistance, cloth area of 4.7 ft.2 and
persistence of maximum reverse pressure for 0.2 sec.
-------
rate of pressure rise was consistent with the greater dust concentration
expected for the lower resistant and more porous filter structure. The
residual dust holding for the filter was also lowest when the pressure
rise rate was the greatest.
Although the cleaning has been described as reverse flow because a low
pressure fan was used in the above tests, the actual results were es-
sentially the same as would have been obtained by the restricted venting
of a compressed air source (slow acting solenoid valve). Despite the
generally higher operating resistance, Test 25, Table 31, suggests
that steady state operation can also be maintained with the lower energy
pulses produced by the reverse flow system.
Discounting for the moment the size of the blower (up to 600 ft. /min.
capacity) used to provide the reverse flow by the mechanisms shown in
Figure 83, the actual volumes used per 0.2 sec. pulse were in the
0.25 to 0.33 ft. range STP. This is in the same range as the air
volumes delivered by the pulse jet system at 70 to 100 psig for typical
pulse durations, Figure 87. If one considers only the cost associated
with the reverse air volume passing through the filter the relative
power costs greatly favor the lower pressure fan by a factor of 50 to
100. Practically speaking, however, the low pressure system (up to
20 in. water) must be kept in operation to provide a fairly rapid rever-
sal of air. Hence, the overall system must be designed as a multi-
compartmented unit, order of 10 or more, so that by fast acting dampers,
a single fan can service several compartments. The principle conclu-
sion to be drawn from the tests summarized in Table 31 is that there
appear to be alternative means to clean felted media that should be
thoroughly investigated.
Frequency of Cleaning, Reverse Flov Only - Limited tests with the re-
verse flow system, Table 32, indicated that extending the filtration
time to 1 minute between cleanings led to undesirably high operating
resistances, ~ 11 in. water. Conversely, decreasing the cleaning
302
-------
0.5
* 0.2
o
UJ
0.1
I I I I I I I
L I I I I 1 I I
20 50 100
RESERVOIR PRESSURE, psig
Figure 87. Air volume ejected per pulse with commercial 1/A-in.
nozzle and solenoid valve assembly
303
-------
Table 32, EFFECT OF CLEANING FREQUENCY ON FILTER RESISTANCE AND
EFFLUENT CONCENTRATION WITH REVERSE FLOW CLEANING. FLY
ASH/DACRON FELT AT 11.6 GRAINS/FT.3 AMD 8.5 FT./MIN.
VELOCITY.
Test
number
29
25
30
Cleaning
frequency
min.
0.25
0.5
1.0
Bag resistance in. water
Residual
7.8
7.5
9.5
Average
7.9
7.8
10.8
Terminal
8.00
8.1
12.0
Effluent
concentration
gr./ft.3 x 103
0.35
0.85
0.61
Average
reverse
pressure
in. water
15
17
17
Pressure
rise rate
in. /sec.
615
750
900
interval to 0.25 min. did not show any improvement in resistance cha-
racteristics relative to a 0.5 min. cleaning frequency. A direct cor-
relation between rate of pressure rise and the amount of dust deposited
upon the filter between cleaning cycles was indicated. Tests reported
in Chapter III showed large differences in the rate of pressure rise
for a 10 to 12 fold range in inlet dust concentration.
Reverse Flow Duration - Reverse Flow Only • The effect of reverse flow
duration was studied using reverse pulse equipment modified by the
addition of the high volume blower. As in previous tests, the blower
was operated continuously and the flow diverted by the damper into the
Dacron filter bag during the cleaning cycle.
In the present test series, the filter was cleaned every 0.5 minute for
time intervals ranging from 0.025 to 2 seconds as shown in Table 33.
According to the pressure/time traces for these tests given in
Figure 88, the wave forms for the 0.2 and 2 second pulse period.
similar. Filter resistance characteristics also were similar but the
effluent concentration was much greater for the 2 second pulse period.
It is not believed that so large an increase is typical for the extended
pulse. Based upon prior tests with conventional (compressed air) pulse
jet systems, Chapter III, effluent concentrations appeared to increase
304
-------
Table 33. EFFECT OF LOW PRESSURE PULSE DURATION ON FILTER RESISTANCE AND EFFLUENT
CONCENTRATION. FLY ASH/DACRON SYSTEM.
u>
o
Test
number
31
32
33
34C
Reverse flow3
duration
sees.
2.0
0.2
0.05
0.025
Bag resistance in. water
Residual
8.4
8.8
9.3
12.6
Average
9.4
9.0
9.9
14.5
Terminal
10.1
9.7
10.6
16.5
Effluentb
concentration
gr/ft.3 x 1C3
4.4
0.61
0.57
—
Reverse pressure
in. water
Maximum
20
20
19
17
Average
16
16
12
9
Cleaning frequency, 0.5 min.
Inlet concentration, 11.6 gr
Air flow unsteady, fan capacity exceeded.
Inlet concentration, 11.6 grains/ft. , filter velocity 8.5 ft. /min.
-------
cr.
UJ
I
40
? 20
UJ
cr
:D
CO
to
UJ
a:
CL
UJ
cr
UJ
UJ
Ul
o:
0 L
1.0 2.0 3.0 4.0
A. 2 SECOND PULSE
40
20
I I I
40
20
I I I I i i
40
20
0 0.2 O.4 0.6 0.8
B. 0.2 SECOND PULSE
0.2 0.4 0.6
C. 0.05 SECOND PULSE
0.8
B&AMA
I I I I I I l j i
02 O.4 0.6 0.8
D. 0.025 SECOND PULSE
Figure 88. Characteristic pressure/time traces for variations in reverse pulse duration (see Table 33)
-------
as the square root of the pulse duration. Hence, with respect to the
2 second pulse in Table 33, an emission of 2 x 10 grain/ft.3 might
be nearer to the correct level. A small increase in filter resistance
was noted for a 0.05 sec. pulse but the data are too limited to warrant
any conclusions even though the pressure differential attained might
suggest less effective cleaning. There does appear to be a lower limit
for pulse duration, ~ 0.025 sec., at which dust removal falls off rapidly
according to the high filter resistance observed. Although the rate of
differential pressure rise was about the same, 700 in./sec., the maximum
and average pressures displayed for the 0.025 sec. pulse were much lower.
This test shows that not only the rate of pressure rise but the maximum
value reached is also important in effective cleaning. The critical
factor is that the impulse applied to the fabric must persist long
enough for the fabric to experience full flexure. Continuation of
pulses beyond this point appears to lead to higher effluent loadings
and greatly increased reverse air demands.
Reverse Flow With Reverse Pulse Cleaning
A special sequence of tests was performed in which various combinations
of reverse pulse and reverse flow cleaning, differing with respect to
time phasing, were evaluated, Table 34. The basic test apparatus was
that shown in Figure 83 except that the conventional 1/4 in. jet noz-
zle and Venturi section were re-installed in the system. Filtration
was performed with a Dacron felt and an inlet concentration of 11.6
o
grains/ft, at 8.5 ft./rain, filtration velocity.
The characteristic wave forms for the bag pressure differentials noted
for these tests are given in Figure 89. Figures 89a and 89b, respective-
ly, show the individual patterns for pulse jet and reverse flow alone,
while Figure 89c indicates the effect of superposition of both cleaning
processes. It should be noted that the combined Pressure gradients are
approximately equal to those of the pulse jet alone and not the alge-
braic sum of that produced by the high and low energy pulses.
307
-------
Table 34. EFFECT OF PULSE JET AND/OR REVERSE FLOW CLEANING ON FILTER RESISTANCE AND EFFLUENT CONCEN-
TRATIONS WITH A FLY ASH/DACRON FELT SYSTEM
Test
nxnr.be r
35
36
37
38
39
40
41
42
43
44
Pulse jet parameters
Pressure,
psig
60
60
60
60
60
. 60
60
60
60
Frequency,
min.~^
1
No pulse
1
1
1
1
1
1
1
1
Duration,
sec.
0.06
0.07
0.07
0.07
0.07
0.07
0.07
0.07
0.07
Reverse flowb
parameters
Duration,
sec.
No reven
0.2
0.2
0.2
0.5
0.5
0.5
0.5
0.5
Phasing,
sec.
>e flow
-
0.09
0.022
0.01
-0.05
-0.05
0.03
0.05
No reverse flow
Max imun>c
differential
pressure
in. water
Combined
19
-
23
22
22
20
22
23
23
22
Reverse
flow
only
_
13
11
11
11
10
10
11
11
-
Bag resistance in. water
Residual
2.7
5.7
3.3
3.5
3.6
3.8
3.7
3.5
3.6
3.9
Average
4.3
6.8
5.4
5.7
5.4
5.3
5.3
4.8
5.0
5.1
Terminal
5.9
8.0
6.5
6.9
7.2
6.8
6.9
6.1
6.4
6.3
Effluent6
concentration
gr./ft.3 x 103
2.0
0.35
0.48
2.5
2.9
2.7
-
3.4
1.5
-
CO
o
oo
Direct pulses
Minus sign ncans Jet pulse initiated first.
cAverage rate of pressure rise, 2300 in./sec. for pulse jet and 760 in./sec. for reverse flow.
Based on inlet loading of 11.6 gr/ft.3 and filtration velocity of 8.5 ft./min.
RDM measurement.
-------
a:
UJ
I
40
20
BSJ^S H
t£/A*»*-i •^^••••••^••iBHW
a, PULSE JET ONLY - 60 PSIG , FREQUENCY
I MIN. , DURATION 0.06 SEC.
20
UJ
tr
I
UJ
-------
Only one test was performed in this series in which reverse flow only
-3 3
was used. The indicated effluent concentration, 0.48 x 10 grains/ft.
was in good agreement with values reported in Table 31, Tests 25 and 26,
and Table 33, Tests 32 and 33. On the other hand, most tests in which
pulse jetting accompanied the reverse flow showed outlet concentrations
5 to 10 times higher. There appears to be a slight indication that
if the jet pulse lags appreciably the reverse flow, Table 34 Tests 37
and 43, a reduction in effluent concentration is obtained. It may be
that prior removal of dust by the preceding reverse flow reduced the
removal of interstitial deposits by the reverse pulse. Hence, the re-
sulting dust/fabric collection surface constitutes a more efficient col-
lector. The observed resistance values, however, do not appear to
support the above theory.
For the most part, the test results summarized in Table 34 show that
sequential pulsing affords no advantages with respect to lowered resis-
tance or reduced emission but does lead to increased consumption of
cleaning air. One can further conclude that it is the initial mechan-
ical impulse given the fabric (and its resulting acceleration) that is
responsible for dust removal. The role of aerodynamic dislodgement
appears as a very minor factor.
CONCLUSIONS TO REVERSE FLOW CLEANING STUDIES
The studies performed under the heading of reverse flow cleaning differ
from those discussed previously in Chapters II and III primarily be-
cause the total effort devoted to this phase of the study was limited.
As point out at the beginning of this chapter, many tests were sim-
plified (and less sophisticated with respect to instrumental techniques).
At the same time the dust/fabric combinations studied were limited and
the experimental equipment arrangements were not, in most cases, typical
of commercial installations. Because of the constraints cited above,
relatively few firm conclusions can be drawn and those that are
310
-------
presented reflect probable relationships among key variables,•rather than
any specific operating guidelines.
General Conclusions
The conclusions set forth below are intended to place the results in
proper perspective, to highlight the major experiments, and to suggest
basic guidelines for further research.
• Limits of data application - Test results should not be
extrapolated beyond those for the specific dust/fabric
combinations studied. As a further constraint because
the experimental apparatus was not necessarily modelled
after commercial installations, no data scaling should
be attempted. Although dimensional similarity was
maintained in some cases, the tests were too few in
number to warrant any statistical analysis.
• Need for further research - The tests performed in the
limited study described here indicate that the factors
leading to the successful application of reverse flow
cleaning are even less understood than those for
mechanical shaking and pulse jet systems. The
results suggest strongly that the unique features
of dust/fabric interactions, which prevent ready
extrapolation to other dust/fabric systems, are
equally important in reverse flow systems.
• Outlet versus inlet concentrations - Limited test
data indicate that the use of reverse air flow to
augment cleaning by mechanical shaking or pulse jet
air has a negligible impact on base system performance.
Thus with mechanically shaken systems, the effluents
are only weakly dependent upon inlet dust loadings.
Conversely, augmentation of conventional pulse jet
cleaning processes, produces emissions that relate
closely to the inlet dust concentrations.
• It is pointed out again that those factors necessitating
very careful interpretation of any fractional particle
size efficiency concepts are equally important with
respect to reverse flow systems.
311
-------
n of Mechanical Shaking by Reverse Flow
The use of reverse flow air alone or in combination with the mechanical
shaking of woven fabric bags forms the basis for the following
conclusions:
• No benefits were noted, resistance- or efficiency-wise,
by reverse flow during the shaking interval even when
an interior support was used to prevent a complete bag
collapse. This appraisal may not apply to the use of
reverse air in the field where failure of isolation
valves to close completely can be compensated for by
a separate source of reverse air. If one does not
use a separate fan to provide a reverse flow (or at
least a zero flow condition) some of the fine dust
loosened during the shaking process may redeposit
and the foward flow will also transport loosened par-
ticles more readily to the clean air side of the
filter system.
• Not only were resistance and efficiency characteristics
unchanged but as near as could be determined no dust
removal was accomplished by aerodynamic action alone.
• The dust removal observed during tests with only reverse
flow, with both sateen weave cotton and Dacroti bags,
was attributed to bag flexure alone. Despite the ob-
served dust removal, the overall performance of the
cleaned filters was highly unsatisfactory with a fly
ash aerosol.
• The main role played by the reverse air appears to be
that of preventing mechanical projection of dust into
the clean air side of the system. Because the gas
flow is stopped to the compartment undergoing cleaning
for periods of about 1 minute before and after shaking,
settlement alone enables most of the suspended dust
to reach the collection hopper.
• Any attempt to use reverse flow air with shaking bag
systems will be unsuccessful unless bag collapse can
be prevented by means of internal support structures
or high bag tensioning. It does not appear practical
to use highly tensioned bags with shaking because other
tests (Chapter II), have shown an adverse effect on
emissions. One can also speculate that greater damage
to fabric structure is also likely. Although the use
of internal supports such as rings can prevent col-
lapse, bag wear may result. More importantly, the
312
-------
nature of the motion imparted to the bag by a shaker
arm and the resulting cleaning action might be dras-
tically altered.
Felt Bags Cleaned by Reverse Flow
Hie tests described in this section, despite their arbitrary classifica-
tion as reverse flow cleaning experiments, do not depict the operation
of standard commercial equipment. Although a conventional blower (thin
scroll centrifugal with a static pressure capacity up to 30 in. water)
was the reverse air source, its delivery characteristics were not
radically different from those of a compressed air jet. Thus, some of
the tests performed with the modified pulse jet apparatus more properly
describe the influence of wave form on transient bag differential pres-
sures.
• Dust Removal by Aerodynamic Drag - A controlled series
of tests in which fabric acceleration was reduced by
roughly 1000 times the level produced by high pressure
jet pulse, indicated that reverse air velocities and
air volumes ranging from 4.3 to 11.4 ft./min. and 0.5
to 2.9 ft.3, respectively, had no effect on the
amount of dust dislodged from the filter. The nearly
constant removal of a few grams of dust was attributed
solely to the flexing (slow dilation and slow con-
traction) produced by each air admission.
• Waveform and Dust Removal - The results of several
tests in which the rate of rise of differential pressure
across the bag was varied indicated that the primary
dust removal was due to the acceleration imparted to
the bag. The actual waveforms were nearly identical
to those produced by high pressure jet pulses except
for a less rapid rise in pressure (and hence lower
fabric acceleration as described in Chapter III).
• Reverse Flow Duration - Except for those cases where
the reverse flow was restricted to intervals less than
0.05 sec., increased pulse duration exerted no effect
on average operating resistance. The only critical
factor for a given rate of pressure rise is that the
valve open time be sufficient for the full pressure
differential to develop and the fabric to experience
its maximum deflection.
313
-------
Sequential Pulsing - A series of tests in which low
energy pulses produced by reverse flow were augmented
by high energy compressed air pulses indicated that
the filter performance with respect to resistance
and efficiency was controlled mainly by the high
energy pulse. At this time, there appears to be no
particular advantage to using this technique which,
for the most part, only increases the demand for
cleaning air.
REFERENCES
1. Billings, C.E. and J.E. Wilder. Handbook of Fabric Filter Tech-
nology. Volume 1: Fabric Filter Systems Study. GCA/Technology
Division. Department A, Clearinghouse, U.S. Department of Com-
merce, Springfield, Va. 22151. Report Number GCA-TR-70-17-G,
APTD-0690, Contract No. CPA-22-69-38, PB-200-648, December 1970.
2. Larson, R.I. The Adhesion and Removal of Particles Attached to
Air Filter Surface.AIHA Journal J.9, 265 (1958).
3. Zimon, A.D. Adhesion of Dust and Powder. Page 112, Plenum
Press, New York (1969).
4. Walsh, G.W. and P.W. Spaite. An Analysis of Mechanical Shaking
in Air Filtration. Journal of the Air Pollution Control Associa-
tion. 12, 57 (1962).
314
-------
CHAPTER V
CONCLUSIONS AND RECOMMENDATIONS
The results of a detailed study of fabric filter cleaning processes used
for industrial gas filtration are presented in this report. The informa-
tion deriving from this program is intended to improve the performance
and to further the application of fabric filters for control of par-
ticulate emissions. Although many factors are involved in assessing
system behavior, the maximizing of gas handling capacity, collection
efficiency, and service life of filter fabrics and other system com-
ponents and the minimizing of power requirements related to gas handling
and filter cleaning operations are considered to be key areas for per-
formance improvements.
The conclusions and recommendations appearing in the following paragraphs
are based mainly on the findings of the present study. The results of
related studies by GCA/Technology Division or other groups, however,
have also been used to support and/or clarify the experimental results
provided in this report.
The conclusions presented in this chapter apply both to fabric filter
systems as a class, irrespective of the method of cleaning, and to the
comparative performance of devices that are cleaned by different dust
removal methods. It is again emphasized that the major work in this
study was with mechanically shaken and pulse jet cleaned fabrics. Gen-
erally, the experiments were designed to describe the behavior of ex^.t-
ing gas cleaning systems and to develop operating parameters or at least
315
-------
guidelines to improve their performance. The limited tests performed
with various approaches to low pressure, reverse-flow cleaning were
essentially exploratory. Therefore, the test results shou!4 not be extra-
polated directly to real field situations.
All tests were carried out at typical ambient conditions which, in this
study, were an air temperature of 70 ^ 2 F and a relative humidity of
40 to 50 percent. It was assumed, therefore, that any observed differ-
ences in collector performance were not attributable to temperature our
humidity changes. It should be noted that large changes in either
variable can affect fabric filter performance significantly.
The terms "equivalent" and steady-state operation have been used in this
report to describe any series of replicate tests in which no perceivable
changes in system performance were observed over short term periods (days).
Under these conditions, several parameters could be investigated, since,
for all practical purposes, the effective residual and terminal drags
remained constant for any set of fixed operating conditions. Over the
long term, however; e.g., weeks to months, a gradual rise in fabric
resistance properties may result from dust accumulating within the fabric
pores.
CONCLUSIONS
Limits of Data Application
Unless indicated to the contrary, the reader should
assume that descriptive and operating parameters
cited for a specific dust and fabric combination
apply solely to that specified dust/fabric combination.
Outlet Versus Inlet Concentrations
There is no simple relationship between typical
outlet and inlet concentrations for fabric filters.
316
-------
The average mass emission and its related particle
size properties for a mechanically shaken filter
when operating in a fixed mode with a specific
dust/fabric combination is nearly independent of
the concentration and size properties of the inlet
dust.
On the other hand, mass emissions from felted fabric
filters cleaned by high pressure, reverse air pulse
show a positive dependency on the inlet dust
concentration.
Fractional Particle Size Efficiencies
The concept of using fractional particle size effi-
ciencies to predict filter performance based upon a
knowledge of inlet dust properties can rarely be
applied to a fabric filter system.
In the case of mechanically shaken filters, the
"apparent" measured efficiencies will depend upon
the filter inlet loading and the instantaneous or
average values of the effluent.
Conversely, the "apparent efficiencies" for pulse
jet systems may be independent of particle size due
to the release of many agglomerated particles during
the cleaning process.
Filter Effluent Concentrat ions
Filter effluent concentrations for the same inlet
dust may range from 10~7 to 10~3 grains/ft.-* for
mechanical shaking systems with woven fabrics and
from 10~3 to 10~^ grains /ft.-* for pulse jet sys-
tems using felted fabrics.
The lower effluent concentrations found with shaken
filter bags is due mainly to the touch greater super-
ficial dust deposit that results primarily from ex-
tended filtering periods (approximately 20 minutes)
compared to approximately 1 minute intervals for
felts cleaned by pulse jet air. At the same time,
the somewhat larger pores and greater depth of the
felts may require a longer period to fill the
interstitial structure.
317
-------
The time required to fill the filter pore structure depends
upon the dust deposition rate, the bulk density assumed by
the interstitial dust deposits and the pore dimensions of
the filter media compared to those of the dust.
Filter Cleaning Action
Dust removal by both mechanical shaking and by high
pressure air impulse is mainly the resulting inter-
action of tensile forces produced by acceleration of
the dust-laden fabric and the characteristic dust
adhesive and cohesive forces.
The cleaning attained by mechanical systems, when
defined in terms of residual resistance and dust hold-
ing capacity, can be related to the maximum accelera-
tion seen by the shaker arm, the latter determined
solely by its amplitude and frequency provided that
the number of individual shakes is sufficient ( 200).
By decreasing shaking amplitude and increasing shaking
frequency a moderate reduction In effluent loadings
(factor of 2) may be obtained while maintaining con-
stant the acceleration level (and also the resistance
and capacity properties).
The cleaning (dust removal) afforded by pulse jet sys-
tems is best defined in terms of the residual and
average filter resistance for a fixed cleaning cycle.
The amount of dust removed during pulse jet cleaning
depends upon the initial felt holding; inlet dust con-
centration; the magnitude, rate of rise and duration
of the reverse air pressure; and the frequency of
pulsing.
The modulation of the pressure pulse waveform affords
one means of controlling effluent concentrations.
Aerodynamic forces play a very minor role in actual
dust removal for pulse jet filter systems and also for
mechanically shaken systems in which low velocity, low
pressure reverse air flow is used as an adjunct to
cleaning. The main role played by low velocity
( 100 ft./min.) reverse air is to flush out dust par-
ticles already loosened by high energy impulse. In
the case of typical reverse jet systems extended pulse
(reverse flow) periods usually had an adverse effect
on effluent concentrations.
318
-------
Energy Requirements and System Capabilities
Requirements for mechanical shaking operations represent
but a small fraction, 10 percent at the greatest, of the
overall system power needs. For comparable gas handling
capacity and filter resistance, typical power needs for
high pressure pulse jet cleaning are roughly equivalent
to those for gas handling.
Pulse jet systems permit filter operation at high air-to-
cloth ratios (three to five times higher than those for
mechanically shaken systems) provide nearly constant ven-
tilation rates at dust sources, adapt more readily to
high loadings with respect to minimizing average filter
resistance, and require smaller physical plants because
of their higher air-to-cloth ratios.
RECOMMENDATIONS
Equipment Application and Operating Parameters for
Mechanically Shaken Filters
Fabric filter systems cleaned by mechanical shaking
should be used whenever the particulate to be con-
trolled is highly toxic. Sateen weave cotton
fabrics or their equivalents in synthetic or mineral
media in terms of fiber diameter, staple, weave and
cloth weight will provide the best performance.
Based upon laboratory pilot tests, a mechanical shak-
ing frequency of 8 cps and a shaking amplitude not
greater than 1-inch are recommended. Field observa-
tions have indicated, however, that some field filter
installations may not have the structural rigidity
to be shaken at this energy level.
To provide improved dust removal and increased hold-
ing capacity with many existing filter systems (some
of which operate at low amplitude and frequency;
e.g., 0.25 in. and 4 cps), fabric acceleration should
be increased by increasing the shaking frequency in-
stead of the shaking amplitude to avoid resonance pro-
blems with the baghouse structure. As a fringe benefit,
shaking at reduced amplitude also reduces particulate
emissions.
Filter bags should be tensioned so that the tensile
force exerted at the top is roughly 0.5 Ib. in excess
of the combined weight of the bag with its residual dust.
319
-------
Approximately 200 individual shakes should be given a
bag during a cleaning cycle to achieve the optimum
dust removal for a fixed combination of bag shaking
frequency and amplitude.
The off-line period for the shaking of a single bag
or filter compartment should represent only about 10 to
15 percent of the time allocated for filtration;
e.g., 4 minutes versus 30 minutes.
Any field measurements to define the average size pro-
perties and the average mass concentration of a filter
effluent must include every phase of the cleaning and
filtration cycles.
No final design criteria should be submitted for a
filter system until all possible interactions of the
dust of interest and the selected fabric are known.
The use of low velocity, reverse flow air as an ad-
junct to mechanical shaking is not ordinarily recom-
mended except where the use of compartment isolation
dampers fails to stop all air flow through the filter
in the direction of normal filtration flow.
Equipment Applications and Operating Parameters for
Pulse Jet Filter Systems
Pulse jet systems should not be selected when it
appears possible that mass emission rates for highly'
toxic materials may be exceeded.
Pulse jet systems are recommended for application
where process control or source ventilation needs
require very constant air or gas flow rates assuming
that emissions criteria are satisfied.
Pulse jet systems are recommended for those applica-
tions in which inlet dust loadings are either variable
and/or particularly high, approximately 15 to 20 grains/
ft.^ assuming that emissions criteria are satisfied.
Pulse jet systems must be operated with fast-acting
valves so that a rapid change in differential pressure
is obtained across the filter during the transient
pulse, approximately 2000 to 4000 in. water/sec. The
above levels are obtained routinely with commercial
apparatus in which a 1/4-inch nozzle is located above
each bag, the compressed air pressure is 70 to 100
psig and the valve opening time is of the order of
0.01 to 0.02 seconds.
320
-------
Approximately 200 individual shakes should be given a
bag during a cleaning cycle to achieve the optimum
dust removal for a fixed combination of bag shaking
frequency and amplitude.
The off-line period for the shaking of a single bag
or filter compartment should represent only about 10 to
15 percent of the time allocated for filtration;
e.g., 4 minutes versus 30 minutes.
Any field measurements to define the average size pro-
perties and the average mass concentration of a filter
effluent must include every phase of the cleaning and
filtration cycles.
No final design criteria should be submitted for a
filter system until all possible interactions of the
dust of interest and the selected fabric are known.
The use of low velocity, reverse flow air as an ad-
junct to mechanical shaking is not ordinarily recom-
mended except where the use of compartment isolation
dampers fails to stop all air flow through the filter
in the direction of normal filtration flow.
Equipment Applications and Operating Parameters for
Pulse Jet Filter Systems
Pulse jet systems should not be selected when it
appears possible that mass emission rates for highly '
toxic materials may be exceeded.
Pulse jet systems are recommended for application
where process control or source ventilation needs
require very constant air or gas flow rates assuming
that emissions criteria are satisfied.
Pulse jet systems are recommended for those applica-
tions in which inlet dust loadings are either variable
and/or particularly high, approximately 15 to 20 grains/
ft.3 assuming that emissions criteria are satisfied.
Pulse jet systems must be operated with fast-acting
valves so that a rapid change in differential pressure
is obtained across the filter during the transient
pulse, approximately 2000 to 4000 in. water/sec. The
above levels are obtained routinely with commercial
apparatus in which a 1/4-inch nozzle is located above
each bag, the compressed air pressure is 70 to 100
psig and the valve opening time is of the order of
0.01 to 0.02 seconds.
320
-------
The compressed air jets must be located far enough
above the bag exit (which may or may not contain a
Venturi section) so that supplemental entrained air
is extracted from the exit plenum and not from the
dirty air side of the bag.
In most cases, the duration of the pressure pulse
should be only long enough to persist through the
complete reverse inflation of the bag, roughly
0.05 second. With few exceptions it is not recom-
mended that the cleaning pulse be longer as it only
increases the compressed air demand and sometimes
the exit dust concentration.
It is suggested that the use of an auxiliary damp-
ing reservoir or a mechanical, electrical, or
pneumatic regulation of the jet nozzle valves be
used to reduce the particulate emissions. Laboratory
measurements indicated that the damped pulses and
the associated gradual venting of air from the bag
at the end of the pulse resulted in a fivefold re-
duction in emissions and about a 20 percent rise
in average filter resistance.
A recommended alternative to the pulse damping pro-
cedure for reducing emissions is to reduce the clean-
ing pressure. In many cases, the lowering of com-
pressed air requirements more than compensates for
the increased resistance (and air moving) power
requirement.
Research and Instrumentation Needs
It is recommended that operating data be obtained
through pilot plant testing and/or actual field
measurements to determine representative "K"
values for several commonly encountered dust/
fabric combinations. This effort should also
include an updating of related field and labo-
ratory experience from as many sources as
possible.
It is recommended that fundamental studies be
performed to determine what key particle and
fiber parameters taken singly or in combination
determine: (1) the K-value for a given dust/fabric
combination and (2) the characteristic effective
residual resistance for the above dust/fabric
combination in conjunction with a specified fabric
. cleaning technique.
321
-------
It is recommended that mechanical shaking systems using
using a spring loaded hanger arm (typical of many
commercial bag suspension systems) be compared to
systems with rigid shaker arms such as used in
this study. The purpose of these tests is to
determine whether the basic acceleration and
dust removal relationships developed for systems
with rigid arms apply to the spring loaded supports.
It is recommended that other mechanical means of
transmitting cleaning energy to the fabric be examined.
Although simple mechanical shaking works, examination
of many bags after shake cleaning showed surprisingly
large amounts of dust adhering to the fabric as
distinct patches. Since the latter material could
often be removed by a very light flick of the finger,
it appears that energy transmission by shaking is
not sufficiently uniform nor as efficient as it
might be.
It is recommended that a simple method be developed
to determine whether proper tension levels are main-
tained in mechanically shaken bags. Newly installed
bags undergo rapid stretching such that frequent read-
justment is necessary during the shake down and early
service periods. Failure to re-tension will lead to
excessive slackness causing bag plugging and physical
damage.
Controlled laboratory pilot testing (and field testing
where practical) should be carried out with high tem-
peratures and humid atmospheres to provide design
parameters for difficult field applications.
Rigorous tests should be performed, initially on a
bench scale and later on a pilot scale, to determine
how the following factors control or may be utilized
to improve the filterability of various dusts:
electrostatic charge, its presence or absence on par-
ticles and/or fibers; particle size distribution with
shape factor constant; humidity control; particle size
versus fabric pore size; surface deposition versus
interstitial deposition; and the use of conditioning
methods such as induced agglomeration.
It is recommended that detailed tests be performed to
determine the effects of variations in inlet concentra-
tion and filtration velocity on the performance of
pulse jet cleaning systems.
Further pilot testing should be performed to determine
whether use of a simple plenum pulse system that elim-
inates individual valving for each bag offers advantages
322
-------
over the pulse-per-bag approach investigated in
detail in the present study.
Definitive tests should be performed to determine
what maximum bag length can be cleaned effectively
by pulse jet systems.
323
-------
APPENDIX A
AUTOMATIC FLOW CONTROL SYSTEM
To enable extended operation of the mechanical shaking equipment without
close attention yet with precisely repeated timing of the cycles, an
automatic control system was developed. This system uses a rotary cam-
type timer to control the following operations:
1. Primary fan ON
2. Aerosolizing compressed air ON
3. Dust feeder ON
4. (Dust is collected for a prescribed time)
5 Dust feeder OFF
6. Aerosolizing compressed air OFF
7. Fan OFF
8. (Shaker Motor ON for prescribed time)
9. (Dust is permitted to settle for a prescribed time)
10. Cycle repeats.
The intervals between steps and the length of the overall cycle are
adjustable to represent any standard shaking cycle. The system includes
manual override switches for operation of the fan, dust feeder, com-
pressed air, and shaker motor in any sequence or combination.
For cases where extended recycling is necessary, making unmanned opera-
tion desirable, certain fail-safe provisions are included in the system.
These are designed to prevent damage to the equipment in case of mal-
function. They are also intended to prevent data loss due to upsetting
the status of the fabric and dust deposit, which may represent several
325
-------
hours of previous operation. These criteria for unmanned operation in-
clude the following:
1. The main fan, dust feeder, and compressor supply to the
air ejector cannot operate during the cleaning cycle,
and vice versa.
2. Reduction of reservoir pressure within the compressed
air tank below some preset level due to compressor fail-
ure; e.g., 90 psi will shut down the entire system.
3. If the flow adjustment damper can no longer control the
flow rate either at the full-open or full-close position,
the entire system is shut down. Failure to operate within
the working range of the control valve indicates either a
plugged bag (faulty cleaning) or a rapid decrease in bag
resistance (excessive dislodgement of dust or holes in
fabric).
4. Excessively high or low inlet dust concentrations to the
collector as monitored by automatic light attenuation
measurement will direct a system shut down.
5. Manual override at any time.
The system with fail-safe provisions is shown schematically in
Figure A-l.
The system is designed to operate under a preset time cycle, recording
continuously the pressure differential across the bag. Alternatively,
with minor modifications, the system could be made to operate on reaching
a preset upper pressure limit, with a varying time cycle. For flow con-
trol, a Bailey Meter Co. pneumatic system is used. This system senses
the flow past a Stairmand disc in the 2-inch duct, amplifies the signal,
and supplies 3 to 25 psig air to a diaphragm control valve in the
4-inch duct. The system also records primary flow and bag differential
pressure on a 2-pen circular 24-hour recorder. Because the Bailey
system is of standard industrial quality, it furnished only marginal
control. The control system was calibrated by pitot static tube tra-
verses believed to have been accurate to within 2 percent. The control
provided by the system was steady from second to second. Over a typical
326
-------
CJ
N>
MAIN
SHUT-OFF
RELAY
COMPRESSOR
DETECT.
04-1
SOURCE
DUST
FEEDER
LIGHT
ATTEN.
RECORD.
CONTROLLER
1HP
BLOWER
SHAKER
TIMERS
SHAKER
MANUAL
OVERRIDE
§.
E
IT
CH
MASTER
TIMER
MANUAL i
STOP
SWITCH
BAG PRESSURE DROP
TRANSDUCER- READOUT
r—i
J. TO
t BAG PRES.
| TAPS
I RECORDER-
^CONTROLLER
! TO
VALVE
• ACTUATOR
« "!•—-TO
FLOW4"''" FLOW
TRANSDUCER- QIFF
READOUT TAPS
FLOW VALVE LIGHT
SWITCHES
Figure A-l. Automatic flow control system
-------
filtration cycle, during which the pressure differential across the fil-
ter increased by a few inches of water, the controlled flow decreased
by a few percent due to an inability of the control system to hold ex-
actly to the set point. This decrease in flow was usually negligible
except when the pressure differential reached the neighborhood of 8
inches of water, at which point the flow began to drop rapidly. To
compensate for this in the few experiments involving such high pressures,
the recorded pressure differentials were adjusted to the preset flow
conditions.
A further limitation of the flow control system was that a brief period
of about 5 seconds was required for the flow to stabilize, following
start-up or a sudden change of pressure differential as in pulse clean-
ing. This was associated with a tendency of the control system to hunt
and overshoot the set point. To minimize any possible effects on the
dust deposit of such an overshoot in pressure or flow, a damping re-
sistance was used in the pressure line to the automatic control valve.
This diminished the overshoot to, at most, a few percent at the expense
of a slightly longer period required for stabilization. The stabiliza-
tion period was only a problem when attempting to obtain true residual
pressure differentials, especially in the case of pulse cleaning with
its short intervals between pulses. The recorded residual pressure
differentials were adjusted to reflect any abnormal flow at the moment
of record.
In general, the control accuracy and precision provided by the Bailey
system was adequate for this study. However, the system required so
much extra attention and data correction that an improved better system
is recommended.
Stable operation of the fabric filter system relies on a steady delivery
of aerosol. A simple light scattering device was constructed (based on
conventional stack instrumentation) in which a simple photocell detector
and light source were mounted on opposite sides of the 2-inch diameter
328
-------
gr
inlet duct. A battery and Rustrak recorder produced a strip chart
record of the photocell signal which was inversely proportional to the
optical opacity of the aerosol stream. The record provided an excellent
check of aerosol concentration stability, although no attempt was made
to make the instrument quantitative. The photodetector and light source
were not precisely flush with the walls of the duct and perhaps as a re-
sult) tended at first to become blinded by deep dust deposits. To mini-
mize this, apertures approximately flush with the walls were placed in
front of both detector and source. Air was blown slowly through both
apertures into the duct at a velocity sufficient to prevent dust migra-
tion into the optical path.
329
-------
APPENDIX B
FORCE AND VELOCITY MEASURING INSTRUMENTATION
FORCE AND PRESSURE MEASUREMENTS
A commercial load cell (Dynisco FTI-10) was used to measure top tension
and bag weight in shaker tests and to measure bag weight in pulse tests*
The mountings for the cell sketched and discussed in the text are shown
here in more detail. The main difficulty with the early mounting in
Figure B-l was that the motion required to cock and partially jam the
slide assembly was of the same magnitude as the motion produced during
load cell compression, less than 0.001-inch. This made the cell un-
reliable for other than static vertical measurements.
The subsequent mounting in Figure B-2 was very nearly frictionless and
thus responded at any angle or rate of shaking almost instantly. The
difficulty with this second mounting was mainly that the left and right
ends of the bag hanger (see side view) had different mechanical advan-
tages to the load cell. Thus any swaying of the bag left and right, or
any left and right motion at the lower end of the bag, appeared in the
signal as tension changes. Nevertheless, the second mounting was pre-
ferred over the first for most measurements. By doubling the second
design into a Watts linkage, it should be possible to produce a fric-
tionless mounting with perfectly vertical motion. Refinements of the
load cell mounting may yield diminishing returns however, because of
the unpredictable mechanical properties of the bags. Because of these
bag properties, the load cell systems were calibrated in situ with a
bag of known weight suspended from the cell and with the bag "shaken
331
-------
FTI-IO DYNISCO
LOAD CELL
SLIDING
FRAME
CEMENTED-ON
STRAIN GAUGES
PRESTRESSED
LOAD LI MITER
BLOCK FIXED ON
SHAKER SHAFT
BAG HANGER
Figure B-l. Load cell mounting for shaken bags. Model 1
232
-------
LOAD
CELL
SIDE (OPEN) VIEW
ALUMINUM BOX, APPROX. 2" x 7
SHAKER
SHAFT
LEVELING
ECCENTRIC
COLLAR
FLANGED
BEARING
PIVOTS
LATERAL ,
FORCE
STRAIN
GAUGES
THIN-WALLED
TUBING
n
£
2>
BAG HANGER
V SHOULDER BOLTS
END VIEW
Figure B-2. Revised design for load cell mounting for shaken
333
-------
In" to a steady configuration. Weights were then attached to the bot-
tom of the bag for calibration.
The same load cell was used in the pulse cleaning studies to weigh the
bag, dust, and mounting cage assembly as shown in Figure B-3. By re-
leasing clamps sealing the upper end of the bag and cage assembly to
the bottom of the upper plenum, the entire assembly was free to move
upward. The external load cell mounting handle raised the load cell
into contact with the bottom of the bag and slight additional lifting
placed the entire weight on the load cell. The top of the bag assembly
was restrained from tipping sideways by light, frictionless wire yokes
(not shown). About the only difficulty with this arrangement was that
other instrumentation attached to the bag and cage (pressure transducer,
accelerometer, ) required wires that affected the weighings. With
accurate tares and care in aligning the wires, however, fairly reliable
weights were obtained. They were checked from time to time by removing
the entire assembly through the top of the upper plenum and weighing it
on laboratory scales.
In all cases, the strain gauge bridge inside the load cell required the
circuit shown in Figure B-A, a simple standard circuit with voltmeter
readout. Oscilloscope readout was also obtained for studying instan-
taneous forces during the shaking cycle. In the latter case the volt-
meter had to be disconnected from the circuit to eliminate a conflicting
signal from the meter's armature assembly. The power supply voltage
and circuit were adjusted to give exactly 1 voltmeter division per pound
of force for convenience in data reduction. The calibration remained
drift-free and constant through the entire program. The load cell-to-
readout precision was about 0.01 pound. The force-to-load cell pre-
cision was not always this fine due to the mounting difficulties noted
above, up to 0.1 pound.
334
-------
CLAMPS
(EXTERNALLY MANAGED)
( UPPER PLENUM)
BAG AND CAGE
ASSEMBLY
LOAD CELL
VERTICAL/PARALLEL
MOUNTING
•GASKET,
BOTTOM-GLUED ONLY
EXTERNAL
HANDLE, TO RAISE
LOAD CELL
FLEXIBLE BOOT SEAL
Figure B-3. Load cell mounting for weighing the pulsed bag assembly
335
-------
-O-
•t-O-
POWER
SUPPLY
FTI-IO DYMSCO
LOAD CELL
FINE CALIB.ADJ
[ *25ft
6-vU-—<>
O.I M 10 K
————\A^- <—W\
VOLTAGE
READOUT
ADJUST.
o UQ "vOLT. READ FOR REFERENCE
V—X ADJUST.
i O O 1
SIGNAL
t
22 K
IM
ZERO
BALANCE
VOLTMETER
K) mV FULLSCALE
Figure B-4. Circuit diagram, signal output from load cell
336
-------
As noted in the figures, strain gauges were cemented on the sides of
the thin-wall tube that formed the bag hanger arm. Two of these strain
gauges,
BLH SR-4 Epoxy Backed BLH Electronics, Inc.
Type FAE - 18 1286 Baldwin Lima Hamilton Corp.
Resistance: 120 ohms Walthatn, Mass. 02154
Gauge Factor: 2.04 617-894-6700
Dimensions: Overall: 3/8" x 9/16"
Active: 1/16" x 3/16"
were attached with Eastman 910 cement on opposite sides of the tube
near the point of maximum bending moment. Thus as the tube bent slightly,
one gauge compressed and the other stretched, the gauges were made
part of an "opposed" bridge circuit, so that bending the tube resulted
in a doubled effect. On the other hand, tension applied to the tube
produced the same stretching in both gauges, and so the effects can-
celled. Thus the circuit was sensitive to bending and insensitive to
tension. Because the signal was small, it was amplified before trans-
mission to the calibrating circuit similar to that in Figure B-4, The
signal was read on an oscilloscope. Calibration was accomplished by
turning the bag hanger arm to the horizontal and hanging known weights
from it. The factor of order 30 mV/lb was drift-free through all
experiments.
Thermal effects in the amplifier produced a wandering zero point, but
since the average lateral force during shaking is zero, this was not a
particular problem. Somewhat better shielding was required for this
system than for the load cell system. With a sufficiently thin-walled
hanger arm, the amplifier might become unnecessary.
If the two strain gauges were connected differently to the detection
bridge the system would be sensitive to tension and insensitive to
bending, taking the place of the load cell, thus, using a switching
arrangement, both lateral force and tension could be obtained with the
same simple system, at least in principle.
337
-------
VELOCITY MEASUREMENT
For precise analysis of the relationship between the position of the
shaker arm at a given instant and the lateral and tension forces at
the top of the bag, an electrical signal related to the arm position
was needed. The first approach, that of using a photoelectric detector
to describe arm motion, did not give an adequate time response. Attach-
ing the rotor of a resistance potentiometer to the shaker shaft did not
produce a reliable signal. Finally, a magnet was attached to the end of
the shaker shaft, and a small stationary coil was positioned close to
the magnet. The coil produced a clean, symmetric signal proportional
to velocity from which the position of the arm was deduced to be 90 de-
grees out of phase with the velocity signal. This signal directly
entered an oscilloscope on which the lateral and tension forces were
being analyzed. Both Lisajou displays (one variable plotted against
another) and time-based displays were studied.
338
-------
APPENDIX C
PHOTOELECTRIC DETECTION OF SHAKEN BAG MOTION
This approach to monitoring the instantaneous position of the bag dur-
ing shaking applies to a single point on the bag. The detector may
be moved to other points, or several such systems may be used simul-
taneously. The system is very inexpensive and accurate and can be
easily calibrated.
The system uses a rapidly responding photoconductor:
Clairex Photoconductor
Type 5H cadmium sulphide Clairex Corp.
Rise time at 23 foot candles: 0.004 sec. 1239 Broadway
Decay time at 23 foot candles: 0.0017 sec. N.Y. N.Y. 10001
Approx. resistance at 23 foot candles: 3,000 ohms 212-684-5512
Light Sensitivity: Approx. 1 percent per 1 percent
Dimensions: Approx. 0.5 in. dia. x 0.2 in.
As sketched in Figure C-l the photocell senses the filament of the
light source not obstructed by the bag. Motion of the bag normal to
the line of sight is detected immediately by the cell, allowing for
the time response limitations indicated above. The simple circuit
shown requires only standard laboratory equipment. Calibration is
accomplished by displacing the bag a known amount and observing the
signal deflection on the oscilloscope. A moderately low level of room
lighting, minimal shielding, and a sufficiently large battery for
light source stability will ensure a fairly clean, steady signal. Ele-
mentary design will produce an essentially linear response. The signal
may be recorded on an appropriate strip recorder for later analysis.
339
-------
PHOTOCELL
120V, 60 CYCLE IS PER-
MISSABLE.BUT DC WOULD BE
MUCH BETTER.
OPTIONAL LIGHT
MASKING
u>
*~
o
MOVING BAG
10V DC
OSCILLOSCOPE
SHOW-CASE 120V LIGHT
BULB, WITH APPROX. T"
LONG FILAMENT. THIS HELPS
PROVIDE A LINEAR SYSTEM
RESPONSE.
Figure C-l. Description of bag motion by interception of light beam
-------
APPENDIX D
THEORETICAL AVERAGE WAVE VELOCITY
In a hanging bag with constant mass per unit length (p) and total
length (L) the tension in the bag is given by
- pg (L - x)
where at the bottom of the bag x is zero and T is whatever tension
is applied at the bottom. Since the velocity of the wave at x is
/p the time required for the wave to travel down the bag is
W
(*L r° r°
t J dtJ .* . /
•o •{, * -i /
-dx
TT - pg (L - X)]
72
(D-l)
where W is the total weight of the bag, pgL. The effective overall
average velocity is therefore:
1
v'^
(D-2)
W W
This compares with the wave velocity in a string under uniform tension:
VrTw"
(D-3)
341
-------
If the bag were assumed to have an average velocity based simply on
the average tension, (T - W/2), the velocity would then be:
sT l" (D-4)
W " 2
Ihe above assumption introduces an error in the exact expression,
Equation (D-2), depending upon the dimensionless ratio (T_/W) :
• 100
TT/W
1.0
1.1
1.2
1.5
2
3
5
VD-2/ ^gL
0.500
0.683
0.772
0.968
1.21
1.57
2.11
VD-4/ 8L
0.707
0.775
0.836
1.00
1.22
1.58
2.11
% Error
41.4%
13.5
8.3
3.1
0.8
0.6
~0
Since the tension at the top of the bag cannot be less than the hanging
weight of the bag, the simpler expression Equation (D-4) results IB *
41 percent maximum error in computing average velocities. As the bag
is tightened at the bottom which increases the ratio of top tension to
bag weight, the error diminishes rapidly. Thus the simpler expre*sion
may be adequate except when working with slack and near-slack bags.
Hie same comment applies in computing resonant frequencies, which
depend on the effective overall average wave velocity:
where f is the Nth resonant frequency of the bag.
Equation (D-5) was tested against a variety of shaken bagi, observing th«
first few resonant frequencies on each bag and the associated
342
-------
The theoretical resonant frequencies were computed from Equation (D-5).
In many cases, the computed f., were about 1 cps lower than the observed
£ . This was attributed to the tacit and incorrect assumption that the
bag behaved as an ideal string. Actually, since the bag has appreciable
thickness while shaking, it also possesses stiffness. Hence, its wave
*
velocity must exceed that for a simple string. Determining the stiff-
ness of a bag while shaking is extremely difficult because the cross
section is neither circular nor flat, but somewhat oblate (Figure IV-4).
Rather than compute the exact wave velocity, therefore, an empirical
correction term was subtracted from the bag's length. This is to say,
the bag behaves as if it were slightly shorter because of the higher
wave speed. Equation (D-5) then becomes
2L ! _
(D-6)
where T^ » TT - W/2, and where the D/L ratio of the bag takes into
account the bag's stiffness to a first approximation. The equation
indicates that the stubbier the bag, the higher the resonant frequency'
The observed resonant frequencies were plotted against those computed
by this method in Figure IV-5. The good agreement furnished a firm
support to the theories of resonant frequency as applied to shaking
bags.
it
The wave velocity of a string with tension but no stiffness depends on
the tension. The wave velocity of an elastic beam with no tension but
with stiffness; i.e., with elastic modulus and bending moment, has a
characteristic wave velocity that depends on the stiffness. A member
with both tension and stiffness has a higher wave velocity due to the
superposition of the separate effects.
343
-------
APPENDIX E
BAG STRETCH AMD TENSION DURING SHAKING
The problem is to determine the bag strain resulting from the added bag
length when shaken according to the general wave form y « Yain (2«x/X).
t- 1
In the figure above,
When reduced,
Stationary Bag
/ Position
Shaking Bag
Position
dx2 + dy2 - (dx + h)2
h - dy/2dx ,
da
dx
dy
dy
(E-l)
provided that hs« 2 dx.
Since the stretched length ds - dx •»• h, the increase of the original
length dx is h and the average strain along this increment of the bag
2 2
is h/dx « dy /2dx . From the wave equation,
¥
(¥)
«-«
345
-------
The average strain along the bag length X/2 can be found by Integration:
x=X/2
x=o
That is, at the instant that the amplitude of the bag is Y, the average
strain along the bag depends simply on Y and on the wavelength. This
assumes that the bag has an integral number of wave nodes; otherwise,
a more complicated dependency would apply.
The strain may vary with time. In a simple string oscillating with a
standing wave
y = (Y sin wt) (sin 2jtxA)
2
the average value of (itY/X) can be shown to be one-half the maximum,
so that the time-averaged value of the strain average along the string
is
.1 /rtYN
2" \T)
2
The motion in a shaken bag , however , is predominantly a downward
traveling wave with the standing wave pattern less pronounce^. In a
traveling wave, the bag is under practically the sairie net elongation
at all times. This is an over-simplification since the bag length is
finite, and Y and /or the wave shape will T>e different at the ends of
the bag. However, to a first approximation,
describes the strain in a shaking bag averaged both ttmewisB and along
the bag. The strain is also defined by the simple relationship
g- Td/ML
where T is the tension difference attributed to elongation^, to the blig
modulus and L the bag length.
346
-------
Standing wave patterns and the process of reflection at the ends of
the bag may be expected to cause variations in this average tension.
Local tension variations will rapidly be distributed along the bag,
however, because longitudinal waves travel much faster than lateral
waves. The longitudinal wave velocity isvML/(p), which for a typi-
cal bag is approximately 750 ft/sec. This is -/ML/I, or about 25
times faster than the lateral wave velocity. Thus, any local modifi-
cation of tension distributes itself so rapidly along the bag, that
the bag is virtually under uniform dynamic tension from end to end.
The weight component of the tension of course still varies from end
to end.
347
-------
APPENDIX F
DAMPING AND AVERAGE BAG AMPLITUDE THEORY
The wave applied to the top of the bag diminishes as it travels down the
bag, such that the maximum displacement seen by successive points down
the bag may be given by A e"^*. A is the wave's initial amplitude at
x a 0 and p is a damping constant that depends on the rapidity with which
the wave energy is dissipated. If p is large, the wave may, practically
speaking, never reach the bottom of the bag, whereas for intermediate to
low damping constants, the remaining amplitude at the bottom of the bag
will be A e"pL.
Since the dynamic shaking tension depends on the square of the bag am-
plitude, it is preferable to determine the average value of the amplitude
squared, rather than the average value of the amplitude:
2fSL
Assuming that the wave reflected from the bottom of the bag has the
initial amplitude Ae"^1; i.e., there is no energy loss during reflection,
then by a similar process of integration it can be shown that the average
value of the amplitude squared for the reflected wave alone is
349
-------
It is assumed here that the waves travel with the same extant mechanics,
and do so independently of one another. By similar assumptions the Y '§
of the tertiary and any subsequent waves can be computed.
Let it now be assumed that the maximum amplitude seen by a given point on
the bag is the sum of all such successive Y's, that is, sooner or later
the waves will all combine constructively, resulting in a temporarily
large bag displacement at that point.
Y2
The same conclusion could be reached directly from Equation (F-l) under
the assumption that there is considerable damping, such that e"2pL « j^
so that the first reflected wave is negligible.
From the definition of the shaking attenuation parameter called a,
A -
For example, an average amplitude attenuation of 0.80 for a 10 foot bag
corresponds to a damping constant 3 of 0.078 per foot. That is, th* bag
motion is expected to diminish at this rate in moving down the bag.
350
-------
APPENDIX G
ENVELOPE PHOTOGRAPHS OF BAG MOTION
In this appendix, a series of photographs taken during a study of bag
tension versus frequency have been hand-copied with scale adjustments to
facilitate distance measurements. The maximum horizontal (envelope)
excursions from the rest or non-shaking condition have been plotted
for several locations over the length of the bag. The identifying
photograph numbers refer to the frequency locations listed on the
curve shown in Figure G-l. Examination of these photos, Figure G-2,
shows that the lateral excursions are maximized and the model loca-
tions clearly delineated at the resonance points.
The above measurements were made with a used, unnapped cotton sateen
bag, 10 ft. long and 6 in. in diameter. The shaking amplitude was
1 in. and the total weight of the bag and residual dust was 1.3 Ibs.
TENSION
(D
PKQUENCY
Figure G-l. Position of photographs relative to resonant frequencies
351
-------
SHAKER ARM DISPLACEMENT
CO
m
K>
MAX. LEFT
BAG DISPLACEMENT
MAX. \RIGHT
BAG DISPLACEMENT
321 0123
FIXED LOWER END OF BAG. in.
Photo no. 1: at first resonance; f *
1.8 cps; initial tension: 2.22 Ibs, (arm
ave.); shaking tension: 3.90 Ibs, time
ave. (max: 5.28; min: 2.52 Ibs.)
321 01 2
FIXED LOWER END OF BAG, in.
Photo no. 2: f » 2.2 cps (anti-resonance);
initial tension; 2.22 Ibs.; shaking tension:
2.86 Ibs. ave. (max: 3.50; min: 2.22 Ibs.)
Figure G-2. Bag displacement versus shaking frequency
-------
u>
en
u>
3 2 1 01 23
FIXED LOWER END OF BAG, in.
32101 2
FIXED LOWER END OF BAG, in.
Photo no. 3: f - 3.3 cps (mid resonance);
initial tension: 2.20 Ibs.; shaking ten-
sion: 4.05 Ibs. (max: 4.37; min: 3.73 Ibs.)
Photo no. 4: £ » 3.95 cps (second resonance);
initial tension: 2.20 Ibs.; shaking tension:
4.94 Ibs. (max: 5.82; min: 4.06 Ibs.)
Figure G-2 (continued). Bag displacement versus shaking frequency
-------
Ol
32 10 I 23
FIXED LOWER END OF BAG, in.
432 10 I 2 34
FIXED LOWER END OF BAG, in.
Photo no. 5: f = 4.1 cps (anti-resonance);
initial tension: 2.15 Ibs.; shaking ten-
sion: 4.13 Ibs.; (max: 4.46; min: 3.80 Ibs.)
Photo no. 6: f = 5.4 cps (mid resonance);
initial tension: 2.15 Ibs.; shaking tension:
4.85 Ibs.; (max: 5.41; min: 4.29 Ibs.)
Figure G-2 (continued). Bag displacement versus shaking frequency
-------
CO
Ul
LSI
1
432 I 0123
FtXED LOWER END OF BAG, in.
Photo no. 7: f - 5.95 cps (third resonance);
initial tension: 2.10 Ibs.; shaking tension:
5.42 Ibs.; (max: 5.82; min: 5.02 Ibs.)
432101 234
FIXED LOWER END OF BAG, in.
Photo no. 8: f • 6.4 cps (anti-resonance);
initial tension: 2.10 Ibs.; shaking tension:
5.26 Ibs. (max: 5.42; min: 5.10 Ibs.)
Figure G-2 (continued). Bag displacement versus shaking frequency
-------
U)
Ul
5210123
FIXED LOWER END OF BAG, in.
43210 I 23
FIXED LOWER END OF BAG, in.
Photo no. 9: f = 8.6 cps (fourth resonance);
initial tension: 2.10 Ibs. ; shaking tension:
6.70 Ibs.; (max: 6.86; min: 6.54 Ibs.)
Photo no. 10: f = 8.9 cps (anti-resonance);
initial tension: 2.10 Ibs.; shaking tension:
6.06 Ibs.; (max: 6.22; min: 5.90 Ibs.)
Figure G-2 (continued). Bag displacement versus shaking frequency
-------
U)
Ln
-J
32(0123
FIXED LOWER END OF BAG, in,
Photo no. 11: f - 11.o cps (fifth resonance);
Initial tension: 2.10 Ibs.; shaking tension:
7.80 Ibs.; (max: 7.95; min: 7.65 Ibs.)
3210123
FIXED LOWER END OF BAG, in.
Photo no. 12: f » 11.4 cps (anti-resonance);
initial tension: 2.10 Ibs.; shaking tension:
6.86 Ibs.; (max: 7.02; min: 6.70 lbs.j>
Figure G-2 (continued). Bag displacement versus shaking frequency
-------
APPENDIX H
DATA SHEETS FOR MECHANICAL SHAKING STUDIES
Because of the large number of original data and/or work sheets gener-
ated during this study, we have included in this section only the
summary tables for test measurements in the following categories:
• Dynamic Tension Tests
• Dust Removal Tests
• Effluent Size Properties
Copies of the original data are on file in the Control Systems Laboratory
of the Environmental Protection Agency.
DYNAMIC SHAKING TESTS
During dynamic tension measurements the bags were not ordinarily loaded
with dust, except for some residual dust holding in certain instances.
Although their weight did not change during these tests, they usually
underwent some stretching.
Total tension at the top of the bag is tabulated against the shaking
frequency. This is the time-averaged tension about which larger and
smaller tensions occur during one complete shake. Usually the first
two entries are:
Initial tension: defined as the zero cps tension at the
top of the bag when it is vertical;
0+1 in.: defined as the zero cps tension at the
~~ top of the bag when it is displaced
left or right of vertical to the full
amplitude being used in that series.
359
-------
The last entry in each series is sometimes a "zero check," which is
merely a re-measurement of the initial tension (vertical) after the
series is completed. An apparent decrease in initial tension indicates
bag stretching during the series.
Side force is also tabulated for some tests. This is the peak lateral
force at the top of the bag directed perpendicularly to the shaker arm;
i.e., the force causing the bag top to assume a sidewise motion. During
one complete passage of the shaker arm, this side force is directed
alternately to the right and to the left such that the average value is
zero. A listing of these data sheets appears in Table H-l.
Dust Removal Tests
The following tabulations of original data describe the loading of a
filter bag and the subsequent dust removal process under specified clean-
ing conditions. The data outlined below are included in the data sheets
summarized in Table H-2. Figure H-l represents a typical data sheet.
Loading Data;
Bag type, size, and general condition
Inlet dust loading and air flow to the bag, plus
relative humidity.
Tension at the top of the bag, with vertical shaker
arm, without inflation; "taut" = with the bottom of
the bag clamped; "loose" = with the bottom undamped;
i.e., nominally the weight of the free bag (pounds).
Time, pressure differential across the bag, and total
tension at the top of the bag due to this inflating
pressure in addition to the previous "taut" value.
Cleaning Data;
Amplitude (one-half the full shaker arm stroke) and
average frequency of the shaker arm;
Shaking Data: Cumulative seconds of shaking in
several intervals; average shak-
ing frequency observed in each
interval; cumulative number of
shaker cycles.
360
-------
Table H-l. SUMMARY LISTING OF DATA SHEETS FOR
TENSION/SHAKING FREQUENCY STUDIES
Test
number
D 1
D 2
D 3
D 4
D 5
D 6
J> 7
D 8
D 9
D 10
D 11
D 12
D 13
D 14
D 15
D 16
D 17
Shaking
amplitude (s)
(in.)
1
1
1
1
1
1
1
1
1, 0.5
2
1, 0.5, 0.25
1
1, 0.25
1
1
1
1
Bag parameters
Fabric
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Dacron
5 02 /yd2
Dacron
5 oz/yd2
Dacron
5 oz/yd2
Dacron
plain weave
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Dacron
5 oz/yd2
Dacron
plain weave
Unnapped
cotton
Dimension
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
5 ft x 6 in
5 ft x 6 in
10 ft x 4 in
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
10 ft x 6 in
Used or
clean
Clean
Used
Used
Clean
Clean
Clean
Clean
Used
Used
Used
Clean
Clean
Clean
Used
Clean
Used
Clean
Weight
(Ibs.)
1.11
1.33
1.33
0.56
0.56
0.56
1.11
1.33
1.33
1.33
0.6
0.6
0.77
1.4
0.56
1.33
1.11
Number
4
4
4
1913
1914
1916
1915
-
-
-
-
-
-
3
-
36
5
361
-------
Table H-l (continued). SUMMARY LISTING OF DATA SHEETS FOR
TENSION/SHAKING FREQUENCY STUDIES
Test
number
D 18
D 19
D 20
Shaking
amplitude(s)
(in.)
1
1
1
Bag parameters
Fabric
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Dimension
10 ft x 6 in
10 ft x 6 in
10 §t x 6 in
U$e..d or
clean
Clean
Clean
Used
Weight
Cl:b-6.)
1.11
1.11
1,4
Nwfcer
a
5
4
Circular cap installed at top of bag.
362
-------
TEST SI
DUST REMOVAL TEST
Date:
Test No.
4 May 71
LOADING DATA:
Bag No.
Bag type: 10'x6" cotton
Well Used
Inlet ,
cone.: 5 gr/ft"
CFM: A3.8
I RH
Initial Tension:8 Taut: 3.10
Loose: 1.90
Retaut:
Time (min)
A p (in..)
Filtering
Tens ion
0
1.0
3.5
9.5
17.5
30
1.15
1.35
1.54
2.10
2.67
3.45
8.05
8.00
2.95
8.05
8.45
8.70
Notes:
CLEANING DATA:
Amplitude: ± 1 inches Frequency: -» 9 cps
Cui.
lecondt
shaken
e
5
10
IS
IS
45
ep«
(Incrtn't
«v«.)
».-
».o
».0
9.0
8.9
Cut.
BO,
•haVea
0
45
90
VIS
224
402
Sh«k« •
tcniloa
r«ng«*
7.3
8.0
-
8.2
8.13
Diut
off
(•'«••)
0
128
12S
3*
10
11
Cut.
duft
oil
0
12«
2S3
2W
2»9
311
e— . t
4wt
•11
0
U.I
3}.l
41.0
42.3
44.0
K«»«U,
A»
<1«.)
3.4}
l.IS
R**vlll*t tnwlM*
T«ut
J.JS
J. 10
2.73
2.(»
2. SO
1.4S
Loan
2.70
2.4*
2.14
2.0*
l.M
1.13
Mtt
O.U
0.71
0.5»
O.I)
O.S2
0. SS
t«t»ut
TM*1
4u«t
M.
70*
«U
letti Vclght Ion bticd on tr» hin(tn| W» w«t|ht« • 1.79 - 1.13 > 0.77 1V«. or ISO |r«M,
SUrCRCLEAHlNC:
11. s
230 «.»-».!
'I uu |i..,ji.,o |o.n| j
*Tension in pounds at top of bag.
Figure H-l. Sample data sheet for dust removal tests
363
-------
Table H-2. SUMMARY LISTING OF DATA SHEETS FOR TYPICAL FILTRATION AND
MECHANICAL SHAKING TESTS
Test
number
SI
S2
S3
S4
S5
S6
S7
S8
S9
S10
Sll
S12
S13
S14
S15
S16
S17
S18
S19
S20
S21
S22
S33
S34
S35
S36
S37
S38
S39
S40
S41
S42
S43
S44C
S45.
S46*
S47*
S48*
S49d
Shaking parameters
Amplitude frequency
in.
1
1
1
1
1
1
1
1
1
0.75
1.15
0.50
1
1.50
1
1.50
2
1
1.5
2
0.50
0.75
1
1.50
1
1
1
1
1
1
1
2
1.5
1
1
1
1
1
1
cps
9.0
7.0
10.0
6.2
8.1
7.2
7.8
7.5
7.7
7.7
8.0
7.9
7.5
7.8
5.9
6.0
5.7
4.7
4.5
4.1
10.9
10.7
7.8
8.0
8.0
7.8
9.1
8.0
8.0
7.6
3.9
3.8
11.3
7.9
7.8
7.7
7.7
7.6
7.9
Tens ion
Ibs.
Shaking
7.9
7.1
8.3
6.8
7.4
6,8
7.1
9.5
7.5
6.0
8.1
4.5
7.2
11.9
6.4
8.4
14.3
5.1
10.4
9.4
3.9
5.4
7.8
13.0
8.8
6.0
6.0
6.3
6.5
6.0
6.3
8.6
4.5
4.9
4.9
6.0
6.3
4.8
5.7
Static
3-1
2.9
2.5
2.2
2.7
2.2
1.9
6.1
2.1
2.9
2.5
2.0
2.1
2.6
2.5
2.0
1.9
2.0
2.0
1.6
1,6
1.6
2.9
2.8
3,1
1.4
3
2.1
2.2
2.2
2.3
2^6
2.4
2-4
2.3
2.1
2.4
1.7
2.1
Filter*
capacity „
grains/ft.
326
246
414
264
312
289
314
289
304
202
350
31
310
336
179
300
348
61
233
197
83
255
372
315
315
197
241
153
135
191
46
71
44
246
213
80
385
49
470
Operating
parameters
Tests 81
through
S37
Unnapped
sateen
weave cot*
ton, equi-
librated,
Bag no. 4
Fly ash
aerosol.
Tests S3«
through
849 as
above but
with bag
dimensii3R«
s>f 10 ft,.
x 4 in.
15 TttitX.
6.0 min.
7.5 min.
60 fl&n..
364
-------
Table H-2 (Continued).
SUMMARY LISTING OF DATA SHEETS FOR TYPICAL
FILTRATION AND MECHANICAL SHAKING TESTS
Test
number
S50
S51
S52
S53
S54
S55
S56
S57
S58
S59
Shaking parameters
Amplitude frequency
in.
1
1
1
1
1
1
1
1
1
1
cps
8
8
8
8
8
8
8
8
8
8
Tension
IKe
4. UO •
Shaking
5.0
5.3
4.0
XLO.O
8.4
>10
8.1
8.6
7.0
7.7
Static
2.7
2.7
2.3
2.3
2.5
2.6
3.1
3.8
2.4
3.1
Filter*
capacity 2
grains/ft.
291
267
300
287
309
327
357
357
308
323
Operating
parameters
Tests S50,
S51 plain
weave Da-
cron, Bag 35
New equi-
librated
After 2 x
10' shakes
Tests S52,
S53 crow
foot Da-
cron bag
72
New equi-
librated
After 2 x
107 shakes
Tests S54,
S55 napped
cotton bag
69
New equi-
librated
After 2 x
107 shakes
Tests S56
through
S59 un-
napped cot-
ton 2 x
107 shakes
Bag 8
Bag 8
Bag 10
Bag 10
365
-------
Table H-2 (Continued). SUMMARY LISTING OF DATA SHEETS FOR TYPICAL
FILTRATION AND MECHANICAL SHAKING TESTS
Test
number
S60
ft
S616
f
S63r
S64
S65
S67
Shaking parameters
Amplitude frequency
in.
1
1
1
1
1
1
cps
8
8
8
8
8
8
Tens ion
Ibs.
Shaking
5.6
4.9
-
6.1
4.2
4.0
Static
2.1
2.2
2.2
1.7
1.3
-
Filter8
capacity 2
grains /ft.
156
313
195
160
158
-
Operating
parameters
Talc aero-
sol
4 grains/
ft. 3
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Napped
cotton
Plain
Dacron
Silica
aerosol
~ 1.5
grains/
ft.3
No equi-
libration,
no useful
data
Filter capacity x 0.0021 = Ibs. dust dislodged by cleaning action. Fly
ash deposition rate per 30 min. cycle at 3.5 grains/ft.3 = 316 grains/
ft. for 10 ft. x 6 in. '
Talc deposition rate per 20 min. cycle at 4.0 grains/ft.3 - 240 grains/
ft. .
Filtration velocity 3.0 ft./min. unless otherwise specified.
Bag dimensions 10 ft. x 6 in. unless otherwise specified.
/•*
Filtration velocity 5 ft./min.
Variable filtration time (and cloth loadings).
e 3
Inlet concentration = 5.2 grains.ft. .
Inlet concentration = 3.2 grains.ft. .
366
-------
Dust Data:
Pressure Drop:
Resulting Tension:
Tension at the top of the bag
during shaking averaged over
complete shaker cycles, usually
with tension variation from start
to finish of the interval (pounds).
Grains of dust removed in each in-
terval; cumulative grams of dust
removed in all intervals. Cumula-
tive dust removed, expressed as a
percent of the total dust initially
on the bag including both the pre-
vious residual and the loaded dust.
(Total given in last column.)
The differential pressure across
the bag, obtained by operating the
fan at the same air flow used in
loading the bag.
Measured at the top of the bag with
vertical shaker arm and without
inflation; "taut" and "loose" the
same definitions as during loading;
"net" • the difference; i.e., the
load applied at the bottom of the
bag; "re-taut" = the tension ob-
tained when the bottom was once
again clamped (pounds).
Supercleaning Data;
The same as the above, during and resulting from a more
intense cleaning, after the above test, in preparation
for the next loading and cleaning test.
Effluent Size Properties
During a filtration period, the B&L light scattering particle counter
was used to count the number of particles penetrating the bag. Data for
a typical test is plotted on Figure H-2. For brief time intervals dur-
ing the filter cycle (0.07 to 0.1 minute) all particles greater than a
pre-set size were counted. For example, 1 minute after filtration
began, during an interval of 0.07 minute, 110 particles larger than
0.5 micrometers were counted by the instrument, using an instrument
sampling rate of 170 cc/min. This is equivalent to a concentration of
367
-------
DATE:
10*
fO
I-'
Lu
V.
Z
H
UJ
O
o
O
u
o
o:
UJ
— .5 5
1.
PARTICLE EMISSION DATA (BBL)
BAG: , FLY ASH,_
5.
PRECONOmONING:
INLET:
TYPICAL PLOT OF ORIGINAL DATA
POINTS, SHOWING MOMENTARY
VARIATION IN APPARENT
CONCENTRATIONS.
GR./FT.3
FPM
•V5 '3 -3
j??5.5\3
1.0
2.0 3,0
FILTRATION TIME
4.0
Figure H-2.
Typical effluent concentrations versus particle size
category and filtration time
368
-------
5 3
2.62 x 10 particles per ft. for particles larger than 0.5 micrometers.
These data are the coordinates for the encircled data point, 1.0 minute
5 3
and 2.62 x 10 particles/ft. , shown on Figure H-2.
Similar measurements were continued over the filtration period, or as
long as particles continued to be emitted. As may be seen from the
resulting plot, the indicated concentrations vary by a factor of 2 to 3
from moment to moment due partially to experimental errors and probably
partially to actual transient variations in concentration. Curves of
best fit have been estimated visually to determine the average instan-
taneous size distribution of particles being emitted. The curves are
5 3
not usually drawn above about 5 x 10 particles/ft, due to a limita-
tion in counting ability of the instrument at high concentrations.
The curves of best fit were used to estimate mass concentrations emitted
as a function of time as reported in the text of the report.
Three groups of data curves are depicted:
• Amplitude-frequency tests, with unnapped cotton sateen
and fly ash
• Life tests, using six fabrics and fly ash
• Talc tests, using three fabrics
A listing of several curves similar to those of Figure H-2, along with
descriptive operating parameters, is given in Table H-3.
369
-------
Table H-3. SUMMARY LISTING OF DATA SHEETS FOR EFFLUENT SIZE AND
CONCENTRATION PROPERTIES FOR VARIOUS DUSTS AND FABRICS
Test
number
Al
A3
A3
A4
A5
A6
A7
A8
A9
A10
Bl
B2
B3
B4
B5
B6
Bag
number
-
-
-
-
-
-
-
-
-
-
8
10
8
10
8
10
Fabric
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Unnapped
cotton
Dust
Atmos.
Fly ash
Fly ash
Fly ash
Atmos .
Fly ash
Atmos.
Fly ash
Atmos .
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Shaking cycle
Freq.
cps
7.15
7.15
7.8
7.8
7.5
7.5
4.3
4.3
11.3
11.3
8.0
8.0
8.0
8.0
8.0
8.0
Amplitude
inches
1.0
1.0
1.0
1.0
2.0
2.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Bag life
Cumulattve
number of
shakes
Tests Al
through
/A10 < 104
shakes
5.8 x 106
5.8 x 106
10.4 x 106
10.4 x 106
16.0 x 106
16.0 x 106
370
-------
Table H-3 (Continued). SUMMARY LISTING OF DATA SHEETS FOR EFFLUENT SIZE
AND CONCENTRATION PROPERTIES FOR VARIOUS DUSTS
AND FABRICS
Test
number
B7
B8
B9
BIO
Bll
B12
B13
B14
C-l
C-2
C-3
Bag
number
8
10
-
-
-
-
-
-
-
-
-
Fabric
Unnapped
cotton
Unnapped
cotton
Napped
cotton
Plain
Dacron
Crow foot
Dacron
Napped
cotton
Plain
Dacron
Crow foot
Dacron
Unnapped
cotton
Napped
cotton
Plain
Dacron
Dust
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Talc
Talc
Talc
Shaking cycle
Freq.
cps
8.0
8.0
8.0
8.0
8.0
8.0
8.0
8.0
8.0
8.0
8.0
Amplitude
inches
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Bag life
Cumulative
number of
shakes
20 x 106
6
20 x 10
< io4
4
< 10*
4
< 10
6
20 x 10
6
20 x 10
6
20 x 10
4
< 10*
4
< 10
4
< 10*
371
-------
APP HJDIX I
THEORY FOR LATERAL FORCES IN SHAKING
The top of the suspended bag is assumed to have an instantaneous force
diagram during shaking such as that sketched below:
where F - the force at right angles to the shaker arm;
T * the tension in the bag;
0 • the angle of the shaker arm to the vertical;
a • the angle of the bag, and also of force T to the vertical.
373
-------
Thus
F «• Tsin (a + 0) (1-1)
The angle 6 is controlled by the shaking mechanism:
6 - sin"1 ^- sin wt (1-2)
where A is the maximum displacement of the arm and w » 2nf, the shaker
frequency.
It is assumed that the pattern near the top of the bag is mainly that of
the downward wave, y - A sin (wt + £ - kx), where 3 allows for the fact
that the bag is usually being pulled at an angle when the arm is vertical,
and k = 2n/X. Then since
dy
tan a = dx at x * '
a - tan"1 [ kA cos (wt + p)] (1-3)
Equations (1-2) and (1-3) permit an evaluation of the lateral force F
from Equation (1-1). Equation (1-1) can be simplified under additional
assumptions:
^
• that in typical shaking Q = r— sin wt, since A/R. is small,
R, A
A
usually ~ 0.17 in the present equipment;
• that a » kA cos (wt + £), since kA is small, usually 0.2 or
less in the present study.
374
-------
Hence,
F/T « sin I ~ sin wt + kA cos (wt + £) } (1-4)
VRA /
Introducing another angle -y • tan kRA, for convenience, Equation (1-4)
can be rewritten:
£ - sin( ^ Jl + (kRA)2 sin (wt + p + Y) I (1-5)
*\.
j
which shows that the instantaneous force ratio is periodic, with ampli-
A / 2
tude equal to sin — / 1 + (kR^) and frequency w.
A A A
Further, since (kRi)"1 « 1 in most cases, and since sin (r—) « r-
RA RA
in most cases,
r: RJ :r— sin (wt + p 4- -y) (1"^)
For estimating the bag power consumption, the peak value of F is needed.
It depends on the variation of T with time. T is approximately equal to
T' (the average tension indicated by the load cell) because the angles a
and 6 are small. Oscilloscope observations indicate that
MA2
T % T' + §£- sin (2 wt + 6) (1-7)
where 6 is another phase angle. Thus from the largest possible value of
T and the largest possible value of F/T, the largest possible value of F
is finally computed to be
375
-------
(1-8)
For example, if the shaking tension average is 5 Ibs., A and RA are 1 inch
and 6 inches, respectively, and M is 16.5 Ibs./in., Fmax is computed to be
0.95 Ibs. If the shaking amplitude is increased to 2 inches, Fmax is
estimated to become 2.6 Ibs., and at 3 inches (many of the assumptions
are weak), 5.6 Ibs. The power consumption is expected to increase in the
same proportions.
The following table lists the maximum lateral forces and average shaking
tensions observed for a variety of bags, amplitudes, and frequencies.
Comparison of their ratio with that predicted by Equation (1-8) indicates
that the actual ratios are of the same magnitude as predicted, but about
50 percent higher. This suggests that the empirical formula be used to
estimate F:
F 1.5 T
max m
376
-------
Table 1-1. COMPARISON OF MEASURED AND PREDICTED LATERAL FORCES
Bag type
5 oz. Dacron
10 oz. plain
Dacron
Unnap. cotton,
10' x 6", new
Same, 5' x 6"
Same, 10' x 4"
f
(cps)
4.9
8.3
6.8
5.2
6.5
8.6
10.0
3.0
8.9
3.5
7.0
10.1
8.0
7.8
8.0
8.1
8.1
A
(in.)
1
1
1
1
1
1
1
1/2
1/2
2
2
2
1/4
1/2
1
1/4
1
x
AVG
(Ibs.)
5.10
9.42
4.30
4.14
4.77
5.65
7.97
2.62
6.21
7.33
9.72
9.90
1.58
2.62
3.42
2.22
3.82
FMAX
(Ibs.)
1.11
5.04
1.25
1.16
1.38
1.54
2.23
0.32
0.74
6.22
3.71
6.61
0.21
0.48
1.58
0.19
0.69
F/T
0.22
0.54
0.29
0.28
0.29
0.27
0.28
0.12
0.12
0.85
0.38
0.67
0.13
0.18
0.46
0.09
0.18
F/T,
Eq. (1-8)
0.27
0.22
0.19
0,19
0.19
0.19
0.18
0.09
0.09
0.46
0.43
0.43
0.04
0.09
0.23
0.04
0.18
F/T,
Eq. (1-9)
0.40
0.33
0.28
0.29
0.29
0.28
0.27
0.13
0.13
0.69
0.74
0.64
0.07
0.14
0.35
0.06
0.28
LJ
-------
APPENDIX J
SYSTEM PRESSURE DIFFERENTIAL VERSUS SINGLE
ELEMENT PRESSURE DIFFERENTIAL
The problem of determining the overall pressure differential flow cha-
racteristics for a multicompartment baghouse has challenged investigators
for some time. Even though the terminal resistance of each compartment
(determined by the length of filtration period used) and the residual
drag of each compartment (a function of cleaning) are completely under
the operator's control, the sequential cleaning of several compartments
means that the resistance of each compartment is different. Conse-
quently, the filter velocity is at all times different in each compart-
ment. Although the pressure differential across every compartment is
essentially the same because they are connected in parallel, this pres-
sure differential fluctuates between cleanings and is as hard to pre-
dict as filter velocity.
Typical variations in filtering velocity, pressure differential, and
drag (or resistance) for a multicompartment baghouse are sketched in
Figure J-l. Since the term T refers to the total time for N compartments
to be cleaned, T/N indicates the time interval between the cleaning of
any two compartments. It is customary to assume on the basis of exper-
imental evidence, that the relationship of drag to dust deposited is
linear. This is acceptable for low air/cloth ratios for which cake fil-
tration is typical. For deep-surface fabrics or high air/cloth ratios
in which cake type filtration may not be well established before clean-
ing, the assumption may not be acceptable. It is also customary to assume
that the inlet dust concentration and the specific resistance coeffi-
cient of the dust are constant at all times and at all locations in the
baghouse.
379
-------
INDIVIDUAL COMPARTMENTS
OVERALL BA6HOUSE
K
UJ
llh
COMR
I
(T/N) H
u
o
UJ
oe
l>
u
o
UJ
UJ
§
•(T/N)-
UJ
o
en
55
u
oc
ui
i
UJ
V)
I
o
i
Ul
o
<
oe
ui
0 o
u z
P
T1M£(t)
TIME (I)
figure J-l. Operating variables in a multicompartment baghouse.
cleaning Is assumed.)
(Five compartments, negligible time for
-------
In one of the first attempts at predicting the overall air flow resis-
tance characteristics of a multicompartment baghouse, Robinson,
Harrington, and Spaite* showed how the overall characteristics could
be computed via a rather detailed process. Later Solbach developed
an analytical process that was simplified by the use of nomograms.
For the special case of a baghouse operating with constant pressure
differential and varying flow, he showed that the filtration velocity
in a single compartment was given as a function of time by
V
2 K C,
Ap
Ap
M^y
_2 U/
1/2
(J-D
where
K
Ci
t
Ap
specific resistance coefficient of the dust
inlet dust loading
filtration time following cleaning
pressure differential (constant)
drag of the filter at time zero. This is the
effective residual drag. The author assumed
that drag increases linearly with amount of
dust deposited, as is customary
Solbach then showed that the filter, velocity averaged for the entire
baghouse over a full operating cycle of duration t' was given by
Ap
K C,
{
2 K C± t'
Ap"
1/2
1/2
(J-2)
He might have simplified this relationship considerably since Ap was
already assumed to be. constant. By equating S to (Ap/V) and re-
arranging terms,
Robinson, J.W., R.E. Harrington, and P.W. Spaite. A New Method of
Analysis of Multicompartmented Fabric Filtration. Taft Engineering
Center, U.S. Public Health Service; Cincinnati, Ohio, paper for pre-
sentation at A.I.Ch.E. 58th Annual Meeting, Dallas, Texas, Feb. 1966.
Solback, W., Derivation of a Computational Method for Multichamber
Cloth Filters on the Basis of Experimental Results, Staub (English
Trans.) 29 (1) 28, January 1969.
381
-------
K C T _ S
V + -7T-
Ap
2 K C. t' Ap S
L__ , e
A 2 A 2
Ap Ap
K C1 t' V2 + 2 Sg V = 2 Ap
f • •.
where W_, is the weight of dust deposited over period t/. This equa-
tion states that a parameter rather similar to the average drag,
(Ap/V), is equal to the residual drag plus half the drag increase from
residual to terminal conditions. That is, ba'ghouse drag is approxi-
mately linear with respect to dust input, just as for a single com-
partment .
The following independent analysis results in somewhat similar conclu-
sions without the assumption that the pressure differential across the
baghouse is constant. At any instant of time, the pressure differen-
tial is equal to the instantaneous product of drag filter velocity,
both selected for any portion of the baghouse filter area:
Ap = S V (J-4)
Under the above assumption of linearity, drag is described only at con-
ditions of loading by:
S = S + KW
e
P- - K TT • K C. V (K, C, - constants) (J-5)
dt dt i i
Using the above expression, the instantaneous pressure differential
becomes
382
-------
K ct dt - I Cj / 13T / CJ-6)
It is implied, therefore, that the instantaneous pressure differential
across any filter having a linear S/W characteristic is a simple func-
tion of S or of W. Since 4p ±. the same everywhere, the instantaneous
drag tinea the rate of increase of drag is the same over every square
inch of fabric operating in a baghouse, regardless of Ap, v, K, or c
The average pressure differential over a time period t to t is-
t «; 12*
S
«;
2 S2
s 2.<: 2
dt 2
K c 2 K
In a baghouse with N identical compartments operating over a total
cycle of T minutes, one compartment is cleaned every T/N minutes. It
is convenient to average the pressure over this period because a com-
plete cycle is depicted. The resulting average pressure is
2?
which is the long-term average for the baghouse as well as that for
any individual compartment.
2
Equation (J-2) says that over the time interval T/N, the quantity S
must be the same for every compartment. The only assumptions are
that K and C be constant, that the S/W relationship be linear for each
compartment, the operation be periodic, and the offline time be negli-
gible during cleaning.
Equations (J-7) and(J-8) also apply to all locations in a single com-
partment, or all stations along a single bag. Since every point in
the baghouse has the same average pressure differential, over any
383
-------
time interval t- to t_, every point in the baghouse undergoes the same
increase with respect to the square of the drag. In this case, the re-
2
suiting distribution of S is not predictable, because of varying
cleaning effectiveness along the bags.
In other words, over time interval T/N between cleanings, the various
compartments undergo the following changes:
2
Net Change of S Resulting S
222 2 1/2
1st: S to S. S. -S = £S S, - (S + DS ) '
elle 1 e
22 2 71/9
2nd: Sl to S2 S2 si = ^ S2 "
-------
A
± J V dt = Ct
Since WM = C, I V dt = C, V T,
o
This equation, which may be compared with Solbach's Equation, (J-3),
cited previously, indicates that a slightly different parameter than
his, but one still similar to the average drag, is still determined
simply by the residual drag plus one-half the residual-to-terminal
drag increase. In the present case, however, the equation applies
to any baghouse, with no assumption of constant operating pressure.
If the plant situation requires filtering at a constant flow rate,
typical of many industrial processes, then the left side of Equation
(J-ll) is equal to (Ap/V), which defines the average baghouse drag.
The equation then says that under constant flow, the average bag-
house drag is exactly half way between the residual and terminal drags
between which each compartment is being operated.
Since each compartment acts as a resistance which is linear with re-
spect to dust deposit, and since the baghouse is a passive assemblage
of these resistances, it is perhaps to be expected that the resis-
tance of the overall baghouse will also be linear with respect to de-
posit. Any passive system composed of linear elements has an overall
linear characteristic. This overall linearity is indicated by
Equation (J-ll)
CONCLUSIONS
1. In each of three situations,
• Constant pressure operation (Equation (J-3))
• Varying pressure and varying flow operation (Equation (J-ll))
• Constant flow operation (Equation (J-ll) using (Ap/V))
385
-------
the long-term average resistance of the baghouse is described by the
same simple expression with respect to dust filtered. It is there-
fore probable that every baghouse having linear S/W compartments,
operating smoothly and periodically, has the same linear character-
istic.
2
2. Over any given time interval, the increase of S of every
portion of the fabric is the same.
2
3. This in turn predicts that S will vary in uniform steps from
compartment to compartment at all times, since all compartments are
cleaned periodically to the same residual level.
4. At any instant, the product of S and dS/dt is the same for
every location in the baghouse.
5. By his control of the cleaning process, the baghouse opera-
tor determines the average operating pressure, via Equation (J-9).
6. The average operating pressure determines the average air/
cloth ratio, via Equation (J-ll).
Mathematical Model of a Five-Compartment Baghouse
To test the above predictions, a five-compartment baghouse was modeled
by digital computation to predict the pressure and flow relationships
for the overall baghouse as well as for the separate compartments.
The following parameters were selected and held constant:
Number of compartments: 5 3
Inlet grain loading: 2.32 gr/ft. (exactly 1/3000 Ib./ft. )
Dust resistance coefficient: 10 in. H20/lb./ft. -ft./min.
Load cycle: 100 minutes overall, or 20 minutes between cleanings
Average air/cloth ratio: 3 ft./min.
Just-cleaned compartment drag: 1.0 in./ft./min.
386
-------
Results
1. With constant inlet flow, making computations at 1-minute
intervals, and cleaning one compartment every 20 minutes until the
overall baghouse displayed a steady pressure cycle (equilibrated):
Compartment
Overall baghouse being cleaned
Ap V S Ap V_ S
Before cleaning: 4.800 3 1.600 4.800 2.41 1.992
Just cleaned: 4.162 3 1.387 4.162 4.16 1.000
Time average: 4.484 3 1.495 4.484 ~3 1.495
2. Next, instead of holding flow constant, a fan curve was used
to allow the flow to decrease into the baghouse as the pressure dif-
ferential increased. The inlet flow-pressure characteristic was
linearized using the points:
3.556
5.173
This fan "curve" supplies flow at about 3 ft./min. when the pressure
differential is about 4.48 in.; i.e., the intention was to keep the
average flow the same as above.
Compartment
Overall baghouse being cleaned
Ap V S Ap V_ S
Before cleaning: 4.645 2.883 1.606 4.645 2.32 1.992
Just cleaned: 4.308 3.106 1.388 4.308 4.308 1.000
Time average: 4.484 ~3 1.495 4.484 ~3 1.495
Thus, within a small margin of error, the average pressure differential
does not depend on the slope of a linear fan curve, as long as the
average operating point is the same. The pressure excursions are
smaller when the flow is allowed to adjust to the pressure. Conversely,
387
-------
the flow excursions are greater into the overall baghouse and also
through each bag.
Incidentally, the drags of each of the five compartments were the sane
with both fixed and variable flows:
C 1 C 2 C 3 C 4 C 5
Before cleaning: 1.253 1.473 1.664 1.836 1.99193
After cleaning: 1.253 1.473 1.664 1.836 1.000
After 20 min.: 1.473 1.664 1.836 1.992 1.253 , etc.
2
Note that £S « 0.565 * constant, as predicted
3. Next, instead of holding the increment of computation at 1.0
minute, both 0.5-and 5.0-minute increments were tried:
No. of calcula-
tions per 100- Overall time-average
minute cycle drag and Ap
50 Ap: 4.47670 S: 1.49223
100 4.48383 1.49461
500 4.48950 1.49650
It appears that with more computation steps, the average drag approaches
the theoretical value 1.50000, which is half way between the initial
and terminal drags for any compartment.
Perhaps the biggest practical limitation to the above is the tacit
assumption that during cleaning of one compartment, the other compart-
ments undergo no filtration. Otherwise, the equations would be more
complicated and the results less clear. For equipment which is
cleaned quickly in comparison with the interval between cleanings, and
which does not utilize appreciable reverse flow volume to assist the
cleaning, the indicated approaches are applicable without modification.
388
-------
APPENDIX K
DUST TRANSPORT DURING PULSE CLEANING
In pulse cleaning, as in any filter cleaning operation, collected par-
ticles must not only be dislodged from the filter medium, but they must
be removed from the air flow system so that redeposition on the fabric
is minimized. The time allowed for settling is small in the case of
pulse cleaning, and the transport of particles by the pulse and airstream
during pulsing and the subsequent resumption of filtration may have con-
siderable impact upon the effectiveness of particle removal.
In this analysis, the three transport processes associated with pulse
cleaning, particle ejection, air flow transport, and settling, are
sketched as vectors in Figure K-l. It is assumed that the three pro-
cesses are additive to a first approximation. The slanting lines to the
right of the filter bag indicate the flow lines taken by the reversal
air, first as the air leaves the bag and blows back toward the hopper,
and then as the same air returns to the bag and is refiltered. The
average angle of these lines depends on the length of the bag and the
proximity of the bag either to a solid wall, to a vertical stagnation
line between two bags being pulsed simultaneously, or to a bag that is
still filtering.
Particles of all sizes are probably ejected from the bag horizontally
because the fabric motion is horizontal, and also because the air passes
through the bag essentially horizontally since the vertical air pressure
gradients inside and outside the bag are small. It is hypothesized
that the particles leave the fabric with a velocity greater than that
389
-------
PULSE
FLOW REVERSAL PHASE
FLOW RETURN PHASE
BAG
WALL
BAG
FILTER BAG
Figure K-l.
Transport effects on particle disposition; s: small par-
ticles, M: medium particles, L: large pargicles. Path
o-a: initial momentum due to pulse; path a-b: air
transport during the reversed flow; path b-c: gravity
settling during flow return; path d-e: air transport
during flow return; point e: final deposition point;
path o-e: net change of position on filter bag.
390
-------
of the accompanying air. With an initial horizontal momentum rapidly
reduced by air viscosity, the particles travel a "stopping distance"
to point (a) in Figure K-l, the distance depending on the particle iner-
tia. During this same time and until the flow reverses, all particles
are transported in effect along the air flow lines through the distance
vectors (a-b). This is the distance that the air moves backwards from
the instant the particles are removed. Again during the same time
period, the particles settle due to gravity a distance vector (b-c).
This is determined approximately by the settling velocities in Figure
*
K-2. Thus under the concurrent action for momentum, aerodynamic trans-
port, and settling, the particles reach points (c) by the time the air
stops moving backwards.
Shortly thereafter, the flow reverses and follows approximately the same
flow lines back to the filter. This time, the net result is the sum of
two distance vectors. Until the particles reach the filter they continue
to settle. Because the larger particles have farther to travel back to
the filter, point (c) is farther from the filter for the larger particles.
They settle even farther (c-d) than during the flow reversal phase. The
transport distances (d-e) are also longer for the larger particles. In
fact, many of them will have reached a point (d) so low in the system
that they miss the filter bag, and continue to fall into the hopper.
On the other hand, the smaller particles are seen to be slightly lifted
up the bag by the cleaning process. This may partially account for the
occasional observation of a very heavy accumulation of dust near the top
of the filter bag.
The foregoing discussion is based on potential laminar -flow. Actually,
the flow is to some extent turbulent, and many secondary air currents
There are additional inertial considerations associated with sudden
changes of the direction of the air, but because the air reverses it-
self, these tend to cancel.
391
-------
may be present. Although Figure K-l indicates the average disposition
of particles, turbulence will cause considerable mixing in the flow
region. Some small particles will be carried even higher on the bag
than indicated in Figure K-l.
Reexamtnation of the five distance vectors in Figure K-l shows that as
long as the air returns via the same flow lines, the settling process
depends only on the free-fall terminal velocity of the particles, VT
and on the time available for fall. This time from the instant the par-
ticles leave the felt surface until they return is approximately equal
to the duration of the pulse, T , plus the time required, T_, to return
p K
to the filter surface. The initial momentum of the particles increases
the distance the particle must travel to return to the fabric surface.
The initial momentum also has an adverse effect, Y , on the net vertical
m
distance, because it drives the particles over to flow lines which
terminate at a higher location on the fabric. Consequently, the net
vertical distance traveled is
y = VT (Tp + TR) - Ym (K-l)
The pulse duration, T , is determined by the design and operating cha-
racteristics of the equipment. The time, TR, is given by T -y2",
K.
where V_ is the average return air velocity, and V is the average pulse
velocity, plus the time, T , gained by virtue of the momentum of the
ejected particle. T is given by Y /V, where V is the vertical component
of the return air velocity. To a first approximation V and V_ may be
equated to each other and the normal upward velocity in the dust collec-
tor; i.e., the filtering flow divided by the cross-sectional area carry-
ing dirty air. The adverse effect of the momentum of the particle, Y^,
is given by m cot 6 where m is the length of the path (o-a) in Figure K-l,
and 0 is the angle made by the flow lines with the vertical. Length, m,
the stopping distance of the particle having initial outward velocity,
Vo, is given by
392
-------
_ _ _ fv ^
m j| (K-2)
using the approximation for V_ given in Figure K-2. With these substi-
r
tutions, Equation (K-l) becomes
V.
|
V _ -I* « _ _^y\ > - —
p «
K.
1
VI
(K-3)
The equation indicates an advantage in having steep flowlines in the dust
collector, such as might be obtained by using baffles between bags not
cleaned simultaneously. An advantage in having a slow recovery of flow
immediately following the pulse is also indicated; i.e., low V_. If the
K
dislodged particles are small (low V /V_ ratio), then a long pulse time
JL M\
may be advantageous. Conversely, agglomerates with a sufficiently high
falling velocity should reach the hopper regardless of the pulse time,
as expected. The hypothesized high initial dislodgement velocity is
advantageous in minimizing redeposition of dust on the fabric.
As an example, consider agglomerates having diameters of 10, 30, 100, and
300 microns, with the same specific density of 1.0. The fall velocities
are respectively, 0.011, 0.09, 0.82 and 3.3 ft./sec. Given the same
ejection velocities of 2 ft./sec, by a pulse 0.06 seconds in duration
in a system with an average upward flow velocity of 0.5 ft./sec. and
& * 1/12 (e.g., 4 ft. bags surrounded by an average 4 in. clearance),
the estimated shifts of particles along the bag during a single pulse are
10 : y = 0.011 (0.06 + 0.24 + (0,67) (0.022-1)) = 0.0039 feet upwards
30 : y = 0.09 (0.06 + 0.24 + (0.67) (0.18-1)) = 0.023 feet upwards
100 : y • 0.82 (0.06 + 0.24 + (0.67) (1.64-1)) = 0.60 feet downwards
300 : y = 3.3 (0.06 + 0.24 + (0.67) (6.6-1)) = 15.5 feet (i.e., 1007.)
The theory thus predicts an intermediate agglomerate size which is lifted,
or at least apt to settle less than either smaller or larger particles.
393
-------
1000 c
(0.02) (0.05) (O.I) (02) (0.5) (1.0)
EFFECTIVE
PARTICLE
DENSITY
(g/cm3)
0.1 I.
VELOCITY , ft/we.
Figure K-2.
Particle (or agglomerate) sizes which are supported by up-
ward air flow. (Spherical particles; equivalent to terminal
settling velocities and sizes.)
394
-------
Also, agglomerates must be larger than 100 microns before the transport
process is very effective.
An effectiveness ratio may be defined as y/L where L is the length of
the bag. Since (cot 0) is approximately given by L/d where d is the
average distance between the bag and the flow stagnation wall, the
Equation can be rewritten:
VT
(
•
Short bags, or bags packed close together, would apparently be cleaned
more effectively than longer or more separated bags. Bags packed close
together must be cleaned simultaneously, however.
Particles falling in clusters or clouds may fall faster than when falling
individually, depending on the cloud density. The possible effects of
this phenomena on the operation of pulse cleaned equipment have not
been investigated.
395
-------
APPENDIX L
FABRIC ACCELERATION IN PULSE CLEANING
In this appendix, estimates are made of the maximum acceleration seen by
the filter bag during the cleaning pulse. The derivation is based on the
mechanical properties of the bag and the pulse pressure differential.
In this study, dust was collected on the outside of the bag, with the bag
supported by a cage composed of ten longitudinal rods 1.25 in. apart.
During filtering, the bag draped inward between the rods under an inward
pressure. The slackness of the bag, its stiffness, and the inward pres-
sure determined the amount of drape; i.e., the distance the fabric was
moved radially inward. This was of the order of 0.3 inches from the
fully outward position. In both the inward and the outward positions,
the bag was a geometric cylinder; i.e., it had negligible curvature in
the longitudinal direction.
When the pressure reversed at the beginning of the pulse, the bag
moved from its initial draped configuration outward until taut, where
in stopping it experienced a sharp deceleration. This was larger
than any other deceleration or acceleration seen at any other time
during the cleaning pulse. To a first approximation, the magnitude
of the deceleration depended on the velocity of outward movement and
on the inelasticity of the fabric that stopped its motion. This
will be clarified in the following paragraphs.
397
-------
MECHANICAL PROPERTIES OF THE BAG
One major characteristic of the bag is the static position that it as-
sumes as a function of pressure differential across the fabric. This
geometry was measured using an auxiliary fan and Variac with the results
shown in Figure L-l. Beginning at zero pressure differential it takes
very little pressure to move the bag either inward or outward to the
extent of the slackness. The pressure only has to overcome the frictional
stiffness of the filter. However, after the slackness is removed, the in-
elasticity of the fabric takes over from the stiffness, and the displace-
ment is greatly retarded. In changing from an increasing to a decreasing
pressure differential, hysteresis is evident, due to friction internal to
the fabric and dust matrix. However, because the cleaning process mainly
involves a one-way movement (inflation), only the lower curve for each
bag is relevant.
The inflation curves of Figure L-l are to a first approximation charac-
terized by a steep slope near zero differential pressure, and a nearly
flat slope at fully inflated pressure. These slopes, which are termed G,
the flexibility, and M, the elasticity of the bags, respectively, are
delineated in the following sketch.
398
-------
0184 »• 10
vo
M
06
0.5
0.4
0.3
0.2
O.I
0.0
T
Deflating Inflating
WOOL FELT »~ ««
OACRON FELT
6 6 4 2 - O+ 2 4 6
INFLATED (reverse air) EVACUATED (filtration)
PRESSURE DIFFERENTIAL , inches water
8
IO
Figure L-l. Static displacement of fabric surface at constant differential pressure, used bags
-------
Flexibility, G Elasticity, M
in/in. H20 ft./(lb./ft.2) in/in. H20 ft./(lb./ft.2)
Wool bag
with dust 0.44 0.0071 0.0033 0.000053
Daeron bag
with dust 0.054 0.00087 0.0020 0.000032
The woolen felt with its residual dust is seen to be much more flexible
and slightly more stretchable than the Dacron felt plus dust.
Dynamics of Bag Motion
In moving radially approximately 0.3 inches and draping between rods
1.25 inches apart, the motion of the bag is essentially one-dimensional.
The equation of motion is:
,2
p S-E + o- = A p(t) (L-l)
The first term is the mass per unit area times the radial acceleration of
the fabric. The second term is the force per unit fabric area which is
resisting motion at an instantaneous bag displacement, r.
-------
accelerations. This computed motion was in good general agreement with
the motions detected by laboratory means.
The point-by-point solution method was laborious and not at all suited
to field application. Therefore, a simplified method of estimating the
peak deceleration of the fabric was developed.
Estimating Fabric Deceleration
The fabric can be viewed as a mass traveling radially outward until it is
stopped by a spring. The stopping deceleration depends on the maximum
velocity acquired in the outward travel. There are several ways of
estimating the maximum velocity, depending on the time taken for the
fabric to reach maximum velocity compared with the pulse rise time. Most
of the fabric's velocity is acquired during the flip-flop, or steep por-
tion of the curve describable by:
p M + G " A P(t)
This has a characteristic oscillation period 2it I Gp seconds, which
means that full velocity is characteristically developed from a dead
start in about 1.6 ^Gp seconds. If this characteristic time is shorter
than or similar to the time required for the pulse to attain maximum
pressure differential, the fabric is either fairly light or flexible and
tends to move with the pulse. If the characteristic time is much longer
than the pulse rise time, the fabric acts sluggish and cleaning may &e
poor until the pulse is lengthened.
STIFF, LIGHT FABRICS
The filter bags tested in this study appeared to be of the type, such
that during the outward travel the position of the fabric was mainly
401
-------
dependent on the stiffness of the fabric rather than on its mass. That
is, the second term of the dynamics equation was much larger than the
first term. Using the Linearized equation (L-2) above, the maximum fabric
velocity will be:
For example, a linear pulse with a rise rate of 1000 in. H20/sec. acting
on the Dacron bag described above, would produce a maximum fabric veloc-
ity of about 54 in. /sec.
At this peak velocity, the bag will move its 'full '0.3 inches in roughly
0.011 seconds. Since the pulse travels the full length of the bag in
about 0.004 seconds (4 feet, at sonic velocity) the entire length of the
bag must inflate almost simultaneously. Much stiff er bags will inflate
as a cylinder, with no difference from top to bottom. Bags having a
faster inflation velocity (minimal stiffness) will inflate in ripple
fashion, with the ripple traveling from top to bottom at approximately
sonic velocity.
Because the Dacron bag is estimated to develop a velocity of 54 in, /sec.
after about 0.0055 seconds, the associated ^acceleration ts of the order
2
of 82.0 ft. /sec. (2.6 g's). Because the weight per unit area was 0.225
2
Ib./ft, for this bag, the first term in the equation of dynamics is only
about 1 in. 1^0 compared with values of the second term of several in.
HoO. This justifies the earlier statement that the Dacron bag behaved
like a stiff, light fabric.
Flexible, Heavy Fabrics
The inflating motion of flexible, heavy fabrics is resisted by the inertia
of the fabric rather than the stiffness. The second term in the equation
402
-------
of dynamics is negligible compared with the first term during develop-
ment of the maximum outward velocity. Consequently,
max
* - (A P T- a rr ) , for A p = A p - at (L-4)
P y o i. i o
describes the velocity obtained by the fabric under a linear pulse after
T seconds. T is limited either by the length of the pulse or by the
fabric reaching the limit of its slackness, whichever happens first.
Moderately Stiff. Moderately Heavy Fabrics
In case neither the first nor the second term of the dynamics equation is
negligible compared to tphe other one, both must be included in the solu-
tion. The most reliable solution is the numerical one already mentioned,
in which point-by-point values of acceleration, velocity, and displace-
ment are computed. A simpler approximation is the piece-wise linear
method, in which the three straight parts of the characteristic curve
are represented by analytic functions.
For the special case in which the characteristic time of the fabric mo-
tion is similar to the pulse rise time, the fabric may be expected to
travel about one half its full displacement in the characteristic time,
giving a rough estimate of the maximum velocity attained of
V - *-f= (L-5)
max 3>2
where s is the distance of travel, approximately 0.3 inches in this case,
403
-------
With an estimate of the maximum outward velocity attained by the fabric
during the beginning of the pulse, it is possible to estimate the de-
celeration of the fabric as it stops in the fully outward position. The
fabric stops when its kinetic energy is fully converted to elastic energy:
1/2 pV2 = 1/2 M
-------
the fabric. In simple harmonic motion, the maximum acceleration occurs
at the end of the stroke, when the velocity is zero. Differentiation
of equation (L-l) and setting the acceleration term equal to zero, re-
sults in:
"""'dt3
In other words, the deceleration is maximum when V = M dA p/dt, ap-
proximately. Using the numerical example above, at the instant that the
deceleration is 279 g's, the velocity is estimated to be 0.167 ft. /sec.
or 10 ft./min. The maximum velocity when the acceleration was zero or
at least too small to remove much dust was 4.5 ft. /sec. Thus, all the
dust is estimated to leave the fabric at velocities between 4.5 and
0.167 ft. /sec. depending on the levels of acceleration necessary to
cause separation. These ejection velocities are referenced to the col-
lector cage, not to the bag which is moving at similar but decelerating
velocities.
In addition to these ejection velocities, reverse air is moving through
the filter at approximately 0.5 ft. /sec. Thus Stokes drag will tend to
change the ejection velocities ultimately to 0.5 ft. /sec.
>-
o
tu
>
2
100 200 300
ACCELERATION, «•«
405
-------
APPENDIX M
SUPPLEMENTARY DATA ON PRECISION OF MEASUREMENT TECHNIQUES
ANDERSEN "OUTSTACK" IMPACTOR
The Andersen impactor design for sampling ambient atmospheres (Particle
Fractionating Sampler) was used for assessing inlet and outlet dust
concentrations during filtration studies. Small duct diameters as well
as the need to collect as much dust as possible precluded use of the
"in-stack" model. Current use of the "outstack" device on another test
program provides some useful statistics on the reliability of this device.
Figure M-l indicates that the results of parallel sampling of a resus-
pended foundry dust (MMD ~ 1 fim) by filters and impactors were in fair
agreement.
Deviations from the regression line are attributed to errors introduced
by sample transfer from probes, weighing errors, and fluctuations in
sampling flow. Practically speaking, these results probably typify the
expected field performance for these units.
Two sets of impactor stages, consisting of Petri dishes lubricated with
a petroleum jelly and weighing about 20 to 21 grams were allowed to stand
in a dessicator for 3 days between successive weighings. No special
precautions other than routine care were exercised in the weighings.
The reported weight differences shown in Table L-l are based upon the
original sample weights.
407
-------
REGRESSION LINE
= O.OI84 + 0.92631
SLOPE = 0.9263 II 1.95%
AT 95% CONFIDENCE LEVEL
O.I 0.2 0-3 0.4 0-5
OUST LOADING, ANDERSEN IMPACTOR , g/m3
Figure M-l. Comparative mass loading measurements with all-glass filters
and Andersen inspector (foundry dust, HMD » «• 1
408
-------
Table M-l. WEIGHT LOSSES FOR LOADED LUBRICATED, PETRI DISH
IMPACTOR STAGES AFTER REPEATED WEIGHINGS
Weight loss milligrams
Stage
1
2
3
4
5
6
set 1
0.3
0.1
0.2
0.0
0.1
0.3
Set 2
0.1
0.2
0.2
0.3
0.2
0.5
Since the estimated standard deviation for the above 12 measurements was
0.13 rag. stage weights should probably be within + 0.26 mg. of the true
weight at least 95 percent of the time.
409
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-650/2-75-009
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Fabric Filter Cleaning Studies
5. REPORT DATE
January 1975
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Richard Dennis and John Wilder
8. PERFORMING ORGANIZATION REPORT NO.
GCA-TR-74-6-G
9. PERFORMING ORGANIZATION NAME AND ADDRESS
GCA Technology Division
Burlington Road
Bedford, MA 01730
10. PROGRAM ELEMENT NO.
1AB012; ROAP 21ADJ-049
11. CONTRACT/GRANT NO.
68-02-0268
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
NERC-RTP, Control Systems Laboratory
Research Triangle Park, NC 27711
:. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
is. ABSTRACTThe pgpQpj gives results of a detailed study of fabric filter cleaning mechan-
isms. A highly instrumented, pilot plant system was built to operate as a single- or
multiple-bag unit for the investigation of cleaning by mechanical shaking, pulse jet
air, and reverse flow air. Four woven bag types (cotton and Dacron) and two felt bag
types (wool and Dacron) were evaluated with resuspended fly ash and talc dusts.
Analysis of cleaning by both mechanical shaking and pulse jet air indicated that the
tensile forces generated by bag acceleration were the main cause of dust removal;
aerodynamic re-entrainment played only a minor role. Residual fabric drag, fabric
holding capacity, and dust penetration characteristics were predictable, based on
such cleaning parameters as shaking frequency, amplitude, pulse jet pressure, and
rate of pressure rise. Based on inlet concentration of 3-10 gr/cu ft, effluent concen-
tration for mechanically shaken, woven fabrics ranged from 10 to the minus 7th
power to 0.001 gr/cu ft, in contrast to 0.001-0.01 gr/cu ft for felted media cleaned
by pulse jet air. Effluent concentrations for both systems decreased significantly as
filtration progressed. Caution should be exercised before extrapolating test results
to dust/fabric combinations other than those investigated, until more data is
available.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air Pollution
Air Filters
Cleaning
Dust
Cotton Fabrics
Wool
Woven Fabrics
Felts
Fly Ash
Talc
Polyester Fibers Kinetics
Air Pollution Control
Stationary Sources
Fabric Filters
Mechanical Shaking
Pulse Jet Air
Reverse Flow Air
13B
13K
13H
11G, 21B
HE, 08G
20K
T8. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
438
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
411
------- |