-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S.
Environmental Protection Agency, have been grouped into seven series.
These seven broad categories were established to facilitate further
development and application of environmental technology. Elimination
of traditional grouping was consciously planned to foster technology
transfer and a maximum interface in related fields. The seven series
are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
This report has been assigned to the INTERAGENCY ENERGY-ENVIRONMENT
RESEARCH AND DEVELOPMENT series. Reports in this series result from
the effort funded under the 17-agehcy Federal Energy/Environment
Research and Development Program. These studies relate to EPA's
mission to protect the public health and welfare from adverse effects
of pollutants associated with energy systems. The goal of the Program
is to assure the rapid development of domestic energy supplies in an
environmentally—compatible manner by providing the necessary
environmental data and control technology. Investigations include
analyses of the transport of energy-related pollutants and their health
and ecological effects; assessments of, and development of, control
technologies for energy systems; and integrated assessments of a wide
range of energy-related environmental issues.
REVIEW NOTICE
This report has been reviewed by the participating Federal
Agencies, and approved for publication. Approval does not
signify that the contents necessarily reflect the views and
policies of the Government, nor does mention of trade names
or commercial products constitute endorsement or recommen-
dation for use.
This document is available to the public through the National Technical
Information Service, Springfield, Virginia 22161.
-------
EPA-600/7-77-084
August 1977
FILTRATION MODEL FOR COAL
FLY ASH WITH GLASS FABRICS
by
Richard Dennis, R.W. Cass, D.W. Cooper, R.R. Hall,
Vladimir Hampl, HA Klemm, J.E. Langley, and R.W. Stern
GCA Corporation
GCA/Technology Division
Bedford, Massachusetts 01730
Contract No. 68-02-1438
Task No. 5
Program Element No. EHE624
EPA Project Officer: James H. Turner
Industrial Environmental Research Laboratory
Office of Energy, Minerals, and Industry
Research Triangle Park, N.C. 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, D.C. 20460
-------
ABSTRACT
A new mathematical model for predicting the performance of woven glass
filters with coal fly ash aerosols from utility boilers is described in
this report. The data base for this development included an extensive
bench and pilot scale laboratory program in which several dust/fabric
combinations were investigated; field data from three prior GGA studies
involving coal fly ash filtratiori with glass fabrics; past GCA studies
of fabric filter cleaning mechanisms and a broad-based literature survey.
Trial applications of the modeling technique to field filter systems
operating at Sunbury, Pennsylvania and Nucla, Colorado indicate excellent
agreement between theory and practice for both penetration and resistance
characteristics. The introduction and experimental confirmation of two
basic concepts were instrumental in model design. The first relates to
the manner in which dust dislodges from a fabric and its subsequent im-
pact upon resistance and penetration in a multichambered system. The
second concept is associated with the relatively large fractions of fly
ash that pass with minimal collection through temporarily or permanently
unblocked pores or pinholes such that observed particle penetrations are
essentially independent of size. Additionally, the quantitation of the
cleaning action with dust removal in terms of method, intensity and dur-
ation of cleaning was essential to the overall modeling process. The
examination of specific resistance coefficient, K2, for the dust layer
in the light of polydispersed rather than monodispersed particle compo-
nents provided improved estimates of K~ although direct measurement of
this parameter and other terms defining the filter resistance (or drag)
versus fabric loading relationship is the recommended approach at this
time.
iii
-------
IV
-------
CONTENTS
Abstract
List of Figures
List of Tables
Acknowledgments
Nomenclature
Special Nomenclature - English and Metric Equivalencies
for Key Filtration Parameters
Sections
I Summary
II Introduction
Description of a Filtration System
Objectives
Outline of Model
Summary of Methodology
The Laboratory Program
III A Review of Fabric Filtration Models
Predictive Models
Robinson, Harrington and Spaite Model
Solbach Model
Dennis and Wilder Model
Page
iii
xii
xxii
xxvii
xxviii
xxx iv
1
5
7
7
9
10
12
14
14
14
18
20
-------
CONTENTS (Continued)
Sections Page.
Noll, Davis and Shelton Model 22
Noll, Davis and LaRosa (1975) Model 23
Stinessen's Approach 24
Fraser and Foley Model 25
Leith and First Model 25
Conclusions 27
IV Laboratory Test Equipment and Measurement Procedures for 29
Determination of Filter Performance
Bench Scale Filtration Equipment 29
Dust Generation Apparatus 35
Pilot Scale Filtration Equipment 35
Test Aerosols 42
Particulate Sampling and Assessment 45
Basic Sampling Equipment 45
Assessment and Interpretation of CNC and B&L 48
Measurements
Tensile Properties 53
V Fabric Structure Studies 57
Introduction 57
Basic Manufacturer or User Specifications 58
Bag Resistance Versus Pore Velocity 61
Simplified Weave Representations 62
Pore Properties g5
Yarn Shape j-.
VI
-------
CONTENTS (Continued)
Sections Page
Pore Type and Area 72
Air Flow Through Pores 76
Physical Properties of Fabrics 80
Tensile Modulus 83
Bag Tension and Permeability 87
Fabric Thickness 92
Initial Dust Deposition Characteristics 92
VI Analysis of Sunbury and Nucla Field Measurements 100
Fabric Dust Loadings 100
Bag Resistance 103
Collection Efficiency • 109
Specific Resistance Coefficient 115
VII Bench Scale Laboratory Tests 117
Fabric Resistance Characteristics 117
Clean (Unused) Fabrics 117
Cleaned (Used) Fabrics 117
Resistance Versus Fabric Loading-Bench Scale Tests 126
Dust Deposition and Removal Characteristics 129
Deposition on Used Fabrics 129
Pinholes and Air Leakage 135
Fabric Appearance After Cleaning 144
Dust Release From Glass Fabrics 154
Filtration With Partially Cleaned Fabrics 157
vii
-------
CONTENTS (Continued)
Page
Sections
Specific Resistance Coefficient L
Effect of Velocity 16°
Effect of Particle Size 163
Dacron Filtration Tests
Collection Efficiency and Penetration 167
Weight Collection Efficiency Measurements 167
Condensation Nuclei Measurements 174
Particle Size and Concentration by Optical Counter 186
Nuclei Versus Mass Concentrations 190
Effluent Concentrations Versus Face Velocity 198
Rating Fabrics With Atmospheric Dust 198
VIII Pilot Plant Tests 205
Introduction 205
Summary of Testing Procedures 205
General Comments 206
Dust Removal Versus Fabric Loading 208
Dust Removal With Successive Filtration and 217
Cleaning Cycles
Dust Removal and Bag Tension 217
Resistance Versus Fabric Loading 219
Dust Penetration Measurements 219
Constant Velocity Tests 21g
Penetration Versus Face Velocity 223
Rear Face Slough-Off
^ ^ J
viii
-------
CONTENTS (Continued)
Sections Page
IX Prediction of Fabric Filter Drag 228
Critique of Linear Drag Model 229
Derivation of Nonlinear (Pore) Model 229
Verification of Nonlinear Drag Model 239
Empirical Correlations 241
Clean Fabric Drag, S 224
Effective Drag, S_ 244
£j
Residual Drag, S 249
R
Initial Slope, K^ 249
Estimation of W* 251
Theoretical Correlations 251
Clean Fabric Permeability - 251
Specific Resistance Coefficient, KZ 252
K Versus Face Velocity 254
K Versus Specific Surface Parameter 260
K Versus Dust Cake Porosity 261
Calculated and Observed K Values, Field and 261
Laboratory Tests
Fabric Cleaning and Filter Performance 270
Resistance (Drag) Versus Dust Distribution on 271
Fabric
Dust Removal Versus Cleaning Conditions 288
Full Scale Applications - Modeling Concepts 306
Pressure Controlled Cleaning 306
IX
-------
CONTENTS (Continued)
Sections
311
Time Cycle Cleaning
315
X Prediction of Fabric Filter Penetration
O -I (1
Particle Capture by Unobstructed Pores
320
Particle Capture by Bulk Fiber Substrates
Particle Capture by Dust Cake (Granular Bed) 327
Fly Ash Penetration for Woven Glass Fabrics 333
Penetration Versus Pore Properties 333
Penetration and Inlet Concentration 336
Penetration Versus Face Velocity 337
Dust Penetration Model 341
XI Mathematical Model for a Fabric Filter System 347
introduction 347
Principal Modeling Relationships 348
Designed Model Capability 352
Basic Modeling Process 353
Program Description
Computational Procedures
Drag Computation
Fabric Penetration
Program Input and Output
Predictive Validation
Introduction
•J / J
System Parameters
374
-------
CONTENTS (Continued)
Sections
Nucla Data Inputs
Sunbury Data Inputs
Nucla Plant - Model Validation
Predicted Versus Actual Resistance
Characteristics
Predicted Velocity Relationships
Predicted Penetration
Sunbury Plant - Model Validation
Predicted Versus Actual Resistance
Characteristics
Predicted Velocity Relationships
Predicted Penetration
Summary of Model Highlights and Direction for
Future Work
References
APPENDIX
A
Effect of Sequential Pore Closure on Shape of Resistance
Versus Fabric Loading Curve
B Input Parameters For Estimating Fiber Efficiency in
Substrate Layer
C Determination of Constants Used in Dust Penetration Model
D Baghouse Computer Program Description
Page
374
381
386
386
390
390
394
394
399
402
406
409
412
414
416
423
XI
-------
LIST OF FIGURES
Page
No. —*-
Q
1 Schematic of n-Compartment Baghouse
2 Flow Chart for Baghouse Model
3 Three-Compartment Baghouse With Sequential Cleaning 15
4 Fabric Drag Versus Loading
5 Schematic of Filter Test Assembly With Exploded View of 31
Fabric Sandwich
6 Bench Scale Filtration Apparatus Showing Inlet Manifold 32
and Test Aerosol Loop
7 Bench Scale Filtration Apparatus 32
8 Size Distribution Measurements for GCA Fly Ash for 33
Particle Density of 2 grams/cm^
9 NBS Type Dust Generator 36
10 All Components of Bench Scale Filtration System 37
11 Schematic of Pilot Scale Fabric Filter System 39
12 Inlet and Outlet Fly Ash Size Distributions for a 41
10 ft x 4 in. Woven Glass Bag, Sunbury Fabric
13 Size Distribution for GCA Fly Ash Entering Bench Scale 43
Filter System, Andersen In-Stack Impactor Measurements
14 Mass Distribution for Sunbury Inlet Aerosol, Field Measure- 44
ments, Based Upon Aerodynamic Diameter. In-Stack Andersen
Impactor
15 Size Properties for Coarse and Fine Rhyolite Dust 46
16 Particle Size Properties for Lignite Ash From Precipitator 47
Hopper
xii
-------
LIST OF FIGURES (Continued)
No. Page
17 Number Size Distributions for Background (Laboratory) Dust 50
Based on B&L Counter Measurements
18 Relationship Between Nuclei Concentrations by CNC Measure- 51
ments and Weight Concentrations Derived From B&L Data
19 Concurrent Measurements for Nuclei Concentrations (CNC) 54
and Particle Concentrations by Bausch and Lomb Counter in
Different Size Ranges
20 Test Apparatus for Measurement of Fabric Tensile Properties 55
21 Resistance Versus Face and Maximum Pore Velocity for Clean 63
(Unused) Sunbury Glass Fabric
22 Textile Schematic Drawing of Sunbury Fabrics A. 1973 Bags, 64
B. 1975 Bags. Circles on Diagonal, Warp Yarn Crossovers,
Indicate Open Pore Locations
23 Schematic of Sunbury Fabric, Filtering Face, 3/1 Twill, 66
Left-hand Diagonal Indicating Pore Locations and Average
Dimensions. No Space Between Warp Yarns Except at Crossing
Points
24 Warp and Fill Surfaces of Clean (Unused) Sunbury Fabric 67
With Substage Illumination (20X Mag)
25 Warp and Fill Surfaces of Clean (Unused) Nucla Fabric With 68
Substage Illumination (20X Mag)
26 Individual Sunbury Warp and Fill Yarns as Seen in Plane of 69
Fabric Showing Maximum and Minimum Dimensions (20X Mag)
27 Individual Nucla Warp and Fill Yarns as Seen in Plane of 70
Fabric Showing Maximum and Minimum Dimensions (20X Mag)
28 Schematic Drawing Showing Alignment, Approximate Form, and 73
Spacing of Yarns and Pores in Sunbury Filter Bags (Menardi
Southern Woven Glass Media)
29 Edge Views of Clean Sunbury Fabric (20X Mag) 74
30 Schematic Drawing Showing Idealized Alignment of Parallel 75
Yarns and Maximum Pore Cross Section (Shaded Area)
xiii
-------
LIST OF FIGURES (Continued)
Page
No. —&-
31 Stress/Strain Relationship for Used Sunbury Media 84
7.6 cm x 45.7 cm (3 in. x 18 in.) Strip With Tension
Applied in Warp Direction
32 Effect of Dust Loading on Tensile Properties of Woven 89
Glass Bags
33 Effect of Bag Tension on Resistance to Airflow, With 91
Conventional Bag Suspension
34 Fabric Thickness Versus Compressive Loading 93
35 Schematic of GCA Fly Ash Filtration at 2 ft/min. Dark 94
Areas Show Dust Deposits. Light Areas Indicate Relatively
Clean Warp Yarns Transmitting Light With Rear Face
Illumination
36 Fly Ash Deposition on Monofilament Screen Versus Filtration 96
Time, Surface Illumination
37 Fly Ash Deposition of Monofilament Screen Versus Filtration 97
Time, Rear and Surface Illumination
38 Fly Ash Deposition on Monofilament Screen Versus Filtration 98
Time, Rear and Surface Illumination
39 Residual Dust Loadings for Bags in 14-Compartment Sunbury 102
Collector. Cycle Interrupted Between Cleaning of Compart-
ments 12 and 13 for Removal and Replacement of All Bags
40 Average Filter Resistance for Sunbury Glass Bags, Normal 104
Field Use After 2 Years Service
41 Filter Resistance and Outlet Concentration Versus Time for 105
Glass Bag Filters at Sunbury, Pennsylvania Power Plant
42 Resistance Versus Time for Old and New Sunbury Bags 106
43 Filter Resistance Versus Time for Successive Filtering, 108
Compartment Cleaning and Reverse Flow Manifold Flushing,
Sunbury Field Test of February 14, 1975
44 Inlet and Outlet Dust Concentrations for Sunbury Field Tests 112
45 Inlet and Outlet Dust Concentrations for Nucla Field Tests 113
xiv
-------
LIST OF FIGURES (Continued)
No. Page
46 Field Measurements of Outlet Concentrations From New and 114
Well-Used Sunbury Bags
47 Filtration Resistance for Unused Sunbury and Nucla Glass 118
Bags, Laboratory Measurements
48 Resistance Characteristics of Used Sunbury Fabrics Cleaned 120
in the Laboratory to Various Residual Dust Holdings, Tests
1 to 5
49 Resistance Characteristics of Used Sunbury Fabrics Cleaned 121
in the Laboratory to Various Residual Dust Holdings, Tests
6 to 8
50 Resistance Characteristics of Used Nucla Fabrics Cleaned in 122
the Laboratory to Various Residual Dust Holdings
51 Fabric Resistance Versus Residual Fabric Loading for Sunbury 124
Bags at 0.61 m/min (2 ft/min) Filtration Velocity
52 Typical Resistance Versus Dust Loading Curves for Fly Ash 125
Filtration With Staple and Multifilament Yarns
53 Resistance Versus Average Fabric Loading for Sunbury Fabric 127
With GCA Fly Ash at 0.61 m/min Face Velocity
54 Resistance Versus Average Fabric Loading for Nucla Fabric 129
With GCA Fly Ash at 0.61 m/min Face Velocity
55 Filtration of Granite Dust (Rhyolite) and Lignite Fly Ash 130
With Sunbury Fabric at 0.61 m/min Face Velocity
56 GCA Fly Ash Filtration With Unused Sateen Weave Cotton 131
(Unnapped) and Dacron (Crowfoot Weave) at 0.61 m/min
Face Velocity
57 Fly Ash Dust Layer on Sunbury Fabric, Laboratory Tests Prior 133
to Removal of 945 grams/m Cloth Loading (20X Mag)
58 Photomicrograph of Sunbury Media Showing GCA Fly Ash Loading 134
With Pinhole Leak and Cracks Induced by Flexure
59 GCA Fly Ash Deposit on Previously Used Sunbury Fabric Showing 136
Crater and Pinholes, 430 grams/nr Cloth Loading
xv
-------
No.
60
LIST OF FIGURES (Continued)
Page
Pinhole Leak Structures, GCA Fly Ash Filtration on Sunbury 137
Fabric
61 Estimation of Pinhole Velocities by Capillary (A) and 141
Orifice (B) Theory for Fly Ash Loaded Sunbury Fabric
62 Checking or Cracking of Deposited Fly Ash Layer on Glass 145
Fabric by Intentional Flexing (60X Mag). Test With Clea^
(Unused) Sunbury Fabric With Cloth Loading of 945 grams/in
63 Dust Cake as Seen in Sectional Views With GCA Fly Ash on 146
Sunbury Fabrics (20X Mag)
64 Before and After Appearance of Dirty and Cleaned Sunbury 147
Fabric With GCA Fly Ash Filtration
65 Pore Appearances for Clean and Dirty Faces of Cleaned 149
Sunbury Fabric With GCA Fly Ash Filtration (60X Mag)
66 Appearance of Previously Used Nucla Fabric Before and After 150
Cleaning. GCA Fly Ash Loading of 1200 grams/m2 (20X Mag)
67 Appearance of Fill and Warp Faces of Nucla Fabrics After 151
Removal of GCA Fly Ash Loading of Approximately 1000
grams/m, Previously Clean (Unused) Fabric (20X Mag)
68 Photograph Showing a Section of Nucla Test Panel From Which 153
Dust has been Dislodged. Roughly 3/4 Actual Size
69 Fly Ash Filtration With Clean and Partially Cleaned Sateen 158
Weave Cotton, Flat Panel and Bag Tests
70 Effect of Filtration Velocity (V) on Specific Resistance 161
Coefficient (K^. Sunbury Fabric With GCA Fly Ash
71 Effect of Face Velocity on K , Sunbury Fabric With GCA Fly 162
Ash i
72 Effluent Concentration Versus Fabric Loading For Unused 175
Sunbury Media With GCA Fly Ash, Test 65
73 Effluent Concentration Versus Fabric Loading for Used 177
Sunbury Fabric (Test 66) With GCA Fly Ash
74 Effluent Concentration Versus Fabric Loading for Used 178
Sunbury Media With GCA Fly Ash, Test 67
xvi
-------
LIST OF FIGURES (Continued)
No. Page
75 Effluent Concentration Versus Fabric Loading for Unused 179
Nucla (Test 68) Media With GCA Fly Ash
76 Effluent Concentration Versus Fabric Loading for Used 180
Nucla Fabric (Test 69) With GCA Fly Ash
77 Effluent Concentration Versus Fabric Loading for Used 181
Sunbury Fabric and GCA Fly Ash With Uniform (Test 71) and
Nonuniform (Test 72) Dust Loading
78 Effluent Nuclei Concentration Versus Fabric Loading for 183
Used Sunbury Fabric With Lignite, Test 83
79 Effluent Nuclei Concentration Versus Fabric Loading With 184
Used Cotton Sateen and GCA Fly Ash, Test 84
80 Effluent of Filtration Velocity on Effluent Nuclei Concen- 185
tration, GCA Fly Ash With New Sunbury Fabric
81 Effluent Concentration Versus Fabric Loading and Particle 187
Size for Used Sunbury Media With GCA Fly Ash, Test 67
82 Effluent Concentration Versus Fabric Loading and Particle 188
Size for Unused Nucla Fabric With GCA Fly Ash, Test 68
83 Effluent Concentration Versus Fabric Loading and Particle 189
Size for Used Nucla Media With GCA Fly Ash, Test 69
84 Effluent Particle Concentration Versus Fabric Loading and 191
Particle Size for Used Sunbury Fabric and Lignite, Test 83,
B&L Measurements
85 Effluent Concentration Versus Fabric Loading for Unused 192
Cotton Sateen With GCA Fly Ash, Test 84, B&L Measurements
86 Calibration Curve - Nuclei and Related Mass Concentrations 194
for GCA Fly Ash
87 Outlet Concentration Versus Fabric Loading at 0.61 m/min 199
(2 ft/min) Face Velocity. GCA Fly Ash With Sunbury and
Nucla Fabrics. Loading Increase Referred to State of
Filtering Cycle
88 Outlet Concentration Versus Fabric Loading for Three Face 200
Velocities. GCA Fly Ash and Sunbury Fabric. Loading
Increase Referred to Start of Filtering Cycle.
xvii
-------
LIST OF FIGURES (Continued)
No.
89 Room Air Filtration With Clean (Unused) Woven Fabrics at 204
0.61 m/min Face Velocity, Inlet (x) and Outlet (o)
Concentrations
90 Dust Removal Versus Fabric Loading and Estimated Distribution 212
of Interfacial Adhesive Forces for GCA Fly Ash and Sunbury
Type Fabric
91 Dust Removal Characteristics for Repetitive Cleaning Cycles, 215
Sunbury Type Fabric With GCA Fly Ash
92 Performance of Sunbury Fabric With GCA Fly Ash With Repe- 218
titive Filtration and Cleaning Cycles
93 Successive Filtration and Cleaning Cycles for Sunbury Fabric 220
With GCA Fly Ash Based on Data of Table 25
94 Single Bag (10 ft x 4 in.) Filtration of GCA Fly Ash With 221
Sunbury Fabric - Three Cleaning Cycles With Variations in
Residual Loading
95 Effect of Face Velocity on Outlet Concentration, GCA Fly Ash 222
With 10 ft x 4 in. Woven Glass Bag (Sunbury Type Fabric)
96 Relationship Between Final and Average Outlet Concentration 225
and Face Velocity for 10 ft x 4 in. Bag and Test Panel With
GCA Fly Ash and Sunbury Type Fabric
97 Effluent Particle Size Parameters From GCA Fly Ash Loaded 227
Sunbury Fabric When Filtering Atmospheric Dust
98 Typical Drag Versus Fabric Loading Curve for a Uniformly 230
Distributed Dust Holding
99 Schematic, Dust Accumulation Below Surface of Fabric With 234
Bulked Fiber or Staple Support
100 Comparison Between Experimental and Predicted Drag 243
Properties
101 Relationship Between Effective (S ) and Clean (S ) or 247
Residual (S ) Drag E 0
R
102 Effect of Previous Fabric Loading on Residual Drag for New 250
and Well Used Fabric
xviii
-------
LIST OF FIGURES (Continued)
No. Page
..... i -.PL,..,,
103 Specific Resistance Coefficient (lO?) Versus Mass Median 257
Diameter and Face Velocity. Data From Table 34
104 Estimated Effect of Face Velocity on K Based Upon 259
Literature Review, Table 6
105 Specific Resistance Coefficient Versus Specific Surface 269
Parameter (S ) for Various Dusts
106 Average Filter Drag With Various Degrees of Dust Removal - 273
Fly Ash Filtration With Woven Glass Fabric
107 Effect of Cleaning Duration on Filter Capacity for Several 276
Shaking Conditions
108 Effect of Shaker Acceleration on Filter Capacity 276
109 Typical Drag Versus Loading Curves for Filters With 278
Different Degrees of Cleaning and a Maximum Allowable
Level for Terminal Drag, S , and Terminal Fabric Loading, W
110 Appearance of Partially Cleaned Fabrics 281
111 Fly Ash Filtration With Completely and Partially Cleaned 284
Woven Glass Fabric (Menardi Southern), Tests 71 and 72
112 Fly Ash Filtration With Completely and Partially Cleaned 285
Woven Glass Fabric (Menardi Southern), Tests 96 and 97
113 Fly Ash Filtration With Completely and Partially Cleaned 286
Sateen Weave Cotton, Unnapped (Albany International),
Tests 84 and 85
114 Resistance Versus Fabric Loading for Partially-Loaded 289
Fabric, Measured and Predicted (Using Linear and Bilinear
Models), Test 72
115 Average Residual Fly Ash Loadings Versus Fabric Type and 293
Number of Mechanical Shakes (8 cps at 1 in. Amplitude)j
Reference 10
116 Estimated Distribution of Adhesive Forces for Woven Glass 298
Fabrics and One Dacron Fabric With Coal Fly Ash
xix
-------
No.
LIST OF FIGURES (Continued)
Page
117 Adhesion of Spherical Fe Particles of 4 ym Diameter to Fe 301
Substrate at Room Temperature in Air as a Function of
Applied Force (From Bohme, et al, Reference 36) and
Reference 1
118 Effect of Particle Size and Relative Humidity on Adhesion 302
for Various Materials (Reference 1)
119 Relationship Between Cleaned Area Fraction and Initial 305
Fabric Loading. GCA Fly Ash and Woven Glass Fabric
120 Efficiency of Sampling an Aerosol From a Variable Velocity 318
Flow Field at a Constant Sampling Velocity of 6 m/sec
121 Filtration Velocity Through Cleaned and Uncleaned Areas of 340
Filter. GCA Fly Ash and Sunbury Fabric
122 Predicted and Observed Outlet Concentrations for Bench 345
Scale Tests. GCA Fly Ash and Sunbury Fabric
123 Effect of Inlet Concentration on Predicted Outlet Concen- 346
trations at a Face Velocity of 0.61 m/min. GCA Fly Ash
and Sunbury Fabric
124 System Breakdown for I Bags and J Areas per Bag 355
125 Schematic Representation of Approach to Steady State Cleaning 357
and Fabric Loading Conditions for a Three-Compartment System
With 50 Percent of Each Compartment Surface Cleaned
126 Baghouse Model Computational Procedure 360
127 Baghouse Simulation Program Flow Diagram 362
128 Pressure-Time Trace for Run Number 1, Nucla Generation 387
Station (Reference 8)
129 Test Run No. 0422 Nucla Baghouse Simulation - Linear: 388
Pressure Versus Time Graph
130 Test Run No. 0422 Nucla Baghouse Simulation - Linear Flow 391
Rate Versus Time Graph
xx
-------
LIST OF FIGURES (Continued)
No. Page
131 Test Run No. 0422 Nucla Baghouse Simulation - Linear 392
Individual Flow Rate Graph
132 Test Run No. 0422 Nucla Baghouse Simulation - Linear 393
Penetration Versus Time Graph
133 Pressure Drop History of Sunbury Baghouse - Run No. 1 395
(Reference 9)
134 Test Run No. 0422 Sunbury Baghouse Simulation - Linear 397
Pressure Versus Time Graph
135 Test Run No. 0422 Sunbury Baghouse Simulation - Nonlinear 398
Pressure Versus Time Graph
136 Test Run No. 0422 Sunbury Baghous Simulation - Linear 400
Individual Flow Rate Graph
137 Test Run No. 0422 Sunbury Baghouse Simulation - Nonlinear 401
Individual Flow Rate Graph
138 Test Run No. 0422 Sunbury Baghouse Simulation - Linear 403
Penetration Versus Time Graph
139 Test Run No. 0422 Sunbury Baghouse Simulation - Nonlinear 404
Penetration Versus Time Graph
140 Estimation of Pore Cross Section in Fiber Substrate Region 415
141 Penetration Versus Loading for Bench Scale Tests 417
142 Steady State Penetration as a Function of Velocity 420
143 Initial Slope of Penetration Versus Loading Curve as a 421
Function of Velocity
xxi
-------
LIST OF TABLES
NO.. Zssi
1 Some Permeabilities and Fabric Weights per Unit Area 23
2 Fabric Properties for Glass Bag Filters Used at Sunbury, 59
Pennsylvania and Nucla, Colorado Coal-Burning Power Plants
3 Special Properties, Sunbury and Nucla Fabrics 60
4 Dacron and Cotton Properties for Fabric Test Panels 62
Studied in Laboratory
5 Characteristic Pore Dimensions for Sunbury (Menardi 77
Southern) and Nucla (Criswell) Glass Fabrics
6 Results of Physical Characterization Tests on Sunbury 81
Fabric Filter Bags
7 Results of Physical Characterization Tests on a Nucla Fabric 82
Filter Bag
8 Tensile Modulus Values for Glass Bags Used for Coal Fly 86
Ash Filtration
9 Properties of Common Woven Fabrics Including Tensile 88
Modulus
10 Residual Fabric Dust Loading for Sunbury Bags as Received 101
From Field
11 Field Performance of Filter Systems with Glass Bags - 110
Sunbury/Nucla Station
12 Measured K Based on Field Tests at Nucla Generating 116
Station
13 Fabric/Dust Combinations Studied in the Laboratory Program 129
14 Filtration Characteristics of New (Unused) Sunbury Fabric 140
With GCA Fly Ash
xxii
-------
LIST OF TABLES (Continued)
No. Page
15 Fabric Resistance Before and After Cleaning by Flexure, 155
0.61 m/min Filtration Velocity
16 Weight Collection Efficiency for Sateen Weave Cotton With 159
GCA Fly Ash
17 GCA Fly Ash Filtration With Crowfoot Dacron, Bench Scale 165
Tests
18 Summary of Bench Scale Filtration Tests 168
19 Summary of Concentration, Efficiency and Penetration 196
Measurements for GCA Fly Ash Filtration With Woven Glass
Fabrics at 0.61 m/min Face Velocity
20 Initial and Average Outlet Concentrations and Related 197
Penetration Data for Fly Ash/Woven Glass Fabric Filters
21 Change in Effluent Concentration With Increasing Fabric 201
Loading for Fly Ash Filtration With Woven Glass Fabrics
22 Atmospheric Dust Penetration With Woven Glass and Cotton 203
Fabrics
23 Effluent Concentration From New (Unused) and Partially 207
Loaded Sunbury Type Fabric With GCA Fly Ash and Atmospheric
Dust
24 Relationship Between Dust Removal and Previous Fabric 209
Loading, GCA Fly Ash Filtration With 10 ft x 4 in. Woven
Glass Bag (Sunbury Type) at 0.61 m/min Face Velocity
25 Repetitive Cleaning and Filtration Cycles With GCA Fly Ash 210
and Woven Glass (Sunbury Type) Fabric at 0.61 m/min Face
Velocity and 50 Ibs Tension
26 Effect of Reduced Bag Tension, 15 Ibs, on Dust Removal and 211
Penetration GCA Fly Ash With 10 ft x 4 in. Bag, Sunbury Fab-
ric, at 0.61 m/min Face Velocity
27 Effect of Several Successive Cleanings by Bag Collapse and 214
Reverse Flow, GCA Fly Ash With Woven Glass Fabric (Sunbury
Type)
28 Effect of Face Velocity on Outlet Concentration, GCA Fly 224
Ash 10 ft x 4 in. Woven Glass Bag, Sunbury Fabric
xxiii
-------
LIST OF TABLES (Continued)
XT Page
No. —£Z—
29 Physical Properties of Fabrics Involved in Model Testing 240
30 Summary of Measured Filtration Parameters for Model Testing 242
31 Clean (Unused) and Effective Drag Values for Commercial and 245
Experimental Fabrics by Draemel With Resuspended Coal Fly
Ash
32 Summary of Experimentally Derived Model Input Parameters 248
Used to Predict Drag Versus Fabric Loading Relationship
33 Corrections Factors for K 255
34 Data Summaries for Estimating K as a Function of Face 256
Velocity and Particle Size
35 Porosity Function for Granular Porous Media 262
36 Measured and Calculated K2 Values for Nucla Field Tests 263
37 Summary of Average K Value From Nucla Field Studies 265
38 Calculated and Measured Values for Specific Resistance 266
Coefficients for Various Dusts
39 Measured and Predicted K. Values 268
40 Relationship Between Cleaned Fabric Surface and Average 274
Filter Drag - Coal Fly Ash Filtration With Woven Glass
Fabric (Predicted)
41 Fraction of Filter Surface Cleaned Versus Dust Separation 291
Force, GCA Fly Ash With Woven Glass Fabric (Sunbury Type)
42 Effect of Number of Mechanical Shakes on GCA Fly Ash Removal 294
From Selected Fabrics
43 Physical Properties and Penetration Data for Woven Fabric 296
Examined for Dust Cake Adhesion
44 Input Parameters for Estimating Bulk Fiber Efficiency. 322
45 Collection Parameters and Initial Efficiency for Woven 323
Fabric Filters for Fiber Phase Collection
xxiv
-------
LIST OF TABLES (Continued)
No. Page
46 Filtration Parameters for Combined Fiber-Particle Collection 324
47 Penetration Estimates for a 2 \im Particle as a Function of 325
Fabric Loading and Inlet Concentration at 0.61 m/min Face
Velocity, Fiber Filtration Phase
48 Parameters for, and Estimation of, Overall Weight Collection 326
for Fly Ash During Fiber Phase Filtration
49 Estimated Values for Diffusion, Interception and Impaction 330
Parameters, Granular Bed Collection
50 Parameters Used to Compute Dust Cake Particle Collection 331
Efficiency
51 Estimated Overall Weight Collection Efficiencies as a 332
Function of Cake Thickness and Inlet Particle Size for
Fly Ash
52 Comparative Penetration Characteristics for Uniformly Loaded 339
and Partially Loaded Fabrics, GCA Fly Ash
53 Simulation Program Input Data 368
54 Sample Program Output With Supplementary Definition of Terms 372
55 Data Used for Model Trials With the Nucla and Sunbury Fabric 375
Filter Systems
56 Normal Cleaning Sequence for Each Nucla Compartment 376
57 Simplified Cleaning Sequence per Nucla Compartment Used in 377
Predictive Modeling
58 Test Run No. 0422 Nucla Baghouse Simulation - Linear 379
Printout of Input Data for Baghouse Analysis
59 Normal Cleaning Sequence for Sunbury Compartment 382
60 Simplified Cleaning Sequence per Sunbury Compartment 383
61 Test Run No. 0422 Sunbury Baghouse Simulation - Linear 384
Printout of Input Data for Baghouse Analysis
62 Test Run No. 0422 Sunbury Baghouse Simulation - Nonlinear 385
Printout of Input Data for Baghouse Analysis
xxv '
-------
LIST OF TABLES (Continued)
No.,
63 Predicted and Measured Resistance Characteristics for 389
Nucla Filter System
64 Comparison of Observed and Predicted Fly Ash Penetration 405
Value, Sunbury Installation
65 Summary Table Showing Measured and Predicted Value for 408
Filter System Penetration and Resistance, Coal Fly Ash
Filtration With Woven Glass Fabrics
66 Data Used to Determine Constants in Dust Penetration Model 419
67 Input Specifications for Various Types of Cleaning Cycles 425
68 Baghouse Simulation Program Listing 426
69 Variables and Arrays Used in Baghouse Simulation Program 442
70 Data Input Format 446
xxvi
-------
ACKNOWLEDGMENTS
The authors express their appreciation to Dr. James H. Turner, EPA Project
Officer, for his advise and technical support throughout the program.
We also wish to acknowledge the assistance of the following GCA personnel:
Dr. Michael T. Mills in the computer program area, Messrs. Mark I.
Bornstein, Lyle Powers and Manuel T. Rei for technical support in the
laboratory program, and Mr. Norman F. Surprenant for his technical
reviews.
xxvii
-------
NOMENCLATURE
Roman
a Proportionality factor, Leith and First efficiency model
a Concentration decay function
a Fraction of fabric area cleaned
c
a. Fraction of fabric area not cleaned
d
a. System constant in Walsh and Spaite model
'a Fraction of fabric area not cleaned
u
a Average acceleration
b Proportionality factor, Leith and First efficiency model
c Proportionality factor, Leith and First efficiency model
c Concentration
cm Centimeters
d Pore diameter
d Collector diameter
c
d,. Fiber diameter
d . Pore diameter at greatest pore depth
mm
d Pore diameter at pore surface
max
d Particle specific surface diameter
o
d Particle diameter
P
d Particle surface mean diameter
s
t
xxviii
-------
NOMENCLATURE (continued)
d Particle volume mean diameter
v
f Shaking frequency
g grams, or gravitational force
k Constant in Equation (25) = (d - d . )/WT
max mm I
k Constant in Equation (50)for determination of average acceleration
k' Parameter defined by Equation (77) for the calculation of efficiency
k" Constant defined in Equation (77)
k. Cleaning period for a single compartment in Solbach model
m Meter
min Minute
n Number of filters or compartments operated in parallel
n Total number of areas in the system
At Cleaning interval for a compartment
Zt Combined operating time for a cycle
p Pressure
t Time
v Gas velocity
x. System constant in Walsh and Spaite Model
A Amplitude
AA, Rate of increase of collector surface area per unit area of
filter cross section
j,t
A. Cloth face area of the i compartment
A Particle surface area based on surface mean particle diameter, d
p s
A Ratio of total projected fiber surface to filter cross sectional
area
xxix
-------
NOMENCLATURE (continued)
A Cross sectional area of a single pore
P
A Particle volume, based on particle volume mean diameter, d
v
C Concentration
C Inlet dust concentration
C Cunningham slip correction factor
C. Inlet concentration
i
C Orifice coefficient
o
C Outlet concentration
o
C_ Residual concentration due to rear face slough-off
R
D Diffusion coefficient
D
E Average collection efficiency in Fraser and Foley model
F Adhesive force
a
F^_ Median adhesion force
I Total number of compartments
J Total number of areas on a bag
K Reciprocal of K» in Stinessen drag model
K? Specific resistance coefficient
K2° Specific resistance coefficient measured at 0.61 m/min and actual
gas temperature
Fabric surface correction factor for K_
K rm Fabric permeability correction factor for K
K . Particle shape correction factor for K
K^ Initial slope of drag versus loading curve
K Velocity correction factor for K
L Filter thickness
xxx
-------
NOMENCLATURE (continued)
L Fiber bed depth
M Average hydraulic radius
N Newton
N Number of pores
N Number of particles in a unit mass of filter bed
N- Limiting number of shakes beyond which no appreciable reduction
of residual drag occurs
NW Limiting number of shakes beyond which no increase in filtration
capacity is attained
P Pressure drop
AP Increase in pressure drop
P Constant system pressure drop
P Pressure drop based on system velocity and effective drag, S
P Specified maximum pressure during a cleaning cycle
Pn Penetration
Pn Penetration at a residual loading
Pn Steady state penetration
S
P Pressure used to determine the fabric loading W
PT: Pressure drop at the average loading, W
w_ c
Q Volumetric gas flow rate
Q Volume flow per pore
j»t_
Q. Volume flow rate through the i compartment
R_ Ratio of particle diameter to fiber diameter
R Pore radius (based on minimum pore area)
S Total filter drag
xxxi
-------
NOMENCLATURE (continued)
S Total system drag
S Cleaned area drag
c
S, Uncleaned area drag
Q
S Effective drag for cleaned filter area
ii
S' Average effective drag
E
t-i"»
S . Overall drag of the i compartment
S Particle specific surface parameter
S New fabric drag
S Average resultant drag of a partially cleaned filter
R
S Residual drag
R
S_, Average drag after cleaning
R
S Uncleaned area drag
u
V Average pore velocity
V Face velocity
V Constant average system face velocity
V. Velocity through the i compartment
V^ Average velocity through the i compartment when one compartment
is off-line
V Reverse flow velocity
R
*
Absolute fabric loading minus the residual loading = W - W
W Constant used in the nonlinear drag model, specified for each
dust/fabric combination
W Weight of clean fabric at the start of filtration Noll David
and Shelton drag model '
Wz Fabric loading at inception of pure cake filtration (linear drag)
xxxii
-------
NOMENCLATURE (continued)
W Fabric loading at the limiting pressure
Wp Corrected value for W
W Average fabric loading for a system operating under continuous
sequential cleaning conditions
W Residual fabric loading; i.e., the loading on a cleaned area
R
W' Average fabric loading after cleaning
R
W Average fabric loading before cleaning
X Deposit thickness, Leith and First efficiency model
Dimensionless
groups
Ref Fiber Reynolds number
Pe Peclet number
Greek
a Ratio of bed packing density to particle density
e Porosity
n Single particle-single fiber efficiency
TV T Interception efficiency
nT Impaction efficiency
ri' Diffusion parameter
pf Discrete fiber density
p Discrete particle density
p Bulk fiber density
a Particle size distribution geometric standard deviation
a
<|> Function specific to fabric and dust
M Gas viscosity
xxxiii
-------
NOMENCLATURE (continued)
ym Micron or micrometer
y- Fluid viscosity
Subscripts
c Cleaned area
i Interval of size distribution
i Number of the time increment
i Refers to the i compartment
j Refers to the j area on a bag
t Refers to time = t
u Uncleaned area
xxxiv
-------
SPECIAL NOMENCLATURE
ENGLISH AND METRIC EQUIVALENCIES FOR KEY FILTRATION PARAMETERS
Filter resistance
Filter drag
Velocity
Volume flow
Fabric area
Areal density
Specific resistance
coefficient
Dust concentration
Units
Metric
N/m2
3
N min/m
m/min
3
m /min
2
m
g/m
N min/g-m
g/m
English
in. HO
in. HO min/ft
ft /min
ft /min
2
ft
lb/ft2
in. HO min ft/lb
3
grains/ft
Equivalency
249 N/m2 = 1 in. water
817 N min/m = 1 in. water min/ft
0.305 m/min = 1 ft /min
0.0283 m /min = 1 ft3 /min
2 2
0.093 m = 1 ft
4882 g/m2 = 1 lb/ft2
0.167 N min/ g-m = 1 in. HO min ft/lb
2.29 g/m3 = 1 grain/ ft3
-------
SECTION I
SUMMARY
The overall objective of this program was to develop mathematical models
to predict fabric filter behavior with emphasis on systems for the control
of particulate emissions from coal-fired boilers. In conjunction with
the development of drag and efficiency models, a laboratory experi-
mental program was conducted to provide insight into critical filtration
parameters (e.g., fabric structure, particulate deposition and removal)
affecting field and laboratory filter performance. The laboratory program
has been directed mainly to the collection of coal fly ash with woven
glass fabrics of the type used at Nucla, Colorado, and Sunbury, Pennsyl-
vania fabric filter installations. The results of both field and labora-
tory testing and research have been utilized in the development of the
model. Further experimental work was carried out on full scale bags to
verify the results of the bench scale program and to test and improve the
models.
The literature with respect to filtration is vast, but the efforts to
model fabric filtration have been limited in number and usefulness.
In fact, the results of a detailed survey suggest that many parameters
are best determined by carefully controlled experiments until an adequate
theory is developed. Modeling approaches have usually depended upon a
linear approximation to define the increase of fabric drag with fabric
dust loading and many fabric collectors have been described as an array
of cylinders. The latter (cylinder) approaches treat particle collection
by concepts developed for bulk fiber filters for which randomly or
-------
preferentially oriented discrete fibers appear to capture particles in
fair accord with "single particle-single fiber" collection theory.
A woven fabric filter, however, is more properly viewed as an array of
pores whose number relate approximately to fabric thread count and whose
boundaries are formed by the intertwined warp and fill yarns. Because of
the low yarn porosity per ses "TO percent, only those fibers constituting
the napped, bulked or protruding staple fibers are available for effective
"single fiber" collection. Conversely, negligible gas flow, and, hence,
filtration, can take place within the yarns because of their very low
permeability.
The fiber fraction that extends into and across the pore openings, which
is fairly uniformly 'distributed in a good filter, actually constitutes
the supporting substrate for initial dust cake formation. (If a filter
fabric is composed entirely of multifilament yarns, the yarn proximity
must be significantly increased before effective particle collection
ensues.)
Considering particle capture to consist first of a bridging over of pore
openings at the substrate level (a process that commences somewhat below
the superficial fabric surface and continues until an appreciable dust
cake has developed) has enabled the development of a new, nonlinear
model. The new model (or assemblage of predictive equations) has the
capability to describe more accurately the filter resistance and particle
capture properties during the initial filtration phase than the simplified
linear model when a large fraction of the filter surface is cleaned.
Bench and pilot scale tests showed that certain portions of the fabric
were cleaned to a very low residual dust level whereas the remaining
surface experienced no cleaning whatsoever. Exploratory tests with two
element systems (dust removed from only 50 percent of the fabric surface)
indicated that filter resistances were significantly lower while pene-
trations were correspondingly higher for nonuriiformly loaded fabrics.
-------
Higher penetrations result from the initial high velocity transients
through the "just cleaned" fabric areas. Under the more rigorous analysis
of velocity distributions afforded by the nonlinear drag model, even
higher penetration levels would be predicted. When the model system is
composed of six or more separate bag compartments in which the degree of
cleaning is like that observed for many collapse or mechanical shaking
systems (~10 to 20 percent) the difference between linear and nonlinear
modeling diminishes.
The drag and efficiency models for a full scale system appear to give
results which are both reasonable and informative. With respect to the
Sunbury and Nucla type fabrics, experiment and theory indicate that by
far the largest fraction of all dust penetrating these filters is that
which passes through unblocked or unbridged pores immediately after filter
cleaning or through oversize pores (pinholes) that fail to close at any
time during the filtering cycle. Since very little dust is separated
from that fraction of the inlet air that passes through a pore and, since
o
pore velocities may exceed cake velocities by a factor of 10 or greater,
the particle size properties of the filter effluent are dominated by the
properties of the inlet dust that passes through the pore. Those changes
in particle size efficiency attributable to either preliminary fiber fil-
tration and subsequent cake filtration are usually completely obscured by
direct dust penetration and/or clean face slough-off components. The
above statement applies specifically to the common woven glass fabrics
used for fly ash filtration.
Over the range of face velocities studied, 0.40 to 4.3 m/min (1.3 to
14 ft/min) velocity was observed to exert a strong influence on mass
penetration while having little impact on size properties. Theory and
confirming microscopic observations of filter surfaces suggested that
complete pore bridging is more difficult tp obtain at higher velocities
due to increased reentrainment of deposited dust.
-------
As the cake built up, the penetration decreased rapidly in exponential
fashion followed by a leveling-off to an asymptotic value determined by
penetration through the pinholes or by seepage and/or slough-off from
the rear face of the dust layer.
The mathematical model developed within this study represents a new and
very effective technique for predicting the average and instantaneous
values for resistance and emission characteristics during the filtration
of coal fly ash with woven glass fabrics.
Two basic concepts used in the model design: (1) the quantitative de-
scription of the filtration properties of partially cleaned fabric sur-
faces and (2) the correct description of effluent particle size proper-
ties for fabrics in which direct pore or pinhole penetration constitutes
the major source of emission, have played important roles in structuring
the predictive equation.
A third key factor in the model development was the formulation of ex-
plicit functions to describe quantitatively the cleaning process in terms
of the method, intensity and frequency of cleaning. By cleaning we refer
specifically to the amount of dust removed during the cleaning of any one
compartment and the effect of its removal on filter resistance and pene-
tration characteristics.
The drag and efficiency models have been combined to form an experimental
computer program for a complete multichamber filtration system. The re-
sults of such modeling are presented for both flow resistance and particle
penetration behavior. The linear drag model will probably satisfy most
practical field filtration applications. However, the nonlinear model,
which also visualizes fabric performance from the pore array concept,
may provide a better fit in those cases where an unusually high dust re-
moval is achieved during filter cleaning. The above (cleaning) process
creates a filter surface that provides not only low resistance to air flow
but also a highly permeable region for dust particles.
-------
SECTION II
INTRODUCTION
The development and evaluation of a predictive model for fabric filtra-
tion with special emphasis on the control of coal fly ash emissions from
boilers are described in this report. The primary goals of this proj-
ect were to relate basic filtration parameters including dust properties,
dust loadings and air-to-cloth ratios to key performance parameters, pres-
sure loss and dust penetration characteristics. Laboratory experiments
conducted as part of this program as well as the results of prior field
and laboratory investigations by many researchers constituted the bases
for the modeling concepts developed under the present program.
The following factors can be expected to influence the efficiency and the
pressure/flow relationship:
1. Dust - chemical composition, particle size distribution,
particle shape, particle phase, particle concentration.
2. Gas - chemical composition (especially moisture content),
temperature, pressure.
3. Fabric - material, weave (including fill and warp counts),
finish, history (especially accumulated dust).
4. Cleaning operations - cleaning type, intensity, frequency,
duration.
Although it is desired that both resistance and efficiency characteristics
be predictable for the conditions cited above, this goal is not easily
attained. In forming a useful model, therefore, one treads a narrow path
between untractable complexity and impractical simplicity, particularly
so in the case of fabric filtration.
-------
Several research programs are being conducted in the fabric filtration
area because of the importance of fine particle removal by air and gas
cleaning processes. By and large, the Environmental Protection Agency
has provided the recent impetus for such activities, either through its
in-house research programs or the sponsorship of outside research.
Completion of the "Handbook of Fabric Filter Technology" in 1970 by
Billings and Wilder along with supporting appendices, bibliography and
recommendations for research under Contract CPA-22-69-38 represented the
first major step to bring together and evaluate available data on fil-
tration technology. The state-of-the-art in filtration technology was
reviewed more recently in a joint EPA/GCA sponsored symposium whose
2
papers appear in the December 1974 issue of the APCA journal. Since
that time, additional in-house and field studies performed by EPA have
q
dealt with filter performance versus fabric structure, Draemel; the
performance of nonwoven (spun bonded) nylon fabrics, Turner; and field
filtration of metal oxide fumes, Harris and Drehmel.
Various EPA contractor groups have investigated the use of fabric fil-
tration with coal-fired industrial boiler effluents, McKenna;^ the
performance of field filter systems for bronze smelting operations,
asphalt concrete production and coal-fired cyclone boilers, Hall and
Dennis, and more recently the performance of commercial glass fabric
filter systems at two coal-fired power plants by Bradway, et al.8 and
9
Cass, et al. The role of fabric filter cleaning mechanisms in control-
ling resistance characteristics and dust penetration has been studied ex-
tensively in the laboratory by Dennis and Wilder.10 Based upon studies
of the type described above, attempts have been made to develop mathe-
matical models for describing fabric filter performance. Although one
can argue that almost all models proposed to date have at best only
limited application, their deficiencies are often due to a lack of relia-
ble field and laboratory data. This situation has prevailed because of
(1) the number of variables encountered in a filter system and (2) the
often complex relationships among these variables.
-------
DESCRIPTION OF A FILTRATION SYSTEM
Figure 1 shows a very simple schematic of a fabric filter installation.
A dust-laden flow of gas enters the filter installation with a volumetric
flow rate Q and a concentration c. The flow is divided among n compart-
ments, the bags within each compartment having a fabric loading of average
dust weight per unit fabric area W.. The pressure drop across the ith com-
partment's bags is given by the equation:
Ap =
in which Q.^ is the volume flow rate, A. is the cloth face area and S. is
the overall drag caused by the fabric and any accumulated dust. The
ratio S./A. is analogous to electrical resistance (with Q. depicting the
current and Ap. the voltage). The total flow, Q, is the sum of the in-
dividual flows, Q. (as long as temperature and pressure corrections are
made) . Usually the compartments are operated in parallel and so con-
structed that the pressure drop at any given time is expected to be es-
sentially the same across all of them.
The relationship between the pressure drop and the volume flow for a par-
ticular installation will depend upon the locus of the intersections of
the system fan curve and the system resistance curve, each of which can
be expressed as volume rate of flow versus pressure drop. Usually, one
of the following conditions holds approximately for the installation:
1. The fan produces a constant volume of flow while the
pressure drop changes with system resistance.
2. The fan produces a constant pressure drop, while the
volume flow changes with system resistance,
OBJECTIVES
The cost of the installation will depend, in part, upon the type and quan
tity of fabric and a major operating cost will be that required to over-
come filter resistance to gas flow.
-------
Q
oo
-^
^
^p
/
}•
V
1
w
Si
1
Q,
W
S2
2
Q2
— — H
W
S3
3
«.
(
w
S4
4
o.
W
Sn
n
Qn
Figure 1. Schematic of n-compartment baghouse
-------
Often the available choice of fabrics will be limited, so that the major
question becomes the area of cloth needed to handle a specified volume
flow rate of gas.
The cloth area required has as its criterion "operation at an acceptable
pressure drop across the cloth for a predetermined cycle."^ One might
add that this assumes that the collection efficiency is adequate under
these conditions. Thus, an important goal is to be able to predict the
pressure drop for a particular dust and fabric combination at a given
air-to-cloth ratio. The drag will depend upon how much dust is on the
filter surface, how it is distributed, the geometrical structure of the
cake, the geometry of the fabric and the viscosity of the gas. A second
and equally important goal is to predict the system emissions.
OUTLINE OF MODEL
A procedure for calculating the pressure/flow relationship and filter effi-
ciency can be developed by first subdividing the fabric area into smaller
homogeneous subunits (compartments, bags, or areas on bags) and then per-
forming the following operations.
1. Calculate the drag (the pressure drop per unit face
velocity) for the subunit.
2. Determine the flow from the drag and the instantaneous
pressure drop.
3. Determine the penetration, or fraction of the particulate
concentration reaching the subunit which penetrates to
the clean air side.
4. Calculate the emissions rate from the subunit (penetration
times inlet concentration times volume flow rate).
5. Calculate the new dust loading of the subunit.
6. Determine the new pressure drop or the new total
flow rate by combining the resistance of the sub-
units according to the law for the addition of
parallel resistances:
-------
n
Q = Ap - = Ap
To develop a time profile of the performance, this procedure must be done
iteratively, with any cleaning taken into account as well.
The program goal was to develop those modeling .concepts as diagrammed in
Figure 2. With such a model one should be able to predict the collection
efficiency and the relationship between flow and pressure drop for fabric
installations for rational combinations of variables relating to dust,
gas, fabric, and cleaning method.
SUMMARY OF METHODOLOGY
The model is built up from individual units:
1. Analysis of the system elements.
2. Analysis of operating modes.
3. Determination of flow through the elements
during these processes.
4. Determination of particulate emissions during
these processes.
5. Calculation of the pressure drop, flow, and
emissions.
The steps involved in obtaining the necessary information to construct
these units have been:
1. Review previous work.
2. Develop working model with regard to drag
and particle removal.
10
-------
IN PUTS'GAS, DUST, FABRIC, CLEANING, FAN (Q OR Ap )
DETERMINE DRAG' ELEMENTS
COMPARTMENTS
SYSTEM
CALCULATE Q.Ap AND CHECK LIMITS
DETERMINE FLOW' ELEMENTS
COMPARTMENTS
SYSTEM
DETERMINE PENETRATION , ELEMENTS
CALCULATE AND SUM EMISSIONS
DETERMINE LOADING FOR NORMAL OPERATION
FOR CLEANING, ADJUST LOADING
« RETURN UNTIL LIMIT REACHED
RESULTS'EMISSIONS Q,Ap VS TIME
USE - COMPARE ALTERNATIVES
STUDY CORRELATIONS, ETC.
Figure 2. Flowchart for baghouse model
11
-------
3. Compare model with existing data.
4. Alter hypotheses where necessary.
5. Identify areas of data deficiency.
6. Obtain needed data through laboratory
investigation.
7. Test model again and modify where necessary.
THE LABORATORY PROGRAM
The laboratory program was designed to assist in the development of the
model through the following investigations:
• A study of those fabric properties expected to in-
fluence fabric filter behavior; e.g., pore structure,
pore area, napped, bulked or staple yarns.
• A review of field operations previously conducted at Nucla
and Sunbury to provide empirical and theoretical insights
into critical parameters.
• A bench scale program to identify and measure critical fil-
tration parameters for inclusion in the model and to test
and validate the model and its possible revisions.
• A pilot scale experimental program to verify the bench scale
program results and to supply additional data for the modeling
effort.
At this time, the modeling process is directed specifically to coal-fired
combustion systems used mainly in electric power production. Therefore,
the results of power plant field measurements performed at Sunbury,
Pennsylvania and Nucla, Colorado with woven glass bags plus supporting
laboratory studies on used and new filter media of the types employed at
the aforementioned field locations are described in this report in the
light of their contributions to mathematical model design.
12
-------
The results of several field measurements at both Sunbury and Nucla have
been presented in earlier reports. It was the aim of the previous tests
to provide as much data on filter system performance as possible. To this
end, mass concentration and particle size information were obtained over
several days of typical power plant operations. Although it is believed
that these measurements described accurately the inlet and outlet dust
properties, the inability to make certain measurements in the field makes
it difficult (1) to ascertain whether in fact certain system components
were operating as intended and (2) to vary basic plant operating parameters
without interfering with electric power production.
Because it was not possible to install instrumentation describing the
performance of individual bag compartments (and bags), most field data
depict average performance characteristics with respect to gas stream
composition, temperature, pressure drop across the baghouse and emission
characteristics. Therefore, although these data should enable reliable
projections for the performance of replicate systems, caution must be
exercised in applying the findings to coal-burning power plant operations
where kW capacity, gas flows, number of compartments and cleaning, cycles
differ. To extrapolate these data for the filtration of noncombustion
aerosols with glass fabric at different temperatures and with other modes
of cleaning could lead to serious errors unless particle/fabric relation-
ships are clearly understood. For the above reasons, several tests were
performed in the laboratory not only to provide supplemental data but
also to make maximum use of field measurements. Past and present field
measurements plus those from carefully-designed laboratory experiments
have provided the base for further testing and improvement of the model.
13
-------
SECTION III
A REVIEW OF FABRIC FILTRATION MODELS
Although filtration processes have been treated extensively in the tech-
nical literature, most data are only tangentially related to fabric fil-
tration. The remaining information more often examines the behavior of
isolated cylindrical fibers or fibers which are part of a high-porosity
matrix, as, for example, deep bed or bulk fiber filters. The literature
describing models for determining pressure drop, flow rate and collection
efficiency for fabric filters is much more limited.
PREDICTIVE MODELS
The efforts of several investigators to develop predictive models for
fabric filtration processes are reviewed in the following paragraphs.
Robinson, Harrington and Spaite Model
One of the first modeling attempts was made by Robinson, Harrington and
13
Spaitex who designed a mathematical model for predicting performance of
a multicompartment, parallel flow baghouse. Their basic equation for
calculating the drag, S, of an individual compartment was:
S = SR + K2V C t
The relationship between the drag of the individual compartments within
equal filter areas and the total drag of a parallel filter system is
given by:
14
-------
(2)
The symbols appearing in Equations (1) and (2) are defined here and
in the following sections as shown below:
S = Total filter drag
= Residual drag
R
« = Specific resistance coefficient
c = Inlet dust concentration
t = Time
n = Number of filters or compartments operated in parallel.
Their model was derived from experimental data obtained on a pilot fil-
tration unit consisting of 3 parallel flow compartments, each with eight
cotton sateen bags cleaned by mechanical shaking.
The air flow distributions during these experiments as determined by
Walsh and Spaite are shown in Figure 3.
u.
9
GRAPH I
COMPARTMENT NO. I
ON STREAM AFTER
CLEANING
^COMPARTMENT
NO- I
COMPARTMENT NO 2
COMPARTMENT N0.3
\
GRAPH 2
COMPARTMENT NO. 3
ON STREAM AFTER
CLEANING
^.COMPARTMENT
^ - NO. 3
COMPARTMENT MO. 2
GRAPH 3
COMPARTMENT NO. 2
ON STREAM AFTER
CLEANING
COMPARTMJEJU. HO. 3
COMPARTMENT NO. I
TIME
. TIME
I 0
TIME
COMPLETE FILTERING CYCLE
Figure 3. Three-compartment baghouse with sequential cleaning.
15
-------
The three curves on each single graph represent the air flow through in-
dividual compartments over the same time interval, whereas the three
graphs indicate the changes in compartment flow over a complete filtra-
tion cycle. Note that each compartment has assumed three distinct flow
characteristics over the complete filtration cycle.
Analysis of the volume flow rate, Q, versus time curve for each compart-
ment indicated the following approximate relationship:
Q1 = a± tXl • (3)
where the numbers 1=1, 2, 3, n refer to the order of cleaning with "1"
indicating the most recently cleaned compartment and the terms a. and x.
are system constants requiring experimental evaluation.
The drag values for the individual compartments were obtained as follows :
K c -t
si = S + -T- I Q dt
1 R A J o
o
Kc
S2 - SR + —
/:
f X Ql dt +f Q2 dt
o o
kiQidt+/>dt+r
where t. = time for one complete filtration interval
t = elapsed time in the current filtering interval
A = filtering area
Although the constants a± and x_L appearing in Equation (3) can be de-
fined in terms of operating variables and a combination of simple and
multiple regression analyses, the overall mathematical manipulations are
16
-------
cumbersome, and, in the long run, provide data outputs that cannot be
safely extrapolated beyond the operating conditions used to calculate
the system constants.
The authors state that the values for the constants will also depend upon
baghouse configuration and inlet concentration. Additionally, correct
evaluation of IL is claimed to be very important. Although not mentioned
by the authors, knowledge of the actual residual dust loadings and the
degree of dust removal attained by various cleaning methods whether it
be mechanical shaking or bag collapse with reverse flow is essential to
any useful extrapolation of their proposed modeling equation.
If the residual drag denotes the drag obtained by the extrapolation of
the linear zone, S , Equation (1) calculates the straight line shown
E
in Figure 4 by the solid line. If their use of the term "residual
drag" refers to SR, then in Figure 4 their equation must be represented
by the dashed line. In either case, Equation (1) considers the linear
portion only, omitting the dust cake repair zone where the drag may exhibit
initially a steep nonlinear rise. The model they present does not predict
collection efficiency or effluent loadings.
SE
SR
FILTRATION PERIOD*!
Figure 4. Fabric drag versus loading
17
-------
Solbach Model
Based on bench-scale tests, Solbach15 derived a simplified model for single
and multicompartment filters. His approach was to extrapolate linearly
the upper portion of the drag versus fabric loading curve so that an effec-
tive drag intercept, £„, was obtained for the residual fabric loading con-
E
dition W . This simplifying concept has been used by many previous inves-
R
tigators. Solbach also assumed that once the filter was conditioned or
stabilized, repetitive values for either !
successive filtering and cleaning cycles.
stabilized, repetitive values for either S_ or S_ would be obtained for
hj K
Solbach has also used the common expression for predicting total filter
drag, i.e. :
S = SE + K2W (4)
where W' indicates the amount of dust added to the filter since resumption
of filtration.
He again uses the classical expression:
dS = K_ cV dt
with the added constraints that K and c are constant for a given operation
and that the operating pressure loss is known and maintained constant
during the filtration process. This enables calculation of the gas velo-
city within a single compartment system.
V =
18
-------
If the cleaning period for the filter compartment is k and n is the num-
ber of the compartments, the total filtration period of each compartment
is:
Thus the average gas velocity through the filter becomes:
n
v -
avg n
or as a good approximation:
V = /Vdt/t (6)
avg *
Solbach obtained an expression for the average gas velocity through, the
multicompartment system by combining Equations (5) and (6):
(7)
The required total filter area for the multicompartment fabric filtra-
tion system is given by:
A - —2- (8)
n V
avg
In order to estimate average face velocity and fabric area by Equations
(7) and (8), it is necessary to select an operating pressure that will
not change appreciably over a filtering cycle and to determine K2 and
S_ by methods described previously.
19
-------
By rearranging Equation (7) a simplified form of this equation can be
obtained in terms of the amount of dust deposited in time t:
K0W'
= s + 2
Vavg
showing that, according to Solbach, the "average" total drag Sa is equal j
to the effective drag plus one-half the drag of the dust cake before
cleaning. Dust penetration characteristics are not considered in the
above models.
Dennis and Wilder Model
A somewhat similar expression for the average total drag of the multi-
compartment system was obtained by Dennis and Wilder-'-0 by an independent
analysis.
At any time, the pressure drop is equal to the instantaneous product of
the filtration velocity and the drag so that the total drag can be
described by the well-known approximate equation:
S = SE + K2W (4)
and the instantaneous increase of the drag with the time is given by:
dS v dW „ „
dF = K2 dT = K2CV
By expressing W as a function of dust concentration, filtration velocity
and time, the instantaneous pressure drop is equal to:
S dS
20
-------
Because the pressure drop is essentially constant across each compartment
and each bag, the instantaneous drag times the rate of increase of the drag
is also the same over any area of the filter.
Thus, the average pressure drop over a time period t« - t-, is:
£. _L
Ap H
By substituting Ap from Equation (10), the average pressure drop Ap is
expressed as follows:
-
^/;;
-------
This form, which is almost identical to that developed by Solbach,
Equation (9) , does not require the assumption of constant operating
pressure.
Because both models use the identical basic equations, it is under-
standable that they agree in predicting the average drag. The av-
erage of a linear function of W' is that function evaluated at the
midpoint of the W interval.
Noll, Davis and Shelton Model
In a model proposed by Noll, Davis and Shelton, the same drag/fabric
loading relationships cited previously are presented in equation form as
s =
Again, K~ is the specific resistance coefficient as described by Williams,
Hatch and Greenberg.1' The term Sc is the increase in drag resulting from
the increase in fabric loading, W' , over the filtration interval. The
authors have used what they refer to as a "triangulation method" to define
the effective drag $„ in terms of K and W , that is:
t / t
SE = K2wf (^)
The term W is defined as the weight of the clean fabric at the start of a
filtration. Although Equation (14) may apply over a narrow test range,
it cannot have broad application because the nature of the fabric weave as
well as its density and the presence of residual dust all exert a signifi-
cant influence on the effective drag. The data presented in Table 1 show
clearly that even clean cloth permeability shows no consistent relation-
ship to fabric areal density, Wf.
22
-------
Table 1. SOME PERMEABILITIES AND FABRIC WEIGHTS PER UNIT AREA5
Fabric description
Frasier permeability,
fpm at 0.5 in. H20
Weight,
oz/yd^
Nomex filament 3x1 twill
Cotton sateen
Spun acrylic 2x2 twill
Nomex filament (combination
cotton-fill) 3x1 twill
Nomex felt
15 - 20
15 - 20
60
30 - 50
20 - 40
4.5
9-10
9.8
4.5
14. - 16.0
1 I Q
These data were obtained from Durham and a DuPont research report.
19
Noll, Davis, and LaRosa (1975) Model
,20
In a more recent paper, Noll, Davis and LaRosai(J evaluated the parameters
K2 and Sg by means of performance tests on clean and conditioned fabrics of
polyester. According to this work, the K« values depend on the proper-
ties of the dust only. It also appears that the earlier concept of ex-
pressing S by the product K0 W. (Equation 14) has been abandoned.
E 2. i
These published data along with numerous tests performed on other types
of fabric filters (glass fiber, Nomex, cotton, polypropylene) may repre-
sent "the potential for producing generalized methods of the performance
prediction - and optimization for application to industrial fabric fil-
ter design." This statement was confirmed by P. J. LaRosa^ from Pollu-
tion Control Division of Carborundum, Environmental Systems, Inc., who
stated that a predictive model, based on these data, has been established
for the strict use of the company.
Although the authors report successful curve fitting, it is emphasized
that their so-called dust loading range in the nonlinear region (0.01 to
2
0.05 Ib/ft ) does not represent the true fabric dust loading. The re-
ported values were obtained by extrapolating data from uniformly loaded
fabrics to full scale bags that had experienced only partial cleaning.
23
-------
Therefore, the models and constants derived from these measurement should
apply only to filter systems that have identical residual dust holdings.
This presumes, therefore, very similar cleaning processes. As stated
1 o
earlier, the original modeling studies by Robinson et al. involve the
same oversights in treating the state and behavior of cleaned fabric
filters in real, commercial applications.
Stinessen's Approach
Stinessen^ has also studied the relationship between filter drag, S,
and the permeability, K, and mass, W, of the dust cake:
AS = AW/K
The term K is the reciprocal of the well known term K« (specific resis-
tance coefficient) that has been defined previously in Equation (1).
Stinessen's equation for estimating total filter drag:
S = SE + — 1 cV dt
J o
uses the effective drag, S , thus avoiding the nonlinearity factor en-
countered in many real filtering applications. Although Stinessen in-
troduces no new concepts, he does correctly surmise that K or K should
depend mainly upon dust cake and fluid properties. Furthermore, he does
point out that misleading values for K will obtain until the cake under-
goes "repair." In effect, a repaired cake is one that displays a nearly
uniform dust deposit density over the entire filter surface. Stinessen
did not include provisions in his model for predicting particle emissions
properties.
24
-------
Fraser and Foley Model
23
Fraser and Foley'' have also presented a predictive model for single bag
or single compartment performance. Their basic equation again assumed
the classical form:
S_ + K« | cV E dt
E
except that average collection efficiency, E, was introduced. The latter
refinement appears unnecessary, however, because few practical fabric
filters operate much below 98 percent efficiency.
At the time of their modeling effort, the authors found it necessary to
depend upon the best available data in the literature (which did not
provide strong support). The major failing in the Fraser and Foley model,
however, is that it attempts to treat a fabric filtration process as a
highly specialized bulk fiber system. Thus, a complex series of empirical
corrections are applied to the filtration theory for high porosity filters
to explain the performance of a woven fabric filter. Considerable effort
is also devoted to determining how much dust must fill the filter void
volume, when in fact most dust captured by a filter resides upon the sur-
face with a relatively shallow interstitial penetration. The net result
is that no successful application of these models can be expected unless
they are applied to situations that replicate the conditions used to de-
velop the modeling equations.
Leith and First Model
By using tagged fly ash aerosols, Leith and First were able to distin-
guish between those fly ash particles which, under laboratory conditions,
penetrated a needled felt fabric filter immediately and those particles
which exhibited a delay in their penetration. These researchers
25
-------
postulated three types of penetration mechanisms: direct penetra 10 ,
gradual seepage of the dust, and the breakage and penetration of plugs
of material in the vicinity of pinholes. To summarize the results of
their work, we quote:^4
"Penetration by 'straight through' dust loss was found to fall
off rapidly after cleaning, to reach a minimum, and then to
increase. 'Seepage' of dust through the fabric was found to
be constant throughout the filtration cycle. Dust loss as
'pinhole plugs' was found to increase after cleaning, to pass
through a maximum, then to decline. The pinholes appear to
open the way for further emission by the 'straight through
mechanism."
The experiments were performed at face velocities from 5 cm/s to 15 cm/s
and for dust cakes up to 60 ym thick. These velocities are higher than
those in normal use for the filtration of fly ash (~ 1 cm/s) and these
cake thicknesses are rather low. The efficiency as a function of par-
ticle size was such that the penetration was found to "remain relatively
constant for particles from 0.3 to 4.0 micrometers in diameter." The
penetration increased with face velocity.
The results quoted from the abstract are for the relative contributions,
rather than the absolute mass flux for the different mechanisms. From
their figures for mass flux versus time (at 10 cm/s), one would conclude:
1. Straight through mass flux seemed to decrease (roughly exponen-
tially) with deposit thickness, but may have gone through a
minimum near 20 ym thickness.
2. Seepage mass flux remained fairly constant with time and
deposit thickness.
3. Pinhole plug mass flux decreased with increasing deposit
thickness.
The fraction of the total penetration which was due to pinhole plugs
and seepage was greater than the direct (straight through) contribution
(at 60 ym thickness) for 15 cm/s but substantially less than the direct
for 10 cm/s, which suggests seepage and pinhole plugs might be very much
less than the direct at 1 cm/s.
26
-------
Leith and First postulated the following functional forms for the three
types of penetration:
-aXb
1. Direct penetration proportional to e , where X is the
deposit thickness.
2. Constant seepage mass flux versus thickness of deposit.
~cX
3. Pinhole plug mass flux proportional to Xe
where a,b,c are proportionality factors. The correlations they found in
using these equations ranged from 0.86 to 0.91 correlation coefficients,
for mechanism-by-mechanism comparison, adding support for their proposed
mechanisms.
Leith and First found only a weak particle size dependence for efficiency
with the efficiency decreasing slightly as particle size increased. The
particle size dependence, the velocity, dependence, and their general
appraisal of the filtration process led them to conclude:
"Because media filtration theory does not describe the trends in
penetration found in a fabric filter, and was not developed for
the operational conditions found there, it should not be used to
predict or interpret the penetration characteristics of fabric
filters."
CONCLUSIONS
1. All but one of the models for fabric filtration reviewed here
used a linear dependence for drag versus fabric dust loading
(in weight per unit area). Such models ignore the possible
effects of the zone of cake repair in the drag versus loading
curves and they lack a means for predicting the effective
drag, SE.
9 "3
2. Except for the models of Fraser, et al. , and, to an extent,
Leith and First, ^ the work thus far has not attempted to
predict collection efficiency. The Fraser model to predict
efficiency relied on a questionable combination of the concept
of effective diameter with an equation for the effects of mutual
fiber interference. The Leith and First model is supported
by evidence from tests with unusually high face velocities,
no
3. With the exception of the work by Fraser, et al. J values of
K.2 and S-g (S-g) were assumed or obtained experimentally rather
than derived from a predictive analytical equation.
27
-------
Improvements in the state-of-the-art should include the following:
1. Further development of models which apply to the nonlinear
portion of the drag versus dust loading relationship.
2. Analysis and prediction of the parameters K2, Sg, SR, So,
based upon at least semitheoretical equations rather than
purely upon correlations.
3. Formulation of collection efficiency relationships starting
from another basis other than isolated fibers in a flow and
including such facets as collection by dust already captured,
dislodgement and flow through pinholes.
28
-------
SECTION IV
LABORATORY TEST EQUIPMENT AND MEASUREMENT PROCEDURES
FOR DETERMINATION OF FILTER PERFORMANCE
BENCH SCALE FILTRATION EQUIPMENT
The laboratory program was designed so that filter performance tests in-
volving fabric resistance and particulate retention characteristics could
be carried out on either bench or pilot plant scales. Although the bench
approach is always attractive, it was recognized that in those cases where
dimensional or dynamic similarity could not be satisfactorily attained
with small scale apparatus, it would be necessary to resort to the pilot
approach in which the filter bags and system operating parameters would be
essentially full scale at least on a single bag basis.
Because the bench approach affords the potential advantages of reduced
testing time, higher measurement precision, less expensive equipment, and
less space, a special test assembly was fabricated for this program in
which the filtration area was reduced to a 15 cm x 23 cm (6 in. x 9 in.)
3
flat test panel and the system air flow rate reduced to 0.0213 m /min
3
(0.75 ft /min) at 0.61 m/min (2 ft/min) filtration velocity. Sufficient
flexibility in fan capacity was provided to operate at air to cloth ratios
up to 6.0.
There was no special reason for selecting a nominal 6 in. x 9 in. filter
area except that stainless steel filter holders used routinely for sus-
pended particulate sampling were available. By fabricating a rigid, steel
picture frame assembly as the actual filter holder (Figure 5), a vehicle
29
-------
was provided by which clean or used filter panels could be easily removed
and subsequently replaced after weight determination or microscopic
observations.
In order to maintain some semblance of similarity with respect to full
scale systems, the filter medium was installed in the normal vertical
field position with no physical support or backing behind the fabric.
Because the air approach to most filter bags consists of a parallel flow
either inside or outside the bag whose initial velocity is characterized
either by the ratio of bag volume flow to bag cross section (or external
separation distance between bags), a flat distribution manifold section
was installed upstream of the filtering surface as shown in Figure 5.
The photographs of the test equipment shown in Figures 6 and 7 provide
more details on the experimental system. The depth of the manifold,
2.5 cm, was reduced as much as possible so that the vertical velocity
component of the entering aerosol would be sufficiently high to support
all fly ash particles less than 30 ym diameter.
With respect to size distribution measurements of the GCA fly ash by
(a) Andersen impactor before any appreciable particle fallout or (b) by
light field microscope examination of an oil resuspension of the parent
dust, Figure 8, it appeared that greater than 99 percent of the dust was
represented by particles less than 30 ym in diameter. Therefore, a
negligible fraction of the particle mass would fail to reach the filter
when the average air velocity at the base of the filter panel is
6.8 cm/sec. The latter velocity corresponds to an air to cloth ratio
of 2/1. At the filter midpoint, the average rise velocity of 3.4 cm/sec
would fail to entrain only those particles greater than 23 ym in diameter
(roughly 2 percent of the entering dust).
On the basis of the above analysis, it does not appear that the somewhat
lower vertical rise velocities of the bench scale system requires special
consideration in data treatment. The dimensions of the hopper beneath
the level of the filter face were selected to provide gas retention times
30
-------
INLET SAMPLER
FILTER OR CASCADE
IMPACTOR
INLET MANIFOLD
VIEW PORTS
STATIC PRESSURE
TAPS
EXIT MANIFOLD
FROM DUST
GENERATOR
•EST AEROSOL LOOP
OUTLET SAMPLER
FILTER
•TO FLOW METER
AND PUMP
TO CNC
TO B AND L COUNTER
HOPPER STORAGE
NOTE'PARTICULATE SAMPLING AT S,,SZANDSS
Figure 5. Schematic of filter test assembly with exploded view of fabric sandwich
-------
Figure 6. Bench scale filtration apparatus showing
inlet manifold and test aerosol loop
Figure 7. Bench scale filtration apparatus
32
-------
100 n-
70 h
50 \-
I
t-
UJ
-J
o
fe
<
CL
a:
UJ
l-
UJ
1 '0
7
5
- i i i ii
CURVE
—i—i 1 r
DESCRIPTION
PARENT DUST REDISPERSION IN OIL
ANDERSEN IMPACTOR MEASUREMENT IN UPSTREAM DUCT-
BEFORE SETTLEMENT LOSSES
\ Si II!
1 It I L
• B
O.I
10 30 50 70 90 95 98 99
PERCENT MASS < STATED SIZE
99.9
Figure 8. Size distribution measurements for GCA fly ash, for particle density
of 2 grams/cm-^
-------
in the 0.2 minute range to simulate field hopper settlement conditions.
This allows for the typical selective removal of the coarser particles
from the air stream. At the present time, roughly 50 percent of the
solids entering the hopper falls to the collection jar located beneath it.
By means of this inlet system, it is possible to obtain a good solid ma-
terial balance. The dust deposits either on the fabric surface or falls'
to the collection jar at the bottom of the hopper. Fortunately, wall dis-
position has proven to be minimal. The manifold geometry allows for
installation of a glass window for observing and photographing the filter
surface during a test and also makes it possible to sample the inlet
aerosol at several locations. The inlet pipe to the hopper was designed
with a diverging section to reduce the chance of particle impaction on the
opposite wall of the hopper.
Attention is called to two singular disadvantages of flat test panels as
compared to the usual cylindrical bag configurations. First, it is
nearly impossible to pre-tension the panels without going to an impracti-
cally complex apparatus. More important, however, is the fact that appli-
cation of aerodynamic pressure causes the panel to dish inward so that a
uniformly sized pore structure cannot be maintained because of the warping.
In contrast, the pores in a tubular configuration such as a bag filter will
undergo simultaneous and uniform changes under a tensile loading generated
by pressure gradient. As pointed out later in this report, the size of
pore openings may actually decrease with increased filter load and an
extreme lack of uniformity in pore sizes may lead to very poor particle
collection.
No problems were encountered in working on the clean air side of the
bench scale filter system. Several probes could be introduced to the
downstream converging section of the filter holder for particulate
sampling or pressure measurements. To measure average effluent concen-
trations (mass basis) the entire fabric filter effluent was passed through
an all-glass filter prior to flow measurement. Ordinarily, several hours
34
-------
were required to collect weighable effluent samples when high fabric
efficiencies prevailed. Therefore, a condensation nuclei counter (CNC)
and a Bausch and Lomb single particle light scattering counter (B&L) were
used to indicate system performance against fine particles over brief,
^ minutes, time periods.
DUST GENERATION APPARATUS
25
An NBS dust generator was constructed to provide an accurately regulated
dust feed to the system at a working range of 0.1 to 2 grams/minute. This
device, Figure 9, consists of a small hopper, ^ 200 grams capacity, that
discharges to a slowly rotating spur gear located below it. By adjusting
the rotation rate (^ minutes) and the clearance between the hopper dis-
tributing plate and the gear teeth, dust is transported to an aspirating
tube leading to a compressed air ejector. A clean, dried compressed air
3
supply of about 3 ft /min at 50 psig, which entrains and shears the dry
dust at sonic velocities within the nozzle, provides the test aerosol
system from which the desired volume is extracted by the fabric filter
pump. Excess aerosol is vented to a waste gas treatment system. By pro-
viding a separate test loop for the aerosol generator, the flow requirements
for the filtration process are uncoupled from the stringent flow regulation
requirements of the dust generator. Because the generator system operates
under positive pressure, it augments the fabric filter fan system such
that the negative pressure behind the fabric filter is seldom more than a
few inches of water. This prevents sampling difficulties with the CNC
equipment which is not designed for sample extraction from negative pres-
sure regions. Figure 10 shows all the components of the bench scale system
as assembled for testing.
PILOT SCALE FILTRATION EQUIPMENT
A pilot scale fabric filter system was used to make measurements which were
impossible or impractical to make on a field scale system. Data to sup-
plement and verify the bench scale tests were also obtained with the pilot
35
-------
Figure 9. NBS type dust generator
36
-------
Figure 10. All components of bench scale filtration system
37
-------
scale equipment. The pilot scale system simulates the full scale system
geometry in that it has a normally tensioned cylindrical bag instead of a
slack panel. In addition, full scale cleaning operations can be performed
on the pilot scale system.
The apparatus was operated at flow rates, dust concentrations and with
fabrics selected to represent typical field applications. To obtain
accurate measurements, the pilot scale system was designed for startup,
normal filtration and shutdown with a minimal disruption of the filter
cake. Test measurements included the following: average mass effluent
loading, instantaneous counting of effluent particles, average size pro-
perties for the effluent, the determination of the mass of particulate
removed from the bag during cleaning and location of dust dislodgement
sites during cleaning. Dust generation was performed with a commercial,
auger type, feeder and a high pressure (90 psig) air ejection nozzle to
attain maximum dispersion of the bulk fly ash.
The basic pilot scale fabric filter system was developed at GCA/Technology
Division during a previous study.10 A schematic of the pilot scale system
as modified for this study is shown in Figure 11. Some of the important
design features of the system are: the by-pass loop which permits the
initiation and termination of flow to the bag with a minimum of system
flow excursions; a Plexiglas cylinder to catch dust dislodged from the bag,
thus permitting determination of the mass of dust removed and the time of
removal; a removable filter housing which was made of from flexible hose
to allow its removal without disturbing the bag and an 8-foot fluorescent
lamp (not shown in Figure 11) that was installed within the bag to allow
for observations of the bag surface. A turnbuckle located between the cap
and spring assembly and the load cell was used to adjust the bag tension.
38
-------
<£> FLOW ME TEA
{§) SAMPLING LOCATION
OJ
vo
MAIM FAN
MANUAL VALVE
AUTOMATIC VALVE
FLY ASH HOPPER ond FEEDER
I FEED CONTROL
110 ft LONG * 4 in 01A BAG
I CLOSED FILTER HOUSING
I REVERSE FLOW FAN
| CAP and SPRING ASSEMBLY
|LOAD CELL
) DUST HOPPER
) CYLINDER TO CATCH DUST AFTER CLEANING
) OUTLET SAMPLING PORT
BY- PASS LOOP
AIR EJECTOR
DUST PICKUP
BY-PASS EXHAUST
BAG PRESSURE
DROP
RECORDER
FLOW CONTROL
and RECORDER
Figure 11. Schematic of pilot scale fabric filter system
-------
Filtration parameters were selected to typify field operations. The pilot
system was operated during the testing program at a constant flow of
'mj
3
o
0.498 m /min (17.6 acfm) which provided a face velocity of 0.61 m/inin
(2 ft/min). The inlet dust loading ranged from 6.9 to 8.0 grams/m'
Q
(3.0 to 3.5 grains/ft ). An Andersen impactor positioned to sample the
dust entering and leaving the bag was used to determine the inlet and out-
let size properties. The cumulative size distribution for both inlet and
outlet were the same with aerodynamic mass median diameter and geometric
standard deviation of 5.8 ym and 2.42 jim, respectively, Figure 12.
The bags studied were manufactured by Menardi-Southern Company from
Teflon-coated fiberglass cloth. Manufacturer's specifications for the
fabric material are listed below:
2
• weight = 9.5 oz/yd
• thread count = 54 x 30
• weave =3x1 twill
r\
• Frasier permeability = 75 cfm/ft
• Mullen burst strength = 595 psi
The dimensions of the bags were 10.16 cm (4 in.) diameter by 304.8 cm
(10 ft) length with five equally spaced antideflation rings.
Bag cleaning was accomplished by reversing the flow through the system,
thereby causing the bag to collapse. The normal reverse flow produced
a face velocity of 0.52 m/min (1.7 ft/min). After the normal filtering
(loading) portion of the test, the dust feed was stopped; valve B of the
by-pass loop was opened and the main flow valve A was closed. Cleaning
was initiated by starting the reverse flow fan and opening valve C over a
period of 2 seconds. Valve C was kept fully open for 56 seconds and then
closed over an additional 2-second period, thus completing the cleaning
cycle. Valve D was preset to give the desired flow rate. The above
cleaning regimen was chosen to replicate the field operating system used
at the Nucla, Colorado plant. Immediately after cleaning, the dust
40
-------
20
10
o
UJ
I-
UJ
5
o
o
o
-------
collected in the catch cylinder was removed and weighed. The resumption
of normal flow was accomplished by slowly opening valve A and slowly
closing valve B. This step eliminated the problem of flow surges in the
system which would have caused abnormal flexing of the bag.
Filtering at several velocities was accomplished by inserting a plug into
the bag which blocked off selected regions of the fabric. Thus, by main-
taining a constant volume flow rate, it was possible to increase the face
velocity to any desired level. Without the plug, the normal face or filtei
velocity was approximately 0.61 meters/min (2 ft/min). This approach was
more desirable than changing the flow in the system because it did not
alter the particle size properties of the inlet dust.
TEST AEROSOLS
The simulant aerosols used during these tests consisted of resuspensions
of a GCA fly ash obtained from a coal-burning power plant, rhyolite,
a type of granite used in shingle manufacture and fly ash obtained
from a lignite-fired power plant. The coal ash, which was recovered from
electrostatic precipitator hoppers, was finer than the usual pulverized
coal product because of the fractionating characteristics of cyclone-
fired boilers. The size properties of the GCA fly ash as dispersed by
NBS type dust generator are given in Figure 13. It appeared that no
significant change in size parameters took place as the dust traveled
from the Si to the 83 sampling stations shown in Figure 5. The mass
median diameter (MMD = 9 pm) and geometric standard deviation (a =3.0)
g
indicated in Figure 13, fell within the band for similar measurements
performed during the evaluation of the Sunbury filter system.^ Although
the Sunbury fuel consisted of a mix of anthracite fines, No. 5 buckwheat
and petroleum coke, its size properties, Figure 14 actually appeared very
similar to the GCA test fly ash. Therefore, it is believed that much of
the test data deriving from the current laboratory studies with GCA fly
ash can be used directly to support the field measurements. It was also
42
-------
IT
UJ
H
UJ
Q
O
2
<
Q
O
a:
UJ
0.5
0.2
DESCRIPTION
SAMPLING POINT, S|
SAMPLING POINT, S2
SAMPLING POINT, S3
'SEE FIGURE
LOCATIONS
FOR SAMPLING
NOTE:SHADED REGION SHOWS RANGE FOR
SUNBURY FIELD AEROSOL SIZE DISTRIBUTIONS.
JL
10 20 30 50 70 80 90
PERCENT MASS < THAN STATED SIZE
95
Figure 13. Size distribution for GCA fly ash entering bench scale
filter system, Andersen in-stack impactor measurements
43
-------
20
10
o
UJ
TESTNO.
• 3b
O 2o
• I4b
n 80
A 3°
i • i ------
m-i
o
o
o
o
-------
noted that the size parameters for the Sunbury effluent were nearly the
same as those for the inlet.
The fly ash size properties for the pilot system were approximately the
same as those for the bench tests. Dust dispersion was accomplished with
an auger-type Acrison feeder in conjunction with a 90 psig air ejector.
No significant differences between inlet and outlet size distributions
were noted as shown in Figure 12. This observation has played an important
role in explaining filter performance.
Similar size measurements by Andersen cascade impactor for the rhyolite
(granite) and lignite test dusts are shown in Figures 15 and 16. Under
normal testing procedures, the redispersed granite and lignite dusts were
slightly coarser with mass median diameters of 15 ym and 12.5 ym,
respectively.
A special, but very simple, extraction technique was used to provide
much finer rhyolite particles. By reversing the manifold extraction
probe (180° from isokinetic) the mass median diameter for the rhyolite
was reduced to 2 ym, Figure 15. The object of this procedure was to
provide radically different size parameters for a specified dust for
which prior analysis had indicated that chemical composition, density,
shape factor (and other physical properties) were essentially invariant
with respect to size. Under the above circumstances, the effect of dust
size parameters alone upon specific resistance coefficient could be
established.
PARTICULATE SAMPLING AND ASSESSMENT
Basic Sampling Equipment
The selection of instrumentation for determining mass concentrations,
efficiencies and particle size properties was based mainly on the equip-
ment used in prior EPA, GCA or other EPA sponsored programs.
45
-------
20
6
4.
cc
UJ
h-
UJ
2
UJ
_l
o
I-
tr
o
5
O
o
DC
UJ
0.5
1—|—-j—TTT
SYMBOL TEST* DESCRIPTION
r—i—r
-0,A 77
— * SEE TABLE 18
I 5 10 30 70 80 90 95
PERCENT WEIGHT< THAN STATED SIZE
99
Figure 15. Size properties for coarse and fine rhyolite dust
-------
10
5r
z
o
f-
2
cc
i-
co
UJ
_J
O
(T-
<
Q.
I —
0.5 -
0.2
^»-
MMD =l2.5/im
=2.5
O NUMBER CONCENTRATION
LIGHT FIELD MICROSCOP
• RUN & 80,ANDERSEN -
IMFACTOR, MASS
DISTRIBUTION
=2.5
NOTE^DASHED LINE SHOWS MASS
DISTRIBUTION DERIVED FROM
INDICATED NUMBER DISTRIBUTION
O.I
0.01 0.1 125 10 20 50 80 95
PERCENT < THAN STATED SIZE (NUMBER)
99
99.99
Figure 16. Particle size properties for lignite ash from precipitator
hopper
-------
Four basic sampling methods have been used:
• All-glass, Method 5 type filters for determination of
both inlet and outlet mass concentrations. The only
disadvantage to this approach is the long-time period
required to collect weighable dust quantities. Thus,
one can seldom detect important changes in concentra-
tions that aid in describing the cleaning process.
• Andersen, in-stack type, cascade impactors for esti-
mation of size properties. For a fixed aerosol
system, this technique affords reasonable precise
estimates of mass distribution for the central, 90
percent region, of the size range. Very high or
very low concentrations present the respective prob-
lems of stage overloading or very long sampling periods.
• Bausch and Lomb Single Particle Light Scattering
Counter (B&L) for number concentration and particle
size distribution. Although its accuracy may be
questioned, this instrument can provide time resolu-
tions down to 0.1 minute insofar as reflecting
changes in number concentrations for particle diam-
eters in the 0.3 to 5 ym range. Prior GCA studies
have indicated that mass concentrations derived from
B&L data are usually lower than those determined by
parallel gravimetric sampling.
• Condensation Nuclei Counter for detecting number
concentration changes in the very fine, 0.0025 to
0.5 pm, diameter range. Although one may dispute
the absolute concentration values, the capability
of this instrument to follow concentration changes
over brief, rv seconds, time periods makes it a
useful adjunct to the B&L system that traces
changes in the ^ 0.5 ym particle size range.
Assessments and Interpretation of CNC and B&L Measurements
ft
A condensation nuclei counter, CNC , and a single particle light scattering
counter, B&L+, were used extensively to delineate the rapid changes in
Model Rich 100 Condensation Nuclei Monitor manufactured by Environmental
One Corporation, Schenectady, New York.
Model 40-1 Dust Counter manufactured by Bausch & Lomb, Rochester New York.
48
-------
effluent mass concentration and particle size distribution that take place
over a filtration cycle (here defined as the period between resumption
and termination of filtration). Periodic measurements were made on the
background laboratory air throughout the testing programs to ascertain
that the instrument performance characteristics remained unchanged.
Additionally, these tests provided background data for nuclei concentra-
tions (which, according to the CNC manufacturer, indicated particles less
than the size range 0.3 ym to 0.5 urn and greater than 0.0025 ym.
The above nuclei always constituted a small fraction of all test aerosols
unless the ambient air underwent special filtration prior to entering the
dust generating system.
Reference to Figure 17 indicates that the size properties for the ambient
atmospheric dust did not change greatly over the testing period. The B&L
measurements showed that the number median diameters, NMD, ranged from
0.3 to 0.4 ym and the geometric standard deviations, a , from 2 to 2.5.
o
These results were in fair agreement with light field microscope sizing
data for atmospheric dust, 0.3 to 0.5 ym NMD and a a value of 1.5 to
O
2.0, depending upon the dust generating activity in the area.
Those measurements depicted within the shaded region, Figure 17, represent
the usual range of size parameters observed over the testing intervals.
The calculated weight concentrations associated with each of these curves
were developed by converting the fractional number concentrations to
their equivalent weights by assuming that the particles were spherical
2
with a density of 1.91 g/cm . The above density was selected so that
in combination with the shape factor of ir/6 for spheres, particle mass
3
in grams would be expressed directly as D .
A separate graphing of the parallel CNC counts versus the matching weight
concentrations derived from the B&L measurements is shown in Figure 18.
Those points that fell outside the dashed envelope lines were, with one
49
-------
10
TEST SIZE CLASS
Oi
o
Ul
s
o
I-
CC
0.5
O
Cf
D
G/
A
A
V
O
x
•
b
94 AVERAGE
81
770-4 "
79 C
89-2
83-2
79A
89-1
83-1
93-1
79-B
84.-I
84-2
96
FINE
COARSE
COARSE
COARSE
COARSE
b
0.2
0.01
O.I
5 10 20 30 50 70 90
PERCENT BY NUMBERS STATED SIZE
95 98 99
99.9
99.99
Figure 17. Number size distributions for background (laboratory) dust based
on B&L counter measurements
-------
100
90
80
70
60
50
40
x
•°E20
u
c 10
h- 9
uj 8
8 6
a 5
i 1—i—i i
C=COARSE,SEE FIG. 17
F = FINE,SEE FIG. 17
NO SUPERSCRIPT
MEANS AVERAGE
CONCENTRATION
RANGE
= |.55)t jQ4n/cm3
I09n/cm3
O
O
3 4 5678910
20
40 60 80 100 322
WEIGHT CONC£NTRATION,i*9/m%
Figure 18. Relationship between nuclei concentrations by CNC measurements
and weight concentrations derived from B&L data
51
-------
exception, associated with tests where the dust was either coarser or
finer than the average background aerosol (shaded region, Figure 17).
The approximate 45° slope displayed by the data points within the envelope
shows that the nuclei concentrations are directly proportional to the
weight concentrations, as they should be, when the size properties of the
atmospheric dust are fairly constant. Because the complete B&L size
spectrum was used to estimate mass concentration values, including the
relatively few coarse particles that exert a large influence on the
sample weight, it is believed that both the CNC and B&L data outputs
were in reasonable agreement, at least on a relative basis. The calculated
weight concentrations derived from B&L measurements were, for the most
part, in good agreement with independent gravimetric measurements, 20 to
3
100 pg/m , in the GCA laboratory areas.
The few unsusually low values for the calculated weight concentrations,
Figure 18, are believed to be in error because of failure to sample the
coarse particles, > 5 ym, in the air stream because of anisokinetic
sampling conditions and/or line losses. For example, the sloughing off
of agglomerates in significant quantities from the clean air face of a
filter may produce a highly bimodal distribution in which large particles
are seldom detected by the B&L instrument.
As a result of extensive comparisons between effluent fly ash concentra-
tions determined concurrently by gravimetric (filter) sampling and CNC mea-
surements, it was concluded that the ratio of nuclei counts to mass con-
centrations was nearly a constant quantity irrespective of the concentra-
tion level. In the case of the previously cited comparisons between CNC,
B&L and filter measurements for atmospheric dust, it was expected that a
fixed proportionality would exist provided that the size distribution of
the ambient aerosol did not change.
It was deduced, therefore, that the entering fly ash aerosol underwent
no change in size properties after passing through the filter. The
52
-------
reason for this behavior, which appears to contradict all classical fil-
tration theory, is discussed in a later section of this report. At this
point, we only wish to point out that the existence of this very conve-
nient proportionality between nuclei and mass concentrations allows CNC
measurements to be used in conjunction with a calibration curve to deter-
mient changes in mass concentrations over brief, ^ seconds, time intervals.
The latter operation is essential if one is to make accurate forecasts of
particulate emission levels from sequentially cleaned, multicompartmental
filter systems.
Comparisons were also made between indicated nuclei concentrations and
B&L measurements with respect to the number concentrations in specific
size ranges, > 0.3 to 0.5 um and > 0.5 urn. The regression lines shown
in Figure 19 indicate a closer correlation between the finer size frac-
tion than that shown for the coarser, > 0.5 ym particles. These data
suggest properly that the nuclei counter, in accordance with its specifica-
tions, probably gives very little response for particles larger than
0.5 pm. The point scatter for both correlations results from the range in
size distributions occurring within the data set.
The sampling procedures described above, in conjunction with pre- and
post-drying and desiccation of samples in accordance with Method 5
protocol, represent standard EPA test methods. By weighing the fabric
test panels before, during, and after filtration tests, accurate estimates
of average inlet dust concentration and fabric loading were obtained.
Temperature and humidity measurements by recording hygrothermograph with
periodic checks by wet and dry bulb sling psychrometer were also included
with the instrumental methods used in this study.
TENSILE PROPERTIES
Figure 20 shows the bench scale apparatus used to determine stress/strain
relationships for the glass fabric under static loading conditions.
Horizontal clamps were secured to the top and bottom of a fabric strip
53
-------
100
Ul
o
§ 5
o
0.8 I
PARTICLE RANGE 0.3^m to O.S^m —-O
(A) LOG CCNC=0.329 LOGBaL+ LOG(6.9 x I03 )
PARTICLE RANGE >0.5^m A
(B) LOG CCNC=0.527 LOGBa|_4 LOG (1.78 x I04 )
NOTE-CONDENSATION NUCLEI
-------
Figure 20. Test apparatus for measurement of fabric tensile
properties
55
-------
(6 in. x 18 in. or 3 in. x 18 in.) so that any applied load would be
exerted uniformly over the width of the strip. A free floating ring
was attached to the loading cable to assist in distributing the load
evenly. Parallel scales on each side of the strip in conjunction with
pointers attached to both sides of the lower clamp furnished replicate
indications of fabric elongation under load. During the present test
series, the maximum applied tension was 380 N or 85 Ibs.
A similar system for static tension measurements was used for full scale,
(10 ft long by 4 in. diameter), glass bags prepared from the Sunbury
(Menardi) and Nucla (Criswell) fabrics. A strain gauge incorporated
within the hanger arm in conjunction with a turnbuckle adjustment allowed
for the determination of bag elongation as a function of applied tension.
This arrangement also permitted precise control of tension levels during
permeability and filtration tests.
56
-------
SECTION V
FABRIC STRUCTURE STUDIES
INTRODUCTION
Classical approaches to modeling filter performance frequently begin with
the clean (unused) fabric which is studied from the perspective of resis-
tance to air flow, dust retention characteristics and interstitial particle
deposits. Although tests with different, unused fabrics permit relative
comparisons, these measurements can seldom be extrapolated directly to
predict overall fabric performance under normal steady state filtration
and cleaning conditions. In the latter case, continued filter usage
followed by periodic cleaning leads to initial and terminal equilibria
for which characteristic filter drag and dust holding levels may be
assigned. The magnitudes of these terms are functions of both specific
dust/fabric relationships and the method of fabric cleaning employed. It
is emphasized that the path (e.g., filter resistance versus fabric dust
holding) by which one progresses from the residual to the terminal states
is seldom a simple linear function.
In addition, the manner in which the total filter dust loading is dis-
tributed over the fabric surface plays a controlling role in determining
the filter resistance/fabric loading relationship. Consideration of this
factor has enabled us to analyze the performance of both mechanical shaking
and bag collapse-reverse flow cleaning systems in terms of the same basic
variables.
A careful examination of fabric structure can provide several insights as
to the probably performance of many dust/fabric combinations. The
57
-------
O O £
previously cited work of Draemel and studies reported by Butterworth
27
and Pedersen consider both the aerodynamic and dust retention charac-
teristics of filters as functions of structure. Fabrics have been analyzed
in terms of free area, thread count, weave, nap, size distribution for
pore dimensions, and the yarn type such as number of strands, twist,
multifilament or staple. Although the correlations deriving from these
(structure) studies are frequently broad, particularly so with respect
to the tighter and denser weaves, they still represent useful data inputs
that can be readily obtained by simple laboratory microscopy.
Basic Manufacturer or User Specifications
Fabric properties as specified by the manufacture and/or user are given
in Table 2 for the Sunbury, Pennsylvania and the Nucla, Colorado power
plants. Despite the differences in fabric treatment, the two woven glass
fabrics are very similar. It was observed, however, that certain of the
descriptive parameters (Table 2) were not always internally consistent
nor the same as those measured by GCA. For example, despite similar
measurement techniques, clean cloth permeabilities appear to vary con-
siderably, ~ ± 30 percent. It is suspected that these differences depend
upon the fabric bolt from which the bag is made, the section of the bolt
from which the test specimen is removed, and the handling of the fabric
before and during the testing process. In view of these differences, it
does not appear advisable to depend heavily on any filtration parameter
derived from clean cloth permeability. Although these differences were
not large, Tables 2 and 3, they may, in certain cases, be important in
determining fabric performance.
The mixture of English and metric units in Table 3 is a result of commer-
cial.fabric descriptions being given in English or specialized textile
units. Thus, replicate GCA measurements are also given in English units.
Where no comparisons are made, however, metric dimensions have been
assigned to such parameters as yarn dimensions and fabric thickness.
58
-------
Table 2.
FABRIC PROPERTIES FOR GLASS BAG FILTERS USED AT SUNBURY,
PENNSYLVANIA AND NUCLA, COLORADO COAL'BURNING POWER PLANTS
Dimension, length x diameter - ft x in.
Fabric weight - oz/yd2
Weave
Warp (w) yarn
Fill (f) yarn
Yarn (thread) count - w/in. x f/in.
Permeability at 0.5 in. - ft3/min
Primary application
Fabric treatment
Manufacturer and fabric designation
SL
Sunbury
30 x 12
9.2
3x1 Twill
150' s 1/2
Bulked 1/4
54 x 30
54. 3b
Reverse flow
Teflon coating
Menardi Southern
601 T(Tuflex)
Nucla
22 x 8
10.5
3x1 Twill
Multifilament
Bulked staple
66 x 30
86.5°
Shaking and
reverse flow
Graphite-
silicone
coating
W.W. Criswell
No. 640048
Seven anticollapse rings.
bGCA test, Perm at 0.5 in. = 42.5 ft3/min.
°GCA test, Perm at 0.5 in. = 112 ft3/min.
59
-------
Table 3. SPECIAL PROPERTIES, SUNBURY AND NUCLA FABRICS
CTs
o
Fabric weight
oz/yd2
Fabric thickness
pirn
Yarn count
per inch
warp/fill
Warp weight
oz/yd2
yds strand
x 10~2/lb yarn
Fill weight
oz/yd2
yds strand
x 10~2/lb yarn
Warp yarns
Max/min diaro, (im
Fill yarns
Max/min diam, n«n
Sunbury fabric
Menardi
Southern
parameters
9.2
-
54 x 30
4.35
-
4.85
-
-
-
GCA
measurements
9.88
400
53.4 x 30.5
4.67
-
5.21
-
450/200
650/200
Menardi
Southern
data, GCA
calculations
9.2
-
56.6 x 31. 6a
4.35
150 - l/2b
4.85
150 x l/4b
-
-
Nucla fabric
W.W. Criswell
parameters
10.5
-
66 x 30
-
-
-
-
-
-
GCA
meas ur eraen t s
9.4
400
66 x 30
4.29
"
5.11
-
375/200
600/200
thread count derived from (b) and weight of 9.2 oz/yd2.
150 x 102 a yards of strand per pound of yarn.
1/2, 1/4 indicate 2 and 4 strands (plys) per yarn.
-------
With respect to clean fabric weight, the differences may be attributable
to variations in protective coating because GCA and manufacturers values
for yarn count were in good agreement. The weight of yarn representing
warp and fill densities was determined for this study by the microbalance
weighing of 50 to 100 individual yarns of 5 cm length to determine the
weight per unit length. These data, in combination with the measured
yarn count, provided the GCA fabric weight values given in Table 3. It
is not clear why the predicted thread counts for the Menardi Southern
fabric (Sunbury) are significantly higher when estimated on the basis of
a 9.2 oz yd2 weight and the GCA yarn parameters (strand weight per unit
length).
Fabric properties for the cotton and Dacron media tested in this study
are given in Table 4; Although the above materials would not be used
for hot fly ash filtration, they have been evaluated in earlier GCA tests
in the form of 10 ft x 6 in. or 10 ft x 4 in. bags with conventional
mechanical shaking. Thus, by conducting similar tests with 6 in. x 9 in.
flat panels it was possible to ascertain whether the bench test geometry
had any significant effect on performance parameters. At the same time,
it was expected that any unique interaction between a given dust and
various fabrics would be revealed.
Bag Resistance Versus Pore Velocity
A special sequence of measurements was made to determine fabric resistance
levels at very high pore velocities, Figure 21. The object of these tests
was to establish reasonable estimates of the maximum pore or pinhole vel-
ocities when the filter pressure loss is relatively high, ~500 to 750 N/m2
(2 to 3 in. HO). In the case of the Sunbury fabric whose free area was
estimated to be about 3 percent, the pore velocity is 33.3 times the face
velocity. Use of Figure 21 in conjunction with the minimum cross sectional
area of a pore or pinhole will indicate the volume of air passing through
a pore at a specified pressure drop.
61
-------
Table 4. DACROH.AND COTTON PROPERTIES FOR FABRIC TEST PANELS
STUDIED IN LABORATORY
Sateen weave
cotton-unnapped
Dacron
crowfoot
Panel dimension - in. x in.
Fabric weight - oz/yd2
Weave
Warp (w) yarn
Fill (f) yarn
Yarn (thread) count
w/in. x f/in.
Permeability at 0.5 in.
H20 - ft3/min
Design application
Manufacture and fabric
designation
9x6
10
Sateen
Staple
Staple
95 x 38
13
Mechanical shaking
Albany International
No. 960
9x6
10
1/3 Crowfoot
Multifilament
Bulked staple
71 x 51
33
Mechanical shaking
bag collapse
Albany International
No. 865B
Simplified Weave Representations
A schematic drawing of the Sunbury fabrics in accordance with textile
conventions is given in Figure 22. The original bags installed in 1973
were characterized by a right-hand diagonal as depicted by the warp
yarn surfaces seen on the filtering face. On the other hand, the replace-
ment bags installed in 1975 were woven with a left-hand diagonal. Although
this variation had no apparent bearing upon filter performance, it required
that care by exercised in interpreting microscopic images with respect to
pore shape and location. Note that warp (vertical) and fill yarn (horizon-
tal) alignments and pore locations are indicated in Figure 22. The bags
used at the Nucla, Colorado power station were also fabricated from a 3/1
twill weave with a left-hand diagonal, as shown in Figure 22,
62
-------
,10'
icr
MAXIMUM PORE VELOCITY
I02 o , I03
5 -
UJ
2 5
UJ
cc
(£.
m
10'
I I I 1 I l T
0.305 m/min ^Ift/min
_._!__ J J J I
10"
jO ^ 5 |0i
FACE VELOCITY, m/inin
Figure 21. Resistance versus face and maximum pore velocity for
dean (unused) Sunbury glass fabric
63
-------
A. FILL (FILTERING) FACE 3/1 TWILL WEAVE,
RIGHT HAND DIAGONAL
Q-
CC
©•
e •/-o
/.
©
FILL
B. FILL (FILTERING) FACE, 3/1 TWILL WEAVE,
LEFT HAND DIAGONAL
Figure 22. Textile schematic drawing of Sunbury fabrics A.
1973 bags, B. 1975 bags. Circles on diagonal,
warp yarn crossovers, indicate open pore locations
64
-------
A simplified version of the appearance of the fill (filtering) face for
the Sunbury media is given in Figure 23. Except for the differences in
pore sizes, Figure 23 applies equally well to the Nucla fabric which is
also a 3/1 twill weave. The average spacing between all fill yarns was
200 um whereas all warps yarns were contiguous except for a 27 ym separa-
tion at yarn crossover points. The locations of the three characteristic
pore types are shown by the encircled areas. Note that type III pores
are blocked by virtue of the contacting warp yarns.
In Figures 24 and 25, photomicrographs of warp and fill faces for the
Sunbury and Nucla fabrics, respectively, are shown at 20Xmag. The warp
faces for both fabrics show clearly the smooth, compact appearance of the
multifilament warp yarns which, with a 3/1 twill weave, occupy approxi-
mately 75 percent of the downstream (clean side) fabric surface. On the
other hand, the bulk staple constituting the fill yarns presents a rela-
tively loose structure in which a large fraction of the individual glass
fibers (about 7.5 to 8.0 ym diameter) are separated from one another.
The graphite in the Nucla surface coating is responsible for the black
metallic luster of the yarns, Figure 25.
The density and porosity is relatively easy to establish for the multi-
filament yarns because the fibers are tightly twisted. Assuming that the
spinning process layers the parallel fibers in a 60 offset array, the
porosity is only about 10 percent. It is apparent that with void spaces
or interyarn porosities of the order of 50 percent, air flow through high
density warp yarns will be inconsequential. On the other hand, the bulked
or fluffy character of the fill yarns as indicated in Figures 24 through 27
provides an extended surface for aerosol permeation and particle capture.
PORE PROPERTIES
Microscopic viewing of the fabrics, Figures 24 and 25, indicated that there
were no spaces between the warp yarns except where they looped over the
fill yarns. Due to distortion of yarns (Sunbury fabric) by stressing at
65
-------
ttl
O_
DC
I
SPACE BETWEEN WARP YARNS
~27 pm AT CROSS OVER ONLY
SPACE BETWEEN
FILL YARNS
NOTE •• CIRCLED REGIONS INCLUDE
TYPES I, H and HL PORES.
Figure 23. Schematic of Sunbury fabric, filtering face, 3/1 twill, left-hand diagonal
indicating pore locations and average dimensions. No space between warp
yarns except at crossing points
-------
A. Warp surface
B. Fill surface
Figure 24. Warp and fill surfaces of clean (unused) Sunbury
fabric with substage illumination (20X mag)
67
-------
A. Warp sux>faee
B. Fill surface
Figure 25. Warp and fill surfaces of clean (unused) Nucla
fabric with substage illumination (20X mag)
68
-------
Warp yarns
Fill yarns
Figure 26. Individual Sunbury warp and fill yarns as seen in
plane of fabric showing maximum and minimum
dimensions (2OX mag)
69
-------
Warp yarns
Fill yams
Figure 27. Individual Nucla warp and fill yarns as seen in
plane of fabric showing maximum and minimum
dimensions (20X mag)
70
-------
crossover points, perceptible openings having an average width in pro-
jection of about 27 ym appeared at these locations. Since there was
a significant separation between adjacent Sunbury fill yarns of approxi-
mately 200 ym, Figure 24, slotted apertures or "see through" regions with
projected cross sectional areas of about 5.4 x 10~3cm2 appeared at each
pore location. Inspection of Figures 24 and 25 shows that open pores
exist only at warp/fill crossings. Hence, the Sunbury and Nucla fabrics
lose 25 percent of the potential pore count in both the warp and fill
directions. The net result is that the number of pores per in.2 appears as
(54-1) (30-1) (0.75)2 = 865
for the Sunbury fabric, and
(66-1) (30-1) (0.75)2 = 1060
for the Nucla media.
The thread counts are corrected by minus one because there always exists
one less pore than the number of bounding surfaces generating the pores.
Inspection of Figure 23 also shows that there are two Type II pores for
every Type I pore.
Yarn Shape
As near as can be ascertained, the warp and fill yarns for the glass
fabrics assume approximately elliptical cross sections typified by the
maximum and minimum diameters given in Table 3. By assuming elliptical
cross sections, however, misleading information are furnished with respect
to the true fabric interstitial volumes and true internal surface area
relative to skin friction. Hence, we have assumed a modified rectangular
cross section in which the ends are depicted as having the minor diameter
for the yarn cross section, Figure 28. Separate micrometer measurements
on the Sunbury media indicated a thickness of about 400 ym. This value
agrees with the thickness estimated by the sum of the minor diameters.
71
-------
The geometry shown in Figure 28 appears to be an acceptable representa-
tion of the actual yarn contacts according to the edge section photo-
micrographs shown in Figure 29. Because the yarns are deformable,
they are brought into intimate contact over large sections of their
surfaces. In the case of the Sunbury fabrics, adjacent warp yarns were
in direct contact except at crossing points as shown in Figure 28.
The average projected pore dimensions cited previously do not describe
the true minimum pore cross section. Actually, the interstitial geometry
is quite complex, even when the presence of protruding fibers and separated
strands and yarns are ignored (which is often the case).
Pore Type and Area
First, according to the fabric weave, there are several possible pore
types. In the case of the Sunbury, (Menardi Southern) fabric, three
distinct pore types are found, Figure 23, two of which, Nos. I and II,
constitute the passageways through which the air flows. The type III
pores represent closed cells or blocked passages for the Sunbury and
Nucla fabrics because there are no open spaces between adjacent warp
yarns except at the previously designated crossing points.
The sectional views shown in Figure 30 provide a better indication of
the effective cross sectional areas for the pores and their respective
orientations. Displacement of the warp yarns as shown for a type I pore
produces an opening between the bounding edges of the fill yarns resemb-
ling two apex-to-apex, truncated triangular openings. Furthermore, the
curvature of the fill yarns creates the additional areas which are con-
cealed beneath the surface of the fill yarns. The estimated cross-
sectional areas per effective pore shown in Figure 30 were attained by
rotation of the actual warped surface generated by the minimum separation
distance between yarns into the same plane. In the present case the
error introduced by this approach for calculating the area of a warped
surface was estimated to be less than 10 percent.
72
-------
W=WARP YARN
F =FILL YARN
P = PORE
SEE FIGURE 30
FOR DETAILS
TYPE II
PORE
FILL FACE
Figure 28. Schematic drawing showing alignment, approximate
form, and spacing of yarns and pores in Sunbury
filter bags (Menardi Southern woven glass media)
73
-------
FU:
Figure 29. Edge views of clean Sunbury fabric (20X mag)
-------
200 pm
200 pm
TYPE I PORE
B'
SHADED REGION IS WARPED
NOTE: PORE SECTION ROTATED
INTO A SINGLE PLANE
200 pm
TYPE H PORE
Figure 30. Schematic drawing showing idealized alignment of
parallel yarns and maximum por£ cross section
(shaded area)
75
-------
The development of the contours for a type II pore followed the same
process. For purposes of simplification, the yarns show an abrupt rather
than a smooth transition as they displace from top to bottom locations.
The additional expansion areas extending beyond the 200 ym gap between
fill yarns have been treated as triangularly shaped elements because of
the difficulty in establishing the true contours.
Air Flow Through Pores
The analysis of pore dimensions and general yarn structure should permit
rough estimates of the probable performance of fabric filters with respect
to clean media resistance to gas flow and particle removal characteristics.
Two approaches were used in conjunction with the fabric measurements dis-
cussed in this section to estimate probable resistance characteristics.
The first was based upon the average pore dimensions shown in Figure 28.
These values were calculated by using the yarn counts and maximum/minimum
yarn dimensions given in Table 3 coupled with the observation that there
are no spaces between warp yarn except at the type I and II locations.
If one assumes that the principal pore length is established by the 200 ym
space between each fill yarn, one can estimate the minimum pore cross
section from the schematic representations given in Figure 30. Because
the assumed pore boundaries appear (approximately) as a triangular and
truncated triangular or trapezoidal shapes, the hydraulic radii have been
computed in lieu of diameters for type I and II pore openings. For the
Sunbury fabric, the hydraulic radii for type I and II pores are 17.8 ym
and 17.9 ym, respectively, Table 5.
According to the Hagen-Poiseuille relationship, the pressure loss through
a cylindrical pore of the Sunbury fabric under laminar flow conditions
can be expressed by the following relation:
4
Ap = 8yQL/10trR
76
-------
Table 5.
CHARACTERISTIC PORE DIMENSIONS3 FOR SUNBURY (MENARDI SOUTHERN)-
AND NUCLA (CRISWELL) GLASS FABRICS
2
Cross-sectional area, ym
Perimeter, ym
Hydraulic radius, (M) ym
Equivalent pore radius, ym
Based on R = 2 M
Based on minimum pore area
f+
Measured resistance GCA tests,
in. water
Calculated resistance
in. water
£
Calculated resistance
in. water
Calculated resistance
in. water
Sunbury
Type I
pore
22,700
1,276
17.8
35.6
85.0
0.024
0.011
0.062
0.016
Type II
pore
19,050
1,065
17.9
35.8
77.8
0.024
0.015
0.074
0.018
Nucla
Type I
pore
24,130
1,368
18.4
36.8
87.5
0.009
0.008
0.045
0.011
Type II
pore
19,800
1,157
17.2
34.4
79.5
0.009
0.012
0.062
0.016
Based on analysis of Figure 30.
Average M values for type I and II pores - 17.8 ym for Sunbury and Nucla fabrics.
£
Measured values, GCA tests.
Calculated from Equation (1), R depicts circular equivalent of pore cross-sectional
area.
Q
Calculated from Equation (2), using M values.
Calculated from Equation (2), with V = V /2 and M = M . /2~
max/ mm
-------
2
where Ap = pressure loss N/m
-4
p = gas viscosity 1.84 x 10 poise
-3 3
Q = volume flow per pore 7.57 x 10 cm /sec
_2
L = filter thickness 4.0 x 10 cm
_ _3
R = pore (capillary) radius 8.5 x 10 cm
(based on minimum pore area)
Use of Equation (15), in conjunction with a pore radius derived from the
circular equivalent of the pore cross sectional area, provides estimates
of filter resistance that agree roughly with measured values. The actual
results for Sunbury and Nucla fabrics, respectively, show predicted values
50 percent lower and 33 percent higher than measures values.
Equation (15) may also be expressed in the form:
Ap = 2yVL/10M2 (16)
where V is the average pore velocity based upon the pore cross sections
given in Table 5 and the pore flow of 7.57 x 10~ cited above and M the
average hydraulic radius.
Estimates based upon Equation (16) showed resistances of 15.5 N/m (0.062
2
in. water) and 18.4 N/m (0.074 in. water), respectively, for types I and
II pores in the Sunbury fabric. GCA measurements with flat test panels
r\
15 cm x 23 cm (6 in. x 9 in.) indicated a pressure loss of 6.0 N/m (0.024
in. water).
Similar calculations for the Nucla fabric at the same air flow rate
32 2
(1.015 cm /sec/cm fabric), a pore count of 164/cm , and the pore
dimensions given in Table 5, indicated filter resistances of 11.1 N/m2
(0.045 in. water) and 15.5 N/m2 (0.062 in. water), respectively, for types
I and II pores. The GCA measured value for the clean fabrics was roughly
2.2 N/m for identical flow conditions. Thus, both the Sunbury and Nucla
estimates were unsatisfactory when the effective radii were computed
78
-------
as twice the hydraulic radii. Although both assume a capillary structure,
which does not describe the filter interstices, one can argue that
Equation (16) offers a better approximation because it takes into account
the highly irregular pore boundaries through the use of hydraulic radius.
In using Equation (16), the assigned values for the hydraulic radii were
computed from the pore geometry shown in Figure 30. Thus, the use of
a maximum value for average pore velocity and a minimum value for hydraulic
radius automatically leads to a high predicted pressure loss.
Since the velocity at the surface of the fabric and the pore inlet is
small compared to that at the throat of the pore, a better estimate of
average pore velocity is one-half the throat value. For continuity of
flow it is then required that the hydraulic radius at the throat be
increased by the /2. When the adjusted values for V and M are substitute
in Equation (16), the computed clean fabric resistance for the Sunbury
2
fabric becomes 3.86 N/m or 0.0155 in. water for a type I pore which is in
good agreement with the GCA measured value, 0.024 in. water.
Similar calculations for type II Sunbury pores and types I and II Nucla
pores are shown in Table 5. Despite the fact that a very simplistic
model of the filter pore structure has been used (basically a symmetrical
Venturi type opening with a minimum circular cross section at the center
and a depth equal to the filter thickness), Equation (16) appears to pro-
vide reasonable values for fabric resistance characteristics when good
estimates of effective pore count and minimum pore cross sectional area
are available.
In applying Equation (16), it must be remembered that average pore velocity
is based upon the gas flow per pore and the best estimate of pore cross
sectional area. On the other hand, the hydraulic radius was computed on
the basis of pore cross sectional area and pore circumference.
79
-------
Except for square or circular cross sections where M = L'/4 and D/4,
respectively, the M value satisfying resistance criteria in Equation (16)
will not, at the same time, define the true pore cross sectional area
and hence true average pore velocity. Therefore, if by successive mea-
surements of filter resistance versus time one desires to estimate the
average or effective open area per pore, it will be necessary to define
the relationship between the M values characterizing resistance and pore
areas, respectively. For example, with respect to a type I pore in the
Sunbury fabric, the hydraulic radius is 17.8 ym for resistance computation
and 42.5 urn for pore area estimation.
PHYSICAL PROPERTIES OF FABRICS
Several measurements of selected physical properties of previously used
*
and new Sunbury and Nucla bags were performed by FRL as part of the
field sampling phase of this project. These data, which are summarized
in Tables 6 and 7, are intended to help explain field performance, in-
cluding resistance, dust retention characterisitcs, and evidence of undue
wear and tear.
Most of the changes shown in Tables 6 and 7 are consistent with what one
expects to see in fabrics with extended field use; i.e., decreased per-
meability due to interstitial dust fill and a corresponding increase in
fabric weight; a detectible reduction in breaking strength and elongation
prior to breaking; and a very pronounced increase in flexural rigidity.
One might infer that decreased permeability will result in improved
dust retention at the expense of higher resistance. However it is also
possible for the permeability to increase due to partial blinding while
at the same time the dust retention characteristics are reduced because
Fabric Research Laboratories
1000 Providence Highway
Dedham, Mass., 02026
80
-------
Table 6. RESULTS OF PHYSICAL CHARACTERIZATION TESTS ON SUNBURY FABRIC FILTER BAGS
oo
Test description
ASTM D 1910, Sample weight, oz/sq yd
ASTM D 1777, Sample thickness, mils
Range
Average
ASTM D 737, Air permeability, cfm/sq ft at V H~0 4P
Range
Average
ASTM D 1602, Breaking strength and elongation
Breaking strength, Ib
Warp : Range
Average
Fill : Range
Average
Elongation at break, percent
Warp : Range
Average
Fill : Range
Average
Average energy to break, inch-lb
Warp:
Fill:
Average :
Flexural rigidity-beam method,
(I
-------
Table 7. RESULTS OF PHYSICAL CHARACTERIZATION TESTS ON A NUCLA FABRIC FILTER BAG3
oo
ASTM D1910, Sample weight, oz/sq yd
range
average
ASTM D1777, Sample thickness, inches
range
average
ASTM D737, Air permeability, cfm/sq ft
range
average
ADTM D1682, Breaking strength and elongation
Breaking strength, Ibs
Warp: range
average
Fill: range
average
Elongation to break, percent
Warp: range
average
Fill: range
average
Flexural rigidity, Ibs (in.)2/in. width
average
New bag
7.4 - 7.5
7.4
0.0135 - 0.0156
0.0147
83.5 - 91.8
86.5
i 168.6 - 210.0
186
82.2 - 116.0
104
8.9 - 11.7
10.7
4.6 - 5.2
4.8
6.26 x 10
-4
Used bag, middle
0.0139 - 0.0158
0.0147
30.8 - 48.2
38.6
Used bag, bottom
I
11,3 - 11.7
11.4
0.0149 - 0.0169
0.0156
30.8 - 48.2
38.6
117.0 - 225.0 102.0 - 135.0
166 I 116
35.1 - 100.5 ! 54.7 - 96.1
66.5 73.1
6.2 - 8.1
6.0 - 8.1
7.6 I 6.9
2.4 - 4.0
3.1
2.0 - 3.7
2.9
-3 -1
1.99 x 10 2.04 x 10
Tests performed by Fabric Research Laboratories for GCA Corporation.
-------
of the loss of the nap or loose staple fibers after extended usage. With
respect to the cleaning of bags by collapse and reverse air flow, those
properties related to flexure may also be related to stiffness and rigi-
dity. If cake dislodgement is more dependent on the rate of flexing than
the actual degree of curvature present when the bag is collapsed, the
rigidity factor may be very important. On the other hand, if curvature
alone determines when the interfacial bonds between particles and yarns
are severed, filter cleanability, and hence resistance properties, may be
less sensitive to rigidity changes.
The apparent spread in the descriptive parameters given in Tables 6 and 7
suggests that caution should be used in developing predictive models
based on limited tests. When one examines field performance tests on the
Sunbury system over a 2-year period, (see Analyses of Sunbury Field
Measurements), it seems reasonable to conclude that the order of the
change and/or variations reported has not highlighted any serious per-
formance defects. In some cases, the main value of the measurements
given in Tables 6 and 7 is relative; i.e., once field experience with
one fabric is well defined, a set of very similar measurements for another
fabric will probably indicate similar field performance.
Tensile Modulus
Tensile properties were determined for several new and used samples of
Sunbury and Nucla fabric in accordance with procedures described in
earlier GCA studies. The present measurements were made by determining
the elongation (warp direction) of 7.6 cm x 45.7 cm (3 in. x 18 in.)
and 15.2 cm x 45.7 cm (6 in. x 18 in.) strips of fabric under applied
static loads ranging from 22 to 336 N (5 to 75 Ibs). The equipment
used for these measurements is described in the section on instrumentation.
A representative loading curve is shown in Figure 31 for a 3 in. x 18 in.
fabric sample, Sunbury plant, from a Compartment 6 bag. During the
loading phase, the tension/elongation relationship followed the path
83
-------
00
2.0
M
O
X
W
O
Ul
z
UJ
O
cc
bJ
Q.
TEST 13 WR = I45 grams/m2, WIDTH (L) =0.0762 m
TENSION/UNIT LENGTH
MODULUS (Mt)
O INCREASING LOAD
A DECREASING LOAD
VA.
STRAIN
(E2-E,)L
= 5.43 x 10* N/m
100
200
APPLIED TENSION (T), N
300
400
Figure 31. Stress/strain relationship for used Sunbury media, 7.6 cm x 45.7 cm
(3 in. x 18 in.) strip with tension applied in warp direction
-------
given by the circles. As the tension was relaxed, however, the return
to original length displayed the same lag (or hysteresis) noted with
many fabrics tested previously by GCA. ° For general characterization
of the elongation properties, the average slope of the curve at 222 N
(a typical applied tension for field installed bags) can be used for the
calculation of the stretch modulus, M. This approach applies to the
utilization of tensile properties for estimating the average acceleration
of mechanically shaken bags. However, when cleaning is dependent upon
bag collapse, the tensile loading rather than the unloading curve appears
to be a better indicator of bag installation and flex properties because
tensile changes brought about by flow cessation and reversal are (a) very
small compared to the installed tension levels 220 N (50 Ibs) and (b) take
place at low frequencies.
Because fabric thickness is often difficult to determine, the modulus for
previous and present fabrics is expressed in terms of the periphery or
width rather than the cross-sectional area of the material subjected to
a tensile load. The tensile properties of the filter fabrics will be
used to define the dynamic behavior of the fabric (acceleration or flex
rate) during the cleaning process.
The results of several measurements are given in Table 8 for fabric
samples from different compartments and with different residual dust
loadings. These modulus estimates were based upon the curves generated
while increasing loads were applied to the fabric rather than the aver-
age of load and unload conditions depicted in Figure 31. As stated
above, it was believed that the former approach would provide a better
indication of the dust holding/tension relationship.
If there is reasonable confidence in the estimation of fabric thickness,
the tensile or elastic modulus values shown in Table 8 can be converted
2 o
to the conventional form, N/m or Ibs/in. , by dividing each by the fabric
thickness in the appropriate units. As discussed elsewhere, the thickness
of the Sunbury and Nucla fabrics, 400 ym, was not difficult to ascertain.
85
-------
Table 8. TENSILE MODULUS VALUES FOR GLASS BAGS USED FOR
COAL FLY ASH FILTRATION
number
12S
13S
145
15S
166
17S
18S
19S
20S
2 IS
22S
23S
28S
30S
313
323
24N
25N
Z6H
27H
33M
34H
35N
32SB
B.gb
compartment
number
6
6
14
14
11
11
3
3
7
10
10
7
3
14
6
-e
2f
lf
lf
2
-e
lf
2f
&
Residual
dust loadc
grams /n
130
145
114
149
235
203
120
162
129
141
102
102
120
115
131
0.0
Unused
15.8
25.9
17.2
12.9
0.0
Unused
2.9
0.0
0.0
Tensile Modulu*
15.2 cm x 45.7 cm
N/ra x 10"5
1.72
-
1.14
-
-
1.58
1.67
-
-
-
1.23
1.10
1.53
1.28
1.47
0.85
0.95
1.06
-
-
0.85
0.86
1.26
0.67
Ib/in x 10~3
1.28
-
1.2:
-
.
1.51
2.10
-
-
-
0.98
0.94
1.40
1.36
1.83
0.96
0.90
1.10
-
-
1.04
0.94
2.12
1.65
7.6 cm x 45.7 Cm
N/o x 10"S
.
3.65
-
4.16
3.22
-
-
3.80
2.52
2.34
-
.
-
.
.
-
.
.
2.78
2.97
.
-
-
Ib/ln x 10*3
.
3.0
-
3.14
4.70
-
-
3.50
1.97
2.47
-
.
.
.
.
-
.
.
3.75
2.10
.
-
.
*S, N refer to Sunbury, Pennsylvania and Hue la, Colorado power plant!.
Fourteen comparttoent baghouse, Sunbury, Pennsylvania.
Resultant loading after laboratory cleaning*
Tension applied in 18 ia. (warp) direction.
Clean, unused bag.
Individual bag number.
Clean Sunbury bag/ tension measured in fill direction.
86
-------
Since the bag axis is usually aligned in the direction of the warp yarns
to provide maximum bag strength, few tension determinations were made in
the fill direction. The results of a single test, Table 8, Test 32SB
show the increased stretch properties of fill yarns (bulked staple)
relative to the warp yarns. A comparison of modulus values for glass
bags with those determined in previous GCA studies, Table 9, for cotton
and Dacron fabrics shows that even with monofilament yarns, the elongation
characteristics are far greater for synthetic fiber yarns than for glass.
Figure 32 indicates that the tensile modulus increases as the inter-
stitial dust deposition increases. This behavior is attributed to the
fact that dust particles within the pores and yarns prevent normal
elongation and contraction which, in turn, reduces the elongation attain-
able per unit tensile force.
It should be noted that the indicated fabric modulus values for 7.6 cm
wide strips were approximately twice those for the 15.2 cm x 45.7 cm
strips, Table 8. Because woven glass fabrics fray badly (and the
lubricated yarns slide over each other quite readily), it was expected
that any contribution to tensile strength from the fill fibers would be
less for narrow strips. If one assumes a constant yarn modulus, the
doubling of the number of warp yarns (the principal support of the
applied load) should show a decrease in elongation for a fixed load.
Our measurements, however, refute this logic. It is expected that modulus
determinations on full size bags, ~ 10 ft x 4 in.j will explain this
anomaly.
Bag Tension and Permeability
Test filters fabricated from new Sunbury and Nucla media were sewn with
conventional stitching and internal support rings in the form of 10 ft.
long by 4 in. diameter filter tubes. The resistance versus air flow rela-
tionship was determined over the approximate velocity range 0 to 1.83 m/min
87
-------
Table 9. PROPERTIES OF COMMON WOVEN FABRICS INCLUDING TENSILE MODULUS
00
oo
Fabric
1. Cotton
2. Cotton
3. Dacron ^-^
. ^ (R)
4 . Dacron v-/
Weight3
10
10
10
10
Weave
Sateen
Sateen
Plain
1/3 Crowfoot
v b
Yarn
count
95 x 58
95 x 58
(Napped)
30 x 28
(Staple)
71 x 51
(Filament)
Perme-
ability0
13
13
55
33
Tensile •
Modulus
Ib/in.
105
105
88.6
466
Mfgr.
No.
960
960C
862B
865B
Mfgr.'se
comment
S
S
S
S,RF
weight: ounces per square yard.
Yarn count: yarns per inch, warp x fill.
o 9
cPermeability: ft /min of air passing through 1 ft of clean, new fabric at 1/2 in. HO
pressure drop.
GCA measurements.
Q
S indicates for shaking, RF indicates reverse flow cleaning.
uuPont trademark.
-------
in A ,
I 4 h
o
X
1 3
CO
o
o
OSUNBURY FABRIC, 15.2 cm. X 45.7cm.
XNUCLA FABRIC, is.acm.x 45.7cm.
NOTE: 1.0 N/m =5.68 X I0~3 Ib./in.
I 'Jnr*
z
UJ
H
100 200
RESIDUAL FABRIC LOADING(Wp),grams/m2
Figure 32. Effect of dust loading on tensile properties of woven
glass bags
89
-------
0 to 6 ft/min for each of several preselected tension values. The effect
of bag tension upon resistance was then determined for a constant filtration
velocity of 0.61 m/min as shown in Figure 33.
Examination of the resistance characteristics of clean glass bags indi-
cates that the resistance to air flow actually undergoes an increase as
bag tensioning increases from 0 (slack) to ~ 60 Ibs (267 N). This behavior
suggests that the effective pore or channel dimensions must decrease as
the load increases. As shown in Figure 33, however, the apparent resis-
tance (or permeability) properties does not change significantly for either
the Nucla or Sunbury media over the expected normal tensioning range,
35 to 60 Ibs (156 to 267 N). The resistance increase is attributed to an
appreciable flattening of the yarns as tension is applied.
A comparison of Curves 2 and 3 suggests that a slightly higher resistance
is encountered when flow measurements are begun with the bag at maximum
tension level; e.g., Curve 2. It should be noted, however, that after
completion of Curve 3, the filter bag was held at 48.5 Ibs (216 N) tension.
The following day, after a tensioning period of about 16 hours, the bag
underwent some stretching such that the tension reduced from 48.5 Ibs
(216 N) to about 45 Ibs (200 N). Although one might expect to see a re-
duction in resistance, it should be noted that the continuous stressing
of the deformable yarns probably produced additional flattening over the
16-hour period. Thus, despite the lowered tensions reflected in Curve 2,
there is a considerable lag or hystereris in the return of the yarn dimen-
sions to its unstressed form.
Because the essentially slack installation condition noted for square
test panels during bench scale tests leads to lower pressure loss, (and
more open pore structure), it is quite possible that some filter media
may show poorer performance as a flat test panel than in the form of a
full scale filter bag. At the present time, there is no practical way
to prepare a small, test panel such that a pre-set uniform tension can
90
-------
CM
I
O
X
(M
E
LJ
O
CO
(O
UJ
(T
O
£t
m
0
0
(?) SUNBURY FABRIC, INCREASING TENSION
(T) NUCLA FABRIC, DECREASING TENSION
NUCLA FABRIC, INCREASING TENSION
NOTE'UNUSED 10ft. x4in. BAG (a>0.6lm/min.
FILTRATION VELOCITY
50
100 150
BAG TENSION, N
200
250
Figure 33. Effect of bag tension on resistance to airflow, with
conventional bag suspension
91
-------
be maintained during filtration. We have also observed that considerable
assymetry in pore structure may ensure when a flat panel is deformed by
pressure stressing. Both factors must be considered before extrapolating
the results of bench scale tests to field conditions.
Fabric Thickness
Although fabric areal density; i.e., its weight per unit area, is readily
measured, the determination of fabric thickness can present difficulties,
particularly with highly napped, woven media or felted fabrics. Standard
thickness gauging is usually carried out in accordance with ASTM proce-
dures (D-1777-64) that involve accurate calipering of the fabric thickness
under known compressive loads. The recommended pressure range for firm
2
fabrics such as asbestos is 0.1 to 10 psi (7 to 700 grams/cm ). A simple
modification of the ASTM method was used in this program to establish
thickness parameters. Glass fabric samples, cut to the dimension of
2 in. x 3 in. glass microscope slides were inserted between two such
slides and compressed by adding various known weights. The distance be-
tween adjacent plate surfaces was then determined by an optical micro-
meter. According to the thickness versus loading curves shown in
Figure 34, minimum thicknesses for the Sunbury and Nucla fabrics, respec-
tively, were reached with loadings of 0.70 and 1.6 psi. The 400 urn
thickness noted for both fabrics agreed well with data in Tables 6 and 7
for used media. Our values for the clean (unused) Sunbury fabric were
significantly higher, however, 400 urn versus 280 ym.
INITIAL DUST DEPOSITION CHARACTERISTICS
rf
A special experiment in which a succession of photomicrographs of the
fabric surface were made during the filtration of fly ash with the Sunbury
fabric suggests that pore closure takes place early in the filtration
process and under conditions where parallel flow appears to predominate.
The appearance of the fabric (shown schematically in Figure 35 for various
92
-------
to
CO
LJ
u
±
H
o
oc
m
If
900
800
700
600
500
400
O UNUSED SUNBURY FABRIC
A VERY CLEAN NUCLA FABRIC
-a-
_L
_L
_L
1.0 2.0
COMPRESSIVE LOADING, psi
3.0
Figure 34. Fabric thickness versus compressive loading
93
-------
SPACE BETWEEN WARP YARNS
~27Am AT CROSS OVER ONLY
NS}
V|.
I. TIME ZERO CLEAN, UNUSED SUNBURY FABRIC
2. TIME 9mrn. W 47.7 g/m2
3. TIME 19 min. W 100 g/m2
4, TIME 37min W 196 g/m2
5. TIME 50 min. W264 g/m2
Figure 35. Schematic of GCA fly ash filtration at 2 ft/min. Dark areas
show dust deposits. Light areas indicate relatively clean
warp yarns transmitting light with rear face illumination
94
-------
surface loadings) indicates that dust first accumulates on and within
the bulked fiber region about the Type 1 and Type 2 pores. Although we
'have referred to the sketches in Figure 1 as photos, they actually rep-
resent standardized and slightly simplified versions of the images seen
by microscopy. As filtration progresses, the deposits spread such that
the remaining surface of the fill yarns become covered, photos correspond-
ing to filtration times of 37 and 50 minutes.
The open areas show the surface of deposit-free, multifilament warp yarns
that transmit light when illuminated from the rear (clean) face. It should
2
be noted that even when the fabric loading has reached 196 g/m , all "win-
dow sections" remained uncovered, thus suggesting a relatively even flow
distribution through the regions of no dust deposit. It is emphasized that
four pores, presumably completely bridged, constitute the boundaries or
corners of the light transmitting region. High local velocities through
these areas preclude dust deposition until the filling is complete above
the underlying bridged pores. Unfortunately, when the filter surface is
aligned normally to the viewing direction, the actual pores are concealed.
However, light transmittancy as viewed by oblique camera angle indicated
that clearly defined openings were present with clean (unused) fabrics.
These open areas were observed to disappear shortly after filtration com-
menced. Finally, complete coverage is attained after 50 minutes. Varia-
tion in apparent "window" size suggests that all pores are not identical
and that some sequential blockage must also take place. The presence of
the uncoated warp yarns, photos after 19 and 37 minutes, do not indicate
. that the pores which act as sinks for these regions are unclosed or
' unbridged.
A second series of special fly ash filtration tests were made with a plain
weave, pressed monofilament screen having a free area of 0.2 and 3120 square
2
(0.025 cm x 0.025 cm) openings per inch. The maximum air velocity through
the clean pore was about 43 m/min, approximately that estimated for the un-
used Sunbury fabric pores. Figures 36 through 38 (representing sequential
tests on a single filter) show that only partial closure of the pores was
95
-------
=.:
Zero minutes
cr.
5 minutes
3 minutes
Figure 36. Fly ash deposition on monofilament screen versus filtration time, surface illumination
-------
5 minutes
*,*,*'« •
7 minutes
8 minutes
Figure 37. Fly ash deposition on monofilament screen versus filtration time, rear and surface illumination
-------
VD
CO
9 minutes
14 minutes 30 minutes
Figure 38. Fly ash deposition on monofilament screen versus filtration time, rear and surface illumination
-------
attained after 5 minutes despite the fact that sufficient dust had ap-
proached or been "seen" by the filter to produce an areal density of
188 g/m had dust retention been 100 percent. Observe, also, that no pore
is completely bridged after 5 minutes filtration although for reasons of
variable pore size, preferred deposition sites via dendrite formation and
statistical randomness, the resulting apertures vary in size. The point
to be emphasized, however, is that had the unobstructed openings between
filaments been in the 10 urn to 25 ym range, the dendritic growth rate
from the bounding filaments would have caused complete bridging well be-
fore a 5-minute filtration period. Thus, the degree of openness after
5 minutes seen in Figures 36 and 37 should probably scale to a time in-
terval of the order of seconds. On the other hand, the distribution of
opening sizes depicts the sequential aspects of pore bridging as demon-
strated by real filters.
After 30 minutes filtration, several well defined pinholes appeared on
2
the substrate which finally attained a loading density of 175 g/m ,
Figure 38. Although the average efficiency over the test period was
21 percent, the relatively high resistance of the blocked pore region
causes most of the flow to pass through the pores or pinholes. No fur-
ther closure of pores is expected; in fact, any slight vibration at this
point in time would dislodge most of the dust. The openings shown in
Figure 38 typify the appearance of many fabric surfaces that develop
pinholes when face velocities are too large or the gradation of, and/or,
absolute pore size is excessive.
If one assumes that pore bridging is accomplished in the very early
period of filtration for a good filter, >99.5 percent efficiency, (thus
excluding either a strict sequential or parallel pore closure process
as the theoretical model), another description of the filtration process
must be sought to explain the form of the resistance-loading curves for
Sunbury and many similar woven fabrics.
99
-------
SECTION VI
ANALYSIS OF SUNBURY AND NUCLA FIELD MEASUREMENTS
FABRIC DUST LOADINGS
Residual dust loadings for several Sunbury bags as received from the field
are shown in Table 10. These bags were removed after 2 years' service so
that all replacements could be installed at the same time. Since the
original guarantee had been only for 1 year there was also concern that
future use might entail costly unscheduled plant shutdowns in the event
of bag failure. Although there was no evidence of physical damage nor any
significant change in collection efficiency (> 99.9 percent), average
filtration resistance at 0.61 m/min (2 ft/min) filtration velocity had
2
risen from 180 to 650 M/m (0.6 to 2.6 in. water).
Filter bags were removed by first unfastening the bottom followed by
placing a large box beneath the bag so that with the top disconnected
the bag could be eased carefully into its container. Although some dust
was undoubtedly lost to the hopper, it is believed that the dust holdings
reported in Table 10 are reasonable estimates.
Examination of the residual dust holdings suggests that compartment 12
was probably cleaned most recently while the next in line for collapse
was compartment 13. Because the individual compartments were cleaned in
a 1 through 14 sequence, the graph of dust loading versus compartment
number, Figure 39, should in theory display an increasing negative slope.
Gross deviations from the curve, which we have attributed to accidental
spills during handling, have been flagged. It is emphasized that despite
100
-------
Table 10. RESIDUAL FABRIC DUST LOADING FOR SUNBURY BAGS AS
RECEIVED FROM FIELD3
Compartment
number
2
3
4
5
6
7
8
9
10
11
12
13
14
Average
Fabric dust loading**
grams /m
610
780
605
206
384
434
527
480
424
449
624
1430
920
580
Weight ratio
dust/bag
1.96
2.50
1.94
0.66
1.23
1.39
1.69
1.54
1.36
1.44
2.00
4.60
-
1.83
aBags removed after 2 years service.
bAverage of two bags sampled per compartment.
101
-------
0.6! m/min,
2.6 inches WATER (S> 2ft./min.
(c) AVERAGE TEMPERATURE, 330° F
(d)ACCIDENTAL SOLIDS LOSS
SUSPECTED
od
od
od
13 14
2345678
COMPARTMENT NUMBER
10 II 12
Figure 39. Residual dust loadings for bags in 14-compartment Sunbury
collector. Cycle interrupted between cleaning of compart-
ments 12 and 13 for removal and replacement of all
102
-------
the point scatter one can see from Figure 39 that the fabric loadings are
considerable. Since it was not determined in the field whether uniform gas
flow prevailed throughout the 14 compartments, the curve shape indicated
in Figure 39 must be considered as speculative. On the average, however,
it appears that the average system fabric loading is in the range of 650
to 700 grams/m . This loading level, in conjunctions with the field pres-
sure measurements to be discussed in the next section will be compared
with laboratory tests on the Sunbury media.
BAG RESISTANCE
Fabric resistance values as determined by GCA during field tests in
March 1975 are shown in Figure 40 as a function of filtration velocity.
No apparent increase was noted over a 35-day test period for which the
3 3
average inlet dust loading was 6.4 grams/m (2.78 grains/ft ). The
clustering of experimental points about the regression line suggests that
variations in mass gas flow rate and not inlet loading were the main
causes of resistance fluctuations shown in Table 11.
Analyses of old pressure charts provided by Sunbury personnel allowed us
to trace the 2-year history of the glass bags that were evaluated by GCA
during their last month of service, Figure 41. Based upon average monthly
pressures, it appears that the main increase in fabric resistance occurs
during the first few months of bag service. Once steady state conditions
are attained, the increase in baseline resistance which is attributable
to a gradual interstitial filling of the pores (which may be partially
compensated by fabric stretching) is approximately 0.5 inches water. An
improved time resolution of the pressure/time traces (daily basis) suggests
that a near-steady state operating pattern may be reached in less than
3 weeks. Despite problems in instrument function and uncertainty as to
system gas flow rates during the shakedown interval depicted in Figure 42,
it appears safe to assume that a very radical increase in fabric resistance
(^ 0.3 to 2 in. water) takes place in at least 3 weeks and possibly sootier.
103
-------
AVG. INLET CONG
= 6.4 grams/m3
(2.78 grains/ft3)
0.2 0.4 0.6 0.8
FILTRATION VELOCITY, m/min
Figure 40. Average filter resistance for Sunbury glass bags,
normal field use after 2 years service
104
-------
750
cu
e
Z
.BOO
UJ
2
^
l-
V)
VI
UJ
»- a
ui o
en
S *s°
u.
o
3.O
u
o
- * 2.0
c
•
UJ
o
Z
2
l/l
V)
UI
o
- g 1.0
u'
O , RESISTANCE v*. TIME (OLD BAGS) © _
A , OUTLET CONCENTRATION vs. TIME °
( 1 ) NEW BAGS - MARCH 1975
(2» OLD BAGS -FEBRUARY 1975 O .
__FABmC0REStSTANCE0 _— o
"- _— i • " ~~ O P> n
O «*** — " O i\ rt ~ """ ^* rt
~ /\ n, ' ' 0 °
/
P
/
/
- /
/
If
1
w
I
U 1
x_
& —
o
O O
—
-v
OUTLET CONCENTRATION 2 2~
/r^
' NOTE: INLET CONCENTRATIONS , 1 TO 4 groins/ft3
FILTER VELOCITY ,~ 2 fl/min
1
I I I 1 1 1 i I 11
5
tn
0
X
w
c
'5
*
Z
o
4 rt P*
^
o:
y
UI
o
3.0 §
u
f.
Ij
2.0g
k o
.0
to
12
14
16
IS
20
24
MONTHS OF FIELD SERVICE (2/73 TO 2/75)
Figure 41. Filter resistance and outlet concentration versus time for
glass bag filters at Sunbury, Pennsylvania power plant
-------
DESCRIPTION
ORIGINAL (OLD) BAGS
FEB. 1973 TO FEB. 1975
REPLACEMENT(NEW) BAGS
FEB. 1975 -
PLUGGED PRESSURE LINES?
FILTRATION VELOCITY =?ft./min. (0.6m/min)
INLET LOADING = 1-4 groins/ft.3 (2. 3-9.2 grams/m3)
30 45
ELAPSED TIME, days
Figure 42. Resistance versus time for old and new Sunbury bags
-------
Limited pressure data for new, replacement bags suggests that the operating
characteristics for old and new bags are about the same. The rather abrupt
rise to the 2 in. water operating level provides a practical guideline for
modeling. In describing the greater fraction of the useful filter operating
life (which Sunbury personnel believe may be as much as 3 years), it appears
that the effective baseline for starting resistance might be considered as
2 in. water.
Although monthly and daily resistance - times curves suffice for practical
estimates of power requirements, they do not provide the resolution neces-
sary to assess the impact of successive cleanings on fabric resistance with
multicompartment systems. A linearized 10-minute time trace from a Sunbury
chart record of February 1975, Figure 43, gives a detailed picture of pres-
sure loss patterns over successive filtering, cleaning, and manifold flushing
cycles. Lowest fabric resistance values, 2.5 to 2.6 in. water, are indicated
when all 14 Sunbury bag compartments are on-line. During the time interval
between the sequential cleaning of compartments, no discernible increase in
resistance was detectable for 13 and 14 chamber operation or during the
admission of reverse flow air. This is readily explained by the fact that
the amount of dust placed on the filtering surfaces during the period between
cleanings represents but a very small fraction, -\- 0.77 percent of the total
estimated filter system dust holding. As soon as a compartment is isolated
for cleaning, the handling of system flow by 13 compartments causes a re-
sistance increase of about 0.25 in. water. With initiation of reverse flow
(roughly 1.4 ft/min), the 13 on-line compartments must accommodate an
additional flow volume (about 5 percent of primary flow) . This leads to
the observed maximum resistance levels of 2.9 to 3.1 in. water. The net
result is that the average working fabric resistances is constrained to a
relatively narrow range. Therefore, the modeling of system performance with
respect to resistance, particulate emissions, and power needs is simplified.
107
-------
3.0
o
$
w
o>
o
•S 2.0
3T-I3R
SI-I3R
3ZH-I3R
0-14
urn-is/
0-14
TIME, minutes
o
00
V)
o
m
u.
3 0
a-u
2.0
2H-I3R
TTTT -!3R
0-14
0-I4F
Figure 43.
6 7 8 9 10
TIME, minutes
NOTES'
I. ROMAN NUMERAL REFERS TO COMPARTMENT BEING CLEANED.
2. 13 OR 14 INDICATES NUMBER OF COMPARTMENTS ON LINE-
3. R INDICATES REVERSE FLOW AIR IN USE.
4. F INDICATES MANIFOLD FLUSHING WITH REVERSE AIR.
5. 0 INDICATES ALL COMPARTMENTS FILTERING OR FLUSHING.
Filter resistance versus time for successive filtering, compartment cleaning and
reverse flow manifold flushing, Sunbury field test of February 14, 1975
-------
COLLECTION EFFICIENCY
The results of prior GCA field sampling of the Sunbury and Nucla effluents
are summarized in Table 11. Inlet and outlet concentrations values, con-
verted to their metric equivalents, are shown in Figures 44 and 45.
After 2 years service, Sunbury effluent concentrations averaged over several
-3 3 3
hours were about 1.7 x 10 grains/ft (3.9 g/m ) DSTP. Identical mea-
surements upon new replacement bags some 10 days after installation showed
—3 ^ ^
slightly higher effluent concentrations, ^ 2.1 x 10 grains/ft (4.89/m ).
Thus, it appears that no appreciable improvements in filtration capabilities
are obtainable once steady state filtration conditions are realized. The
fact that the emissions during the first day of use were significantly
greater is consistent with the correspondingly lower filter resistance
during the early shakedown period. There appears to be a rather good cor-
relation between effluent concentration and fabric resistance properties
according to the data shown in Figure 46. On the other hand, the outlet
concentrations for both new and old bags show no significant dependency upon
inlet concentration, Figure 44. This observation agrees with test results
reported by GCA and others which indicate only weak correlations between
influent and effluent concentrations for fabric filter systems.
The Nucla test data graphed in Figure 45 indicate essentially the same
dust removal characteristics as shown by the Sunbury fabric. The significant
difference between the two Nucla data sets resulted from the replacement of
many faulty bags. It should be emphasized that the bag failures resulted
from an air flow distribution problem that caused severe bag erosion. A
modification in thimble design after the shake down test period corrected
this problem.
Generally, a comparison of field and laboratory data for Siinbury and Nucla
fabrics indicated comparable performance. Thus, it appears acceptable to
extrapolate directly the results of many laboratory tests to estimate key
modeling parameters.
109
-------
Table 11. FIELD PERFORMANCE OF FILTER SYSTEMS WITH GLASS BAGS SUNBURY STATION'
Run
No.a
1
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25_
26
27
28
29
30
31
Mean
Standard
deviation
Inlet
concentration,
gr/dscf
3.6296
4.1235
2.6851
2.5243
3.1661
2.2977
2.4250
3.2926
2.6678
2.0891
2.6020
2.8845
2.6728
2.4403
2.5058
1.8291
2.8042
2.2016
1.6694
1.3822
3.2646
2.0503
3.0946
2.3859
1.3477
3.0022
2.0174
2.0843
2.2181
2.5328
0.6346
Outlet
concentration,
gr/dscf
0.0022
0.0013
0.0017
0.0014
0.0014
0.0014
0.0015
0.0016
0.0033
0.0017
0.0020
0.0015
0.0016
0.0013
0.0016
0.0013
0.0016
0.0018
0.0019
0.0031
0.0028
0.0029
0.0025
0.0022
0.0022
0.0022
0.0023
0.0020
0.0022
0.0020
0.0006
Penetration,
%
0.06
0.03
0.06
0.06
0.04
0.06
0.06
0.05
0.12
0.08
0.08
0.05
0.06
0.05
0.06
0.07
0.06
0.08
0.11
0.22
0.09
0.14
0.08
0.09
0.16
0.07
0.11
0.10
0.10
0.08
0.04
Inlet
nund ,
urn
5.8
7.0
4.6
4.7
5.5
5.1
4.4
4.8
11.9
7.2
11.0
6.5
9.1
5.6
6.1
8.0
3.2
5.9
3.4
8.2
5.4
7.0
5.6
9.6
8.0
6.8
9.2
6.7
7.5
6.4
1.4
Outlet
mmd,
|jm
7.1
7.7
3.7
4.5
4.4
5.6
10.4
6.6
6.1
3.6
3.4
6.6
5.0
6.1
10.0
6.4
7.5
6.6
7.4
6.4
3.1
5.0
5.8
11.5
12.0
5.9
2.6
2.4
4.4
6.1
2.5
Fuel
moisture ,
7.
2.9
3.1
3.0
2.6
3.4
2.9
3.2
3.0
2.5
2.1
2.6
1.7
3.0
2.7
3.2
2.4
2.8
2.6
1.8
2.3
3.5
3.6
4.1
3.5
2.7
3.2
3.6
2.7
3.3
2.9
0.5
Fuel
ash,
%
18.5
25.1
23.6
21.1
31.6
29.5
22.6
23.0
19.7
16.0
18.8
18.7
22.2
20.6
23.5
19.0
21.6
22.2
21.7
20.7
22.3
22.6
20.6
23.2
18.3
21.1
23.8
23.1
22.0
22.0
3.1
Fuel
sulfur ,
%
2.1
1.7
1.6
2.2
1.8
1.5
2.2
1.4
2.2
3.2
1.6
1.7
1.3
1.2
1.6
1.5
1.5
1.2
1.4
2.1
1.8
1.8
2.4
1.6
2.1
2.1
1.6
1.5
2.0
1.8
0.4
Steam
flow,
1000 Ibs/hr
400
395
400
410
410
400
400
370
360
325
325
310
390
390
375
400
400
380
375
370
380
410
380
400
400
410
370
390
400
384
26
Face
velocity ,
ft/min
2.02
2.07
2.18
2.21
2.03
2.05
2.07
2.08
1.88
1.82
1.69
1.64
2.05
2.05
1.98
2.07
2.45
2.36
2.01
2.10
2.02
1.96
2.01
2.05
2.22
2.15
1.95
1.99
2.05
2.04
0.16
Baghouse
pressure
drop
in. H20
2.8
2.6
2.8
2.8
2.7
2.7
2.6
2.6
2.3
2.4
2.0
2.0
2.7
2.7
2.7
2.7
3.6
3.5
2.8
0.4
0.5
0.6
0.6
0.6
0.7
0.7
0.7
0.7
0.7
2.0
1.0
Compartments
cleaned
per hour
28
28
28
28
28
28
28
28
28
28
28
28
28
14
14
28
28
28
28
28
28
28
28
28
28
28
28
28
28
27.0
3.6
aRuns 1 through 22 - old bags with 2 years' service
Runs 22 through 31 - new bags, no prior service
-------
Table 11 (continued). FIELD PERFORMANCE OF FILTER SYSTEMS WITH GLASS BAGS - NUCLA STATION'
8
Date
9/21/74
9/22/74
9/23/74
9/24/74
9/25/74
9/26/74
9/27/74
9/28/74
9/30/74
10/1/74
10/2/74
10/3/74
10/4/74
10/5/74
10/6/74
10/7/74
10/22/74
10/23/74
10/24/74
10/25/74
10/26/74
10/27/74
Run
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Inlet mass loading
grains/dscf
Method
5
2.0759
2.371?
1.9753
1.7021
1.6768
1.7995
1.8516
11.4446
2.3878
1.6873
1.7422
2.1112
2.2693
1.7751
1.3572
2.1779
2.1098
2.0669
1.9828
1.7791
1.9502
2.0572
Andersen
A
0.4984
1.5078
1.4014
1.7092
1.4819
1.3426
1.3144
1.6248
1.6636
1.4206 •
1.0294
1.5900
1.8991
1.6593
2.4579
2.3232
1.8337
1.5351
1.8120
2.9943
1.5053
1.9528
Andersen
B
_
1.4610
1.7176
1.1793
1.4382
1.1600
1.9251
2.0318
1.9608
1.3540
1.4893
1.3091
2.0574
1.4318
1.6854
1.5909
-
1.6651
1.7094
1.6683
1.3352
1.7008
Outlet mass loading
grains/dscf
Method
5
0.0044
0.0049
0.0045
0.0063
0.0042
0.0047
0.0045
0.0016
0.0016
0.0010
0.0015
0.0092
0.0040
0.0029
0.0007
0.0019
0.0022
0.0010
0.0015
0.0017
0.0015
••
Andersen
north
0.0101
O.OOG9
0.0034
0.0043
0.0031
0.0048
0.0033
0.0053
0.0021
0.0021
0.0035
0.0563
0.0034
0.0047
0.0039
0.0042
0.0025
0,0024
0.0030
0.0025
0.0028
0.0036
Andersen
west
0.0031
0.0034
0.0028
0.0021
0.0030
0.0051
0.0025
0.0015
0.0020
0.0034
0.0046
0.0796
0.0035
0.0154
0.0036
0.0037
0.0025
0.0022
0.0021
0.0025
0.0023
0.0035
Mass
efficiency
(percent)
99.7880
99.7743
99.7722
99.6299
99.7495
99.7338
99.7570
99.9360
99.9330
99.9407
99.9139
99.5642
99.8237
99.8366"
99.9484
99.9128
99.8957
99.9516
99.9244
99.9045
99.9231
•"
Baghouse
operation
Norr.al
Normal
Kornal
Normal
Cor.t. cleaning
Cont. cleaning
Normal
Long repressure
Long repressure
Normal
No cleaning
No cleaning
Norr.al
No repressure '
No repressure
Normal
Noraal
Long repressure
Normal
No shaking
No shaking
Nor&al
-------
N>
E 7
<
UJ
o
o
o
RUN
O 1-21
* 22-31
DESCRIPTION
OLD BAGS
NEW BAGS
o
o o
OUTLET CONCENTRATION, g/m
Figure 44. Inlet and outlet dust concentrations for Sunbury field tests
-------
6.0
ro
(9
z
Q
.1-
UJ
5.0
4.0
3.0
x
X
RUN
0 1-7
x 8-21
DESCRIPTION
FAULTY BAGS
BAGS REPLACED
O
O
O
4 6 8 fO
OUTLET LOADING, g/m3 xlO3
12
14
16
Figure 45. Inlet and outlet dust concentrations for Nucla field tests
-------
3.0
to
o
X
10
to
c
'5
o>
x
2.0
111
-J
H
3
O
1.0
\
r\
\
X NEW BAGS, MAR. 1975 (FIRST 10 DAYS)
O OLD BAGS, FEB.-MAR. 1975 (LAST 35 DAYS)
\
\
XXX. X
0 5 10
ELAPSED TIME, days
A-
20 22 24
ELAPSED TIME, months
Figure 46. Field measurements of outlet concentrations from new and
well-used Sunbury bags. See Table 11
-------
SPECIFIC RESISTANCE COEFFICIENT
The operating mode for the fabric filter system used at the Sunbury Station
did not allow the direct estimate of K2 values because of a continuous
cleaning schedule. Reference to Figure 43, for example, indicates that
the interval between cleanings is too brief to detect any significant
resistance versus loading trends. Additionally, any change in slope,
AS/AW, reflects the integrated effect of a parallel flow through fabric
surfaces of unequal dust loading. Therefore, without a complex differen-
tiation process, the true K... values cannot be estimated.
On the other hand, many of the Nucla tests were carried out with very
lengthy, 2 to 4 hour-filtering periods between cleanings. Hence it was
possible to make determinations of K~ for typical field aerosols. These
results in both English and metric units are summarized in Table 12. A
very detailed analysis of Nucla data relative to determining how reliably
K9 values can be predicted on the basis of dust and flow parameters is given
in Section IX.
115
-------
Table 12. MEASURED K2 BASED ON FIELD TESTS
AT NUCLA GENERATING SECTION
Run
number
1-1-A
1-2-A
1-3-A
11-AB
14-AB
15-B
16-AB
16-B
19-1-AB
19-2-AB
Measured K£a
in H20 min ft/lb
3.18
9.85
4.46
6.03
6.80
7.05
6.76
6.76
5.65
5.95
N min/g m
0.531
1.64
0.745
1.00
1.13
1.18
1.13
1.13
0.943
1.16
Based on actual face velocity of 2.76
ft/min (0.844 m/min), a flue gas temper-
ature of 124°c, and an assumed dust cake
porosity of 0.59.
Note: See Section IX and Table 36 for de-
tailed analyses.
116
-------
SECTION VII
BENCH SCALE LABORATORY TESTS
FABRIC RESISTANCE CHARACTERISTICS
Clean (Unused) Fabrics
Resistance measurements were performed on several samples of new and used
glass bags from the Sunbury and Nucla power plants. These tests were made
on 11 in. x 8 in. cloth panels which were clamped securely in the filter
2 2
holder shown in Figure 5. An unsupported cloth area of 348 cm (54 in. )
(9 in. x 6 in.) was exposed. The results of tests on unused filters,
*
Figure 47, were used to calculate Sunbury and Nucla permeabilities, 42.5
3 +
and 112 ft /min, respectively. Independent measurements by FRL, gave
3
corresponding values of 54 and 86.5 ft /min. Since GCA and FRL used the
same ASTM test methods, the differences are believed to result from the
normal variability in fabric properties.
Cleaned (Used) Fabrics
Resistance measurements were also performed on several test panels removed
from used Sunbury and Nucla filter bags shipped to the GCA laboratories.
Because of handling, shaking, possible moisture absorption or chemical de-
gradation, it is recognized that the laboratory measurements may not
&
Volume flow (or air-to-cloth ratio) at 0.5 in. water filter resistance.
FRL, Fabric Research Laboratories
1000 Providence Highway
Dedham, Mass. 02026
117
-------
i.o -\
00
O CLEAN (UNUSED) SUNBURY MEDIA, AP = 9.63V
CLEAN (UNUSED) NUCLA MEDIA. AP= 3.31V
N07E'
0.5 In. water = 125 N/me
Ift./mln =0.305m/mln
FILTRATION VELOCITY (V),m/min
Figure 47. Filtration resistance for unused Sunbury and Nucla glass bags,
laboratory measurements
-------
represent true field conditions. However, it is believed that laboratory
evaluation of the field media represent a useful supplement to the earlier
field tests.
Prior to testing the used Sunbury and Nucla bags, a standard preparation
and cleaning process was developed. The fabric test panels were first
shaken by hand about 20 times at 1 cps to remove all dust that would fall
off during normal handling of the filter. The clean air face was vacuumed
to remove dust deposited during shipping, approximately 28 grams/m .
During the latter process, there was negligible dust loss from the dirty
air side. The three levels of cleaning applied to each test sample were
arbitrarily defined as dirty, moderately cleaned, or well cleaned; The
designations corresponded roughly to residual dust holdings of 135, 75 and
2
45 grams/m .
Figures 48 and 49 show resistance curves for fabric samples from the
center sections of Sunbury bags removed from different bag compartments.
Although test velocities were extended to the 10 m/min range with only
minor deviations from a linear AP-V relationship, the plotted data were
restricted to the probable filtration rates expected in the field. The
code letters, T, C, and W appearing on Figures 48 and 49 refer, respectively,
to the test number, bag compartment number, and the final fabric dust load-
ing after cleaning the fabric panesl. Pressure velocity curves for the
Nucla bags, Figure 50, indicate that the residual dust holdings for bags
stated to have seen prior field service were exceptionally low. It was
pointed out, however, that these bags had been stored in the open for some
time such that considerable dust had washed off. In the case of bag No. 2,
the resistance after cleaning returned to the same level observed for an
unused (No. 3) bag.
Although ithese data are too limited to quantify, one might infer that the
graphite-silicone treatment on the Nucla bags provides a greater dust un-
loading capacity than that for the Sunbury bags. This feature does not
119
-------
400
300
I
UJ
o
V)
V)
Ul
o:
ce
UJ
j-
200
100
BOILER lA.SUNQURY, PENNSYLVANIA
T=TEST NUMBER
C=BAG COMPARTMENT
W = FABR!C DUST HOLDING,
groms/m'
249 N/m2 = I in. H20
0.61 m/min. = 2 ft,/min.
3- 11-65
5-6-65
4 - 14-58
3-11 -43
0.5 l.O
FACE VELOCITY, m/min
Figure 48. Resistance characteristics of used Sunbury fabrics cleaned
in the laboratory to various residual dust holdings,
Tests 1 to 5
120
-------
-------
60
50
CJ
E
.7 40
o
z
BOILER 2, NUCLA, COLORADO
T=TEST NUMBER
B = BAG NUMBER
W = FABRIC 'DUST HOLDING,
249 N/m2 = |in. H20
0.61 m/min. =2 ft. /min.
* NUMBER I AND 2,USED BAGS
NUMBER 3,UNUSED BAG
T B
W
II- I - 27
II- I - 14
II - I - 0.0
9 - 2 - 7.6
9-2-4.3
10-3 - 0.0
9-2 - o.O
FACE VELOCITY, m/min
Figure 50. Resistance characteristics of used Nucla fabrics cleaned in the laboratory
to various residual dust holdings
-------
necessarily represent an advantage because it is the residual and deposited
dust that provides the dust retention properties of the fabric. The effect
of residual dust loading upon filter resistance is shown in Figure 51 for
Sunbury media at 0.61 m/min (2 ft/min) filtration velocity. Despite the
point scatterm the relation given by the empirical equation:
P = -72 + 1.68 WR (17)
is in fair agreement with independent field measurements. For example,
a resistance of 820 N/m (3.3 in. water) is predicted by Equation (1) in
2
contrast to an observed value of 670 N/m (2.7 in. water) as shown in the
field data of Figure 41. In Equation (17), P is expressed in Newtons per
meter^ and W in grams per meter^.
R
Because of several unknown factors in field handling and the problems of
simulating superficial and interstitial dust deposits by the laboratory
shaking and vacuuming procedure, the point scatter noted in Figure 51 is
not surprising. The minimal point scatter with low residual deposits
suggests that the "most difficult to detach" particles must have rather
specific alignment patterns and deposition sites.
It was not determined whether or not partial or complete blinding of
some of the fabric pores had taken place during field use. On the other
hand, if the hand cleaning and vacuuming of the field fabrics had not
lead to uniform dust removal, a large point scatter would have been ex-
pected from fabric to fabric. This problem, which is treated in detail
in later sections of this report, is described briefly in the following
discussion. Reference to Figure 52 shows the form of resistance/fabric
loading curves for GCA single bag filtration with cleaning by mechanical
shaking. In both cases, dust removal was highly nonuniform with the
actual surface consisting of two distinct regions, the first from which
slabs of dust were separated from the fabric-dust interface, leaving a
relatively clean area below and the second from which no dust was removed.
123
-------
TESTS
200
CM
6
OL
<3
tu
o
z
to
LJ
a:
a:
CD
if
- O 3
A 4
X 5
D 6
07
1-98
BAG
COMPARTMENT
II
14
6
IO
7
3
100
NOTE: 0.61 m/min =2ft./min
AP =-72 + 1.68 WR
40
60 80 100 120 140
RESIDUAL FABRIC LOADING (WR), groms/m2
160
ISO
Figure 51. Fabric resistance versus residual fabric loading for Sunbury bags at 0.61
m/min (2 ft/min) filtration velocity
-------
10-
8
g
x
CM
E
•x
z
III"
S
00
kJ J,
tr 4
u
oc
CD
Se
WR,
CURVE DESCRIPTION
I MULT1 FILAMENT DACRON
2 COTTON SATEEN
FILTRATION AT 3ft./min.
MECHANICAL SHAKE CLEANING
200 WR2 400
FABRIC LOADING, (W), grams/m2
600
Figure 52. Typical resistance versus dust loading curves for fly ash filtration with
staple and multifilament yarns
-------
The net result is a parallel flow system in which each element has ini-
tially a different air flow (and dust loading) rate. As the filtration
process continues, the flow and deposition rates through the cleaned and
uncleaned areas converge and the fabric loading becomes more uniform.
Resistance Versus Fabric Loading-Bench Scale Tests
Typical results of filtering GCA ash with clean and used Sunbury and
Nucla fabric test panels are shown in Figures 53 and 54. Limited mea-
surements indicate that the characteristic interstitial plugging arising
from lengthy field service leads to higher filtration resistance.
According to Figure 53, the base resistance for Sunbury fabric had in-
2
creased by 0.75 in. water (185 N/m ) after 2 years of field service.
The above measurements are in good agreement with data shown in Figure 41
2
wherein an approximate 125 N/m gain was observed for the full scale
Sunbury field system. The more rapid rise in resistance, coupled with
the higher initial resistance, suggests that some partial or complete pore
blinding has occurred as the result of extended field service.
Comparative data for Nucla bags, Figure 54, which show a much smaller re-
2
sistance increase for the used fabric (approximately 50 N/m ) reflect a
shorter service life plus an undetermined amount of dust removal caused
by bag storage in an unsheltered area after removal from the baghouse.
Generally, the results of several tests, Figure 53, upon new Sunbury media,
indicated that over short intervals of repeated cleaning and reuse, the
initial change of pressure with respect to fabric loading and the resistance
difference between used and clean media was similar to that observed for
the Nucla tests described in Figure 54. It also appears that a solid cake
formation has developed for both Nucla and Sunbury fabric after the fabric
2 2
dust loading reaches about 175 g/m (0.036 Ib/ft ) because slopes of the
resistance-loading curves undergo no further change. The estimated K
126
-------
1200 -
TEST PRIOR SERVICE
66 (A) 2 YEARS
67 (V) 2 YEARS
~6 HOURS
~6 HOURS
200
400 600 800 1000
AVERAGE FABRIC LOADING ,g/m2
1200
Figure 53. Resistance versus average fabric loading for Sunbury fabric
with GCA fly ash at 0.61 ra/min face velocity
127
-------
1200 h
0
68 (A)
69 (O)
PRIOR SERVICE
UNUSED
~6 months
200 400 600 800 1000
AVERAGE FABRIC LOADING ,g/m2
1200
Figure 54. Resistance versus average fabric loading for Nucla
fabric with GCA fly ash at 0.61 m/min face velocity
128
-------
value in metric units, about 1.6 N min/gm (9.6 in. H20 min ft/lb) was
slightly lower than that reported previously for the filtration of GCA
fly ash with cotton, 1.85 N min/gm. However, as will be discussed later,
present tests indicated that corrections.for differences in filtration
velocity (0.92 m/min in prior GCA tests) 'converted the K value of 1.6
to 1.95 N min/ g m.
Filtration tests were also performed with other dust/fabric combinations,
Table 13- Resistance/fabric loading curves and tabulations of key param-
eters deriving from these measurements are given in Figures 55 and 56 and
Table 18, respectively. It is emphasized that the tests described in
Figures 53 through 56 typify the behavior of uniformly loaded filters.
Once the filter undergoes a partial cleaning, a decidely nonuniform load-
ing condition prevails as mentioned previously.
Table 13. FABRIC/DUST COMBINATIONS STUDIED IN
THE LABORATORY PROGRAM
Fabric
Sunbury (Menardi)
glass bags, 3/1 twill
Nucla (Criswell)
glass bags, 3/1 twill
Cotton sateen
Dacron crowfoot weave
Dust
GCA fly ash
Lignite
Rhyolite
GCA fly ash
Lignite
GCA fly ash
GCA fly ash
DUST DEPOSITION AND REMOVAL CHARACTERISTICS
Deposition on Used Fabrics
The appearance of dust-laden and cleaned Sunbury fabrics was observed
directly and microscopically to provide improved assessments of the overall
filtration process. The photomicrographs prepared during this phase of
the study answer several important questions as to (1) the disposition
129
-------
1200
1000
800
W
E
LJ
O
z
ac
o
K.
CO
2 400
200
TEST
77 (A)
79(x)
84-(O)
DUST
FINE RHYOLITE
COARSE RHYOLITE
LIGNITE FLY ASH
200
400 600 800 1000
AVERAGE FABRIC LOADING,g/m2
1200
Figure 55. Filtration of granite dust (rhyolite) and lignite.fly ash
with Sunbury fabric at 0.61 m/min face velocity
130
-------
1200
IOOO
TEST
98 (A)
84 (O)
N
UJ
o
CO
o
o:
00
800
600
400
200
—I r
FABRIC
OACRON
COTTON
SUNBURY GLASS
FIGURE 53
200 400 600 800 IOOO
AVERAGE FABRIC LOADING,g/m2
1200
Figure 56. GCA fly ash filtration with unused sateen weave cotton
(unnapped) and Dacron (crow foot weave) at 0.61 m/min
face velocity
131
-------
of the dust on the filter following cleaning, (2) the manner in which
dust is detached, (3) the description of the dust cake per se, (4) the
physical appearance and probable location of leak points on the filter,
and (5) the appearance of the clean air side (warp surface) of the new
filter fabric after 8 hours of fly ash filtration. The samples of filter
fabric discussed in this section were removed from 6 in. x 9 in. test
panels installed in a bench-scale system, Section IV, Figure 5. A re-
suspended coal fly ash aerosol described previously was filtered at a
velocity of 0.61 m/min (2 ft/min) at inlet concentrations ranging from
Q 3
2 to 3 grams/m (1 to 1.5 grains/ft ).
Figure 57 shows the appearance of the filtering (fill) surface after de-
2
positing a fly ash loading of 945 grams/m on a previously used and
cleaned Sunbury bag. At 20X magnification, the surface is relatively
smooth with only a minor indication of the clean fabric surface pertur-
bations. Grain detail is discernible down to approximately 2.5 vim.
After inducing fabric flexure by depressing the clean side, a character-
istic checking or cracking results which, under normal field bag collapse,
is a prelude to dust release. The general appearance of this cracking
resembles a highly polished and etched metal specimen showing crystal
boundaries. In the case of the fabric, Figure 57, the cracking pattern
conforms roughly to the maximum continuous length of warp yarns (500 ym)
and fill yarns (approximately 2000 ym) as exposed on the filtering (fill)
face. The far greater curvatures at warp yarn crossovers appear to
represent cake failure zones where fabric curvature is altered. Because
of this checking process, detachment of the dust layer from its fabric
interface is hastened. By noting the curvature of the supporting fabric
matrix during flexure, it is suggested that tensile, shear, and compressive
properties of the dust layer might be estimated.
Figure 58 shows another fabric sample with a terminal fly ash holding of
only 430 grams/m . Although the basic appearance is unchanged (see
Figure 57, the shadowing technique indicates clearly the ridges or raised
132
-------
A. Duct surface prior to cleaning
B. Checking or cracking induced by flexia?e
Figure 57. Fly ash dust layer on Sunbury fabric, laboratory
tests prior to removal of 945 graras/m2 cloth
loading (20 X magnification)
133
-------
•V
iir
Figure 58. Photomicrograph of Sunbury media showing GCA fly
ash loading with pinhole leak and cracks induced
by flexure
134
-------
diagonal portions caused by the underlying fill fibers. It is concluded
that the presence of the supporting matrix is probably detectible except
for very high surface loadings, approximately 1200 grams/m2. A full-size
and a three times enlargement of the same filter shown in Figure 58 with
430 grams/m surface loading, indicate clearly in Figure 59 the ridge
patterns mentioned previously. Both figures also indicate pinholes or
punctures that can contribute significantly to dust penetration.
t
Pinholes and Air Leakage
Figure 59 reveals several surface perturbations whose true details
are better evidenced in Figure 58. Aside from small depressions
caused by the impact of occasional massive particles (probably agglom-
erates) of the order of 200 ym diameter, several "ant-hill" type mounds
appear on the fabric surface. Substage microscope illuminations reveal
these structures to be associated with open pores with diameters ranging
from 100 to 200 ym. The circular ridges of dust surrounding these open-
ings (pinholes), estimated to be about 1 mm high, result from the inertial
separation of particulates as the aerosol changes direction and accelerates
to flow through the apertures. The presence of these surface deposits
points out that at least some of the dust is collected from that fraction
of the air that leaks through these pores. A pore (pinhole) count indi-
cated a coi
Test 65-F.
2
cated a concentration of 2500 per m of the pore type shown in Figure 60A,
Inspection of Figure 60B shows that a pinhole may act as a focus for the
cake cracking process. This bears out an earlier observation that crack-
ing or checking is initiated at the points of maximum yarn curvature;
i.e., the yarn crossing points. Although it did not appear at first that
2
an observed pinhole concentration of 2500 per m would affect signi-
ficantly the filter behavior, a subsequent analysis indicated that the
pinhole flow was significant. It was noted that all pinholes were located
above Type I and Type II pore openings in the fabric. Thus, the projected
areas of the pinholes were essentially as described in Section V, Figure 28,
135
-------
Full scale
Enlarged 4/1
Figure 59. GCA fly ash deposit on previously used Sunbury fabric
showing crater and pinholes, 430 grams/m2 cloth loading
136
-------
Pinkole leak, filtration surface, showing
characteristic mound,, substage lighting
(20X magnification)
Pinhole leak as focus for radiating checking.
Pore accentuated by substage lighting (BOX
magnification)
Figure 60. Pinhole leak structures, GCA fly ash filtration on
Sunbury fabric
137
-------
while the effective cross section defining the air flow should appear as
shown in Figure 30. If one assumes that the interstitial air flow remains
well within the laminar range, N_ = 100 to 200, Equation (18) may be
Ke
used to calculate average (pinhole) velocity when the pore dimensions and
filter resistance are known:
V = 10 APM3/2yL (18)
2
The resulting calculation for a pressure differential of 1000 N/m (4 in.
water), pore depth (filter thickness) of 0.04 cm, and an average pore area
/ o
of 2.27 x 10 cm gives a pore (pinhole) velocity of 3230 cm/sec
(6350 ft/min). The average pore cross-sectional area is that based upon
best estimates of minimum pore dimensions, Figure 30 and Table 5. A
second estimating procedure is to treat a pinhole opening as a sharp edge
orifice so that the velocity is defined by the relation:
V, ., v= C 128.3/Ap (N/m^) (19)
(cffiYsec) o
Because the pinhole diameter is infinitely small relative to the flow
cross section on either side of the filter, the orifice coefficient C
o*
is approximately 0.62, irrespective of the flow type. By means of a
trial calculation, the N value was redefined, thus providing an im-
proved estimate of 0.66 for C . The final outcome was a slightly lower
value, 2650 cm/sec (5206 ft/min) for pore velocity. Even when the more
conservative (lower) velocity was used to determine the fraction of the
total filter flow that passed through the observed pore area (roughly
-2 2
1.89 x 10 cm for the 83 pinholes in the panel), the calculation in-
dicated that nearly 14.1 percent of the air passed through these pinholes.
Therefore, if dust removal were 99.5 percent or better for the undisturbed
cake and zero percent for the pinholes, one would expect a weight collec-
tion efficiency of 86.5 percent. Actually, the sharp convergence of the
streamlines for that fraction of the flow passing through a pinhole re-
sults in considerable dust collection as can be seen in Figure 60. Based
138
-------
on the angle of the incident surface illumination, approximately 45
degrees, and the shadow dimensions, the average height of the larger
pinhole mounds is about 1 mm. By simple geometric approximations and as
an estimated deposit density of 0.82 grams/cm , the amount of dust sur-
rounding each pore is estimated as about 6.2 x 10~ grams. Details for
determining the density of the superficial dust layer are discussed in
the next section. In the case of the test results illustrated in Fig-
ure 60 very few pinholes were visible when the test panel was removed for
weighing after 33 grams of dust had been deposited. As the result of
accidental jarring and flexure, however, it is postulated that the dust
layer was cracked causing the pinhole leaks observed at the completion
of the dust loading tests. The final dust deposit of 42 grams was equiv-
2
alent to a cloth loading of approximately 1200 grams/m . Until the
apparent damaging of the filter layer, the mass efficiency measurements
over the filter loading process had ranged from 99.26 to 99.88 percent
with an average value of 99-67 percent. However, during the pinhole
leak period the average efficiency reduced to 96.67 percent based upon
gravimetric analyses of filter samples.
A summary of the filtration parameters shown above is given in Table 14.
There appears to be no positive time trend in either outlet concentration
or collection efficiency until the last phase of the filter loading.
Here, as indicated previously, the dust layer must have experienced
severe internal damage, including the rupture of particle-to-fiber bonds
at the dust/fabric interfaces.
At reduced pressure differentials during the early phase of filter loading,
the predicted pore velocities are much lower if the concept of capillary
flow is assumed, Figure 61. It is difficult, however, to state which
geometric concept applies to the actual pinholes. If our interpretation
of Figure 60A is correct, it would appear that the pore consists of a
bell mouth entry converging over a depth of 2000 pm from a diameter of
139
-------
Table 14. FILTRATION CHARACTERISTICS OF NEW (UNUSED)
SUNBURY FABRIC WITH GCA FLY ASHa
Fabric dust
loading
grams /m
130
210
250
345
960
1,200
Pressure drop
N/m
170
210
240
320
950
1,200
In. H20
0.68
0.84
0.96
1.28
3.80
4.80
Outlet concentration0
grams /m
x 103
9.2
1.8
5.3
2.1
3.5
99.0
3
grains/ft
x 103
4.00
0.783
2.31
0.913
1.52
43.0
Weight collection
efficiency
percent
99.26
99-88
99.52
99.93
99.88
96.67
a 2
Test No. 65 A-F performed on flat test panel, 0.0348 m (9 in. x
6 in.) at 0.61 m/min (2 ft/min). Average inlet concentration
2.6 grams/m3 (1.1 grains/ft3).
Indicated fabric loading based on weighing test panel and its
holder.
Q
Outlet samples collected on all-glass, Method 5 type filters.
140
-------
rO
I
O
x
o
in
E
u
o
O
UI
UJ
_J
o
Du
UJ
CC.
O
Q.
(Eq.
B V=C0l28.3,N/Ap (Eq. 195
1.0
0.5,
_L
Q2 0.5 1.0
FABRIC RESISTANCE, N/m2 xlO"3
Figure 61. Estimation of pinhole velocities by capillary
(A) and orifice (B) theory for fly ash loaded
Sunbury fabric
141
-------
roughly 750 ym to channel diameter of 350 ym. The channel depth is the
sum of the mound height (1000 ym) and dust cake thickness (1000 ym).
Within the fabric structure itself, the channel seems to decrease abruptly
in cross section to that of the Type I or Type II pore openings (approxi-
mately 200 ym) shown in Figure 30. The fact that the channel diameter
decreases from about 350 to 200 ym over a distance of 100 ym, Figure 30,
suggests that the orifice approach might be more appropriate for velocity
estimates, at least at the higher resistance levels. Conversely, at lower
resistances, viscous losses rather than inertial factors may dominate the
flow picture.
The fact that the measured weight collection efficiency during test 65-F
was 96.7 percent as compared to a predicted value of 86.5 percent based
upon estimated average pore velocity and pore cross-sectional area is
attributed to the following factors: overestimation of pinhole velocity
and/or cross-sectional area and neglect of particle removal from the
actual volume of air passing through the pinhole.
The results of previous permeability measurements on clean Sunbury fabric,
Section V, Figure 21, indicated that laminar flow conditions persisted
until the pore velocities reached 1000 m/min. The former results are in
rough (factor of two) agreement with the data presented in Figure 61.
It is suspected that appreciable dust collects about a pinhole or oversize
pore when gas velocities through the opening are not excessive. However,
in the case of a sudden dislodgement of the dust mass bridging a pore,
high,~ 900 to 1500 m/min velocities rapidly lead to a steady state
collection/reentrainment condition such that net particle removal from
the gas penetrating the pore is negligible.
A distinction should be made between those openings or pinholes that are
present when filtration is initiated through new or cleaned fabric areas
and those openings that can be seen during the later stages of filtration
when lowered efficiencies have been observed.
142
-------
When filtration begins, the actual number of open pores is determined by
the weave or thread count and yarn proximity. Because there are a large
number of openings, the air velocity through any one pore is not suffi-
ciently large to prevent a gradual accumulation of dust in the form of
bridging over the pore entrance. Provided that the range in effective
pore diameter is not excessive (a necessary fabric characteristic for
high efficiency filtration), complete pore bridging can usually be accom-
plished before particle reentrainment rate equals or exceeds particle de-
position rate.
If the range in pore (interyarn spacings) is too large, complete bridging
of most of the pores may take place while a few still remain open. Due
to the much lowered resistance to air flow compared to the caked-over
region of the filter, the air velocities through the remaining openings
as described for test 65-F become too high to allow particle deposition.
Hence, a permanent opening(s) remains and increased particle penetration
takes place. It is emphasized that rough handling, shock, or vibration
may also dislodge dust blocking a pore(s) so that the same problem arises;
i.e., no further chance of sealing the opening until the filter is cleaned
for the next cycle.
In any case, it appears reasonable to assume that a few pinholes may
contribute significantly to effluent loadings when fabric loadings (and
fabric resistance) are high. Therefore, it becomes very important to
determine which factors cause pinhole formation in industrial practice.
One can postulate that poor quality control in fabric manufacture or
careless handling during bag sewing, shipping, and system installation
can lead to breaks, discontinuities or oversize pore openings. In addi-
tion, gross mechanical vibrations, fan pulsation, sticky dampers or
other accidental disturbances may lead to unintended cake flexure that
initiates penetration at critical pore openings.
143
-------
Fabric Appearance After Cleaning
Previous microscope observations, Figures 57 through 60, have revealed
the filter surface characteristics before any dust removal took place
such as the texture of the dust layer, the presence of pinhole leaks,
and particularly the cracking or checking of the dust cake induced by
flexure. It appears that this cracking process, shown again at 60X mag-
nification in Figure 62, is a necessary precursor to cake detachment.
The photomicrographs of Figure 63 provide informative sectional views
of the checking process as the loaded fabric is viewed from the edge.
In Figure 63A a crack can be seen developing above the warp yarn (end
view) that overlaps the fill yarn upon which much of the visible cake
lies.
The density of the surface dust layer was estimated by carefully excising
measured slabs of dust (length, width, and depth) followed by weighing on
an analytical balance. This process indicated an apparent bulk density of
2
0.82 grams/cm . If one measures the actual depth of the dust layers shown
in Figure 63 based upon 20X magnification, they are seen to be roughly
consistent with the average fabric loadings determined for the filter
prior to preparation of samples for photomicrographing.
When the loaded fabric is gently flexed, the dust falls as flakes or
slabs as shown in Figure 64. It has been stated previously that the
main point of separation takes place at the dust/fabric interface.
Although the focal depth at 60X manification does not allow for clarity
at the fabric surfaces when the outer surface of the dust is in sharp
focus, the use of substage illumination shows very clearly the light
transmitted through the cleaned portions of the filter. By focusing
upon the resultant fabric surface after detachment of the dust layer
Figure 64, it can be seen that the warp yarns (the light regions) re-
tain relatively little dust while the fill yarns hold much larger quanti-
ties within the bulked staple material.
144
-------
Figure 62. Checking or cracking of
deposited fly ash layer
on glass fabric by
intentional flexing
(60X magnification).
Test with clean (unused)
Sunbury fabric with
cloth loading of
945 grams/m2
145
-------
Edge view warp yarns seen on end, cloth load-
ing order approximately 1200 grams/m^
Edge view showing checking and fill yarn
cloth loading approximately 440 grams/m^
Figure 63. Dust cake as seen in sectional views with GCA fly ash on
Sunbury fabrics (20X magnification)
146
-------
Filtering face immediately presiding cake dis-
lodgement. Bright, out-of-foaus regions3 are
aleans warp yarns (BOX magnification)
Filter surface after aake dislodgement shaving
relatively clean (bright) warp surfaces and
residual dust on fill surfaces (20X magnification)
Figure 64. Before and after appearance of dirty and cleaned Sunbury
fabric with GCA fly ash filtration
147
-------
To put things in proper perspective, the terminal fabric dust loadings
2
during current tests ranged from 500 to 1200 grams/m whereas the cleaned
2
surfaces retained only about 50 to 80 grams/m . Thus, when a section of
the fabric is cleaned by bag collapse and reverse flow, the surface im-
mediately beneath the detached slab contains very little dust per unit
area in contrast to many fabrics cleaned by mechanical shaking. Prior
tests with Dacron, cotton and glass fabrics, for example, showed average
2
residual dust holdings in the 54 to 300 grams/m range, the latter also
representing the approximate areal density of many common filtration
fabrics including the Sunbury and Nucla glass media. The reasons for
these variations in fabric loadings and their impact upon filter perform-
ance are discussed in the section on weight collection. The photomicro-
graphs of Figure 65 show the appearance of the cleaned surface and the
downstream face at 60X magnification. By means of substage illumination,
the bounding yarns for a Type I pore, approximately 120 ym diameter, are
shown as well as the nearly particle-free (bright) warp yarns. Figure 65
shows the presence of dust in a Type II pore opening as viewed from the
clean face of the filter. Generally, the clean air faces of the filters
loaded in the laboratory for the first time gave very little indication
of the encroaching penetration shown in Figure 65.
Despite the fact that the graphite-silicone coating on the Nucla fabric
tended to mask the true character of residual dust deposits, the residual
dust loading for the Nucla bags closely approximated that of the Sunbury
media. The undisturbed loaded fabric, Figure 66 (approximately 1200
2
grams/m of GCA fly ash) looked the same as its Sunbury counterpart. It
can also be seen, Figure 66, that the residual dust is concentrated on
or within the fill yarns. Although there was actually very little differ-
ence in residual dust holdings for samples shown in Figures 66 and 67 (top),
it is believed that the fuzzier appearance of Figure 67 is due to the
extra dust retentivity of many protruding fill fibers on the new fabric
which are eroded or broken off after a filter has seen 2 years of field
148
-------
••tt-Jr
.<**\* U
Cleaned filtering surface, bright spot is
substage lighting, Type I pore
Clean, warp face showing dust seepage at
Type II pore
Figure 65. Pore appearances for clean and dirty faces of cleaned Sunbury
fabric with GCA fly ash filtration (60X magnification)
149
-------
Filter surface after cake dlslodgement
light-dust deposit on warp yarns
Filter face before cleaning
Figure 66. Appearance of previously used Nucla fabric befire and after
cleaning. GCA fly ash loading of 1200 grams/m2 (20X
magnif icat ion)
150
-------
Filter surface after aake dislodgement. In-
focus granular areas are warp yarns with
light dust coating
Clean (downstream) surface after dust
removal. Minimal indication of particle
penetration at pore locations
Figure 67. Appearance of fill and warp faces of Nucla fabric after removal
of GCA fly ash loading of approximately 1000 grams/m2, pre-
viously clean (unused) fabric (20X magnification)
151
-------
service. Although the reflection from the graphite flakes obscure some
structural details, one can detect little evidence of dust penetration on
the clean air side of the filter shown in Figure 67.
The unmagnified photograph, Figure 68, shows a 150 cm x 100 cm section of
a Nucla test panel in which the dust has been removed from the center
section. Although it is not apparant in this photograph, transmitted
light can be readily seen if the "cleaned" portion is viewed obliquely
at about 45 degrees.
Although dust layers in excess of 1 mm in depth are extremely thick
relative to what one encounters with many mechanical shaking systems
(approximately 0.2 to 0.3 mm at most), it should be remembered that the
fabric acceleration level and that of the adhering dust must reach
o O
approximately 4 to 5 g's (4.4 x 10 cm/sec ) before the dust layer can be
dislodged by tensile or shearing forces. The forces necessary to over-
come dust layer-to-fabric layer adhesion, estimated by Zimon28,29 to range
2
from 100 to 300 dynes/cm , require therefore that the products (mass) x
(fabric acceleration) or (mass) x (gravity acceleration) attain the 100
2
to 300 dynes/cm level during the cleaning action. Since the bag collapse
process usually involves low acceleration, gravity alone at the 1-g level
must be augmented by a correspondingly larger dust mass to exceed the ad-
hesive forces cited above.
The result of this analysis suggests that where 4.5 g's is sufficient to
bring about dust removal by mechanical shaking when the dust layer is
roughly 0.25 mm thick, the dust layer when subjected to gravity forces
alone must attain a thickness of 1 mm or larger before separation occurs
with a simple, reverse-air-supplemented, bag collapse process. Prior
GCA measurements and the present tests indicate that terminal dust
holdings for the glass bags used in the field are in the 500 to greater
than 1000 grams/m range (approximately 0.5 to greater than 1.0 mm
thickness).
152
-------
Figure 68. Photograph showing a section of Nucla test
panel from which dust has been dislodged.
Roughly 3/4 actual size
153
-------
Dust Release From Glass Fabrics
Because of the excellent dust release properties of the glass fabrics, it
is possible to remove a large fraction of the dust deposit by repeated
flexing. The dust detaches in the form of flakes or slabs from the dust/
fabric interface such that the residual dust beneath the detached slab
constitutes about 10 to 20 percent of the total bag fabric weight. It is
emphasized that in normal filtration practice no more than a small frac-
tion of the fabric surface is ever stripped during collapse and reverse
flow cleaning. Therefore, the relationships indicated in Figures 53
through 56 can be applied directly only to the cleaned areas of field
systems and only when the filtration velocity is constant.
As far as fly ash filtration with glass fabrics is concerned, the loading-
curve appears to be the important one from an operating viewpoint.
According to our microscopic and weighing observations of dust removal by
mechanical (flexural) dislodgement of dust, it appears that once suffi-
cient force is applied at the dust/fabric interface to detach the dust
layer, the separation process appears to leave approximately the same
amount of residual dust, Table 15.
The above effect is not unexpected because the dust layer, irrespective
of its physical properties, detaches from the interface region between
the fabric yarns and the dust cake. Even though the sateen weave cotton
has a more pronounced nap structure because of its all-staple yarn con-
struction, its residual fly ash holding was roughly the same as that for
the glass fabrics. In the absence of more detailed information, it
appears that the assumption that the residual dust holdings and residual
resistances for many dust fabric combinations will not vary greatly in
magnitude may be a good first approximation. It is again emphasized that
the residual levels cited above are those for the fabric surface beneath
the detached "slab" of dust.
154
-------
Table 15. RESIDUAL LOADING AND RESISTANCE AFTER FABRIC CLEANING
Ui
Test
No.
66
67
65
71
98A
99B
69
68
83A
Test
Aerosol
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Fly ash
Lignite
Filter fabric
Q
Sunbury fabric
2 years' service
c
Sunbury fabric
2 years' service
Sunbury fabric,
new, cleaned
Sunbury fabric ,
new , cleaned
Sunbury fabric,
new, cleaned
Sunbury fabric,
new, cleaned
Nucla fabric0
<6 months ' service
Nucla fabric
new, cleaned
Sunbury fabric
new , cleaned
Terminal dust
loading
grams /m
432
1011
1220
660
390
660
1000
1000
1200
Residual
dust loading
grams /m
31.0
29.0
66.0
32.0
47.0
56.0
11.0
40.2
63.0
Residualb
resistance
N/m2
31.0 '
67.0
17.4 (2.5)
7.5 (2.5)
10.0 (2.5)
15.0 (2.5)
82.0 (1.2)
18.7 (2.5)
7.5 (2.5)
-------
Ui
Table 15 (continued). RESIDUAL LOADING AND RESISTANCE AFTER
FABRIC CLEANING
Test
No.
81
82
85
95
Test
Aerosol
Lignite
Lignite
Fly ash
Fly ash
Filter fabric3
Nucla fabric
new, cleaned
Nucla fabric
new, cleaned
Cotton sateen
new, cleaned
Dacron crows foot
previously used
Terminal dust
loading
grains /m
1200
1200
920
210
Residual
dust loading
grams /m
92.0
63.0
42.0
16.0
b
Residual
resistance
N/m2
25.0 (2.5)
7.5 (1.2)
56.2 (20)
6.2 <2.5)
All fabrics used at least once prior to cleaning and retesting.
Term in parentheses indicates clean (unused) resistance at 0.61 ra/min
(2 ft/min).
'Fabric previously used in field application.
-------
The information presented in Table 15 does not indicate the actual energy
required to dislodge the dust. It should not be assumed that because of
similar residual resistances and fabric dust loadings that all dusts are
detached with equal ease.
Filtration With Partially Cleaned Filters
Several tests were performed in which roughly half of the fabric dust
holding was removed from test panels before resuming filtration. Tests
84 and 85 in Figure 69 illustrate, respectively, the difference in resis-
tance properties for uniformly and nonuniformly loaded fabrics. The re-
sistance versus fabric loading relationship is also indicated for the
same fly ash/cotton fabric combination when evaluated on a pilot mechanical
shaking system.
The appearance of the partially cleaned media has been shown in Figure 68.
It was also pointed out that dust separation took place at the dust/fabric
interface. Thus, if the filter initially bore a uniform dust layer, the
partially cleaned unit would display two characteristic regipns, the first
from which no dust was detached and the second, a cleaned region having a
2
uniformly distributed residual loading of the order of 50 to 75 g/m ,
Table 15.
The results of three tests in this category, which are reviewed in detail
in Section IX, were instrumental in the validation of modeling concepts
developed under this program.
In Table 16, weight collection efficiencies are reported for various
uniform dust loadings on the cotton fabric for different time intervals,
Curve 1. In Curve 2, the loading process was repeated except that the
test began after about 50 percent of the original fabric dust holding
f\
800 grams/m was removed. The net result was that the cleaned fabric
area held 42 grams/m2 and the uncleaned section about 800 grams/m2 at
the start of filtration.
157
-------
Ln
00
1.2
1.0
fl)
'o °-8
x
CM
^
z 0.6
I
I
« 0.4
CO
UJ
a:
w 0.2
iZ
DESCRIPTION
BENCH TEST 84, TABLE 18
)0 BENCH TEST 85, TABLE 18
(T)A SINGLE 40ft. x 6 in. COTTON BAG WITH MECHANICAL SHAKING
REFERENCE 10 , PAGE IO2 , TEST Z
NOTE : BENCH TEST RESISTANCES CORRECTED TO
O.SIroAmn. VELOCITY and K2 £J O.9I m/min.
tOO 200 300 400 500 6OO
FABRIC LOADING, grams /m2
700
800
900
Figure 69. Fly ash filtration with clean and partially cleaned sateen weave
cotton, flat panel and bag tests
-------
Table 16. WEIGHT COLLECTION EFFICIENCY FOR SATEEN WEAVE
COTTON WITH GCA FLY ASH (SEE FIGURE 69)
Time interval,
min
Average fabric
loading ,
g/m2
Weight collection
efficiency
percent
a b
Curve 1 ' - Uniform fabric dust loading
0-15
15 - 180
0 - 180
0-67
67 - 800
0 - 800
99.8711
99.9990
99.9871
a b
Curve 2 ' - Nonuniform dust loading
0-10
10 - 90
0-90
400 - 417
417 - 775
400 - 775
99.9732
99.9970
99.9940
aCurve 1 Cinlet =7.1 g/m3, Curve 2
=7.7 g/m3
Average filtration velocity, 0.61 m/min (2 ft/min)
In both cases, the overall efficiency levels are seen to increase as
more dust deposits on the fabric surface. In the case of Curve 2, the
residual dust holding of the cleaned portion of the fabric, 50 to
100 grams/m2, presented a more efficient collection surface than the
unused fabric because of partial plugging. Overall performance for the
fly ash/cotton system was approximately the same as reported in earlier
GCA mechanical shaking studies.10
Figure 69 also allows a comparison between flat panel (Curve 2) and full
scale bag filtration tests with the fly ash/unnapped sateen weave cotton
combination. Curve 3. It is indicated that by plotting bag resistance
versus absolute fabric dust holding (rather than against the amount of
dust added during a steady state filtration cycle) the curve assumes a
form very similar to that for the partially cleaned test panel.
159
-------
Examination of Curve 3 shows that its slope differs considerably over most
of its length from that depicting the true K value (Curve 1). Unless
the filtration process is carried out far enough so that dust accumulations
on the previously cleaned and uncleaned surfaces are almost the same, it
is not possible to estimate KO directly from either field or laboratory
measurements unless (1) the fraction of cleaned and uncleaned areas can
be determined and/or (2) the drag properties of these respective areas have
been determined,
SPECIFIC RESISTANCE COEFFICIENT
Effect of Velocity
Limited bench tests, Figure 70, were performed with Sunbury glass fabrics
and the GCA fly ash aerosol to determine the effect of average filtration
velocity on the specific resistance coeffcient. Filtration velocity was
varied from about 0.38 to 1.52 m/min (1.3 to 5.0 ft/min), the approximate
range over which glass fabrics appear to operate most effectively when
filtering hot flue gases.
Because of the spherical nature of the fly ash particles (which should
assist in establishing a reasonably stable bed structure), it had been
assumed that the K2 values (specific resistance coefficient) would not
change appreciably over a moderate range, < 1.52 m/min (5 ft/min). Test
results, however, showed a consistent increase in K~ with velocity, Fig-
ure 71, which can be described quite accurately for the fly ash/glass
fabric system by the relationships:
•^
K2 (N min/g m) = 1.8 V2, V in m/min (20a)
K2 (in. H20 min ft/lb) = 5.95 V^, V in ft/min (20b)
160
-------
1200 -
96 (O)
99 (x)
98 (Q)
200
400 600 800 1000
AVERAGE FABRIC LOADING,g/m2
1200
Figure 70. Effect of filtration velocity (V) on specific resistance
coefficient OL). Sunbury fabric with GCA fly ash
161
-------
.E
z
10
9
8
7
6 •
UJ
E
u,
UJ
o
0
UJ
O
z
CO
co 2
UJ
oc.
o
u.
UJ
a.
CO
T~—I \ T
K2=|.8
5 10
FACE VELOCITY,V, m/min.
15 20
Figure 71. Effect of face velocity on K9, Sutibury fabric with
GCA fly ash i
162
-------
No electrical charge effects were studied during the above test series.
However, air temperature and relative humidity were maintained nearly
constant at 70°F and 55 percent R.H.
Effect of Particle Size
The effect of particle size on filter resistance coefficients was also
examined via tests on selected size ranges of a rhyolite (granite type)
dust with the Sunbury fabric. These measurements were performed with
rhyolite because density and shape factor appeared to be nearly inde-
pendent of size.
The simple procedure of reversing the extraction probe from the aerosol
loop provided the finer of the two size distributions shown previously
in Figure 13. The resistance/fabric loading curves for the coarse and
fine ryholite are shown in Figure 55.
The results indicated that the K values for the two rhyolite size distri-
butions, 1.4 and 9.9 N min/ g m, varied inversely as their respective
mass mean diameters (MMD). The Carman-Kozeny theory indicates, however,
an inverse square relationship, i.e.:
for granular beds comprised of uniformly sized particles.
Therefore, the inverse relationship, KZ = 6 (MMD)~ , determined for the
rhyolite is a purely empirical function resulting from the choice of mass
median diameter as the characterizing parameter.
It should be noted that the Carman-Kozeny relationship can also be ex-
pressed in the form:
>2 = A (A r2 (2D
-------
where S is the specific surface parameter. In the case of a monodisperse
o
particle system, S is a simple linear function of 1/d . When the system
is polydisperse, it appears reasonable to express S as the ratio of par-
ticle surface area to the volume occupied by the particle, A /A . The
P v
terms A and A , respectively, are calculated from the characteristic sur-
P v
face mean and volume mean diameters, d and d , for the size distribution
s v
of interest.
If the size parameters can be computed from a logaritmic normal distri-
bution, the following relationships obtain:
2
log, = log HMD - 4.605 log a (22)
ds S
log = log MMD - 3.454 log2 a (23)
Use of the above equations in conjunction with the size parameters shown in
Figure 13, gave an S ratio of 2.52 for the coarse and fine ryholite dusts.
° 2
Thus, one would expect the respective K values to differ by a factor S
or 6.35. Actually, the ratio of measured K values for the fine and coarse
dust was 7.07 suggesting that the calculation of the specific surface term
provides a better estimate of K values for polydisperse systems than use
of the MMD value alone. In Section IX, the results of a detailed analysis
of the relationship between K- and S are presented for the field and lab-
/ o
oratory measurements conducted during or related to the present study.
Dacron Filtration Tests
t i . . i .. — — .— . ii i in i ii
Additional tests were performed to determine why the collection efficiency
for crowfoot (1/3) Dacron media was so low, ~ 80 percent, with GCA fly ash
compared to prior measurements, " §9.5 percent, with full scale filter
bags (10 ft x 6 in.). Summaries of all Dacron bench tests are given in
Table 17 and Figure 56.
164
-------
Table 17. GCA FLY ASH FILTRATION WITH CROWFOOT DACRON
BENCH SCALE TESTS '
Test
no.a
92 N
92 N
92 N
93 N
93 N
94 N
95 U
Fabric
resistance
N/M2
200
350
726
196
298
188
284
Fabric
loading
grams /n\2
(range)
0-100
100-352
352-726
0-100
100-325
0-184
45-195
Filtration
surface*3
Warp
Warp
Warp
Warp
Warp
Fill
Warp
Dry bulb
temperature
°C
23.3
24.5
25.0
22.0
22.0
23.2
23.5
Relative
humidity
40C
42C
38^
16C
16
40
23
Weight
collection
efficiency
percent**
69.9
78.7
79.8
76.4
78.5
79.3
80,5
N = new (unused) fabric; U = used fabric.
Warp yarns constitute 75 percent of upstream (filtering) surface.
Fill yarns constitute 75 percent of upstream (filtering) surface.
clndic#tes poor electrical ground. All other tests well grounded.
Reported efficiencies apply to indicated fabric loading range.
2
Note: DACRON - Globe Albany No. 856B, 10 oz/yd , 1/3 Crowfoot,
71F x 51F (bulked) thread count, 33 perm.
All tests indicated a slight increase in efficiency as fabric loading in-
creased but hardly at the levels needed for effective performance. Although
one expects to see some differences due to relative humidity, the degree of
electrical grounding, or the direction of air flow through the fabric (warp
or fill face as the filtering surface), the data in Table 17 showed con-
sistently poor and uniform performance irrespective of test conditions.
Because of anomalies in attempted charge measurements (possibly due to a
defective instrument) electrical charge per se either on the dust particles
and/or on the Dacron fabric could not be related to the efficiency results
although is suspected to be involved.
165
-------
Earlier resistance tests, Figure 33, suggest, however, that the differ-
ence in fabric pore structure between full scale bags and small, flat
test panels may explain the gross differences in dust collection. Clean
fabric resistance measurements showed a significant increase in resis-
tance (and hence reduced permeability) with full scale bags in contrast to
equivalent measurements with flat panels. The results indicate, therefore,
that reduced pore dimensions occur with a tubular (bag) configuration as
well as a distortion feature which should represent a "one way" rather
than a "two way" stretch with a nearly square flat panel. Although the
present rationale is qualitative, it appears possible that the larger pore
openings coupled with less uniformity in pore dimensions, may explain the
poor performance encountered with bench scale tests. Several other
dust/fabric combinations including lignite, GCA fly ash and granite dust
with glass fabric (Sunbury and Nucla) and GCA fly ash with sateen weave
cotton encompassed the same range of humidity and inlet loading levels
while showing efficiencies ranging from 99.9 to 99.999 percent.
2
It was noted that a large pinhole population, ~ 5300/m was present at the
2
end of test 92. Because the resistance, 625 N/M , was comparatively high,
there was probably little chance that these pinholes would ever have be-
2
come blocked. On the other hand, the pinhole density of 9000/m observed
for test 93, might have undergone some reduction if the test had been con-
tinued because of the lower air velocity through the pores at test
termination.
If one assumes that those pores that remained unbridged for Dacron media
are about 1.5 times larger than the corresponding pores for the Sunbury
fabric (as indicated by preliminary microscope observations), the estimated
pore areas/cm for tests 92 and 93 are 10 cm and 1.7 x 10 cm , re-
spectively = In conjunction with estimated pore velocities of 4000 and 2000
ft/min, respectively, for tests 92 and 93 and assumed 0 percent dust
removal for air passing through the pores, overall weight collection ef-
ficiencies of 83 and 85 percent are predicted. The observed values
Table 18, were in the 80 percent range.
166
-------
Aside from demonstrating the extent to which a comparatively small pinhole
area can contribute to dust penetration, the analyses of the above data
also suggest that one might use penetration data in conjunction with the
observed pinhole count and filter resistance to estimate the effective
cross-sectional areas of the larger pores.
COLLECTION EFFICIENCY AND PENETRATION
Weight Collection Efficiency Measurements
Weight collection efficiencies, inlet and outlet dust concentrations,
fabric loadings and other relevant test parameters are summarized in
Table 18 for bench scale filtration tests with various aerosol and fabric
combinations. All tests were performed in the filter assembly shown in
Figure 5. During some tests, the filter panels were removed at intermediate
loading levels for visual inspection and determination of fabric dust
holdings. For all practical purposes, the increase in filter dust holding
divided by the air volume passed through the filter provided an accurate
measure of true inlet dust concentration when efficiencies exceeded 99
percent. Otherwise, Method 5 type filter samples were collected upstream
to establish inlet load levels. All outlet concentrations were deter-
mined by the Method 5 type sampling of the test airstream as well as by
supplemental condensation nuclei counting (CNC), Bausch and Lomb (B&L)
single particle light scattering measurements and Andersen cascade impac-
tor samples.
According to Table 18, effluent loadings and collection efficiencies were
about the same for clean (unused) samples of the Sunbury and Nucla filter
fabrics. Except for Tests 65A and 65F, there appeared to be no dependency
upon fabric dust holding. However, inspection of CNC and B&L measurements
in the next section shows a rapid decrease in particle number concentra-
tion up to the point when a fabric dust loading of about 150 grams/m has
167
-------
Table 18. SUMMARY OF BENCH SCALE FILTRATION TESTS
Test No.
65 A
65 B
65 C
65 D
65 E
65 F
65 A-F
66 A
66 B
66 C
66 D
66 A-D
67
68
69
70 A
70 B
70 C
70 D
70 E
70 C-E
70 A-E
71
72 A
Test
dust3
FAC
FA
FA
FA
FA
FA
FA
FA
FA
Fabric
tested
New Sunbury
New Sunbury
New Sunbury
New Sunbury
Sew Sunbury
New Sunbury
New Sunbury
Used Sunbury
Used Sunbury
FA j Used Sunbury
FA
FA
FA
FA
FA
Used Sunbury
Used Sunbury
Used Sunbury
New Nucla
Used Nucla
FA • New Sunbury
FA INev Sunbury
FA
FA
FA
FA
FA
FA
FA
New Sunbury
New Sunbury
New Sunbury
New Sunbury
New Sunbury
Used Sunbury
Used Sunbury
Fabric loading,
g/m5
Initial
0.0
130
210
250
-
960
0.0
31
77
100
280
31
29
0.0
11
0.0
95
230
380
510
230
0.0
32
340
Final
130
210
250
-
960
1200
1200
77
100
280
430
430
1000
1000
1000
95
230
380
510
640
640
640
660
450
Resistance,
N/m2
Initial
2.5
160
180
220
320
750
2.5
31
170
290
410
31
67
2.5
82
2.5
150
250
400
560
250
2.5
7.5
25
Final
170
210
240
320
950
1200
1200
190
290
500
670
670
1200
1200
1200
150
240
370
540
700
700
700
630
370
Dust concentration
Influent ,
g/m3
1.24
1.53
1.10
2.96
2.96
2.99
2.33
0.63
0.31
2.02
1.41
1.09
8.05
6.36
6.25
5.18
7.38
-
-
8.00
7.03
6.84
Effluent,
g/m3 x 103
9.2
1.8
5.3
2.1
3.5
99
20.6
58
41
35
39
40.3
68
6.9
6.9
41
9.2
-
-
2.3
11.2
4.58
"
Height
efficiency,
percent
99.26
99.88
99.52
99.93
99.88
96.69
99.12
90.79
86.77
98.27
97.23
96.31
99.16
99.89
99.89
99.21
99.88
--
-
99.97
99.84
99.93
"
Comments
Pinholes detected, 2500/m2 of fabric.
Fabric from Sunbury Steam Electric Station No. 1 A
Baghouse, cleaned.
Pinholes detected 680/m2 of fabric.
Fabric from Sunbury Steam Electric Station No. 1 A
Baghouse, cleaned, piuholes detected.
Fabric from Nucla Generating Station No. 2 Baghous<
cleaned .
Fabric from test 70, cleaned.
Fabric from test 71. partially cleaned to residual
dust holding of 340 grams /m2 of fabric. Pinholes
detected, 402/a2 of fabric.
00
-------
Table 18 (continued). SUMMARY OF BENCH SCALE FILTRATION TESTS
Test No.
72 B
72 C
72' A-C
Test
dust*
FA
FA
FA
77 A RF
77 B RF
77 C
77 B
77 E
77 A-E
79 A
79 B
79 C
79 D
79 C-D
79 A-D
81 A
RF
RF
RF
RF
RC
RC
RC
RC
RC
RC
L
81 B L
81 C ; L
81 A-C ; L
82 A ! L
82 B L
82 C
L
82 B-C L
82 A-C j L
83 A
83 B
83 C
83 A-C
L
1
L
L
Fabric
tested
Used Svmbury
Osed Sunbury
Used Sunbury
New Sunbury
New Sunbury
New Sunbury
New Sunbury
New Sunbury
New Sunbury
Kew Sunbury
New Sunbury
New Sunbury
New Sunbury
New Sunbury
New Sunbury
New Nucla
New Nucla
New Nucla
New Nucla
New Sunbury
New Sunbury
New Sunbury
New Sunbury
Fabric loading,
g/m2
Initial
450
540
340
0.0
13
30
53
64
0.0
0.0
57
180
300
180
0.0
0.0
550
870
0.0
0.0
540
790
540
Sew Suabury 0.0
Used Sunbury
Used Sunbury
Used Sunbury
Used Sunbury
63
130b
480
63
Final
540
750
750
13
30
53
64
79
79
57
180
300
390
390
390
550
870
1200
1200
540
790
1200
1200
Resistance,
N/m2
Initial
370
520
25
2.5
140.
220
320
410
2.5
2.5
75
160
290
160
2.5
1.2
300
550
1.2
2.5
290
500
290
1
1200
130b
480
760
760
2.5
7.5
87
280
7.5
Final
520
710
710
150
220
Dust concentration
Influent ,
g/m3
-
-
6.73
0.34
0.34
320 0.34
410
560
560
75
160
290
390
0.34
0.34
0.34
1.65
1.65
-
_
390
390
1.65
1.65
Effluent,
g/m3 x 103
-
-
90.83
18.53
2.29
0.23
0.69
0.92
Weight
efficiency,
percent
-
_
98.65
94.55
99.33
99.93
Comments
•
99.80 j
99.73
3.66 98.92
5.72
0.69
-
_
99.65
99.96
-
_
!
0.09
1.14
330 9.08 10.07
620 8.63
980 8.35
980 8.76
310 9.83
520
7.78
24.02
13.3
4.80
-
910
910 10.11 16.47
910 9.98 ' 11.0
87 ; 7.43 : 1.37
280 7.43
570
570
7.85
7.66
1.14
0.92
1.14
99.99
99.93
99.88
99.91
99.71
99.85
99.95
-
-
99.84
99.89
99.98
99.98
99.99
99.981
Pinholes detected, 287/tn2 of fabric.
Pinholes detected, 29/m^ of fabric.
Fabric from test 82, cleaned.
-------
Table 18 (continued). SUMMARY OF BENCH SCALE FILTRATION TESTS
Test No.
84 A
84 B
84 C
84 D
84 B-D
84 A-D
85 A
85 B
85 A-B
89 A
89 B
89 A-B
92 A
92 B
92 C
92-A-C
93 A
93 B
93 A-B
94
95
96
Test
dust3
FA
FA
FA
FA
FA
FA
FA
Fabric
tested
Sew Cotton
New Cotton
Nev Cotton
Nev Cotton
Nev Cotton
Nev Cotton
Used Cotton
FA jUsed Cotton
FA Used Cotton
L
L
Used Sunbury
Used Sunbury
L ! Used Sunbury
1 *
FA
FA
New Dacron
Nev Dacron
PA ;New Dacron
FA JNew Dacron -
!
FA
FA
New Dacron
New Dacron
FA Sew Dacron
FA 'Sew Dacron
j
FA 'Used Dacron
i
i
FA
New Sunbury
Fabric loading,
g/m2
Initial
0
70
280
540
70b
0
400
450b
400
340
410b
340
0.0
100
350
0.0
0.0
100
0.0
0.0
16
0
Final
70b
280
Resistance,
N/m2
Initial
20
100
540 | 210
780 ; 400
780
780
450b
820
820
410b
870
870
100
350
740
740
100
320
320
180
210
400
100
20
110
260
110
17
190
17
5.0
200
340
5.0
3.7
150
3.7
2.5
6.2
11.2
Final
100
220
420
580
580
580
260
700
700
190
600
600
200
350
620
620
190
290
290
190
280
1370
Dust concentration
Influent,
g/ra3
7.10
_
-
-
7.10
7.10
7.69
7.69
7.69
7.60
7.60
7.60
7.33
8.53
8.72
8.43
7.48
7.48
7.48
6.18
6.50
5.37
Effluent,
g/m3 x 103
9.15
_
-
0.07
0.92
2.06
0.23
0.46
21.05
12.58
13.7
2208
1818
1767
1850
2081
1605
1762
1275
1269
58.34
Weight
efficiency,
percent
99.87
_
-
-
99.999
99.99
99.97
99.997
99.991
99.72
99.83
99.82
69.88
78.69
79.74
78.05
72.18
78.54
76.44
79.37
80.48
98.91
Comments
Fabric from test 84, partially cleaned to resid-
ual dust holding of 400 grams/m2 of fabric.
Tests 89 A-B, fabric from test 83, par-
tially cleaned to residual dust holding of
340 grams/m2 of fabric.
Pinholes detected, 86 m/2 of fabric.
Tests 92 A-C, outlet side of fabric very dirty
after filtration.
Pinholes detected, 5223/m2 of. fabric.
Pinholes detected, 5310/m2 of fabric.
System well grounded, pinholes detected, 8897/m2
of fabric.
Air flow through fabric opposite normal filtering
direction. Pinholes detected but not countable.
Fabric from previous lab studies, pinholes detect
but not countable.
Filtered at 1.51 meters/minute,, pinholes detected
603/m2 of fabric.
-------
Table 18 (continued). SUMMARY OF BENCH SCALE FILTRATION TESTS
Test Ho.
97
98 A
98 B
98 A-B
99 A
99 B
99 A-B
Test
dust3
FA
FA
FA
FA
FA
FA
FA
Fabric
tested
Used Sunbury
Used Sunbury
Used Sunbury
Fabric loading,
g/mz
Initial
270
47
90b
47
56
140b
56
Final
390
90b
620
620
140b
660
660
Resistance,
H/m2
Initial
162
5
62
5
15
130
15
Final
1100
62
260
260
130
540
540
Dust concentration
Influent ,
g/m3
4.60
8.09
8.09
8.09
8.40
8.40
8.40
Effluent,
g/m3 x 103
562.8
16.7
1.83
3.20
43.01
1.60
6.86
Weight
efficiency,
percent
87.77
99.79
99.98
99.96
99.49
99.98
99.92
Comments
Fabric from test 96, partially cleaned to residua
dust holding of 270 grams/m2 of fabric, filtered
at 1.52 meters/minute, pinholes detected, 3588/m2
of fabric.
Fabric from test 97, cleaned, tests 98 A-B fil-
tered at 0.39 meters/minute.
Fabric from test 98, cleaned, tests 99 A-B fil-
tered at 0.60 meters/minute.
FA = fly ash; RF = Rhyolite fine; RC = Rhyolite coarse; L = Lignite.
Fabric loading estimated from inlet concentration, flow rate and time; i.e., W = CQt.
GCA fly ash was filtered at 0.61 meters/minute for all tests except those indicated. GCA fly ash HMD - 9 ym; ag • 3.0.
-------
accumulated. Test 65A indicates a high effluent concentration during the
early filtration phase. A very significant decrease in efficiency was
observed for Test 66F as the result of severe pinhole formation or punc-
ture damage.
It was concluded that accidental overstressing or vibration of the filter
cake before or during reinstallation for the final filtration sequence was
at last partly responsible for the pinhole formation. Details on the
appearance, number, approximate dimensions and the total volume of air
flow diverted through the pinholes have been discussed previously. Tests
on one previously used Sunbury fabric, Test 66, indicated the presence of
pinholes throughout the entire measurement interval. Semiquantitatively,
the estimated total pinhole area in the latter case appeared consistent
with the reported effluent concentrations and efficiencies, ~ 88 to 99
percent.
By not disturbing a second sample of the used Sunbury fabric, Test 67, an
overall test efficiency of 99.15 percent was obtained. Based upon these
data, it appears possible that the less effective performance of the used
Sunbury fabric may have resulted in part from its 2 years of field service.
On the other hand, since previously reported field data indicated fairly
8 9
high efficiency levels, ~ 99.5 percent or greater, ' it is suspected that
bag removal and shipment to our laboratories plus subsequent cleaning and
handling were the main contributors to poorer performance.
Little can be said for the Nucla tests, Nos. 68 and 69, except that high
efficiency levels were observed for both used and unused fabrics. However,
the Nucla media had seen less than 6 month's field service.
New Sunbury fabric did not indicate relatively high efficiencies, ~ 99.9+
percent for all test conditions. When the fabric surface was partially
cleaned, Test 72A-C, the overall efficiency during the dust reloading
process was 98.65 percent. It is again pointed out that resumption of
172
-------
filtration with a partially cleaned filter surface leads to transient high
velocities, about 1.5 m/min (5 ft/min) in the present case, that lower the
collection efficiency. In most cases, the effluent concentrations from
the glass fabrics were in the range of 2 to 5 x 10~3 grains/ft3 (~ 4.6 to
10 mg/m ).
The fact that very thick dust cakes deposited on the glass fabrics (~ 0.5
to 1 mm) compared to those for mechanically shaken cotton fabrics (~ 0.2 mm)
does not necessarily imply equivalent performance.
If the pore structures for the cotton and glass fabrics are examined, it
is seen that the volume of fiber obstructing the pores is greater and the
distance between individual fibers is less for the sateen weave cotton.
Therefore, the particulate emissions are expected to be lower for the
cotton fabric for two reasons: (a) the "tighter" weave provides a firmer
support upon which to develop a uniform, unbroken dust layer, and (b) the
greater number of interlacing fibers obstructing the pores will reduce
agglomerate slough-off from the rear face of the filter. The above re-
lease is not to be confused with the "pinhole plug" emissions described
24
by Leith et al. . In the latter case, an open channel or pore is created,
perhaps 50 to 200 vim in diameter, through which the upstream aerosol passes
with very little dust removal. As a result, the "pinhole" plug effluents
are also expected to reflect the upstream concentration level. On the
other hand, rear face slough-off without breakthrough is expected to consist
of a low order emission that may depend upon face velocity but not neces-
sarily relate to the inlet loading. Limited cascade impactor measurements
suggest that the mass median diameter for the slough-off material is
roughly half that of the inlet dust, 3 versus 6 ym. As far as the fly ash/
woven glass fabric combinations are concerned, it is suspected that the
effluent loadings reflect both rear face slough-off and direct penetration
through pinholes.
Because the size parameters for the rear face slough-off are not radically
different from those of the inlet duct and because the former effluent
173
-------
ordinarily represents only a small fraction of the total effluent, the
dust fraction that penetrates the filter directly is the one that controls
the overall effluent particle size properties.
In some cases, the confirmed presence of pinhole leaks provided a ready
explanation for the observed penetration values. However, it is very im-
portant to note that a careful inspection of the dirty surface of a filter
often revealed no pinholes despite the observed downstream emissions.
Thus, it was assumed that periodic sloughing-off of dust from the dust
layer/fabric interface region was the major source of such emissions,
once all pores were completely bridged.
It should be noted, however, that many mass samples were collected over
long averaging periods. Thus, the greater part of the dust collected
downstream probably penetrated the fabric before the pore bridging process
was completed. Unfortunately, although CNC measurements provided excel-
lent time resolution for effluent loadings, the instrument sensitivity
3
usually precluded detection of concentrations less than 0.5 mg/m . The
net result was that instrument limitations often prevented any sharp dif-
ferentiation among the factors contributing to the filter effluents.
When the effluent concentrations for a sateen weave cotton are compared
with those for the glass fabrics, it is seen that the former are as much
as 50 times less, Test 85b, Table 18. As stated previously, it is believed
that the much higher fraction of staple fibers in the cotton fabric pro-
vides a stronger bridging mechanism over the pore regions so that less
material is sheared off by aerodynamic drag. The more substantial pore
bridging with the cotton fabric also reduces the chance for cracking and
pinhole formation in the overlying dust cake.
Condensation Nuclei Measurements
CNC concentrations were observed to undergo extreme fluctuations during
Test 65, Figure 72, whenever the filter panel was removed and reinstalled.
Subsequently, the removal difficulties were eliminated by maintaining air
174
-------
10
- NOTES
NUMBER CONCENTRATION BY
CONDENSATION NUCLEI COUNTER
WEIGHT EFFICIENCY BY UP-AND
DOWNSTREAM GRAVIMETRIC SAMPLING
(ALL GLASS FILTERS)
-X - X
*•
1
RANGE
A
B
C
D
E
F
WEIGHT
EFFICIENCY
99.26
99.88
99.52
99.88
99.83
95.67
_L
_L
_L
200 400 600 800 1000
FABRIC LOADING (W), groms/m2
1200
Figure 72.
Effluent concentration versus fabric loading for unused Sunbury
media with GCA fly ash, Test 65
-------
flow through the filter until it was rotated into a "dust face up" posi-
tion. Since the fabric surface was already bowed downward as it would be
under the normal static dust load, stopping the air flow produced only
minimal filter flexure (and minimal cake compression).
From the practical perspective, the early handling difficulties revealed
that accidental jarring and vibration could have a serious impact upon
field performance if applied at the wrong time. During subsequent test-
ing it was difficult to detect when the filter panel was removed as the
result of improved experimental techniques. On the other hand, the
Test 66 filter (used Sunbury media), Figure 73, showed consistently
erratic nuclei emissions at a concentration level about 30 times higher
than noted for the unused fabric during its stable operation phase,
Test 65E, Figure 72. A replicate test, No. 67, on Figure 74, indicates
an outlet concentration level which was not much lower than that for
Test 66. Thus, CNC results as well as those for weight collection effi-
ciency suggest that the Sunbury fabric has undergone some form of
degradation.
Inspection of Figures 75 and 76 for the used and unused Nucla fabric show
much lower and nearly identical results. It is emphasized that the final
uniform nuclei concentration levels do not necessarily represent a level-
ling off at a concentration of 800 x 10~" n/m^. According to the manufac-
turer, the true minimum nuclei concentration might be anywhere from zero to
the actual value when a constant low concentration level is indicated.
In later tests, the lowest recorded value decreased to approximately
200 x 10"6 n/m3.
The effect of filtration with a uniform and a nonuniform fabric dust
loading is shown in Figure 77. The effluent nuclei concentration data
are based upon Tests 71 and 72 with GCA fly ash and a relatively new
Sunbury fabric test panel. In the case of Test 72, approximately half
of the dust was removed prior to resuming filtration. As stated previously,
that area from which the dust cake had been removed performed initially as
a completely cleaned fabric with a W value of the order of 50 grams/m2
176
-------
NUCLEI CONCENTRATION, No./m3xlO"6
6 5 6
o» •** . <•
— NOTE: MEASUREMENTS BY CONDENSATION
- NUCLEI COUNTER
I x x
_ „ X WEIGHT
X RANGE EFFICIENCY
-XXX %
X Y X A 88'6
- v XX x * X B 94-7
XX X C 98.6
yX D 97.0
XX X XX
rr x x
I X
X
.«>•*
-^ /-v "^O]** O ^|^ l^/1 ...i- ,-.-.. ^j
1 1 1 1 1 I 1 1 1 1
100 200 300 400 500
FABRIC LOADING, groms/m2
Figure 73. Effluent concentration versus fabric loading for used Sunbury
fabric (Test 66) with GCA fly ash
-------
I05c
u>
i
O
10
E
Z
O
\-
(T
UJ
O
O
o
s io'
o
NOTE'MEASUREMENTS BY CONDENSATION NUCLEI
COUNTER (CNC)
X X
„ XX X
X
X
99.12% AVERAGE WEIGHT EFFICIENCY
1 1
200 400 600 800 1000
FABRIC LOADING, (W),grams/m2
1200
Figure 74. Effluent concentration versus fabric loading for used
Sunbury media with GCA fly ash (Test 67)
178
-------
ID
O
X
10
E
10'
u
u
z
O
O
ID
NOTE' MEASUREMENTS BY CONDENSATION NUCLEI
COUNTER (CNC)
-99.89% AVERAGE WEIGHT EFFICIENCY
200
j_ i_
400 600 800 1000
FABRIC LOAD1MG, (W), grams/m2
1200
Figure 75. Effluent concentration versus fabric loading for unused
Nucla (Test 68) media with GCA fly ash
179
-------
I03-
(D
I
O
o
Ul
o
I I03
2
o
z
10'
WX
1
1
5
NOTE: MEASUREMENTS BY CONDENSATION NUCLEI
COUNTER (CNC)
vvw
x»x
xxxx xxxxx
X3KX X XXX
3.89 70 AVERAGE WEIGHT EFFICIENCY
^
200 400 600 800
FABRIC LOADING, (W), grams/rri2
1000
1200
Figure 76. Effluent concentration versus fabric loading for used
Nucla fabric (Test 69) with GCA fly ash
180
-------
I06C
\- x
o
x
10
"s
C
-------
and a filtration velocity roughly 2.5 times the average velocity. Inspec-
tion of the nuclei concentration changes with respect to increased fabric
loading, Figure 77, indicates that the effluent concentrations were from
10 to 60 times higher during the reloading of the partially cleaned fabric.
During Test 65, the filter panel was removed twice for check weighings
2
at fabric loadings of 450 and 540 grams/m . It appears that the filter
cake was disturbed in both cases leading to the formation of pinholes.
The fact that nuclei concentrations eventually displayed a tendency to
decrease after each perturbation suggests that the filter performance
might have improved with additional dust accumulation.
Effluent nuclei concentrations for lignite fly ash filtration with the
Sunbury fabric are shown in Figure 78. It was observed that the effluent
concentrations dropped to low values once the fabric loading reached the
2
150 g/m level. The same conditions were noted when the GCA fly ash was
filtered with sateen weave cotton, Figure 79. Because the lower limit
of sensitivity was reached by the CNC device during some tests, the true
instantaneous values for outlet concentrations could not be estimated.
The CNC measurements for GCA fly ash filtration with Sunbury fabric shown
in Figure 80, indicate clearly that particle penetration increases
with increased filtration velocity. Curves 3 and 4 indicate that effluent
nuclei concentrations were generally 30 to 40 times lower at reduced
filtration velocities, 0.37 to 0.61 m/min (1.3 to 2 ft/min) as compared
to filtration at 1.51 m/min (5 ft/min), Curve 1. Again after partial
cleaning of the fabric, the resumption of filtration leads to even higher
effluent concentrations during the transient high velocity flow period
through the previously cleaned area of the filter. Analyses of these data,
in conjunction with measurements of effluent mass and nuclei concentrations
have been used to model concentration versus velocity relationships dis-
cussed in Section X.
182
-------
10
- X
ID
I
O
X
W6
•x
c
O
Z
tu
a
O
2
to
10'
_ x
n
\
200 400 600 800
FABRIC LOADING (W), gromc/m2
1000
Figure 78. Effluent nuclei concentration versus fabric loading for
used Sunbury fabric with lignite fly ash, Test 83
183
-------
10
o
z
Ul
O
8
itf
10
- X
x—x—x.
200 400 600
FABRIC LOADING(Wl.qroms/m2
800
Figure 79. Effluent nuclei concentration versus fabric loading
with used cotton sateen and GCA fly ash, Test 84
184
-------
10'
10s
g
H
c
*>
i
gio'
o
8
10
10'
NOTE: MEASUREMENTS BY CONDENSATION
NUCLEI COUNTER (CNC)
TEST FILTRATION VELOCITY
1.51 m/min
1.51 m/min
0.37 m/min
0.60 m/min
96
97*
98
99
PARTIALLY CLEANED FILTER
100
200
300
400
500
600
'00
FABRIC LOADING, (W),grams/m*
Figure 80. Effect of filtration velocity on effluent nuclei concentra-
tion, GCA fly ash with new Sunbury fabric
185
-------
Particle Size and Concentration by Optical Counter
Particle sizing measurements performed by B&L particle counter are given
in Figures 81 through 83. Test 67 data, Figure 81, show a close parallel
to the corresponding CNC measurements in that a slight increase for all
particle sizes occurred over the testing interval. The presence of par-
ticles in the 2 to 5 ym range suggest strongly that the emissions were
mainly the result of pinhole leaks that allowed many coarse particles to
penetrate. Because of electronic choking (excessive particle numbers in
the size range < 2 ym) , the only measurements possible; i.e., those for
the > 2 ym size categories have qualitative value only. For example, if
we assign an average volume diameter of 3.5 ym for all particles in the
> 2 ym to > 5 ym range, the predicted mass concentration based on a
7 3
number concentration of 6.5 x 10 particles/cm and spheres of density
3 3
2.0 becomes 2.9 x 10 grams/m . Reference to Table 18 shows the above
concentration to be about 24 times less than that computed by Method 5
sampling. Loss of coarse particles that contribute heavily to mass and
the neglect of the fine fraction probable explain the discrepancy.
It was indicated in Section IV that there was good agreement between
atmospheric dust concentrations determined by gravimetric methods and
those computed by converting B&L number concentrations to their equivalent
mass values. This would appear to strengthen the argument that the much
lower mass concentrations calculated from B&L effluent measurements (20 to
100 times lower) is the result of the inability of the B&L device to collect
and efficiently detect particles (or agglomerates) much larger than 5 ym
in diameter. If this is true, it is quite likely that only those number
concentrations reported for particles less than 5 ym are reasonable
approximations.
Bausch and Lomb measurements for the Nucla fabric tests, Figures 82 and 83,
indicate both lower effluent concentrations and finer size distributions
relative to the Sunbury tests. Despite the fact that the absolute values
of CNC and B&L tests may be subject to question under some test conditions,
they are not only consistent with each other but also follow the same
186
-------
10'
AA
_ A
AA
<0
o
x
evi
E
Q.
cc
u
o
o
o
o
10'
NOTE: MEASUREMENTS BY
BAUSCH a LOMB
SINGLE PARTICLE COUNTER
io
- D
D
SYMBOL
O
A
O
SIZE ft
>Z
>3
>5
D
D
to-'
I
I
200 400 600 800
FABRIC LOADING (W), groms/m2
1000
1200
Figure 81. Effluent concentration versus fabric loading and particle
size for used Sunbury media with GCA fly ash, Test 67
187
-------
NOTE: MEASUREMENTS BY
BAUSCHaLOMB
SINGLE PARTICLE COUNTER
10
-X
to
i
o
K>
A
X
X
O
o
u
_l
o
10
A
X
SYMBOL
A
X
6
A
D
10
1
1 1
200 400 600 800
FABRIC LOADING (W), groms/m2
1000
1200
Figure 82. Effluent concentration versus fabric loading and particle
size for unused Nucla fabric with GCA fly ash, Test 68
188
-------
<0
I
o
N.
a
z
o
z
IU
o
||0C
UJ
_l
O I-
io-'
X"""
NOTE: MEASUREMENTS BY
•A BAUSCH a LOMB
SINGLE PARTICLE COUNTER
D A
O
o
*^
: x A
- 0 A
X X A
- . ^ A A n
r x
I °
o
1 1 1 1 I 1 ! !
. .
•
SYMBOL SIZE >i.m
• >0.3
A >o.5
X >I.O
O >2.0
A >3.0
D >5.0
X
1 I I | 1
200 400 600 800 1000 1200
FABRIC LOADING (W), grams/m2
Figure 83. Effluent concentration versus fabric loading and particle
size for used Nucla media with GCA fly ash, Test 69
189
-------
trends set by the gravimetric measurements. Therefore, we believe that the
present program again demonstrates the value of both techniques as means
for detecting and/or explaining rapid changes in filter system function.
The B&L measurements of Figure 84 show that the filtration of a lignite
fly ash with Sunbury fabric (Test 83) produces a finer and lower concentra-
tion effluent than that obtained with the GCA fly ash and the Sunbury and
Nucla fabrics. Effluent nuclei concentrations also decreased to low levels
2
once the fabric loading reached the 150 grams/m level and greater,
Figure 78. The calculated mass effluent concentrations based upon B&L
measurements were about 200 times lower than the gravimetrically deter-
mined levels, Tests 83B and 83C, Table 18. Again, it is believed that
the slough-off of large particles capturable by Method 5 type sampling
but not detected by the B&L sensing system, account for the extreme dif-
ference in estimated effluent concentrations.
Figures 85 and 79 indicate B&L and CNC measurements for a GCA fly ash/
sateen weave cotton system for the same test conditions used in Test 83.
The key observation is that the effluent concentrations as determined by
both the B&L and CNC devices are quite similar to those noted for the
lignite/Sunbury fabric combination described in Figures 84 and 78. On the
other hand, gravimetric measurements showed a greater than 10 times reduction
in effluent concentration for the cotton system, Table 18. Our interpretation
of these data is that the B&L measurements provide a reasonable estimate
of the particle concentrations in the 0.3 to 1.0 pm size range. It is sus-
pected that the staple fiber configuration of the cotton fabric provides an
intercepting, loose fiber mesh above the pore openings that greatly diminishes
the chances of pinhole development as well as the tendency for aerodynamic
reentrainment of agglomerates from the rear (downstream) face of the dust
layer.
Nuclei Versus Mass Concentrations
An extensive review and assessment of condensation nuclei measurements
performed during GCA laboratory and pilot tests have revealed that these
190
-------
10'
(0
,0°
O
I
_
o
o:
Id2
A
O X
A
SYMBOL SIZE.jum
D > 2.0
O l.O
* 0.5
X 0.3
200 400 600 800 1000
FABRIC LOADING (W),grams/m2
Figure 84. Effluent particle concentration versus fabric loading
and particle size for used Sunbury fabric and lignite.,
Test 83, B&L measurements
191
-------
Figure 85.
200 400 600 800
FABRIC LOADING (W), gram»/ra2
1000
Effluent concentration versus fabric loading for
unused cotton sateen with GCA fly ash (Test 84)
B&L measurements
192
-------
data,when used properly, provide an excellent record of the rapid changes
in mass effluent concentrations that take place during a typical filtration
cycle. Average nuclei concentrations, which were determined by conven-
tional graphical integration methods, embraced the same fabric loading
intervals used to establish the corresponding average mass concentration
by filter sampling.
A. graph of the average nuclei concentrations observed or calculated for
the filter effluents versus concurrent mass measurements by all-glass
filters indicates a linear relationship over the concentration ranee
-333
~ 10 g/m to 10 g/m (see Figure 86). Although one can question the
absolute nuclei counts as displayed by a CNC device, one expects a properly
functioning unit to provide reproducible measurements. Thus, with respect
to any aerosol the nuclei population should bear a constant relationship
to the corresponding total mass concentration.
The fact that a linear relationship previals for the effluent measurements
has some interesting implications. First, any significant downstream
nuclei concentration must arise from penetration through the filter, either
through open pores (or pinholes) or through a "thin" freshly formed dust
layer. No nuclei-sized particles should be generated by particle slough-off
from the filter because adhesive forces preclude the release of anything
but agglomerates or discrete particles in the 5 to 10 ym range or larger.
Secondly, the existence of the same proportionality between nuclei and mass
concentration in the filter effluent indicates that the observed mass pene-
tration is principally that which escapes through unblocked pores during
the early filtration period o£ through large pores (pinholes) that fail to
bridge over at any time during the filtering cycle. One can also infer that
the persistence of this proportionality means that very little dust is
removed from that fraction of the gas volume which passes through these
unblocked pores.
The end result is that the dust that passes through the open pores prior
to their bridging dominates the size properties of the filter effluent.
Accordingly, when we examine the comparative size properties of inlet
193
-------
AVERAGE INLET NUCLEI CONCENTRATION, Ni
A AVERAGE OUTLET NUCLEI CONCENTRATION, N0
X AVERAGE OUTLET NUCLEI CONCENTRATION, 10 ft. x 4 in. BAG
(I) N^ = N0 /PENETRATION - ASSUMES 100% NUCLEI PENETRATION
THROUGH P1NHOLES AND 0% PENETRATION THROUGH
DUSTCAKE
AVERAGE VALUE OVER MEASUREMENT PERIOD
10
MASS CONCENTRATION, g
Figure 86. Calibration curve - nuclei and related mass concentrations
for GCA fly ash
194
-------
and effluent dusts we see very little difference in size distribution
measurements by cascade impactor provided that all phases of the filtra-
tion cycle are properly represented. For this reason, practically all
field and laboratory tests performed with identical and properly function-
ing sizing apparatus show essentially the same collection efficiency for
all but the largest particles, 30 ym or larger. The basic problem is that
so much more of the inlet dust passes through the pores with little or no
dust removal as compared to that which passes through the dust cake, that
the real fractionating potential of the dust cake is completely obscured.
The inlet nuclei concentrations were estimated by the following indirect
method. It was assumed that the observed outlet nuclei concentrations
were attributable to the direct penetration of nuclei through pores or
pinholes. It was also assumed that the total dust concentration penetrating
a pinhole is directly proportional to the nuclei concentration because dust
removal from the fraction of air passing through the pinhole is negligible
for glass fabrics of the Sunbury or Nucla types. Therefore, if mass mea-
surements are available to define inlet and outlet concentrations, the
penetration value can be applied to the outlet nuclei concentration to
estimate the corresponding inlet value; i.e., C. = C /P . As stated
i on
previously, this step appeared reasonable because size distribution mea-
surements for upstream and downstream aerosols are nearly the same.
The relationship between nuclei and mass concentrations shown in Figure 86
actually constitutes a calibration curve. These data have been used to
transpose point values for outlet nuclei concentrations determined under
several test conditions to their equivalent mass values.
A summary of concentration, efficiency and penetration data for fly ash
filtration with glass fabrics is given in Table 19. Information on the
state of the filters during these tests with respect to pinholes (if any),
prior service of the filter and its dust holding range during each test
are also presented. A similar data tabulation, Table 20, compares inlet
195
-------
Table 19. SUMMARY OF CONCENTRATION, EFFICIENCY AND PENETRATION MEASUREMENTS FOR GCA
FLY ASH FILTRATION WITH WOVEN GLASS FABRICS AT 0.61 m/min FACE VELOCITY
\o
lest no.
65 A
65 B
65 F
66 A
66 C
66 D
67
69e
70 A
70 B
96f
97f
98 A8
98 B
99 A
99 B
71
72 A-C
Inlet
concentration
gravimetric
1.24
1.53
3.00
0.63
2.02
1.41
8.05
6.36
6.25
5.18
7.38
5.37
4.60
8.09
8.09
8.40
8,40
6.84
6.73
Outlet concentration
g/m3 x 103
Gravimetric
9.2
1.8
99
58
35
39
68
6.9
6.9
41
9.2
58.3
563
16.7
1.8
43
1.6
4.6
90.8
CNCb
28.0
1.6
96.0
80.0
32.0
28
33
8.0
8.0
95.0
1.5
65.0
220
120
0.4
89
0.9
5.5
60.0
Fractional penetration
Gravimetric
and CNC
0.023
0.0010
0.032
0.126
0.016
0.028
0.0041
0.0013
0.0013
0.018
0.0002
0.012
0.048
0.014
0.00005
0.010
0.0001
0.0008
0.0089
Gravimetric
0.0074
0.0012
0.033
0.102
0.017
0.022
0.0084
0.0011
0.0011
0.008
0.0012
0.011
0.122
0.0021
0.0002
0.0051
0.0002
0.0007
0.0135
Fractional efficiency
Gravimetric
and CNC
0.977
0.9990
0.968
0.874
0.984
0.978
0.9959
0.9987
0.9987
0.982
0.9998
0.988
0.952
0.986
0.9999
0.990
0.9999
0.9992
0.9911
Gravimetric
0,9926
0.9988
0.967
0.908
0.983
0.972
0.9916
0.9989
0.9989
0.992
0.9988
0.989
0.878
0.9979
0.9998
0.9949
0.9998
0.9993
0.9865
Comments
Initial filter
state0
new
loaded
loaded
just cleaned
loaded
loaded
just cleaned
new
just cleaned
new
loaded
new
just cleaned
just cleaned
loaded
just cleaned
loaded
just cleaned
just cleaned
Pinhold leaks
-
-
2,500
yes
680
680
yes
yes
yes
yes
?
603
3,588
yes
no
yes
no
yes
402
Fabric loading
range - g/m
0 - 130
130 - 210
750 - 1200
31 - 77
106 - 280
280 - 430
29 - 1000
0 - 1000
11 - 1000
0 - 128
95 - 230
0 - 400
270 - 390
47 - 90
90 - 620
56 - 140
140 - 660
32 - 660
340 - 750
Gravin-.etric refers to Method 5 filter sampling or direct weighing of test panels.
CNC measurements converted to equivalent mass concentration based on calibration curve.
CNew indicates first use of filter; loaded designates intermediate test; just cleaned refers to partial or complete cleaning.
j 2
Observed number of pinholes per m ; ye_s_ means pinholes detected but not counted; all new or cleaned filters.
• eN-jclo fabric for Tests 68 and 69; Sunbury fabric for other tests.
Face velocity = 1.52 m/min (5.0 ft/tnin).
sFace velocity = 0.38 m/min (1.3 ft/min).
-------
VO
-si
Table 20. INITIAL AND AVERAGE OUTLET CONCENTRATIONS AND RELATED PENETRATION
DATA FOR FLY ASH/WOVEN GLASS FABRIC FILTERS
Test No.3
65 A
65 B
65 F
66 A
66 C
66 D
67
68
69
70 A
70 B
96
97
98 A
98 B
99 A
99 B
71
72 A-C
Inlet concentration
nuclei/cm^
CNCb
1.0 x 105
3.4 x 105
9.2 x 105
4.3 x 105
2.8 x 106
3.5 x 106
1.3 x 106
Calibration
5.6 x 105
1.36 x 106
3.0 x 105
9.0 x 105
6.5 x 105
3.8 x 106
3.0 x 106
2.9 x 106
2.4 x 106
2.5 x 106
2.1 x 106
3.8 x 106
3.9 x 106
3.2 x 106
3.1 x 106
Initial outlet*1
concentration
Nuclei/cm
1.5 x 105
1.5 x 105
3 x 105
2.5 x 105
2.7 x 105
1.5 x 105
1.5 x 105
2 x 105
1.8 x 105
7.2 x 104
2.6 x 105
g/m3e
0.335
0.315
0.636
0.537
0.580
0.322
0.428
0.150
0.565
Fractional penetration
Initial
0.27
0.500
0.100
0.086
0.112
0.060
0.053
0.022
0.084
Average
0.0074
0.102
0.0011
0.0011
0.008
0.011
0.0021
0.0007
0.0135
Average outlet
concentration
Nuclei/cm
1.27 x 10"
7 x 102
4.67 x 104
3.6 x 104
1.6 x 104
1.25 x 104
15 x 104
3.7 x 103
3.6 x 103
Initial fabric
loading^
g/B2
0.76
0.38
3.9
3.8
4.3 x 104 i 3.2
0.7 x 103 |
2.9 x 104
3.3
1.0 x 105 |
5.2 x 104
2.5 x 102
4.9
3.8 x 104
3.9 x 102
2.5 x 103
2.7 x 104
4.2
4.1 .
See Table for additional concentration and penetration data.
b
Estimated from outlet nuclei concentration and filter penetration computer from gravimetric
Nuclei penetration assumed to reflect air volume passing through open pares or pinholes that
dust.
Estimated from gravimetric equivalent and calibration curve.
First measurable CNC data; assumed to relate to added dust increment of 4 g/tn (roughly one
after initiation of filtration).
Equivalent mass concentration from calibration curve.
Average penetration over test interval, see Table 19.
gCorresponds to initial fractional penetration.
measurement.
collect no
-------
and outlet concentrations on the basis of nuclei counts. Initial penetra-
tion values are based upon the initial measurements of condensation nuclei
concentrations.
The "initial" value depicts the nuclei concentration about one minute
following resumption of filtration after which time the flow has stabilized.
If the average inlet concentration is assumed to be about 6.5 g/nH, the
2
average fabric holding after 1 minute is about 4 g/m . However, in the
subsequent development of the relationship between effluent concentration
and filtration velocity, the actual fabric loadings at 1 minute (last
column, Table 20) were computed based upon the observed inlet loadings.
Effluent Concentrations Versus Face Velocity
Data for Sunbury and Nucla fabrics are shown in Figure 87 for fly ash filtra-
tion at 0.61 m/min face velocity and for Sunbury fabrics at three filtration
velocities in Figure 88. The coordinates for the origin of each curve are
the inlet dust concentration and the increment of fabric loading added fol-
lowing initiation of filtration. It is expected that the true "instantaneous"
effluent concentration is about one half that of the inlet value as will be
discussed in Section X. Additionally, it was also expected that the initial
effluent concentration would increase, although not necessarily linearly,
with the inlet concentration. Data points for these curves are summarized
in Table 21 for both nuclei and mass concentrations, as a function of time.
Mass concentrations were estimated from point values of nuclei counts and
the calibration curve, Figure 86.
Rating Fabrics With Atmospheric Dust
Most woven fabric filters perform poorly when filtering atmospheric dust
only because there is no solid dust layer for particle removal. Only
after several months does sufficient dust accumulate to provide effective
filtration. For this reason, precoating or flpcking techniques have been
198
-------
IW I"1 — • -
J
2
5
10 y
e
•v.
^
z' 10"'
o
h-
_ —
o^ ~
\ :
\ ° 0
\ o •
i^^^ ^ o -
— » (
Q "^^-^ r
x f
•j
-
i i i i > i
20 40 60 80 100 120 14
AVERAGE FABRIC LOADING, g/m2
Figure 87. Outlet concentration versus fabric loading at 0.61 m/min
(2 ft/min) face velocity. GCA fly ash with Sunbury and
Nucla fabrics. Loading increase referred to start of
filtering cycle
199
-------
10
O
\
V
CD v
• — i —
TEST
T— |
FABRIC
SUNBURY
CONDITION
USED (L)
rp^c.e VEU
"0.39"!
CITY
10
z" 10
o
10"
IO'3
10"
AVG.
96
97
SUNBURY-NUCLA USED"/NEW 0.61
SUNBURY NEW 1.82
SUNBURY USED(L) 3.35
© +
©X)
0 D
NOTE'L INDICATES LABORATORY USE ONLY -=
04 20 40 60 80 100 120
AVERAGE FABRIC LOADING,
140
Figure 88. Outlet concentration versus fabric loading for three face
velocities. GCA fly ash and Sunbury fabric. Loading in-
crease referred to start of filtering cycle
200
-------
Table 21. CHANGE IN EFFLUENT CONCENTRATION WITH INCREASING FABRIC LOADING FOR
FLY ASH FILTRATION WITH WOVEN GLASS FABRICS
Co (Outlet concentration) g/m3
w
g/m2
oa
4
10
20
40
60
80
100
120
130
140
Test 65
N/cm3
g/m3
1.24
1.5xl05
3.5x10"
1.2x10"
4.2xl03
2.3xl03
2xl03
1.6xl03
8xl02
6xl02 b
0.34
0.08
0,027
0.0092
0.005
0.0045
0.0036
0.0018
0.0013
Test 68
N/cm3
g/m3
6.36
3x10 5
8x10"
6xl03
1x10 3
9xl02
8xl02
0.64
0.18
0.013
0.0022
0.0020
0.0018
Test 69
N/m3
g/m3
6.25
2.5xl05
7x10"
4xl03
1x10 3
9xl02
9xl02
8xl02 b
0.55
0.15
0.009
0.0022
0.0020
0.0020
Test 70
N/m3
g/m3
5.18
2.7xl05
8x10"
2.5x10"
4xl03
8xl02
6.5xl02
5xl02
4xl02
3.5xl02 b
0.60
0.18
0.055
0.009
0.0018
0.0015
0.0011
0.0009
Test 96
N/m3
g/m3
5.37
1.5xl05
8x10"
4.5x10"
1.4x10"
1.2x10"
1x10"
9xl03
1.4x10"
2.5xl02 b
0.34
0.18
0.10
0.031
0.027
0.022
0.020
0.031
Test 66
N/m3
g/m3
0.63
l.SxlO5
3.2x10"
2x10"
1.1x10"
3x10"
3x10"
3x10"
6x10 2 b
0.34
0.07
0.044
0.025
0.065
Test 72
N/m3-
g/m2
6.73
2.6xl05
2x10"
9x10"
7xl03
4.7xl03
5x10 3
0.56
0.045
0.020
0.015
0.010
0.011
0.018
Test 97
N/m3
g/m3
4.60
l.SxlO5
7.5x10"
7x10"
8.2x10"
1x10 5
9.5x10"
0.34
0.17
0.15
0.18
0.22
0.21
alndicated concentrations are inlet values. True outlet concentration at time zero should be less than inlet value.
Apparent lower detection limit for CNC during measurement period.
-------
employed when fabric filters have been selected to remove low concentra-
tions of highly toxic particles from the atmosphere. It was believed,
however, that considerable insight might be gained as to the ultimate
performance of many woven fabric filters if their atmospheric dust collec-
tion characteristics could be observed over the short term.
The rational behind this testing procedure is that many fabrics which
possess essentially the same pore structure will display the same clean
permeability characteristics even when there are differing amounts of
loose fiber extending into the pore zone. Although the loose fiber sub-
strate obstructing a pore may make a negligible contribution to (clean)
resistance to air flow, the subsequent accumulation of particles upon it
will change this picture radically. Interlaced fibers with dust accumula-
tion now effectively subdivided a single pore into several smaller areas
as well as causing an appreciable reduction in pore cross section. The
net result is that a significant increase in filter resistance is expected
within a short time when filtering industrial aerosols whose concentrations
are typically 10 to 10 times greater than ambient dust concentrations.
In contrast, the absence of a fiber substrate within a pore limits early
particle removal to the inlet and wall surfaces such that extended time
periods are required before appreciable blocking and resistance increases
can take place.
30
Although there are several choices of test aerosol generators and materials,
ambient dust affords the major advantage of availability at no cost.
Therefore, various fabric test panels were mounted in the bench scale ap-
paratus, Figure 5, so that alternate measurements of particle concentra-
tions as determined by condensation nuclei and B&L optical counters could
be performed immediately upstream and downstream of the filter. Minimal
length sampling lines ran to a glass switching valve so that upstream or
downstream samples could be directed to the sensing areas of the CNC and
B&L units. Approximately 2.5 minute intervals were allowed between up-
stream and downstream to allow for flush out and equilibration.
202
-------
Figure 89 shows inlet and outlet concentrations for the two woven glass
fabrics (Sunbury and Nucla types) and a sateen weave cotton. Whereas the
fill fibers alone produce the discrete fiber phase of the glass filters
the cotton yarns are spun from staple fibers such that there are many more
free fibers as evidenced by the napped appearance. Over the brief testing
periods, < 50 minutes, it is unlikely that sufficient dust is deposited
upon fibers to alter their base collection efficiency. Therefore, the
temporal changes merely reflect normal variations in ambient dust levels.
Data summaries in Table 22 indicate that the cotton fabric is the more
efficient fine particle collector. Hence, one ultimately expects that
better overall performance will be afforded by the cotton fabric insofar
as efficiency and effluent characteristics are concerned. Many prior mea-
surements confirm the above observation. ' ' ' '
The failure of B&L and CNC measurements to display a constant proportion-
ality is due to significant variations within the coarse cost particle
fraction of the ambient dust. Thus, the nuclei concentrations are rela-
tively stable because only the coarse particle concentrations have
increased.
The data presented here are too limited to allow prediction of the probable
residual drag levels for a fabric impregnated with a specific dust when
its clean (unused) permeability is known. However, the measuring technique
is so simple that it is believed that examination of several fabrics whose
structure and fiber array were known could develop this approach into a
useful quantitative tool.
Table 22. ATMOSPHERIC DUST PENETRATION WITH WOVEN GLASS
AND COTTON FABRICS
Instrument
CNC
B&L
Average penetration
Sunbury fabric
0.64
0.45
Nucla fabric
0.73
0.76
Sateen weave cotton
0.40
0.38
203
-------
O 4
K>
E 3
-v
CO
lai
1C
/9t//V /(?(?4 (SUNBURY)
B-f-L > 1.0
\.
10 20 30 40 50
RUN IOOB (NUCLA)
B +L >I.Gpm
0 10 20 30
RUN IOO C (COTTON)
B +L >/fta
0 10 20 30 40
o
24
x
M
h x
RUN 100 A (SUNBURY)
CNC
10 20 30 40 50
RUN IOOB (NUCLA)
CNC
0 10 20 30
FILTRATION TIME, minutes
\
n
(COTTON)
10 20 30 40
Figure 89.
Room air filtration with clean (unused) woven fabrics at 0.61 m/min face velocity
inlet (x) and outlet (o) concentrations '
-------
SECTION VIII
PILOT PLANT TESTS
INTRODUCTION
Although the several bench scale tests described previously in this report
have played a major role in providing a data base for model development,
there were some areas where extrapolations from bench to full scale systems
entailed considerable risk. It was pointed out, for example, that flat,
unsupported test panels would experience more distortion in pore dimensions
than a cylindrical bag because of the warping introduced in the former case.
Additionally, bench scale tests afforded no acceptable means to simulate
the collapse and reverse flow (or mechanical shaking) operations normally
used to clean a woven fabric filter. However, the final state of a cleaned
fabric panel was very accurately simulated by bench scale flexing and/or
tapping such that the modeling concepts deriving from those measurements
could be directly extrapolated to full scale units.
Field or pilot scale measurements, therefore, furnish the only practical
means to properly relate the cleaning process and its associated energy
input to the resultant cleaning. Here by resultant cleaning is meant how
much dust is removed and what fraction of the fabric surface is exposed
after application of the cleaning action.
SUMMARY OF TESTING PROCEDURES
Several tests were carried out with the pilot filter system described in
Section IV, Figure 11. The filter bag was sewn from Sunbury type fabric
205
-------
into a 10 ft x 4 in. tube with five, equally-spaced internal rings to pre-
vent complete bag collapse. Bag tension was adjusted by a turnbuckle built
into the hangar arm attached to the cap section closing off the top of the
bag. Filtration velocity 0.61 n/m (2 ft/min) and inlet dust concentration
o 3
7.16 g/m (3.13 grains/ft ) were maintained constant throughout the test
series unless otherwise specified. Bag cleaning was accomplished by divert-
ing the air flow from the bag through a by-pass loop followed by the use
of reverse flow air at 0.48 m/min for a period of 1 minute. Dust dislodged
during cleaning was collected in a special, readily-removed hopper for
transfer and weighing. Rigorous bag cleaning was accomplished by hand
shaking for those tests where it was desired that filtration begin with
nothing but the "limiting," uniformly-distributed residual loading, W .
K
The latter value was determined by separate weighings of the new and
cleaned bags on a triple beam balance. Particle sizing measurements were
performed with the same instrumentation used for bench tests; i.e.,
Andersen impactor, condensation nuclei counter (CNC) and optical counter
(B&L). Inlet and outlet mass concentrations were determined by a combi-
nation of Method 5 type filtration, dust dislodgement and material balance.
Special interior lighting for the bag (8 ft fluorescent tube) was installed
so that dust removal patterns could be documented photographically.
GENERAL COMMENTS
The pilot test results are presented in several tables in which the pilot
(P) plant measurements have been grouped, whenever possible, according to
the specific purpose of each test series. Test P-l-1, Table 23, was a
shake down operation after which it was ascertained that the bag was not
properly sealed. However, an important observation during this test was
that vibration of the baghouse structure by persons working on the elevated
platform while making measurements led to erratic and excessive dust dis-
lodgement from the bag. This problem was eliminated by relocation of
instrumentation and sampling locations. This test demonstrated clearly
that vibrations or shocks induced by heavy equipment operation, damper
206
-------
NJ
O
Table 23. EFFLUENT CONCENTRATION FROM NEW (UNUSED) AND PARTIALLY LOADED SUNBURY TYPE
FABRIC WITH GCA FLY ASH AND ATMOSPHERIC DUST
Run No.
P-l-la
P-6-2
P-7-la
P-9-13
Fabric
loading at
beginning of
run,
g/m^
0.0
743
491
0.0
Fabric
loading
before
cleaning,
g/m2
1253
743
843
461
Fabric
loading
after
cleaning,
g/m2
113
491
147
461
Dust
removed
by
cleaning,
g/m2
929
205
567
0.0
Percent
dust
removed
by
cleaning
91. Ob
33.9
82.6
0.0
Outlet
concen-
tration3
g/m3
0.0559
0.00063°
0.0319
0.0903
Average
penetration
percent
0.87
-
0.45
1.26
Inlet concentration = 7.16 g/m , fly ash.
Dust removed by hand-shaking.
C* Q
Inlet concentration " 0.00005 g/m , (atmosphere dust)
-------
closings or fan pulsations can contribute to dust removal in the field.
However, from the point of view of field validation or laboratory measure-
ments, it would be difficult to quantitate their role in the cleaning
process.
A second factor that could lead to a significant difference between field
and laboratory performance was the presence of pinholes and the evidence
of fabric distortion or stretching apparently arising from bag sewing
operations. Tests performed with new (unused) fabrics, Table 23, and
used bags, Tables 24 and 25, showed generally higher emission levels
(2 to 3 times) than noted for the bench scale test panels. However,
normal variability in fabric properties aside from sewing factors may also
account for differences in performance.
DUST REMOVAL VERSUS FABRIC LOADING
The tests summarized in Tables 23, 24, 25 and 26 and Figure 90, indicate
the amount of dust dislodged from a filter as the result of a single
cleaning by collapse and reverse flow. Bags were tensioned at either 50
or 15 Ibs and the reverse flow velocity and duration were identical at
0.49 ft/min and 1 minute, respectively. The amount of dust removed was
observed to depend upon the prior dust holding of the fabric. This be-
havior appears to confirm the hypothesis that the dust separating force
must increase as the deposit areal density increases while the opposing
interfacial adhesive force depends upon the specific dust/fabric combi-
nation but not the areal density of the dust layer. The above factors
are treated in detail in Section IX.
In appraising the dust removal relationship shown in Figure 90, it should
be noted that several factors may influence dust removal. Generally
speaking, one set of variables determines the adhesive forces which are
controlled mainly by the specific dust and fabric properties and the re-
lated environmental effects of temperature, humidity and electrical charge.
208
-------
Ni
O
Table 24. RELATIONSHIP BETWEEN DUST REMOVAL AND PREVIOUS FABRIC LOADING, GCA FLY ASH
FILTRATION WITH 10 ft x 4 in. WOVEN GLASS BAG (SUNBURY TYPE) AT 0.61 m/min
FACE VELOCITY
Run No.
P-2-1
7-2-2
P-2-3
P-2-4
P-4-lb
P-4-2
P-4-3
P-4-4
P-4-5
Fabric
loading at
beginning of
run,
g/m2
113
327
387
498
85.9
274
382
476
549
Fabric
loading
before
cleaning,
g/m2
937
422
545
723
696
429
536
631
704
Fabric
loading
after
cleaning,
g/m2
345
387
498
598
274
382
476
549
550
Dust
removed
by
cleaning,3
g/m2
592
35
48
126
422
47.3
49.1
67.0
125
Percent
dust
removed
by
cleaning
63.3
8.4
8.7
17.4
60.6
11.0
11.2
13.0
21.8
Outlet
concen-
tration
g/m3
0.0501
0.0311
0.0230
0.0296
0.0272
0.0281
0.0235
0.0286
0.0300
Average
penetration
percent
0.70
0.43
0.32
0.41
0.38
0.39
0.33
0.40
0.42
Cleaning by bag collapse and reverse flow.
P-4 series also used to demonstrate appearance of cleaned bag surface by means of
light source inside the bag.
-------
Table 25. REPETITIVE CLEANING AND FILTRATION CYCLES WITH GCA FLY ASH AND WOVEN GLASS
(SUNBURY TYPE) FABRIC AT 0.61 ft/rain FACE VELOCITY AND 50 Ibs TENSION
N)
(—»
O
Run No.
P-3-1
P-3-2
P-3-3
P-3-4
P-3-5
P-3-6
P-3-7
P-3-8
P-3-9
P-3-10
P-3-11
P-3-12
P-3-13
P-3-14
P-3-15
P-3-16
P-3-17
P-3-18
P-3-19
Fabric
loading at
beginning of
run,
g/m^
576
513
552
563
602
630
633
623
615
623
657
668
663
667
588
609
625
613
629
Fabric
loading
before
cleaning,
g/m2
725
662
701
713
751
780
782
772
764
772
806
817
812
817
737
758
775
763
778
Fabric
loading
after
cleaning ,
g/m2
513
552
563
602
630
633
623
615
623
657
668
663
668
616
609
625
613
629
657
Dust
removed
by
cleaning,
g/m2
211
111
138
111
121
147
159
157
142
115
138
154
145
201
128
133
161
134
121
Percent
dust
removed
cleaning
29.2
16.7
19.6
15.6
16.0
18.9
20.4
20.3
18.5
14.9
17.2
18.9
17.8
24.6
17.4
17.6
20.8
17.6
15.6
Outlet
concen-
tration,
g/m3
0.0353
0.0259
0.0144
0.0137
0.0279
0.0236
0.0227
0.0227
0.0252
0.0243
0.0213
0.0215
0.0206
0.0215
0.0190
0.0162
0.0144
0.0137
0.0121
Penetration
during
run,
percent
0.49
0.36
0.20
0.19
0.39
0.33
0.31
0.32
0.35
0.34
0.29
0.30
0.28
0.30
0.27
0.23
0.20
0.19
0.17
-------
Table 26. EFFECT OF REDUCED BAG TENSION, 15 Ibs, ON DUST REMOVAL AND PENETRATION GCA FLY
ASH WITH 10 ft x 4 in. BAG, SUNBURY FABRIC, AT 0.61 m/min. FACE VELOCITY
Run No.
P-5-1
P-5-2
P-5-3
P-5-4
P-5-5
P-5-6
Fabric
loading at
beginning of
run,
g/m2
550
501
537
554
553
559
Fabric
loading
before
cleaning,
g/m2
705
656
692
709
708
714
Fabric
loading
after
cleaning,
g/m2
501
537
554
553
559
579
Dust
removed
by
cleaning,
g/m2
167
97.0
112
127
122
110
Percent
dust
removed
by
cleaning
29.0
18.2
19.9
21.9
21.1
18.9
Outlet
concen-
tration*
g/m3
0.0305
0.0234
0.0158
0.0224
0.0225
0.0190
Average
penetraion
percent
0.43
0.33
0.22
0.31
0.31
0.26
-------
o
40
O
0)
o
-------
The latter factors were considered to be constant during the laboratory
studies. Except for the effect of gas temperature on viscosity and local
gas velocities, the above items were not observed to cause any distin-
guishable performance differences between laboratory measurements and
field filtration tests with coal-fired boilers.
The second set of variables relates to the description and quantitation
of dust dislodgement effects. Prior mechanical shaking studies10 and the
sequence of tests described in Table 27 and Figure 91 show that repeated
cleaning action removes additional dust although in rapidly diminishing
quantities. One infers, therefore, that use of a single collapse and
reverse flow cleaning cycle not only dislodges a specified quantity of
dust but also alters the distribution of interface adhesive forces for
the dust remaining on the fabric. Hence a second application of the same
cleaning process will dislodge an additional increment of dust and so forth
until further removal becomes negligible.
It was assumed that all dust removed from the fabric was attached with
an adhesive force less than the applied dislodging force, the latter de-
fined as the product of the fabric loading and the local gravitational
acceleration. Therefore, if the curve shown in Figure 90 represents the
results of a single bag cleaning at each of the indicated load levels, it
is expected that a smaller slope would be displayed if multiple cleanings
were performed at each fabric loading. The rationale for this statement is
that at very high fabric loadings (1200 to 1500 g/m ) ' as much as 90
percent of the dust cake can dislodge. Thus, even with repeated cleanings,
the maximum increase in percent dust removal could not exceed 10 percent.
On the other hand, at lower fabric loadings a very significant increase
in dust removal is possible by repetitive cleanings.
Percent dust removal was graphed on logarithmic probability paper because
the data presented in Figure 90 also describe the distribution of adhesive
forces over the interfacial region of the fabric. The estimated curve
213
-------
Table 27. EFFECT OF SEVERAL SUCCESSIVE CLEANINGS BY BAG COLLAPSE AND REVERSE FLOW,
GCA FLY ASH WITH WOVEN GLASS FABRIC (SUNBURY TYPE)
Number of
cleanings3
1
2
3
4
5
6
7
8
9
10
11
Fabric
loading
before
cleaning
WT
g/m2
778
657
609
584
564
554
546
540
533
527
522
Cumulative
dust
removed by
cleaning
WT* - WR
g/m2
121
169
194
214
222
232
238
245
251
256
259
Fabric
loading
after
cleaning
%'
g/m2
657
609
584
564
556
546
540
533
527
522
519
Cumulative
dust
removed
by
cleaning
percent
15.6
21.7
24.9
27.5
28.5
29.8
30.6
31.5
32.2
32.9
33.2
Cleaned area
fraction'3
ac
0.167
0.232
0.266
0.299
0.309
0.319
0.328
0.337
0.345
0.352
0.356
Uncleaned
area
fraction0
au
0.833
0.768
0.734
0.701
0.691
0.681
0.672
0.663
0.655
0.648
0.644
These tests represent a continuation of the cleaning process with the first cleaning
corresponding to Run P-3-19, Table 25.
ac = ! ~ au
cau = WR - WR/WJ, - Wj^, where W^ is the residual uniformly distributed loading on
the cleaned fabric surface and WTi is the cloth loading before the first cleaning.
-------
N>
I—*
Ln
16
1 I4h
o
w.
J '2h
i l0^-
UJ
(£
8
4
2
0
—i r
CURVE
1 1—
DESCRIPTION
O — Ci) FABRIC LOADING BEFORE CLEANING
x—© percent DUST REMOVAL
NOTE= CLEANING INITIATED WITH TESTP-3-19,
TABLE
j_
_L
JL
X
JL.
-X
_i_
E
800 *
o
z
700 z
UJ
600 °
500
400
UJ
oc.
UJ
CD
300 §
200 o
a:
ao
100 if
24 6 8 10 12
NUMBER OF COLLAPSES AND REVERSE FLOW CLEANINGS
Figure 91. Dust removal characteristics for repetitive cleaning cycles, Sunbury type
fabric with GCA fly ash
-------
derives from a subsequent replotting of the data points on a log-log scale
to simplify the curve fitting mechanics, see Figure 119. It should also be
noted that the point designated as P-3, which represents an average of 19
tests, exerts considerable influence on the curve path. According to many
prior measurements, the variability of adhesive forces about some central
tendency is statistically distributed whether the system be particle to
1 28
particle, particle to fiber or dust layer to fabric. ' A logarithmic prob-
ability distribution was chosen in the present case only because the particle
size parameters were best defined by the above distribution. According to
Figure 91, it appears that the amount of dust removed by a single cleaning
2
for an initial fabric loading of 778 g/m is roughly 67 percent of that
which can be removed by several repetitive cleanings. Because of the ex-
tended times associated with repeated collapses (and the loss of working
fabric surface), any advantage of successive collapses is probably lost
after a few cleanings.
Extrapolation of the removal versus fabric loading curve of Figure 90
suggests that practically all of the fabric dust loading should be dis-
lodged with a single collapse when the areal density is allowed to reach
2
the 1200 to 1500 g/m . It should be noted, however, that even with a
single collapse per cleaning interval, the surface of the fabric from which
no dust has previously been dislodged has undergone several flexures once
steady-state operating conditions have been attained. This condition is
reflected for most data points shown in Figure 90. The exception is the
single point for one collapse only of a heavily laden fabric surface. It
is expected that repetitive flexings would have led to increased dust
removal. On the other hand, the form of the curve indicates that there
is probably a lower level for areal density at which even repetitive col-
lapse and reverse flow cleanings will accomplish little cleaning. If it
is assumed that the adhesive force is always less than the dislodging
force, one can infer that the range of adhesive bonding for the GCA fly
ash/Sunbury fabric system should range from roughly 50 to 150 dynes/cm.
The above force values are associated with fabric loadings of 510 and
2
1530 g/m , respectively, in conjunction with a normal gravity field.
216
-------
Although the preceding data analyses are considered to be correct from
the qualitative viewpoint, it is recognized that more testing is needed
to strengthen their quantitative value. With respect to a coal fly ash-
woven glass fabric system, however, these data have provided very useful
guidelines for the modeling discussed in Section IX.
DUST REMOVAL WITH SUCCESSIVE FILTRATION AND CLEANING CYCLES
Successive filtration and cleaning tests, Table 25, were carried out at
representative field operating conditions to determine how many cycles
would be required before achieving steady state conditions with a single
bag. Reference to Curves 1 and 2, Figure 92, indicates that after 5 to 6
operational cycles, the dust deposition and removal rates become equal.
Dust penetration values for essentially constant inlet concentration show
a consistent downward trend, however, suggesting that progressively more
dust is accumulating within the filter pore structure. Field measurements
at the Sunbury Plant, Section VI, Figure 42, indicated that 10 to 12 days
of operation were required before a relatively constant emission rate was
reached with a multicompartment system.
DUST REMOVAL AND BAG TENSION
A limited test sequence, Table 26, indicated that reducing bag tension
from 50 to 15 Ibs had little effect on dust removal and penetration charac-
teristics. The above tension range encompasses the values commonly used
in the field with glass bags used for fly ash filtration. Prior measure-
ments showing the effect of bag tensioning on clean cloth permeability,
Section V, Figure 33, also indicated that there was little change in fabric
permeability over the 15 to 50 Ib tension range. The Table 27 tests also
showed that dust removal appeared to level off after five to six succes-
sive cleaning cycles. This finding seems to corroborate the test results
of Figure 92 which show that five to six repetitive cleanings of the fab-
ric between loading intervals is sufficient to reach a practical limiting
level.
217
-------
CURVE DESCRIPTION
Qo FABRIC LOADING PRIOR TO CLEANING
(2)x FABRIC LOADING AFTER CLEANING
§A FRACTION DUST REMOVED BY CLEANING
EJ percent DUST PENETRATION
800
700
600
CO
E 500
«^
a>
z
o
o
an
03
400
300
200
00
_L
0.50 o
0.40
LU
CO
uu UJ
S°-
0.30 s
u *•
o: c
0.20
3
Q
Z
o
tr
u.
1.10
0 5 10 15
NUMBER of SUCCESSIVE FILTRATION and CLEANING CYCLES
Figure 92. Performance of Sunbury fabric with GCA fly ash with repe-
titive filtration and cleaning cycles
218
-------
RESISTANCE VERSUS FABRIC LOADING
The resistance versus fabric loading curves for the tests summarized in
Table 25 are shown in Figure 93. Approximate steady-state conditions
appear to have been reached after 8 to 9 successive cleaning and filtra-
tion intervals. The discontinuities indicated in Curves 1 through 9 re-
sulted from a flow regulation problem that was subsequently corrected.
It is emphasized that the slopes of these curves do not enable computation
of K2 values because the filtering intervals were too brief to allow for
regeneration of a uniform thickness dust cake.
On the other hand, the extended filtering times used for the tests
described in Table 23 and Figure 94 show that the resistance versus fabric
loading curves eventually approach the slope obtained when the dust de-
posit is uniformly distributed. The estimated K value for the linear
section of the curve is 1.35 N min/g min, which is fair agreement with
K- values determined previously for GCA fly ash.
DUST PENETRATION MEASUREMENTS
Constant Velocity Tests
Figure 95 indicates that short-term changes in filter emissions are de-
fined by condensation nuclei concentrations are quite similar to correspond-
ing bench tests performed with test panels. When the average nuclei
concentrations were computed for each of the 19 tests listed in Table 25,
their equivalent mass concentrations derived from the calibration curve
of Figure 86, Section VII, were in close agreement with values determined
by concurrent gravimetric sampling. One can infer, therefore, that the
test aerosol properties for the pilot system were very similar to those
of the bench tests.
219
-------
2
x
o
I
c
u
m
£
x>?xx
X
X
X
X
X
X
X
X
xxx
X
X
XX
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
*
X
4567
FILTRATION CYCLE NUMBER
7 -
X
- x
X
X
X
— «;
(/> S
UJ
IE
X'
X/
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X*
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
XX
X
X
X
X
xx
X
X
X
X
X
X
x.
X
NOTE I.) LOADING INCREMENT PER FILTRATION CYCLE (AW) 150
2.) DAMPER REGULATION PROBLEMS PRIOR TO CYCLE 9
J_
10
12 13 14 15 |6
FILTRATION CYCLE NUMBER
17
18
Figure 93. Successive filtration and cleaning cycles for Sunbury
fabric with GCA fly ash based on data of Table 25
220
-------
8 -
x 7 -
6 -
CM
O
I
iu
OC
^ 3
IT
CD
TEST
O P-2-2
A P-2-3
D P-2-4
NOTE
FACE VELOCITY 0.61 m/min
INLET CONCENTRATION 6.4 g/m3
CLEANING BY COLLAPSE AND
REVERSE FLOW
'D
,Q
'
-m'
0'
O
0
100
200
300 400 500
FABRIC LOADING, g/m2
600
700
800
Figure 94. Single bag (10 ft x 4 in.) filtration of GCA fly ash with Sunbury fabric -
three cleaning cycles with variations in residual loading
-------
io3
2
"> 10
o
X
"k
.0°
n
o
A ° O 0 o
x A
A
A
A
A
A " A " A
NOTE: RESIDUAL BAG LOADINGS (WR)
RANGE FROM 153 TO 173 g/m2
1U
o
2
O
O
x x
b
o
10
X X
x x
X X
RUN FACE VELOCITY m/min.
x P-6-4 0.61
A P-6-5 1.23
O P-6-7 1.98
0 P-6-8 2.67
• P-6-9 4.26
J I _1 1 !_..._
40 80 120 160
INCREASE IN FABRIC LOADING, g/m2
Figure 95. Effect of face velocity on outlet concentration, GCA fly
ash with 10 ft x 4 in. woven glass bag (Sunbury type
fabric)
222
-------
PENETRATION VERSUS FACE VELOCITY
Previous measurements with bench scale equipment, Section VII, Figure 88,
indicated that filtration velocity had a very significant effect upon
effluent concentrations. In order to reduce the chance of serious scaling
errors, a second series of tests was performed over the velocity range
0.61 to 4.25 m/min, each test starting with essentially the same fabric
2
loading (150 to 175 g/m ), Table 28. Manual shaking was used to remove
the dust. These tests confirmed the adverse effect of increased face
velocity on effluent concentration. Figure 95 shows plots of effluent
concentration versus fabric loading for several face velocities.
Figure 96 shows the relationships between average and final outlet con-
centration and face velocity for the bag tests described in Figure 95 and
the panel tests discussed previously in Section VII and Figure 88. Final
concentration refers to the essentially constant outlet concentration that
follows the rapid decay phase. According to the curves of Figure 96, bag
and test panel average concentrations appear to increase as the 2.22 power
of the face velocity whereas the final or limiting concentrations increase
as the cube of the velocity. The indicated exponential relationship applies
fairly well for face velocities less than 2.5 m/min. At higher velocities
a marked decrease in slope is observed. Again, the main impression gained
from these data is that high air-to-cloth ratios even if acceptable from
the point of view of operating resistance, may lead to excessively high
dust emissions.
REAR FACE SLOUGH-OFF
Only one pilot test was run to establish the approximate magnitude of
particulate emissions when room air alone was passed through a previously
loaded bag, Run P-6-2, Table 23. The indicated outlet concentration was
0.63 mg/m3, about six times greater than the estimated inlet atmospheric
dust concentration. The spurce of the emission was the slough-off or
223
-------
N>
Table 28. EFFECT OF FACE VELOCITY ON OUTLET CONCENTRATION, GCA FLY ASH
10 ft x 4 in. WOVEN GLASS BAG, SUNBURY FABRIC
Run No.
P-8-1
P-8-2
P-8-4
P-8-5
P-8-6
Face
velocity,
m/min
0.62
1.23
1.98
2.67
4.26 .
Fabric
loading at
beginning of
run,a
g/m2
155.9
153.5
173.3
165.1
.131.4
Fabric
loading
before
cleaning,
g/m2
302.5
360.5
368.5
403.2
507.6
Fabric
loading
after
cleaning ,
gM2
153.5
158.6
165.1
131.4
-
Outlet
concen-
tration^
g/m2
0.0355
0.1615
0.7128
1.0868
1.2750
Average
penetration
percent
0.50
2.25
9.95
15.15
17.80
Bag hand shaken to attain indicated residual loading.
Inlet loading constant at 7.16 g/m^.
-------
10s
M
o
10
o
t-
x
z
»-
UJ
o
I0
10-
I I I
CURVE DESCRIPTION
0 t) BAG, AVERAGE CONCENTRATION
O BAG, FINAL CONCENTRATION
© ^ TEST PANEL, AVERAGE CONCENTRATION.
@ A TEST PANEL, FINAL CONCENTRATION
i . , i . . i I
0.2 0.5 1.0 2.0
FACE VELOCITY, m/min
10
Figure 96. Relationship between final and average outlet concentra-
tion and face velocity for 10 ft x 4 in. bag and test
panel with GCA fly ash and Sunbury type fabric
225
-------
detachment of particle agglomerates from the rear of the pore regions
caused by air reentrainment and perhaps aided by random mechanical vibra-
tions in the system. The particle size distribution for the above emission
source is shown in Figure 97. It is emphasized that such dust releases do
not mean that pinholes have developed in the fabric, although they might
ultimately lead to pinhole formation.
226
-------
10,00
5.0
or
uu
LU
u
_l
O
o
z
o
o
oc.
2.0
1.0
0.5
O.2
NOTE •• AMBIENT OUST
CINLET ~
MMD = ~
erfl = ~l.l
10 30 50 70 90
PERCENT MASS < STATED SIZE
95 96
Figure 97. Effluent particle size parameters from GCA fly ash loaded
Sunbury fabric when filtering atmospheric dust
227
-------
SECTION IX
PREDICTION OF FABRIC FILTER DRAG
A new model for predicting the change in fabric drag (S) in terms of
the fabric areal dust loading (W) is described in this section. The
model is based upon a concept discussed by Billings and Wilder in which
filtration is considered to take place through an assemblage of pores
or channels bounded by the warp and fill yarns, rather than through an
assemblage of isolated fibers such as found in felted media or high
porosity bulk fiber beds. It is further assumed that several discrete
fibers from staple or bulked yarns protrude into the interyarn region
to form a substrate for dust cake growth. Observation of clean and used
woven glass fabrics under low power, 4x to lOx, magnification appears to
substantiate the above assumption. Dust collection is assumed to result
from two processes; first, the rapid blocking of the interspersed bulk
fibers by an essentially superficial dust layer and secondly, the
development of a dust layer or cake upon this substrate that results in
particle removal by direct sieving. In the ensuing dust collection pro-
cess, the characteristic rate of resistance change with dust loading for
the glass fabric appears to conform to the pattern suggested by fabric
geometry and classical fluid dynamics.
Empirical equations have been developed that simplify calculating procedures
although rational physical processes that explain observed filter behavior
can be postulated in most cases. The above statement applies to the mathe-
matical model developed to describe the typical drag versus fabric loading
relationship noted for the fly ash/woven glass fabric systems.
228
-------
CRITIQUE OF LINEAR DRAG MODEL
The linear model of fabric drag (S = S£ + K2 W) discussed in the pre-
ceding review section has the advantage of simplicity. The drag is
modeled as increasing from the artificial value S£ with a constant slope
K2- The extension of this line is superimposed upon the linear section
of the curve at the latter's juncture with the curvilinear section. The
disadvantage of the linear model is that it becomes increasingly incorrect
as W decreases from 175 to 0 g/m2 (0.03 to 0 lb/ft2. The consequence of
this error is that the linear model is most incorrect when the flow and
the emissions through the bag are greatest. The extent of the error,
however, depends upon the difference between the S and the S,, values, the
E R
number of compartments operated in parallel and the amount of dust removal
during cleaning.
The development of a nonlinear model that provides a good approximation
of the actual performance curve and a means by which S can be evaluated
when a linear approximation suffices are presented in the next section.
DERIVATION OF NONLINEAR (PORE) MODEL
The curve shown in Figure 98 shows the typical form assumed by a drag ver-
sus fabric loading curve for a fly ash/woven glass fabric filter after
several repetitive cleaning cycles. It is very important to note that
complete cleaning has taken place such that the residual dust holding,
W , is only that retained within the loose fiber structure obstructing
R
the pores. Careful observation of the region from which a dust layer or
element of the dust cake has been dislodged shows that separation occurs
principally at the interfacing between the dust cake and the fabric. Anal-
ysis of adhesive and cohesive forces suggests that dust loss through sur-
face spallation should be minimal because the cohesive forces within the
dust cake exceed the adhesive bonds between the dust and fabric surface.
229
-------
si
CO
O
/dS_\
/KRFVdW/wR
Figure 98. Typical drag versus fabric loading curve for a
uniformly distributed dust holding
-------
Therefore, if a filter with a uniformly distributed dust deposit, W, is
cleaned at some pre-selected intensity, the cleaned filter will display
two distinct surfaces; the first, the unchanged or uncleaned region
with its original surface loading, W, and the second, the region from
which the surface layer has been detached that now is characterized
by the residual loading, W . Several measurements and observations
during this study have shown that with more intense cleaning, the total
cleaned area is increased but the surface loading upon the cleaned
regions is uniformly distributed at a near constant areal density irre-
spective of cleaning intensity. Residual fly ash loadings for Sunbury
and Nucla type glass fabrics generally fell within the loading range
50 to 100 gram/m2. Additionally, limited tests with other dust/fabric
combinations indicated that W^. values generally fell within the same
K
50 to 100 gram/m2 range although the amount of dust dislodged by a fixed
energy input was strongly dependent upon the individual dust and fabric
properties.
The cleaning process as it bears upon filter drag and dust penetration
characteristics will be discussed in more detail in succeeding sections
of this report. The key factor to be noted at this time is that a cleaned
element of the filter surface is one from which the surface dust cake is
completely detached. The resulting surface with its residual loading,
W , presents the same pore array present in the clean (unused) filter
K
except that dust particles (essentially irreversibly retained within
the loose fiber substrate blocking the pores) lead to an increase in the
residual filter drag.
The residual drag, SR, for a uniformly cleaned filter is associated with
the characteristic residual fabric loading, WR. Over the surface load-
ing interval W -W,,, the rate of change of drag with fabric loading,
I R
dS/dW, gradually decreases from its initial value of ^ at WR,until it
assumes a final constant rate, KZ, for all surface loadings in excess
of Wr The term W indicates the fabric loading at the point where the
curve assumes a linear path. The effective drag, SE, is shown as the
231
-------
lower linear extrapolation of the drag curve. Because the clean drag of
the filter, S , is seen only once in any practical filter application, it
is of interest only to the extent that it may aid in predicting the be-
havior of the fabric with new applications.
Several physical mechanisms were considered in an attempt to provide a
rational physical explanation for the path of the drag/fabric loading
curves studied in the course of this program. Although it appeared rea-
sonable to consider the curve path between WD and W as the result of a
K 1
gradual reduction in unobstructed pores, several measurements and tests
described earlier in this report indicate that nearly all pores ~ 99.99
percent or greater must be completely blocked within a very brief period
of filter use. Otherwise, the extremely high permeability of open pores
would cause most of the air to vent through them. Furthermore, a com-
pletely sequential pore blocking process over the W to W interval would
K. I
dictate a concave upward curve form rather than the path shown in Figure 98
as discussed in Appendix A. In the case of filtration with heavily napped
cotton fabrics, one may encounter a concave upward resistance versus fabric
loading relationship due to a gradual compression of the more porous dust
layer as the resistance increases.
One comparatively simple explanation for the observed curve shape lies with
the fact that once initial bridging is accomplished (which is greatly en-
hanced by the presence of bulked yarns or staple) the dust layer develops
gradually, first below and finally above the fabric surface. Although the
depth of the dust penetration within the fabric structure is restricted by
the location of the fiber substrate, there still remains the possibility
of an appreciable reduction in pore cross section for the subsurface regions.
Under these conditions, two factors contribute to a rapid increase in
filter resistance when filtration commences. First, if the porosity of
the deposit is assumed to be constant irrespective of deposition site,
the first increments of dust collected below the filter surface will
exhibit a greater depth per unit of mass because their cross sections
232
-------
are reduced. Thus, under laminar flow conditions the resistance per unit
mass will be larger because of the increased depth. Secondly, the flow
cross section is reduced for the initial deposits requiring that the
velocity increase proportionally to maintain continuity of flow. Both
the depth per unit increment of deposit and the velocity through the
deposit decrease as the surface of the fabric is approached. Conversely,
once dust fill reaches the filter surface level the cake depth is directly
proportional to unit mass of deposit and cake velocity is constant (assuming
no porosity changes due to cake compression).
Figure 99 depicts a fabric pore with a low density bridging of discrete
fibers within the gap separating the yarns. The latter structure consti-
tutes the principal supporting substrate for the dust layer. Particle
penetration into the bulked fiber mass is relatively small compared to
the surface deposition. The pore cross section is seen to increase as
the surface of the filter is approached. In the simplified model of the
pore structure in Figure 99, the convergence is treated as a truncated
conical section. This allows the pore diameter between the surface of
the fabric and the bottom (or start) of the dust layer to be defined by
a simple linear equation.
where d . is the cake diameter at its greatest pore depth, d the
mm max
cake diameter at the surface of the pore, W the average surface loading
at the inception of cake filtration, and d the cake diameter at any
average fabric loading W. The development of the above approach results
in the following expression for the change in drag AS over the loading
range ¥ to W ;
R I
233
-------
i
APPROACHING
GAS STREAM
I
W
-P-
DUST LAYER
J7
* .000 ^
FILTER SURFACE
dmox.
|
YARN -
BULKED FIBERS
SKETCH OF FILTER SECTION, DUST
LAYER BUILDUP ON SUPPORTING
BULKED FIBER SUBSTRATE
SIMPLIFIED PORE FORM,
TRUNCATED CONE
Figure 99. Schematic, dust accumulation below surface of fabric
with bulked fiber or staple support
-------
AS
-, w
WR
where k =
d d .
max - mm
K2 / U
3k \ max/
1 J
(d . + kW)
\ mm /
(25)
J W
R
According to microscopic inspection of the fabric, and examination of
Figure 28, it appears that the ratio of d to d . should be in the
max mm
range of 1.5 to 1.7. The above values allow the development of a drag
versus fabric loading curve using Equation (25) that describe our lab-
oratory measurements.
The calculation of S over the range W to W follows the standard relation
AS
W
W
(26)
Unfortunately, Equation 25 is cumbersome and the constants d and d
Ttlt3.X ITlXtl
are difficult to determine. Additionally, the complete filtration range
must be defined by two separate equations, each with its specific limits.
Therefore, a simpler approach was sought to define the curvilinear
relationship shown in Figure 98. To satisfy the mathematical, if not
the physical picture, the model should reflect the following:
S = SR at W = WR
dS/dW = Kn at W = W
R K
and
dS/dW = K2 for W
S = SCW.J.) at = W.J..
235
-------
Such a model would display the correct initial and final slopes, K^ and
K , respectively, while satisfying the experimental values at WR and for
W _> Wj.
The above terms were examined in the derivative form; i.e.,
~ = f(W) K_ + g(W) K9 for W ^W
QW Iv <*- K-
which shows that the following conditions must prevail in the successful
model.
f (W) = 1 and g(W) = 0 at W = WD
R
f (W) = 0 and g(W) = 1 at W ^ W
In order to approximate the physical situation discussed previously, that
is, the effective reduction in specific resistance coefficient from K to
R
K0 over the fabric loading range W to W an exponential decay process
/ K I
was selected. The reason for this approach is that the necessary increases
and decreases in the functions f(W) and g(W') can be accommodated by a
single equation. Here W refers to W-W so that the curve path is traced
R
from its true origin (W_, S,,) .
K K
If f(W"*) and g(W') are arbitrarily defined by the following equations
f(W') = exp (-W/W*)
g(W') = 1 - f(W) = 1-exp (-W/W*)
the differential equation defining the drag versus loading relationship
appears as
~ = KR exp (-W/W*) + K2 [(1-exp (-W/W*)] (27)
236
-------
Upon integration, Equation (27) reduces to the form:
S - S + K W- + (K-K ) W* 1-exp (-W/W*)
(28)
In later sections of this report, approximate methods for estimating such
parameters as K2> SR and SQ are given. Until the state-of-the-art advances
well beyond our present understanding of the several factors defining the
above variables, however, the direct experimental determination of these
parameters is strongly recommended.
The terms S , SE> K^ and K2 are readily determined by the graphical anal-
ysis of fabric loading curves of the type shown in Figure 98. Such curves
can be generated from comparatively simple laboratory or field testing pro-
cedures with the specific dust/fabric combination and air-to-cloth-ratio
of interest.
The term W is a system constant whose value is best derived from the
direct graphical measurements of S , S , K^ and K?.
** =
Alternatively, W* also appears to be closely related to W based upon
examination of data described later in this section; i.e.,
W* - 0.35 W (30)
At the present time, it appears preferable to treat the residual drag -
residual loading coordinates as the starting point for the modeling
process. Aside from the fact that the clean (unused) fabric drag is
encountered but once, there is no existing relationship to determine how
the clean and effective drags are related for specific dust/fabric
systems.
The concave form for the drag/loading curve, Figure 98, has been attributed
to the higher velocities and greater cake depth per unit mass of dust for
237
-------
dust deposits below the fabric surface. If the fabric is very highly
napped, however, there will be less chance for interstitial dust pene-
tration. Hence, as noted for napped cotton sateen filters, the initial
(K ) and final (K ) slopes are nearly the same. In fact, if the napped
medium is at all compressible, the resistance increase with fabric loading
may display a concave upward shape as the porosity of the dust/fabric
mass decreases. The same phenomena are observed when the compaction is
apparently increased by filtration at higher velocities as discussed in
a later section.
It has been emphasized that the filtration model presented here involves
oversimplification of some very complex interactions. The assumption
has been made that the pores or interyarn spaces are identical in shape
and dimensions. The weave characteristics alone, however, indicate that
at least two distinct pore geometries are encountered with a 3/1 twill,
Figures 23 and 28, Section V.
Additionally, a certain lack of uniformity in pore dimensions arises
directly from the weaving process while rough handling and improper
installation can also contribute to an undesirable spread in pore sizes.
It was also assumed that loose loops or free fiber ends extending into
or across a pore cross section presented a fairly uniform substrate.
In practice, however, oversize pores can be found that may or may not be
bridged over during the filtering cycle. Thus, there exists a limiting
pore size beyond which a fabric ceases to be a highly effective filter.
GCA measurements suggest that open pore area must be reduced to the order
of 10 5 times that of the total filter surface before good filtration
can ensue; i.e., effluent concentrations in the 10~3 g/m3 range.
Although 100 percent mult ifilament weaves were not investigated with
respect to coal fly ash filtration, it should be noted that the absence
of bulk fiber fill in the interyarn region will reduce particle collection
significantly unless the interyarn spacing is greatly reduced. Tests
performed with a plain weave plastic screen in which the velocity through
200 urn diameter pores simulated that for the -100 to 150 ym pores in woven
238
-------
glass fabric showed that complete bridging was impossible to attain with-
out a supporting fiber structure. Performance of the plastic screen
described in Figure 36 suggests that pore diameters should be of the
order of 10 to 20 |im to achieve collection comparable to that attained
with the 50 to 150 (im diameter pores for common woven glass fabrics.
Note that the adverse effect of oversize pores can be counterbalanced
by the bulk fibers that constitute the substrate for cake formation.
In the preceding analysis it is assumed that all pores are identical with
respect to cross section, depth and quantity of fiber dispersed within
the pores. Thus, aside from any randomness resulting from the spatial
variability of the inlet dust concentrations, pore bridging and the
development of a dust layer should proceed as "n" parallel filtering opera-
tions where "n" is the effective pore count per unit filter cross section.
Should the degree of dust accumulation increase at any point on the filter,
the concurrent increase in resistance would tend to redistribute the dust
laden gas to areas of less resistance. Thus, minor deviations from pore
dimension uniformity, which typifies a useful woven fabric, would not
seriously hamper the bridging process. However, should there be too
large a range in pore diameters, there exists the probability that com-
plete pore bridging or blockage might never be attained. Hence, unsatis-
factory performance may be encountered in the field for the above reason
due to damage or improper fabric selection.
VERIFICATION OF NONLINEAR DRAG MODEL
The experimental performance curves for five different fabric filters
were selected to evaluate the curve fitting capability of the nonlinear
model. Fabric descriptions and test data sources are listed in Table 29.
It was assumed that the fabric dust loadings were uniformly distributed
upon the filters and that the filters had been especially cleaned down
to their minimum W, values, -50 g/m2. Although subsequent investigations
K
239
-------
Table 29. PHYSICAL PROPERTIES OF FABRICS INVOLVED IN MODEL TESTING
to
JN
o
Test
number
1
2
3
4
5
Type
of
fabric
Glass fiber
Polypropylene
Dacron
Cotton
Polyacrylester
Weight,3
oz/yd2
9.06
4.30
10.0
?•
10.0
9.8
Weave and yarn count,
yarns per inch
3/1 crowfoot, filament
55 x 58
3x1 twill, filament
74 x 33
Plain, staple
30 x 28
Unnapped sateen
95 x 58
2x2 twill, spun
39 x 35
Frasier
permeability,
ft /min j.,
@ 0.5 in water
7.9
15.0
55.0
13.0
60.0
Reference
Spaite and
Walsh13
„ , 15
Durham
Dennis and
Wilder16
Dennis and
Wilder16
Durham
al oz/yd2 =33.9 g/m2.
bl in. water = 250 N/tn2.
-------
suggested that the WR values were larger and that the dust was not distrib-
uted uniformly upon the filters after cleaning, the validation of the non-
linear model was in no way affected because the curve fitting process re-
lates only to the operating conditions assumed for each curve. Thus, in
testing the model, KR is the initial curve slope for the coordinates S
WR; W = W-WR is the amount of dust added to the filter following the
filter cycle; and W.j.-WRis the dust deposit required before the drag versus
loading curve assumes its linear form with its characteristic slope of K
The values for K2> 1^, SR, W^. and W* and relevant operating information for
the test fabrics are shown in Table 30. The values for these constants
were determined by the graphical analysis of pressure versus loading curves
of the type shown in Figure 98. These data, in conjunction with Equa-
tions (27) and (28), were used to compute the curve trajectories for the
different fabrics, Figure 100. Comparisons of the predicted and experi-
mental results show excellent agreement over the range of input parameters
tested.
It is therefore concluded that model Equations (27) and (28) are appropriate
for describing nonlinear drag versus fabric loading relationships.
EMPIRICAL CORRELATIONS
If the terms appearing in Equation (28) were easy to define, the modeling
of any filter system would be a comparatively simple process. Unfortu-
nately, except by the avenues of direct measurement or system replication
it is not yet possible to determine such parameters as K^, K^, SE> SR and
W* with the desired degree of accuracy.
In the following sections, data from several sources have been analyzed
to determine their potential usefulness. The close inspection of filter
performance statistics appearing in the literature often shows that
critical data are not available. The most serious omission is the
absence of true residual dust holding data for a single element (or bag)
241
-------
Table 30. SUMMARY OF MEASURED FILTRATION PARAMETERS FOR MODEL TESTING
Test
number
1
2
3
4
5
Fabric type
Glass fiber
Polypropylene
Dae r on
Cotton
Polyacrylester
°Rb
3
N min/m
689
287
66
410
41
N min/m
943
779
246
558
205
KK°
N min/gm
67.2
22.7
15.7
12.1
4.42
*2C
N tnin/gm
2.69
1.02
2.08
2.52
0.77
"s
g/m
17.57
65.9
32.2
36.6
146
w*d
g/m
3.9
22.0
13.2
15.1
44.9
Dust type
Wet ground
mica
Fly ash
Fly ash
Fly ash
Fly ash
Filtration6
velocity
V
m/min
0.61
1.22
0.92
0.92
1.22
Type of
cleaning
Shaking
Shaking
Shaking,
reverse air
Shaking,
reverse air
Shaking
Reference
Spaite and
Walsh13
Durham
Dennis and
Wilder16
Dennis and
Wilder16
Durham
to
.p-
CO
Refer to Table 2 for fabric properties.
~h 3
SD, S I in. water min/ft = 820 N min/m .
/•
K^, K, 1 in. water min ft/lb = 0.168 N min/gm.
^fj, W* 1 Ib/£t2 = 4882 g/m2.
eV 1 ft/min = 0.305 m/min.
-------
to
-p-
u>
\5OO -
GLASS FIBER (
POLYPROPYLENE (#2)
x
(I) SOLID LINE IS EXPERIMENTAL DATA
(2) x DENOTES COORDINATES PREDICTED BY
MODEL
(3) REFER TO TABLE 3
DACRON(#3)
x'"
POLYACRYLESTER
200
AVERAGE FABRIC LOADING, (W)
300
400
Figure 100. Comparison between experimental and
predicted drag properties
-------
within a filter system. This problem is encountered with many laboratory
and field measurements. Additionally, the cleaning operations are usually
defined as vigorous, moderate or typical but without regard to the precise
energy input and/or the amount of dust removal.
Data reviewed in the following paragraphs provide some insight as to
probable range of values for the critical terms appearing in the modeling
equations. With reference to S , S and K^ values the correlations are
strictly empirical for want of basic measurements. On the other hand,
the estimation of specific resistance coefficient, K_, can be undertaken
on the basis of existing theory.
Clean Fabric Drag. S.
' " *"" i*.*.-'--.--. Q
The clean fabric drag, which depicts the permeability of the unused
fabric, is related to the Frasier permeability. In the English system, it
is given as the volume flow rate per unit fabric area that produces a
resistance to air flow of 0.5 in. water. In this report, the clean
fabric drag, SQ, is simply expressed as fabric resistance, P, divided
by the filter face velocity, V- Because S can be determined quite
easily and inexpensively, it is hardly justifiable to resort to any
involved theoretical approaches to determine its numerical value. However,
because there is a rational although rather complex process by which the
use of a modified filtration theory enables reasonable predictions for
SQ, methods for evaluating SQ are discussed later in this section.
Effective Drag, S-T,
ii
As part of a comprehensive study of the effects of fabric weave on filter
3
performance, Draemel performed tests with several experimental and com-
mercial fabrics in the form of conventional filter bags and flat test
panels, Table 31. Steady state filtration parameters are depicted for
mechanically-shaken bags whereas single tests are described for unused
244
-------
Table 31. CLEAN (UNUSED) AND EFFECTIVE DRAG VALUES FOR COMMERCIAL
ANDT EXPERIMENTAL FABRICS BY DRAEMEL3 WITH RESUSPENDED
FLi ASH
Fabric type
Weave and yarn count
Clean fabric
in H20/fpm
drag, so
N miti/m.
Single bag - Mechanical shaking*5
Effective drag, Sg
in H20/fpm
N mi n An 3
Dacron 1-39703
Dacron 39707
Dralon 3039577
(acryleater)
Spun-acrylic 4-4589
Polypropylene 5-33106
Dacron 6-39704
Spun rayon 7-884
(cellulose)
Polyester 8-4388
Notnex 9-4400
3x1 twill
78 x 65
3x1 twill
68 x 54
3x1 twill
78 x 70
3x1 twill
76 x 51
3x1 twill
67 x 53
3x1 twill
67 x 58
sateen
96 x 86
3x1 twill, comb, fill-spun
77 x 77
plain - spun
46 x 38
0.027
0.0085
0.043
0.015
0.0038
0.0106
0.0034
0.026
0.013
22.1
7.0
35.3
12.3
3.2
8.7
2.8
21.3
10.6
1.29
0.36
0.79
0.50
0.22
0.49
0.21
0.76
0.54
1058
295
648
410
180
402
172
623
443
Test panels - 1 ft , one-filtration cycle
Dacron Oil
Dacron 020
Dacron 015
Dacron 038
Dacron 088
3x1 twill, filament
77 x 63
sateen, filament
76 x 63
3x1 twill
76 x 82
3x1 twill, staple
76 x 73
3x1 twill, staple
76 x 82
0.014
0.046
0.033
0.009
0.0056
11.5
37.7
27.1
7.4
4.6
0.83
0.83
0.80
0.51
0.18
681
681
656
418
148
Fly ash, HMD = 3.7 urn, = 2.42.
Repetitive filtration at steady state operation with commercial fabric.
Single tests on new test panels, experimental fabrics.
245
-------
filter test panels. In both cases, a redispersed coal fly ash aerosol
was used. Most test fabrics were 3/1 twill weaves of Dacron or related
2
synthetics. Their areal densities were about 206 g/m as compared to
2
about 312 g/m for the woven glass fabrics evaluated in the present study.
The estimated fiber surfaces of the Dacron and glass media, however,
were roughly similar because of the much lower Dacron density (1.4 g/cm3
3
versus 2.2 g/cm for glass). Based upon prior GCA experience with Dacron
fabrics cleaned by mechanical shaking, Draemel's single bag measurements
are assumed to reflect relatively low residual dust holdings. Therefore,'
his reported values of effective drag, S , are assumed to be approximately
correct. In Figure 101, Draemel's data from Table 31 and the results of
the present study, Table 32, have been graphed to determine whether effec-
tive drag, S , might be predicted on the basis of clean fabric drag.
E
Because test dusts, basic fabric properties and length of fabric service
were quite similar it appeared reasonable that clean fabric permeability,
(which reflects among other things the degree of openness or pore area),
should exert a significant effect on the ultimate filter effective, S ,
£j
drag or the residual drag, S . It is emphasized, however, that as pointed
R
out in Draemel's studies, several factors other than clean fabric perme-
ability influence the working drag parameters for a filter.
These variables include the size, amount, and location of bulk fiber
collecting area within the pore structure; the number of effective pores
per unit area, the actual pore geometry and the size distribution of the
particles to be collected.
Therefore, in using SQ alone as the key parameter, a fairly wide spread
in data points should be expected, Figure 101. Effective drag values
for new fabric can be estimated by the relationship
SE (N min/m3) = 189 + 18 S (31)
246
-------
ro
E
z
tr
a
o
Ul
U. Z
10'
SE and SQ
VGCA TESTS, COTTON TABLE 32 (
<3>GCA TESTS, DACRON TABLE 32
O SINGLE BAG, MECHANICAL SHAKE,
MISCELLANEOUS SYNTHETICS TABLE 31
A TEST PANELS,- DACRONS TABLE 31
DGCA TESTS, GLASS FABRICS,
<6 hours SERVICE TABLE 32
x GCA TESTS, GLASS FABRICS
6 months TO 2 years SERVICE
TABLE 32
I GCA TESTS, ALL FABRICS
SE/SR
5 I01 2 5
CLEAN (So) OR RESIDUAL DRAG ( S R ) , Nmin/m3
Figure 101. Relationship between effective (S^,) and clean (S ) or residual (S ) drag
hj L) R
-------
Table 32. SUMMARY OF EXPERIMENTALLY DERIVED MODEL INPUT PARAMETERS USED TO
DRAG VERSUS FABRIC LOADING RELATIONSHIP
PREDICT
ro
JN
oo
a
Test no.
65, 70
71, 99
66, 67
68
69
98b
96C
84
92
77
79
84
Test dust
GCA
fly ash
GCA
fly ash
GCA
fly ash
GCA
fly ash
GCA
fly ash
GCA
fly ash
GCA
fly ash
GCA
fly ash
Rhyclite,
line
Rhyclite
coarse
Lignite
fly ash
Fabric
Sunbury
glass
Sunbury
glass
Nucla
glass
Nucla
glass
Sunbury
glass
Sunbury
glass
Sateen
weave
cotton
Dacron
crowfoot
weave
Sunbury
glass
Sunbury
glass
Sunbury
glass
Service
life
<6 hours
2 years
Unused
~6 months
<6 hours
Unused
Unused
<6 hours
Unused
Unused
Unused
Drag 3
N min/m
S
0
4.1
4-1
4.1-
4.1
2.5
15
32.8
6.6
4.1
4.1
4.1
SR
18
80.3
-
134
13.2
-
*°
18.8
-
-
-
SE
115
352
205
434
lit
60
49.2
188
-
-
-
Specific resistance
coefficient
N min/gm
*R
2.65
7.54
6.56
5.85
2.84
-
2.32
6.23
-
-
-
K2
1.60
1.60
1.60
1.60
1.08
2.06
1.14
1.11
12.3
1.39
1.26
W*
g/m
72.9
45.7
-
60.5
34.5
-
13.9
-
-
-
-
Residual
fabric leading
g/m
44
30
0.0
11
47
0.0
0.0
16.0
-
-
-
V
g/m
175
150
-
175
110
-
50
-
-
-
-
W*/^
0.42
0.38
-
0.35
0.32
-
0.28
-
-
-
-
aFace velocity is 0.61 m/min unless otherwise indicated
Face velocity =0.38 m/min
CFace velocity » 1-52 m/rtiin
Fabric loading at inception of linearity
-------
Residual Drag. Sfa
K
It was expected that the relationship between effective drag,.s , and
residual drag, SR, would parallel that for the previously discussed
effective drag versus clean drag. This follows from the fact that an
increase in SR, which is the result of increased particle entrapment within
the fiber blocked pores, should lead to a higher starting resistance for
the cleaned filter. Figure 101 appears to support this hypothesis despite
the limited data.
Examination of Figure 102 also indicates that the residual drag does not
show any clearcut dependency on the dust/fabric combination. It does
appear, as expected, that extended filter usage increases the residual
drag. In comparing the behavior of filters that have seen very limited
use, there seems to be a slight correlation between the amount of dust
on the filter before cleaning and the residual loading- Since the
resistance across the filter is loading dependent, it is fair to assume
that increased loading may cause increased compression of the residual
dust/fabric substrate. This could account for the higher residual
loadings shown in Figure 102. Aside from calling attention to these factors,
however, it should be noted that there are not yet sufficient data avail-
able to develop the resistance properties of any dust/fabric combination
to the point where they constitute a reliable data input for predictive
models.
Initial Slope, Kj,
The initial slope, K^ of a nonlinear drag versus dust loading curve is
best estimated by careful experimental measurements. Although the early
changes in slope, dS/dW, are logically expressible as functions of weave
characteristics (which determine interstitial deposit geometry) and
intrinsic dust properties (which determine cake permeability), current
249
-------
100
i 1 1 r~
O NEW GLASS OR DACRON FABRICS
A USED GLASS FABRICS, ~2 YEARS
80
A
c
"i
- 60
CD
<
cc
Q
A
A
A
A
O
If)
LJ
cr
40
A
A
20
200
400 600 800 1000
PREVIOUS FABRIC LOADING, g/m2
1200
Figure 102. Effect of previous fabric loading on residual
drag for new and well used fabric
-------
analyses indicate that it would be difficult to make, accurate determina-
tions of the necessary input parameters in this relationship:
- dS/dW = 0(W)(K9)
*
Estimation of W
Given the situation where experimental measurements of S , S , K and
ERR.
K2 are available (or can be readily obtained) it was indicated previously
*
that the term W could be determined empirically as:
E R 2 R R £• (.£?)
An alternative approach is to define W , the constant appearing in Equa-
tions (13) and (14) in terms of the fabric areal density, W , characterizing
the start of the linear portion of the drag versus fabric loading curve.
A
According to test parameters summarized in Tables 30 and 32, W may be
estimated by the following expression:
W* = 0.35 W, <30>
It is emphasized that Equation (30) should be used only as a guideline and
never as a substitute for actual test measurements.
THEORETICAL CORRELATIONS
Clean Fabric Permeability
A detailed examination of fabric properties in which microscopic observa-
tions played a large role provided several insights as to the probable
performance of many dust/fabric combinations. In Section V of this report,
it was shown that the number, type and approximate shape of pore openings
inwoven fabrics could be established by simple geometric considerations.
251
-------
If the filter pores are treated as capillaries, the Hagen-Poiseuille
relationship provides an approximate means to calculate resistance
characteristics.
Ap = 8yQL/10irR
2
where Ap = pressure loss N/m
U = gas viscosity poise
3
Q = volume flow per pore cm /sec
L = filter thickness cm
R = pore (capillary) radius cm
(based on minimum pore area)
Reasonably good agreement was found between measured and observed resistance
values (50 percent lower and 33 percent higher, respectively, for Sunbury
and Nucla fabrics). However, the determination of the minimum pore cross
section Figure 30, Section V by a combination of geometric and microscopic
analyses represents considerable effort.
The value of R appearing in Equation (15) is based upon the circular equiv-
alent of the minimum pore cross section. Since the pores vary in cross
section and present tortuous rather than straight channels, several fabric
weaves should be studied with special attention directed to pore geometry
before any version of the Hagen-Poiseuille equation can be applied with
confidence. In almost every case, direct measurement of clean cloth
permeability (a very simple procedure) is the preferred approach.
Specific Resistance Coefficient, K
The specific resistance coefficient, K2 , has been discussed extensively
in the filtration literature. It is directly calculable from the true
linear portion of the drag versus fabric loading curve where K = dS/dW
252
-------
which is constant for a specified dust/fabric cognation. In accordance
with the Carman-Kozeny theory,1 ^ can also be predicted by the relationships'
K2 = k y S/ (1 - 0/p ,3 (3U)
or
K2 = k y ^ (1 - e)/p £3 (31b)
d v
P
in which the terms are defined as follows:
k = Carman-Kozeny constant, frequently assumed to be 5.0.
V = gas viscosity
S = ratio of particle surface to particle volume
d = particle diameter with a monodisperse system
e = dust cake porosity
p = particle density
P
Equation (31a) was developed for use with granular beds composed of uniformly
sized spheres in which porosity, e, would ordinarily range from roughly
0.3 to 0.7. If the porosity is very high, ~ 0.9, and/or the particle shape
deviates significantly from the spherical, Equation (32a) has little predic-
tive value. The same can be said for those circumstances in which the par-
ticle sizing data are incomplete or incorrect for the dust of interest.
A detailed review of the filtration literature by Billings and Wilder
revealed no reliable means for predicting K2 values except for direct
experimental measurements. Their attempts to correlate data from several
sources were not successful because of the absence of many critical data
inputs. Additionally, the common failing of reporting filter drag as a
function of the dust increment added during the filtering cycle rather
than on the basis of total fabric dust loading, makes it impossible to
interpret correctly most field and laboratory data.
253
-------
Only if one assumes that there are lengthy filtration periods without
interruption for cleaning can the K2 values be considered as approximately
correct. In those instances where a drag versus fabric loading curve in-
volves a nonuniform dust distribution upon the fabric, the true K value
cannot be determined.
K0 Versus Face Velocity - Attempts to correct or modify K values in
2 z
accordance with changes observed when particle shape factor, fabric surface
properties and clean cloth permeability differed for a new set of dust and
operating parameters have been cited by Billings and Wilder in the form
of tabulated correction factors, see Table 33.
The term K is defined roughly by the expression
K = V(ft/min)/3 (32)
32
Borgwardt et al have indicated that K can be defined as
a V1'5 (33)
where a is a characteristic constant for the dust in question. One can
infer from the above that K~ varies as V to the 1 to 1.5 power.
The correctness of these relationships, however, is seriously questioned
because of measurement techniques and lack of critical data. As stated
previously, the nature of the cleaned fabric surface is seldom defined
and the dust is often characterized by a single parameter only such as
the mass mediam diameter.
Experimental data from various laboratory sources,"*"' '"^'^ Table 34 are
graphed in Figure 103, in order to estimate the impact of particle size and
face velocity upon Kr If the probable variations in the physical proper-
ties of the dusts (i.e., size, distribution, shape and density) and the
face velocities are assumed to balance one another the point array in Fig-
ure 103 suggests that K2 varies nearly inversely with mass median diameter.
254
-------
Table 33. CORRECTIONS FACTORS FOR K.
Particle shape, K
sh
Dust material
Crushed
Ash
Irregular
Collapsible
Fumes
K
sh
10.0
4.
3.
0.2
0.05
Fabric surface, K.
Fs
Fabric
Smooth
Napped
Felts
Va
1.0
1/2
1/4
Fabric permeability,
K
perm
sa
o
10
20
30
40
50
60
70
80
90
K
perm
1.3
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5
aClean (unused) permeability
CFM/ft2 at 0.5 in.H20
These correction factors are intended for use in the empirical equation;
= 1000
255
-------
Table 34. DATA SUMMARIES FOR ESTIMATING
VELOCITY AND PARTICLE SIZE
K2 AS A FUNCTION OF FACE
Ln. H,0 min ft/lh
57.0
42.6
40.0
37.0
37.0
28.0
27.0
24.4'
23.0
22.5
21.4
21.3
21.2
16.5
16
16.3
15.0
15.0
l./i . 7
14,4
12.4
11.6
11.2
9.6
9.6
9.6
8.
7.8
7.7
l.'i
7. 1
Type of filter
Dacron
Class fiber
Dacron RP
Dae ron AN
Rlnss fiber
Dacron DN
Dacron EN
Glass fiber 1
Glass fiber 1
Glass fiber 1
Glass fiber 3
Glass fiber .1
Dacron A
Glass fiber 3
Glass fiber 3
Dacron RC
Dacron B
Nomex A
Nomex A
Dacron B
Dacron RP
Nomex filament
Dacron RP
Glass fiber N
Glass fiber S
Glass fiber S
Nomex B
Glass fiber
Dacron C
JiicriMi C
>acrim C.
Weave yarn count
No data
3x1 Crowfoot
SSxin
Plain ytaple
30x28
3x1 Twill
78x69
3x1 Twi 11
53x51
3x1 Twill
79x81
3x1 Twill
42x28
3x1 Crowfoot
55x50
3x1 Crowfoot
55x50
3x1 Crowfoot
55x50
3x1 Crowfoot
55x58
3x1 Crowfoot
55x55
3x1 Twill, £11.
82x62
3x1 Crowfoot
55x58
3x1 Crowfoot
55x58
3x1 Crowfoot
71x51
3x1 Twill
82x76
3x1 Twill
96x78
3x1 Twill
96x78
3x1 Twill
82x76
Plain stnple
30x2H
3x1 Twill
96x78
Plain staple
30x28
3x1 Twill
66x30
3x1 Twill
54x30
3x1 Twill
54x30
3x1 Twill, spun comb.
95x58
3x1 Twill, fil. bulk
54x30
3x1 Twill
77xHl
•)xl Twi 1 1
77x81
3*1 Twi 11
77xHl
Dust type
size, MMD,
urn
As. salts
3.3
Mica
r,.o
Fly ash
5.0
Tiilu
5.1
Fly ash
3.5
Talc
5.1
Talc
5.1
Mica
6.
Mica
6.0
Mica
6.0
Mica
fi.O
Mica
6.0
Mica
6.0
Mica
6.0
Mica
6.0
Fly ash
8.0
Mica
Fly ash
15.0
Fly ash
15.0
Mica
6.0
Fly ash
8.0
Fly ash
15.0
Fly ash
8.0
Fly ash
9.0
Fly ash
9.0
Fly ash
9.0
Fly ash
15.0
Fly ash
18.0
Fly ash
15.0
!•' 1
15.0
VI y Huh
15
Gas
velocity ,
ft/min
1.0
6
3
J
3
3
3
6
4
2
4
2
2
2
2
3
2
4
4
2
3
4
3
2
2
2
4
2
4
'
,,
Ref.
no.
i
14
10
.'li
14
20
20
14
14
14
14
14
14
14
14
10
14
18
18
,,
16
18
10
'Jg. 54
I'lR. 53
I'lg. 53
IH
32
18
III
III
256
-------
50
£ 20
c
E
O
CO
X
CVJ
10
O
ft./min,C>
4 ft./min ,x
3 ft. /min.O
2ft./min, A
5 10 20
PARTICLE MASS MEDIAN DIAMETER >Ax.m
50
Figure 103. Specific resistance coefficient (K^) versus mass median
diameter and face velocity. Data from Table 34
257
-------
K2 = (d ) (34)
Although, this observation appears to contradict theory, which indicates
that the diameter exponent should be -2 (see Equation (31b)), it should be
noted that Equation (31b) assumes a monodisperse and not a polydisperse
particle system.
Inspection of the data also indicates that the larger K2 values are
associated with the higher face velocities. According to the estimated
constant velocity contours, which are conceeded to be speculative, it
appears that the effect of velocity upon K« may be less than that currently
reported in the literature.
Plotting of Ko values for two particle sizes against the face velocities
shown in Figure 104 suggests that the velocity effect might be better repre-
sented by the following
K2 = 0 (V°-5) (35)'
with the velocity exponent ranging between 0.5 and 1.0 for many commonly
encountered dust/fabric combinations.
In a series of experiments performed during the current program, the effect
of filtration velocity upon K2 was studied with the GCA fly ash/Sunbury
fabric system at three filtration velocities. Because these tests were
carried out under carefully controlled conditions, there seems little
reason to question the approximate square root relationship shown by the
dotted line on Figure 104, at least with respect to fly ash and closely
related dusts. For this reason, we have elected to define the effect of
face velocity on K2 by an expression such as Equation (35) to correct for
K2 variations during real filtration processes involving coal fly ash and
woven glass fabrics. With reference to a specific dust/fabric system KZ
should probably be defined as
258
-------
50
E 20
c
CM
10
I A
2 o
3
PARTICLE MMO_,fim
6.0
15.0
GCA DATA
DATA SOURCE
FIGURE 103
FIGURE 103
FIGURE 71
2 5
AVERAGE FACE VELOCITY, ft./min.
A
Figure 104. Estimated effect of face velocity on K2 based upon
literature review, Table 6
259
-------
K2 - a V0-5
where the constant, a, is determined from the actual measurement of K
at any velocity within the expected working velocity range for the system.
K? Versus Specific Surface Parameter - It was stated previously that a
major limitation of the Carman-Kozeny relationships is that they apply to
ideal structures; i.e., beds composed of spherical particles, uniform
with respect to size and physical properties and bed depth. However, the
fact that the present study provided more details on particle characteris-
tics and other relevant filtration parameters than usually available sug-
gested that their predictive capability be re-examined for nonideal
situations.
The first step involved determination of a specific surface parameter,
S , that more clearly describes the pore properties-channel cross sections
and wall surface area - associated with polydisperse distributions Thus
the term A was considered to define total superficial (or envelope)
surface for all particles constituting the dust cake and V to describe
the total particle volume. Thus, for spherical particles, S is then
defined as
2 IT * 6d«2
so= yvp - N * ds /N -g- v= -f- (36)
d
v
where dg and dy are the surface and volume mean diameters, respectively,
and N the number of particles in a unit mass of filter bed.
The characteristic diameters cited above are easily determined from the
linear approximations to logarithmic-normal mass distributions for inlet
fly ash aerosols; i.e.
260
-------
log d = log HMD - 4.605 log2 a
8 g
log dv = log MMD - 3.454 log2 ag
Such measurements were performed both in the field and in the laboratory
with the Andersen impactor, a commonly used device to determine mass
size distributions.
K2 Versus Dust Cake Porosity - A second critical parameter appearing in
the Carman-Kozeny equation is bed porosity, e. In the case of coarse
granular materials, Dalla Valle reports that particles > 10 ym form
moderate porosity beds, ~0.3 to 0.7, whereas powders in the 1 to 10 ym
range may have larger void volumes, 0.5 to 0.9.33 Two approaches for
estimating porosity were used in this study. Actual filter cakes deposited
under normal filtration conditions upon woven all glass fabrics, Section VII,
were excised by micro manipulation so that their volume and weight could
be determined. These tests showed a bulk density of 0.82 g/cm3, which
when related to an assumed discrete particle density of 2.0 g/cm3, indi-
cates a bed or filter cake porosity of 0.59.
A second approach for estimating porosity was to determine the bulk density
of test dusts prior to re-aerosolizing. Generally the "as received,"
moderately shaken or vibrated, and shaken and heated samples showed that
the bulk densities were roughly one-half the assumed fly ash density of
2.0 g/cm3.
It is emphasized that highly accurate estimates of e are necessary before
K2 can be predicted with any high degree of confidence. Reference to
Table 35 shows that small, ~10 percent, variations in porosity lead to
large differences, -50 percent, in the porosity function, (1 - e)/e .
Calculated and Observed K2 Values, Field and Laboratory Tests - Measure-
ments at the Nucla power station were analyzed to determine the probable
261
-------
value of K~ for the field aerosols. Several tests were reviewed,
Table 36, in which lengthy filtration periods (1 to 4 hours) were main-
tained between the cleaning cycles. The Nucla operating procedure usually
involved continuous cleaning of all six compartments over a 25-minute
period once the cleaning cycle was pressure actuated. Because of the long
filtering periods, the characteristic lack of uniformity in fabric dust
loadings from one compartment to another immediately after cleaning de-
creased greatly as filtration progressed. Therefore, it is justifiable to
estimate the specific resistance coefficient, K-, directly from the re-
sistance change, AP, noted at constant velocity, V, for the change in
fabric loading, AW, over the measurement period. It was assumed that the
inlet loadings, filtration velocities and temperatures were constant over
the indicated averaging periods, although some variations were apparent as
evidenced by the change in slope of the resistance versus time chart traces.
Table 35. POROSITY FUNCTION FOR GRANU-
LAR POROUS MEDIA
Porosity
e
0.90
0.85
0.80
0.75
0.70
0.65
0.60
0.55
0.50
0.45
0.40
0.35
0.30
0.25
1-e
£3
0.14
0.24
0.39
0.60
0.88
1.27
1.85
2.70
4.0
6.0
9.4
15.1
25.9
48.0
Particle
diameter ,
Vim
1.0
1.5
1.5
2.0
2.5-3.0
3.5
5.0
8.0
10-12
20
25-30
30
30
30
262
-------
Table 36. MEASURED AND CALCULATED K2 VALUES FOR NUCLA FIELD
TESTS8
to
Run no.
1-1-A
1-2-A
1-3-A
11 -AB
14-AB
15-B
16-AB
16-B
19-1-AB
19-2-AB
Measurement
period ,
minutes
110 j
90 !
80 >
140
100
225
60
60
141
72
Inlet dust b
concentrations ,
g/nr* dstp
4-76
3.98
4.05
3.07
4.99
4.99
4.53
4.53
Particle size parameters0
HMD,
Vim
14.8
10.2
14.1
10.6
11.3
12.7
11.3
11.3
a
g
5.2
2.68
3.33
3.58
3.55
3.27
2.5
2.5
' * 3
&/cm
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
So2 d
cm" 2
2.37 x 108
6.98 x 107
7.79 x 107
2.13 x lOH
2.28 x 108
9.03 x 107
6.51 x 107
6.51 x 107
Measured K2 >
N min/g m
0.531
1.64
0.745
1.00
1.13
1.18
1.13
1.13
0.943
1.16
Calculated KZ >
N min/g m
3.39
3.39
3.39
0.962
1.72
3.39
2.02
1.33
0.843
0.843
rirst number refers to test; second number to different, non-overlapping measurement periods during test; A or B
refers to separate Andersen impactor analyses; and AB to the average of analyses A and B.
Average dust concentration by Method 5 type sampling.
Andersen impactor estimates of mass distribution parameters by log-normal distribution with assumed particle
(discrete) density of 2 g/cm^.
d 2 2 232
So computed from So = (6 ds /dv ) for assumed spherical particles with indicated surface (ds) and volume (dv)
diameters computed from mass distribution parameters.
EMeasured K2 for actual filtration velocity, ~0.84 m/min and a gas temperature of ~124°C.
Ko computed by Carman-Kozeny relation K2 = °~^"E~^^ ' wnere e = 0.59 and (e 3) = 1.9-
Note 1: &in no. 1 - Sizing data suspect, poor agreement between Method 5 (4.76 g/m-*) and Andersen impactor
(1.14 g/m3) loadings
Run no. 1 - Average measured K£ at operating conditions = 0.974 N min/g m.
Hote 2: Tor convert K2 (metric) to K2 (English) multiply N min/g m by 6.0 to obtain in H20 rain ft/lb.
-------
In performing these analyses, the cake porosity, e, was estimated to be
0.59 on the basis of laboratory bulk density measurements on a filter
dust layer, 0.82 g/cm3, and a discrete particle density for fly ash of
2 g/cm3.
The results of these calculations, Table 36, showed an average predicted
K? value about two times greater than the measured value. The apparent
agreement with theory is surprisingly good in view of the acknowledged
limitations of input parameter measurements.
For example, linear extrapolations beyond the observed size classes for
cascade impactor size distributions may not afford an accurate description
of all size properties. Additionally, one is usually compelled to assume
that the particles depositing on the various impactor stages are discrete
particles having the density of the parent material. Actually, there may
be agglomerates present to the extent of 10 to 15 percent of the total
number count when compressed air is used to redisperse dry powders. Most
real gas streams also contain agglomerated particles. Thus, conversion
of aerodynamic size to actual size may give erroneous results for estimates
of surface and volume mean diameters even if all particles are spheres.
Another potential problem is to decide whether a population of agglomerated
particles will produce a deposit whose porosity is at least partially
controlled by the external dimensions of the agglomerated particles.
Were this to be true, a system composed of agglomerates, each of stable
structure and having a porosity of 0.5, might conceivably form a dust
layer with an interagglomerate porosity of 0.5 and an overall porosity
of 0.75. At this time, it does not appear that a precise definition of
the above conditions is possible. In lieu of rather difficult and time
consuming laboratory measurements where sections of dust cake are excised
for analysis, it appears that a practical measure of cake porosity may
be obtained by noting bulk density values for the loose dust under a
variety of tamping (vibration) and heating conditions. Average values
264
-------
for all Nucla tests are summarized in Table 37. Note that the K values
for ambient conditions include corrections for gas viscosity and'filtra-
tion velocity.
Table 37. SUMMARY OF AVERAGE K VALUE FROM NUCLA FIELD STUDIES
Measured K«
Calculated K2
Calculated S
o
Test conditions
124°C, 0.844 m/min
N min/g m
1.05
2.09
1.28 x 108 cm~2 (Average
Ambient conditions
21°C, 0.61 m/min
N min/g m
0.75
1.49
of all tests, Table 36)
Further indication of the degree of conformity found between measured and
predicted K. values (the latter calculated from the Carman-Kozeny relation-
ship) is shown in Table 38 for several past and current GCA tests with fly
ash and other test dusts. In these tests, the porosity values for coal fly
ash deposits were taken as 0.59 based on GCA laboratory tests. Porosity
values for lignite fly ash, talc and granite dust were based upon bulk
density measurements on the dry dust using graduated containers and a
laboratory balance. The first set of size parameters listed for any dust-
fabric combination, Table 38, is the original analysis of size distribution
curve. These (original size parameters) were used to calculate the indi-
2
cated S0 values.
In the case of tests with coarse granite dust, supplemental trial estimates
were made to ascertain what impact variations in estimated size parameters
(HMD and Oft) might have upon So2. The variations in size parameters repre-
O '
sent different visual estimates of the best linear fit to the size distri-
butions shown in Figure 15, Section IV. The same exercise was carried out.
265
-------
Table 38. CALCULATED AND MEASURED VALUES FOR SPECIFIC RESISTANCE COEFFICIENTS FOR VARIOUS DUSTS
Test dust
Coal fly ash
Public Service
Co. , NH (GCA)
Coal fly ash
Public Service
Co., MH
Coal fly ash
Detroit Edison
(EPA)
Coal fly ash
Public Service
Co., NH (GCA)
Co.al fly ash
Nucla, CO
lignite fly ash
Texas Power
and Light
Dust parameters
MMD,a
vm Og
4.17(1) 2.44
5.0 (M) 2.13
6-38(1) 3.28
3.8 (I) 3.28
3-2 (M) 1.8
2.42(M) 1.77
11.3(1) 3-55
8,85(1) 2.5
8.85(1) 2.5
-8.85(1) 2.78
Particle
density
g/cm3
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.4
2.4
2.4
S°-2
cm *•
4.73 x 108
2.. 58 x 108
3.55 x 108
9.94 x 108
4.78 x 108
8.49 x 108
1.28 x 108
1.06 x 108
1.06 x 108
1.30 x 108
Cake
porosity ,
E:
0.59
0.59
0.59
Filtration
Parameters
Velocity ,
m/min
0.915
0.915
0.605
i
0.59
0.59
0.59
0.59
0.46
0.42
0.46
0.823
0.915
0.915
0.851
0.605
0.605
0.605
Temp.,
'21
21
21
138
21
21
124
21
21
21
Filter fabric
Napped cotton.
sateen weave
Glass,
3/1 twill
Glass,
3/1 twill
Napped cotton,
sateen weave
Napped cotton,
sateen weave
Glass,
3/1 twill
Glass,
3/1 twill
Glass,
3/1 twill
Glass,
3/1 twill
Test
scale
Pilot
Pilot
Bench
Field
Pilot
Pilot
Field
Bench
Bench
Bench
Measured K.2 ,
Test
conditions
2.29
2.29
1.40
6.35
1.22
2.17
1.05
1.34
1.34
1.34
Amb i en t
condi t ions
210C
0.605 m/min
1.85
1.85
1.40
4.45
1.00
1.77
0.75
1.34
1.34
1.34
Calculated
K2.
21°C
5.72
3.74
5.14
14.4
6.19
11.0
1.84
3.67
5.16
4-49
Ratio,
calc. K2
meas- K.2
3.09
2.02
3.67
3.23
6.18
6.20
1.98
2.78
3.86
3.36
ON
-------
Table 38 (continued). CALCULATED AND MEASURED VALUES FOR SPECIFIC RESISTANCE
COEFFICIENTS FOR VARIOUS DUSTS
Test dust
Granite dust
Talc
Dust parameters
MMD,a
ym
9.21(1)
9.21(1)
9.21(1)
8.1 (I)
9.84(1)
9.21(1)
1-23(1)
2.77(1)
2.77(1)
2.77(1)
°8
4.83
4.55
4.05
3.88
4.32
4.83
2.38
2.9
2.9
2.9
Particle
density ,
g/cm3
2.2
2.2
2.2
2.2
2.2
2.2
2.2
2.2
2.2
2.2
So2
9
cm ^
5.05 x 108
4.13 x 108
2.88 x 108
3.24 x 108
3.44 x 108
1.01 x 10B
5.10 x 109
1.51 x 109
1.51 x 109
1.51 x 109
Cake
porosity,
E
0.68
0.68
0.68
0.68
0.68
0.60
0.68
0.84
0.82
0.73
Filtration
Parameters
Velocity,
m/min
0.605
0.605
0.605
0.605
0.605
0.605
0.605
0.915
0.915
0.915
Temp . ,
°C
21
21
21
21
21
21
21
21
21
21
Filter fabric
Glass,
3/1 twill
Glass,
3/1 twill
Glass,
3/1 twill
Glass ,
3/1 twill
Glass,
3/1 twill
Glass,
3/1 twill
Glass
3/1 twill
Cotton,
3/1 twill
Cotton,
3/1 twill
Cotton,
3/1 twill
Test
scale
Bench
Bench
Bench
Bench
Bench
Bench
Bench
Jilot
Pilot
Pilot
Measured K2,
Test
conditions
1.38
1.38
1.38
1.38
1.38
1.38
12.3
5.76
5.76
5.76
Ambient
conditions
21°C
0.605 m/min
1.38
1.38
1.38
1.38
1.38
1.38
12-3
4.71
4.71
4.71
Calculated
Ratio,
K2, ! calc. K?
21°C
2.64
2.15
1.50
1.70
1.69
5.28
26.7
2.35
2.72
5.78
meas- K2
1.92
1.56
1.09
1.23
1.22
3.84
1.94
0.50
0.58
1.23
OS
(I) refers to
(M) refers to
cascade impactor sizing -
microscope sizing (light field, oil immersion).
-------
with respect to both size parameters and porosity for the lignite tests
and with respect to porosity alone for the talc measurements.
The relationship between K values and the specific surface parameters,
2
S , Figure 105, indicates that grouping of data points by type of dust
(and/or type of measurement) shows a strong linear correlation between
2
K2 and S as postulated the Carman-Kozeny theory. It is emphasized that
the difference between the MMD value for a highly polydisperse distribution
and the diameter that characterizes the term So may be considereable. In
the case of the coarse granite dust, the MMD was 9.21 ym whereas the single
diameter used to compute SQ was 2.65 ym. Since the SQ term is squared in
calculating K » a twelvefold difference in the estimate of K£ would result.
Of particular interest to the present program is the fact that bench, pilot
and laboratory tests with the same fly ash type (Public Service Co. of
New Hampshire) as well as field tests with a similar (sizewise) Nucla
2
stoker fly ash show surprisingly good agreement with the K?-S correlation.
At the same time, the predicted K_ values are consistently high based upon
the data summaries given in Table 39.
Table 39. MEASURED AND PREDICTED K2 VALUES
Fly ash
Public Service
Co., N.H., coal-
cyclone boiler
Nucla, Colorado
coal-stoker-fired
Texas Power and
Light lignite
Test scale
{Pilot
Bench
Field
Field
Bench
Predicted K2
5.72
5.14
14.4
1.84
4.44
Measured K2
1.85
1.40
4.45
0.75
1.34
K.2 pred./meas.
3.09
3.67
3.23
1.98
3.31
268
-------
10
E
9
>»
e
i
(SI
*
o n
w 10°
v>
1,F and P REFER TO
and PILOT TESTS.
, FIELD
10'
/A 8
DUST
0 COAL FLY ASM
N.H POWER
SERVICE CO.
CAKE
POROSITY
0.59
x COAL FLY ASH 0.59
OETRIOT EDISON
O COAL FLY ASH 0.59
NUCLA, COLORADO
V LIGNITE FLY ASH 0.46
TEXAS POWER a
LIGHT
A GRANITE OUST 0.68
0 TALC OUST
0.04
SIZING
METHOD •
ANDERSEN
IMPACTOR
MICROSCOPE.
ANDERSEN
IMPACTOR
ANDERSEN
IMPACTOR
ANDERSEN
IMPACTOR
ANDERSEN
IMPACTOR
10'
SPECIFIC SURFACE PARAMETER (So)2, cm-2
10
,10
Figure 105. Specific resistance coefficient versus specific surface param-
eter (S ) for various dusts
269
-------
As far as the tests with three different fly ashes are concerned, the
ratios for predicted and measured values appear to range between 2 and 4.
Thus, if compelled to estimate K£ without resorting to experimental measure-
ments, one would have to accept possible errors of at least + 100 percent.
As stated previously, the sensitivity of K£ to the porosity function,
1-E/e3 mitigates against a high level of accuracy. However, because K9 can
be readily measured with simple testing apparatus either in the field or
in the laboratory it would be impractical not to use measured K values
as a starting point for most modeling applications.
Although the data are limited, it does appear that once a K^ value is
established for a specific dust and a specified size distribution, it is
possible to determine K9 for other size permutations of the same dust on
1 2
the basis of the specific surface parameter, SQ .
FABRIC CLEANING AND FILTER PERFORMANCE
The preceding discussions provide the necessary data inputs for modeling
the resistance (or drag) versus fabric loading relationship for a specified
dust/fabric system in which the dust is deposited uniformly upon the
fabric surface. The above conditions prevail when filtering with a new
(unused) fabric or with a used but completely and uniformly cleaned
fabric. However, real fabric filter systems ranging from single to multi-
compartmented, sequentially cleaned units almost invariable see only par-
tial cleaning of the fabric surfaces, regardless of the method, intensity,
frequency or sequencing of cleaning. Therefore, it is imperative to exam-
ine very thoroughly the state of the fabric surface after cleaning and its
impact upon system resistance and emissions characteristics. At the out-
set, it was recognized that gas flow rates and emission characteristics
would vary from point to point throughout the collection system because
of local variations in filter drag.
270
-------
Resistance (Drag) Versus Dust Distribution on Fabric
The results of the GCA fabric filter cleaning study10 indicated that the
actual removal of dust from a fabric by mechanical action usually took
place as a spallation process in which the dust separation occurred at
the interface between the dust layer and the fabric. Except for unique
circumstances, the resistance to tensile or shear forces at this boundary
is much less than within the cake itself.
Examination of the forces needed to dislodge a dust cake by collapse or
mechanical shaking has indicated that shearing or tensile forces in the
100 to 300 dynes/cm2 range are required to cause cake detachment.28 In
the case of bag collapse systems, a 0.1 cm layer of fly ash having a
bulk density of 1 g/cm3 exerts a shearing force of roughly 100 dynes/cm2
in a gravity field of Ig. On the other hand, the acceleration levels im-
parted to the dust layer in a mechanical shaking system are in the 5 to
6g range for a shaking frequency of 7 cps and a 1-inch shaking amplitude.
Therefore, a tensile force of the order of 100 to 300 dynes/cm2 is gen-
erated at the dust fabric interface with a 0.02 cm layer of dust. One
infers that mechanical shaking will remove considerably more dust than
simple bag collapse. The above line of reasoning also suggests strongly
that the physical behavior and ultimate performance of both bag collapse
and mechanical shaking cleaning systems can be treated in similar fashion.
Although the same approach should be applicable to pulse jet systems, two
important factors should be kept in mind. First, estimated accelerations
imparted to the fabric by reverse pulse air are much higher, c-200 g, such
that the areal dust deposit density needed to achieve separating forces in
the 100 to 300 dynes/cm2 range is very low, approximate micrometers. Be-
cause of the napped character of most felts used in pulse jet systems, it
appears unlikely that a distinct, fiber-free layer can develop in most
filtration applications. Second, the felted media presents many more pores
with much smaller diameters and greater depths than encountered with most
271
-------
woven fabrics. Hence, the basic substrate is a much more effective dust
arrester than the typical woven fabric.
Analytical complications had been anticipated in applying the dust separa-
tion concept used for collapse and shake cleaning systems because of the
difficulty in determining which fractions of the dust were interstitially
or superficially deposited for a given set of operating variables. Sub-
sequent laboratory tests, Section VII, indicated that these and other
critical measurements could be made with ease.
By means of laboratory measurements, it was possible to estimate filter
performance by two different approaches.
• The drag values for loaded and cleaned filters in
conjunction with the fraction of dust removed (or
the fraction of cleaned filter surface exposed)
allowed computation of all intermediate system
resistance values as well as the variations in areal
dust deposit density with time.
• The measurement of total system drag in conjunction
with the fraction of surface cleaned by flexure at
two specific levels of cleaning, provided a direct
mechanism for calculating residual and terminal
drag values for the system.
Examination of Figure 106 shows how extreme the changes in systems resis-
tance or drag are when filter cleaning is achieved by the dislodgment of
dust layers from the dust/fabric interface rather than as a uniform sur-
face spallation. The numbers used in developing Figure 106 and Table 40
relate closely to the drag values measured in actual laboratory tests.
The average drag values after cleaning, S_, have been calculated from
K.
the following relationship:
272
-------
FRACTION OF CLEANED FILTER SURFACE
0-80.6 0.4 0.2
FILTER DRAG
15 N min/m3 CLEANED SURFACE
IOOO N min/m3 UNCLEANED SURFACE
AVERAGE FILTER VELOCITY 0.61 m/min
100 200 300 400 5OO 600 700
AVERAGE FABRIC LOADING,grams/m3
Figure 106. Average filter drag with various degrees of dust removal
fly ash filtration with woven glass fabric
273
-------
where a and a, are the fractions of cleaned and uncleaned fabric,
c d
respectively. S and S, represent filter drag values for cleaned and un-
cleaned regions with estimated values of 15 and 1000 N-min/m3, respectively.
To keep within the working range of coal fly ash/woven glass fabric
filter systems, the fabric loading prior to cleaning has been assumed to
be 700 g/m2. By assigning various levels of fractional cleaning, for
which the average residual loading is assumed to be directly proportional
to the cleaned filter surface, the actual system drag values at the re-
sumption of filtration are shown to be highly sensitive to the fraction
of freshly cleaned surface when only a small fraction has been cleaned.
Table 40. RELATIONSHIP BETWEEN CLEANED FABRIC SURFACE
AND AVERAGE FILTER DRAG - COAL FLY ASH FIL-
TRATION WITH WOVEN GLASS FABRIC (PREDICTED)
Average
SR
N-min/m3
1,000
603
432
234
132
70.8
48.3
36.7
29.6
24.7
21.6
18.7
16.6
15.8
Surface area
fraction
Cleaned
0.00
0.01
0.02
0.05
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
0.95
Uncleaned
1.00
0.99
0.98
0.95
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.05
Average
residual
dust
holding
grams /m2
700
693
686
665
630
560
490
420
350
280
210
140
70
35
Cleaned drag = 15 N/min/m3.
Uncleaned drag = 1000 N-min/m3.
Since the fabric drag resulting from successive flexing as depicted in
Figure 106 rapidly approaches the cleaned fabric drag as a limiting value,
one should consider the situation where the flexing process has been
274
-------
stopped after 50 percent of the dust has been removed. Within the ex- '
pected accuracy limits for such measurements, there would appear to be no
advantage, in terms of resistance, to continued flexing beyond the
350 g/m2 load level. However, were flexing continued, much more cleaned
fabric area would become available with an attendant increase in filtra-
tion capacity (the loading present at the maximum allowed pressure drop)
during the next filtration cycle. We do not imply it is best that woven
glass bags cleaned by collapse and reverse flow be flexed until nearly
all of the dust is removed. There would be little reduction in resistance
and there would be a probable penalty in terms of increased dust emissions.
The main objective for the calculations illustrated in Figure 106 is to
show how closely the process relates to the data presented several years
ago by Walsh and Spaite in Figures 107 and 108. For a specified mechanical
shaking system (defined in terms of amplitude and frequency), there was
a limiting number of shakes, N , beyond which no appreciable reduction in
s
residual drag was attainable. There was also a limiting number of shakes,
N , beyond which no increase in filtration capacity could be attained.
w
The latter number of shakes, N , always exceeded the number required to
reach a practical minimum resistance. More recent shaking studies per-
formed by GCA indicated that no appreciable increase in dust removal was
obtainable after about 200 shakes.
According to Walsh and Spaite, the additional number of shakes, NW - Ng,
required to reach a maximum holding capacity for a specified shaking mode
was assumed to re-orient or restructure the cake such that discontinuities
were minimized. Based upon the behavior of fabrics cleaned by collapse
and the other analyses presented in this discussion, it appears more likely
that significant dust removal and additional cleaned surface is gained
during the N - N shaking interval with a negligible decrease in drag as
w s
shown in Figure 106.
275
-------
1.800
"{-1.000
u.
X,
z
< 800
0
£ eoo
o
o *°°
a
u
|j ZOO
1 1 1 1 1 1 I
T|ST CONDITIONS
FILTERS -COTTON SATEEM FABRIC
— ausi - ELUTRIATED FL.Y ASM
FILTER VELOCITY - jo fpm
TERMINAL FILTER DRAG* 10 IN HjO'rpm
" ' JL
/ B---'^
~ f fS CURVE AMPLlTuft FREClJENC
4 / A 4 INCHES JI5cpm
If C .' e | INCHES 4
Lx . ,../... T""1"""
71. " 0 _..-• — i -----
!^._ Jjr-'' ! 1 1 1 1 1
1 1
—
f
J
> ACCf LtHATlON ~
7 enc^iN/uiN/Min
1 1
20
40 60 80 100 120 140 160
CLEANING DURATION (NUMBER OF STROKES)
ieo
?oo
Figure 107. Effect of cleaning duration on filter capacity for
several shaking conditions11+
1000
TEST CONDITIONS
fILTERS- COTTON SATEEN FABRIC
OUST - ELUTRIATED FLY ASH
FILTER VELOCITY - 3 0 Ipm
TERMINAL FILTER DRAG-10 IN H
Wm=8IO(l-«"a )
Q * ACCELERATION in g<
l.385xO6
2 4 t e 10 I! 14 16 is
SHAKER ACCELERATION X KT* UNCHES/MINVM IN )
EFFECT OF SHAKER ACCELERATION ON FILTER CAPACITY
Figure 108. Effect of shaker acceleration on filter
capacity
276
-------
In the case of many mechanical shaking systems, the energy transmitted via
the shaking process is sufficient to dislodge only the more loosely bonded
or thicker (and heavier) sections of the dust layer. Hence, a limiting
residual holding is reached by a path resembling an exponential decay pro-
cess. Most measurements reported by GCA in their study of fabric filter
cleaning mechanisms10 showed that after 200 individual shakes, only about
5 percent more of the potentially dislodgeable dust (assumed to be equiva-
lent to an additional 5 percent of cleaned filter surface) could be re-
moved with the specified cleaning mode.
The curves of Figure 109 illustrate why many past modeling efforts have
not been successful. Curve 1 depicts the cleaned condition described
previously where complete dislodgment of the overlying dust layer has
been accomplished by hand cleaning. Curves 2, 3 and 4 describe the
characteristic drag versus loading curves that result when the fabric
surface has undergone partial cleaning. Note that whereas the abscissa
denotes the average areal dust loading, the actual filter surface dis-
plays two characteristic regions at the resumption of filtration, the
first from which the areal density has been reduced to the W level and
R
the second which retains the former uncleaned fabric loading, W . Thus,
for the fraction of cleaned and uncleaned surfaces relating to Curve 3,
the average starting areal density is 0.5 W in the case of large
terminal loadings (Wm) and small residual loadings (W ), the ratio,
1 K
W AL, is an approximate measure of the fraction of uncleaned area.
K 1
Reference to the literature indicates that filter performance is often
characterized by curves such as shown in Figure 109 except that the zero
point on the abscissa refers to the residual dust holding which may be
0.25, 0.5, 0.75 or any other fraction of the terminal loading, WT> depend-
ing upon the intensity of cleaning. Since the cleaning intensity and the
actual residual dust holding (which is very difficult to measure) are sel-
dom indicated, it is possible to draw several distinct conclusions from
such drag versus fabric loading relationships, most of which will be erro-
neous. For example, if Curves 2, 3, and 4 are graphed so that the abscissa
277
-------
CO
EC
O
UJ
b
NJ
~-J
00
CK
UJ
DESCRIPTION
MAXIMUM POSSIBLE CLEANING
HIGHLY EFFICIENT CLEANING
AVERAGE CLEANING RANGE
(MECHANICAL SHAKING)
AVERAGE CLEANING RANGE
COLLAPSE WITH REVERSE
FLOW
0 W
f T \St ^ TV T-
AVERAGE FABRIC LOADING,W
Figure 109. Typical drag versus loading curves for filters with different degrees of cleaning and a
maximum allowable level for terminal drag, S , and terminal fabric loading, W
-------
refers to the dust increment deposited during a typical filtration cycle
and the system drag is constrained to values < ST, it is apparent that the
upper, nearly linear portion of Curve 2 will display the same slope, K ,
shown by Curve 1. The latter value, dS/dW, gives the correct KZ value for
the dust at the specified filtration velocity. However, as the residual
dust holding increases, the near linear sections of Curves 3 and 4 no
longer display the same slope and a reduction in the permissible S value
further accentuates this difference.
The high degree of initial curvature in Curves 2, 3, and 4 results from
a constantly changing air flow (and dust deposition rate) for the initial
high (cleaned) and low (uncleaned) permeability regions of the fabric sur-
face. Since the dust accumulation is most rapid on the "just cleaned"
regions, the areal densities for both elements of the fabric surface will
converge, thus leading to the dS/dW or K_ relationship shown for a
uniformly loaded fabric. The net result is that one cannot use curves of
the type shown in Figure 109, to determine K and S for any generalized
/ E
modeling procedures. Only if the parameters deriving from any of the
Figure 109 curves are applied to replicate filtration conditions will the
empirically based equations provide useful data.
The problems discussed above can be avoided if the curves of Figure 109 are
correctly recognized as reflecting the results of rapidly changing, parallel
flows through fabric regions of changing permeability. The latter concept
is frequently described in the literature with respect to sequentially
cleaned, multicompartment filters.l»13>16 The compartment approach, however,
fails to take the behavior of individual bags into consideration.
Several experiments were performed to test the hypothesis that fabric fil-
ter performance could be defined by analyzing the behavior of partially
cleaned fabrics after filtration was resumed. The starting assumptions are
reiterated below to make clear the ground rules for the modeling process.
When a uniformly loaded filter has undergone partial cleaning, the resul-
tant surface is composed of two distinct areas; the first from which no
279
-------
dust has been removed and the second which is cleaned down to its charac*-
teristic W value.
Figure 110 indicates the actual appearances of (1) a woven glass fabric
in the form of a 10 ft * 4 in. bag that was cleaned by collapse and re-
verse flow under normal field conditions, and (2) a partially cleaned
filter panel cleaned by hand-flexing. Both photographs show that the
dust has dislodged as slabs or flakes from the interface region with
little indication of spallation from the surface layer. The special
fluorescent tube mounted within the bag reveals the high degree of light
transmittancy (and the minimal residual dust holdings) in those areas
from which the dust has dislodged. Although the use of surface rather
than transmitted light does not permit the same sharp light contrast,
the presence of two distinct surfaces is indicated and the weave struc-
ture is clearly displayed on the cleaned, central section of the panel.
As shown earlier, the residual dust holdings are small, uniformly distribu-
ted and not strongly dependent upon the type of dust or woven fabric. The
uncleaned portion represented by the area fraction a has a drag value of
S based upon the filter resistance just before cleaning. The cleaned
fraction, a , displays the characteristic residual drag, S which, for
purposes of simplification, may be defined by S rather than S . There-
E R
fore, given the initial and final filter dust holdings or_ the fraction of
cleaned filter area, the average effective drag, S' for the two element
fii
system immediately after cleaning can be expressed by the equation:
^ + /=(?) <38>
c u
Since K2, in theory, depends only upon particle and fluid properties it
should not vary with a fixed dust/fabric system. However, tests performed
during this study and many past studies have demonstrated that K, may often
increase with filtration velocity. The increase in K is attributed mainly
280
-------
Fly ash dislodgment from 10 ft x 4 in. woven
glass bag with inside illumination showing
cleaned (bright) areas
x
Partial fly ash removal from woven glass (9 in.
6 in.) test panel with surface illumination showing
cleaned central region
Figure 110. Appearance of partially cleaned fabrics
-------
to a decrease in cake porosity that results from higher particle momentum
when the particle strikes the filter. For the fly ash/glass fabric system
investigated in this study, K can be expressed by the empirical equation:
1/2
K = 5.95 V (English units) (39)
If the specific resistance coefficient, K2, is defined as a function of
velocity, Equation 40, a simple iterative solution based upon the following
equations can be used to predict the fabric resistance/fabric loading
relationship. Using the subscripts c and u to denote cleaned and uncleaned
surfaces, respectively, and t to depict the system parameters at the time
equals t:
P = S V + 5.95 (V )1'5 W
c c c c c
t t t t
P = S V >• 5.95 (V )1-5 W (41)
ufc u u^ ut ut
V = a V , + a V (42)
c c u u
PC is always equal to PU and average filtration velocity, V,
inlet dust concentration, C, and the characteristic drag terms, S and S
c u
are system constants.
The average fabric dust loading after a small time change At (~ 1 to 5 min)
can be approximatedby the following equations:
Wc * Wc + Vc C At
282
-------
Then the equations listed below will indicate the new fabric resistance
at the end of the time interval At :
Pc = Sc Vc +5.95/V 5W (45)
Ct + At C °t + At ct + At ct + At
"t -f At ut + At \ ut + At/ ut + At
/— \
By substituting V = (V - V a \/a , and equating Equations
Ut + At \ Ct + At C / U
(45) and (46) the relationships between effective pressure drop and
velocity and dust holding for the two fabric surfaces are readily com-
puted for successive time increments by a simple programming operation.
The system of equations described above is suitable for describing the
drag versus fabric loading relationship for a partially cleaned, single
bag or a two bag system in which one bag is completely cleaned.
The performance of a large, multicompartment filter system can be deter-
mined in similar fashion by introducing as many equations for the pressure
and fabric loading terms as there are compartments and/or different filter-
ing surfaces in the system. In a generalized form
s = ( E ys-j) A (47)
where S refers to system drag, A, to the area of the j element and A,
to the total filtration area.
The modeling concepts described above were applied to the experimental data
shown in Figures 111 through 113. In each instance, fabric test panels
283
-------
1,000
OO
-P-
800
CM
E
v.
2
CO
CO
Ul
CE
o
CD
Jf
600
400
200
-1
—A WOVEN GLASS- FABRIC, COMPLETELY CLEANED
Q WOVEN GLASS FABRIC, PARTIALLY CLEANED
---x PREDICTED CURVE, PARTIALLY CLEANED
WOVEN GLASS FABRIC
TEST PARAMETERS
PARTIALLY CLEANED FABRIC
V=0.6I m/min
C0 = 6.9 g/m3
ac =0.485
au =0.515
Sc =102.4 N min/m3
Su = 1033 N min/m3
K2 =1-80 V/2N min/g m
0
200
400 600 800 1,000
AVERAGE FABRIC LOADING, g/m2
1,200
Figure 111. Fly ash filtration with completely and partially cleaned woven glass
fabric CMenardi Southern), Tests 71 and 72"
-------
2,000
-A WOVEN GLASS FABRIC,COMPLETELY CLEANED
N3
00
Ln
LU
O
UJ
o:
o
CK
00
Q
x
1,600 -
WOVEN GLASS FABRIC, PARTIALLY GLEAMED
PREDICTED CURVE, PARTIALLY CLEANED
WOVEN GLASS FABRIC
1,200
800
400
TEST PARAMETERS
PARTIALLY CLEANED FABRIC
V = 1.53 m/min
C0 = 4.6g/m3
ac =0.50
au =0.50
Sc =39.3 N min/m3
Su =820 N min/m3
K2 =1.142 V*/2 N min/g m
200 400
AVERAGE FABRIC LOADING, g/m2
600
Figure 112. Fly ash filtration with completely and partially cleaned
woven glass fabric (Menardi Southern), Tests 96 and 97
-------
600
—A USED SATEEN WEAVE COTTON, COMPLETELY CLEANED
—-Q USED SATEEN WEAVE COTTON, PARTIALLY CLEANED
--x PREDICTED CURVE, PARTIALLY CLEANED
SATEEN WEAVE COTTON
t-o
oo
CO
E
UJ
u
CO
co
UJ
cc
o
IT
CD
400
200
TEST PARAMETERS
PARTIALLY CLEANED FABRIC
V = 0.6I m/min.
Co =7.6 g/m3
ac =0.487
au =0.513
Sc =65.6 N min/m3
Su =832 N min/m3
K2 =1.48 V(/2 Nmin/g m
_L
-i.
J_
0
200
800
Figure 113.
400
AVERAGE FABRIC
Fly ash filtration with completely and partially cleaned sateen
weave cotton, unnapped (Albany International), Tests 84 and 85
600
LOADING, g/m2
-------
that had been uniformly loaded with fly ash were partially cleaned so that
approximately half the filter surface was stripped of its dust layer. The
subsequent experimental loading curves followed much steeper paths and only
at the higher average cloth loadings did the slope of each curve, K
approach that of the uniformly loaded fabric.
The drag value for the uncleaned area, S , is that based upon the filter
resistance, face velocity, and fabric loading immediately before cleaning.
Conversely, S is the drag value for the cleaned area only which is deter-
mined by removing completely the overlying dust layer from the fabric. It
(S ) is associated with the residual dust holding, W , for the cleaned
C R -
portion of the fabric. The fraction of cleaned area, a , and uncleaned
area can be determined by actual measurement of the cleaned and uncleaned
areas. However, it is simpler to use the following mathematical relation-
ship when the magnitude of the fabric loading before cleaning (W ) the
average dust loading (AW) added to the filter over the filtration cycle
and the true fabric residual dust holding W are determinable; i. e.,
R
W AW _ w
Wp - WR
= 1 - a (48)
From Equation (48) the uncleaned area fraction is computed as indicated.
2
When W is very large, approximately 1000 g/m , the relationship (Wp - AW)/
W provides a good approximation of a .
The curves designated by "X" on Figures 111 through 113 were generated by
the modeling equations cited previously using the input parameters shown
on each figure. The fact that the theoretical and experimental curves
agree as well as they do suggests that the hypothesized filtration process
is essentially correct.
Although it suffices for modeling purposes to treat the preliminary sub-
strate plugging and subsequent cake growth on the basis of parallel flow
through the pore array, it should be realized that normal statistical
287
-------
variations in pore dimensions and discrete fiber distribution will cause
some pores to bridge over more rapidly than others. In the event of
gross differences in pore size (or excessive filtration velocities) there
is a real possibility that complete pore bridging will never occur. The
later factor is responsible for high dust penetration and, in extreme
cases, erroneously low estimates of K^.
The modeling presented in Figures 111 through 113 is based upon the simpli-
fying assumption that the nonlinear section of the drag curve can be ig-
nored. A trial test was made, however, in which the drag versus loading
relationship was broken down into two straight lines. The initial, curvi-
linear section was approximated by a straight line having a steeper slope
than the normally linear portion of the curve. Reference to Figure 114
shows a slight shift of the predicted drag curve during the early loading
phase. Despite the fact that the model is improved, it does not appear
that much has been gained insofar as predicting average resistance is
concerned.
Dust Removal Versus Cleaning Conditions
It has been determined previously that resistance characteristics for par-
tially cleaned fabrics can be readily computed once the state of the fil-
tration surface is established in terms of cleaned and uncleaned areas.
From an operating viewpoint, however, it is also necessary that the method,
intensity and duration of the fabric cleaning process be directly relatable
to the state of the fabric surface. This means that the dust separating
forces generated by the cleaning process and the adhesive forces that
oppose dust dislodgment must be defined quantitatively.
Dust separating forces have been discussed for both bag collapse and re-
verse flow cleaning, and mechanical shaking. In the former case, it has
been assumed that the shearing force exerted at the interface between the
vertically aligned fabric and the dust cake is equal to the product of
cake areal density, W, and the local gravitational constant, g. The force
288
-------
1000
CO
CJ
E
UJ
o
I
to
tO
UJ
cc
800 -
600
400
E 200
CD
> I I I I > i
& WOVEN GLASS FABRIC, PARTIALLY CLEANED
o PREDICTED CURVE, LINEAR MODEL
Q PREDICTED CURVE, BILINEAR MODEL
200 400 600
FABRIC LOADING (W), g/m2
800
Figure 114. Resistance versus fabric loading for partially-loaded fabric, measured and
predicted (using linear and bi-linear models) Test 72
-------
causing dislodgment is also equal to the tensile force (W x g) exerted at
the interface when the dust deposit is attached to the underside of a
horizontally aligned filter. It is assumed that the interfacial adhesive
force is approximately equal to the separating force at the instant of
cake detachment. Thus, measurement of the areal density of a dust deposit
at its dislodgement location constitutes a simple method to estimate inter-
facial adhesion levels.
If the areal density and the interfacial adhesive forces were uniform over
the fabric surface, all dust would dislodge as soon as the areal density
reached the critical level. Actually, all laboratory and field measurements
indicate that only partial dust separation is attained for a fixed separat-
ing force. Therefore, one concludes that for a multiplicity of reasons
the adhesive forces are distributed in some statistical fashion over the
fabric surface . Furthermore, there is reason to expect that the applied
separating forces are not distributed uniformly over the fabric surface.
Qualitative observations during the current test program indicated that
a vertical gradient in areal density existed with a slightly denser deposit
on the lower surface of the fabric. It is expected that this gradient will
increase as the range of particle sizes (or a ) increases for the entering
o
aerosol.
What is actually determined by laboratory measurements is an "effective"
gradient for the distribution of interfacial adhesion forces. Computations
are given in Table 41 showing the equivalent dust separating force for
each of the tests summarized in Tables 24, 25 and 26, Section VIII. The
separating force for each fabric loading is the product of fabric loading
f\
before cleaning (W ) and the local acceleration (980 dynes/cm ).
The fraction of cleaned surface area, a , associated with each dust removal
c
value has also been calculated for each test in accordance with the
expression:
290
-------
Table 41. FRACTION OF FILTER SURFACE CLEANED VERSUS DUST SEPARATION
FORCE, GCA FLY ASH WITH WOVEN GLASS FABRIC (SUNBURY TYPE)
Run
No.
P-2-1
P-2-2
P-2-3
P-2-4
P-A-1
P-4-2
P-4-3
P-4-4
P-4-5
P-5-1
P-5-2
P-5-3
P-5-4
P-5-5
P-5-6
)ust separation
force,3
dynes /cm2
92
41
53
71
68
42
53
62
69
69
64
68
69
69
70
Fraction of
fabric surface
cleaned ,b
ac
0.67
0.09
0.10
0.19
0.6
0.12
0.12
0.21
0.23
0.32
0.20
0.21
0.24
0.23
0.20
Run
No.
P-3-1
P-3-2
P-3-3
P-3-4
P-3-5
P-3-6
P-3-7
P-3-8
P-3-9
P-3-10
P-3-11
P-3-1
P-3-1
P-3-1
P-3-1
P-3-1
P-3-1
P-3-1
P-3-1
P-3-1
to
P-3-1
Xist separation
force,3
dynes/cm
71
65
69
70
74
76
77
76
75
76
79
80
80
85
72
74
76
75
76
avg. 75
Fraction of
.abric surface
cleaned, k
ac
0.31
0.18
0.21
0.15
0.17
0.20
0.22
0.22
0.20
0.16
0.18
0.20
0.19
0.26
0.19
0.19
0.23
0.19
0.17
0.20
aDust separation force = (W)(g) prior
sive force when dust layer detaches.
bDust detached from cleaned area held
arating force.
to cleaning. Equal to interfacial adhe-
by adhesive force less than applied sep-
291
-------
.
o = 1 —
WT-WR
where W and W refer to the average fabric loading before and after
T R o
cleaning and W_ is the characteristic residual loading (50 g/m ) for the
R
fly ash/glass fabric systems.
The statistical nature of adhesive force distributions has been demon-
strated by many present and past tests ' in which successive repetitions
or continuations of collapse-reverse or mechanical shaking has led to
increased dust removal. Test results for various dust and fabric combina-
tions show that a limiting removal level is attained after about six re-
petitive collapse and reverse flow treatments or 360 individual mechanical
shakes, Figure 115 and Table 42. In the latter case, the bag was shaken
at a frequency, f, of 8 cps with a 1 in. amplitude, A, (horizontal dis-
placement) for the shaker arm such that the approximate maximum acceler-
2
ation imparted to the dust layer was 5 gs (~4900 cm/sec ).
It should be noted that the separation forces generated by mechanical
shaking are also dependent upon fabric loading, W. However, the "g"
factor, which is now governed by the shaking parameters, is much greater
than that afforded by gravity separation. Average acceleration a was
estimated by the relationship:
a = k4TT2f2A (49)
where k ranges from 0.7 to 0.8 for the previously cited amplitude and
frequency conditions.
The most important observation with respect to multiple cleanings is that
beyond a fixed number of collapses (or flexes) or a fixed number of
shakes no further dust removal is attained for a specified energy input.
292
-------
UJ
111
TOTAL NUMBER OF COLLAPSES
6 8
CURVE FABRIC
1,2
3,4
5,6
NAPPED SATEEN WEAVE COTTON
PLAIN WEAVE DACRON
CROWFOOT DACRON
N = NEW, < 10* SHAKES,0 = OLD,2xlOT
SHAKES
WOVEN GLASS (SUNBURY TYPE)
I I I
100 200 300 400
TOTAL NUMBER OF SHAKES
500
Figure 115. Average residual fly ash loadings versus fabric type and
number of mechanical shakes (8 cps at 1 in. amplitude),
Reference 10
293
-------
Table 42. EFFECT OF NUMBER OF MECHANICAL SHAKES
ON GCA FLY ASH REMOVAL FROM SELECTED
FABRICS
10
Sateen weave cotton
Number
of
shakes
0
40
80
120
200
360
Cleaned area
fraction,3-
ac
New
_
0.11
0.22
0.32
0.37
0.43
Old
_
0.31
0.45
0.48
0.51
0.54
b
Comments
! Initial dust loadings, (WT) ,
Slew, 729 g/m2
Did, 635 g/m2
lesidual loading
tfR = 70 g/m2
Crowfoot Dacron
0
40
80
120
200
360
-
0.70
0.80
0.83
0.84
0.86
-
0.75
0.80
0.85
0.90
0.93
/Initial dust loadings, (W ) ,
iNew, 361 g/m2 T
Sold, 341 g/m2
\Residual loading
IWR = 70 g/m2
\
Plain weave Dacron
0
40
80
200
280
360
-
0.32
0.47
0.60
0.65
0.67
i J
-
0.60
0.70
0.80
0.83
0.86
(Initial dust loadings, (W ),
New, .475 g/m2 T
Old, 360 g/m2
Residual loading
W., = 70 g/m2
K.
ac = Fraction of surface cleaned to W-R level.
T-i
Cleaning accomplished by mechanical shaking of
8 cps with 1 in. amplitude. Bag acceleration
= 5 g s (4900 cm/secz).
294
-------
Only by increasing the thickness of the dust layer or by increasing the
dust layer with a concurrent increase in acceleration by inducing an
oscillating motion can a further increase in dust removal be attained.
With respect to dust dislodgement by the collapse and reverse flow process,
the precise nature of the dust separation process is difficult to describe
except for the simplified system in which the dust cake "hangs" from the
underside of a horizontally mounted filter (not a conventional field
procedure).
Figure 115 shows that the type and service life of a fabric affect signi-
ficantly the degree of cleaning for a fixed energy input. For immediate
reference, the relevant properties of all fabrics discussed in this section
have been summarized in Table 43. The presence of bulk or staple fiber
enhances the interfacial adhesion, thus making dust release more difficult.
Additionally, the gradual "shedding" of staple with extended filter
usage appears to decrease the adhesive bonding as suggested by the "new"
and "old" values for average residual loadings. It is emphasized, however,
that reduced average residual loadings may not indicate lowered filter
resistance and decreased penetration. Examination of cleaned fabric shows
that a large fraction of the bulk staple is attached to portions of the
fill yarn that do not enter into the filtration process because of yarn
proximity. Thus, shedding of the superficial staple reduces surface load-
ings in this area without any change in the interstitial region which may,
in the long term, experience reduced flow cross section due to gradual
plugging. It is necessary to assume first that the fabric loading is
already at the level where it produces a separation force equal to that
of the local adhesive force. When air flow is diverted from the bag, a
bending ensues that produces cracking or checking of the surface because
the bending moment of the dust layer has been exceeded. As reported
earlier, many repeated flexings produce a crack pattern whose boundaries
relate closely to the weave structure, Figure 28, Section V. Observations
295
-------
NJ
Table 43. PHYSICAL PROPERTIES AND. PENETRATION DATA FOR WOVEN FABRIC EXAMINED FOR
DUST CAKE ADHESION
Fabric
Woven glass
Woven glass
Woven glass
Woven Dacron
Woven Dacron
Cotton
Weight,
oz/yd2
9.2
10.5
8.4
10
10
10
Weave
3x1 Twill
3x1 Twill
3x1 Twill
1/3 Crowfoot
Plain
Sateen weave
napped
Yarn count,3
w/in. x f/in.
54x30b
66x30b
53x51b
71x51b
30x28
staple
95x58
staple
Frasier
permeability,
ft3/min at
0.5 in. HaO
42.5
86.5
45-60
33
55
13
Mfgr. and
code number0
MS, 601 Tuflex
WWC,
No. 640048
AI
Q53-875
AI
No. 865B
AI
No. 862B
AI
No. 960
Application
Field9
Sunbury , Pa .
Field8
Nucla, Colo.
Field7
Bow, N.H.
Laboratory,
GCA
Laboratory,
GCA
Laboratory ,
GCA
Average
penetration,
percent
0.08
0.16
0.38
0.07-0.29
0.05-0.23
<0.001
Yarn count warp (w) yarn/in, x fill (f) yarns/in.
Multifilament warp yarns, bulked fill yarns.
CMS - Menardi Southern
WWC - W.W. Criswell
AI - Albany International
-------
of the dust dislogement process indicated that collapse alone led to rela-
tively low release rates compared to the amount detached after reverse
flow was initiated (5 to 10 percent). The role of the reverse air flow
appears to be that of applying a mechanical thrust to a slab or flake of
dust whose bonds to the fabric have already been severed by shearing action.
Since the dust layer is vertically aligned in commercial filter systems,
it is necessary to assume that local curvature of the fabric surface be-
tween anticollapse rings (if used) coupled with a statistical distribution
of adhesive forces is sufficient to initiate the dust separation process.
Once a preliminary release takes place, a cascading or avalanche effect
appears to take place until the maximum removal is obtained for a fixed
set of cleaning parameters.
Based upon the dust removal data presented in Tables 24 through 26 and
Figure 115, the fraction of the fabric surface cleaned (a ) and the esti-
mated separating forces, F , have been computed for these tests, Tables 41
s
and 42. As stated previously, it has been assumed that all dust dislodged
from the fabric was held by a force less than or equal to the applied
separating force. The dust removals noted for the collapse and reverse
flow tests actually reflect the results of several collapses for each
element of the fabric surface. For example, both the P-3 and P-5 test
series indicated that after five or six filtration cycles the dust depo-
sition and dust removal rates came into equilibrium. This finding is con-
sistent with the results of the special tests shown in Figure 91, Section
VIII, that indicate no appreciable gain in dust removal after six succes-
sive collapses between filtration intervals. It is assumed that a layer
of dust that has not separated until the sixth filtration cycle, Table 27,
has essentially the same adhesive properties as those for a similar dust
layer that has experience six successive collapses.
Figure 116 shows a graph in which the fraction of cleaned area, a^ is
plotted against the dust separating force, Fg, immediately before cleaning.
The fraction of cleaned area also represents the fraction of the fabric
surface for which the interfacial adhesion, F^ is equal to or less than
297
-------
80 50 100 200
INTERFACIAL ADHESIVE FORCE, dynes/cm2
500
Figure 116. Estimated distribution of adhesive forces for woven glass
fabrics and one Dacron fabric with coal fly ash
298
-------
the applied separating force. Therefore, the abscissa can also be inter-
preted as the interfacial adhesive force. The results of all current
pilot tests, Tables 24 through 26, as well as those for GCA field measure-
ments at coal burning utility boilers located in Sunbury, Pa., Nucla, Colo.,
and Bow, N.H. are presented.
Two additional data points are given that are based upon laboratory measure-
ments with a fly ash/woven Dacron fabric system. Similarities in weave,
fabric density, bulk fiber content and penetration characteristics, Table 43,
suggest that the Dacron behavior at the indicated adhesive force level
might simulate glass fabric performance. Unfortunately, there was not
sufficient time within this program to carry out a rigorous study of dust
dislodgment phenomena. Hence, we have used as much peripheral information
as possible to support the existing measurements.
Noting that the field tests represent independent observations, it appears
that laboratory pilot tests with single bags provide a very reasonable
estimate of field performance insofar as dust removal is concerned. It is
also concluded that mechanical shaking and collapse systems can be treated
in similar fashion just as long as the acceleration imparted to the dust
cake can be defined. For example, if one elects to initiate cleaning at
2
the Nucla station after the average fabric loading has risen to 850 g/m ,
the curve shows that 38 percent of the cleaned compartment surface will
2
have been cleaned to its true residual level of 50 g/m . The predicted
area fraction cleaned for the Sunbury and Bow operations based upon the
measured average residual dust loadings, also fall within a few percent
of the actual values.
It must be remembered, however, that these correlations apply only to fly
ash/glass fabric systems. The magnitude and distribution of estimated
interfacial adhesive forces for other fabrics are indicated in Figure 115.
Although one can make qualitative predictions as to what adhesive proper-
ties might be anticipated for various dust/fabric systems, there do not
299
-------
exist sufficient data or working theory to make any generalized predic-
tions. The problem of predicting adhesive properties, even for single
element systems; i.e., particle to particle, particle to fiber or particle
to plane is a highly complex one because several factors acting in con-
cert such as particle, fiber and gas properties in the presence of external
field forces contribute to adhesion and cohesion.
A fairly extensive review of particle adhesion phenomena as applied to
fabric filtration was prepared by Billings and Wilder. In all but a few
cases, the major research in this area was restricted to analyses of the
adhesive or cohesive forces betweeh a single particle and other objects;
i.e., particles, fiber or plane surfaces. The rather discouraging aspect
of the many reported measurements is that the use of radically different
instrumental approaches coupled with a lack of clarity in defining what
fraction of adhering material is removed by a given force and the doubtful
nature of the "monodispersity" of some particle distributions makes dif-
ficult any quantitative comparisons among the various studies. Many in-
vestigators indicate that the range of measured adhesion for uniformly
sized particles can be described by a logarithmic-normal distribu-
O Q O /, O fi
tion ' with perhaps a 20 to 100 fold difference in force between the
o £
1 and 99 percentiles. Data excerpted from a study by Boehme are
presented in Figure 117. Atmospheric humidity has been shown to exert
a significant effect on adhesion with respect to large ~ 100 urn
1 f\ *3 / *3C *3 7
particles. ' ' ' It appears, however, that the observed increase
in adhesion over the 50 to 100 percent R.H. range is relatively small
for particle diameters less than 15 ym, Figure 118. Examination of
Figure 118 also suggests that the physical nature of the particles
and/or fiber also have a strong influence on adhesion. As far as natural
charging is concerned, the magnitude of the image forces arising from
100 electrons per 10 ym particle appear to be many orders of magnitude
i
less than the noncharge-related adhesive forms. Charged to their
maximum potential, the electrical attraction is only roughly the same as
that for natural adhesion forces, approximately 0.5 dynes.
300
-------
Co
O
cc
% 1
3 I
U. S
O Itl
r-i K
bo
i 4
P H
O CC
< <
a: a
10 ao
ADHESIVE FORCE FQ, mi Hi dynes
Figure 117. Adhesion of spherical Fe particles of 4 ym diameter to Fe
substrate at room temperature in air as a function of
applied force (from Bohme, et al., Reference 36) and Reference 1
-------
(A
o>
c
•o
E
~o
t"6
UJ
c_>
£t
£ 4
0
CO
UJ
I 0
Q ^
< <
<
O Q
% c
ADHESION, dynes
p o r .- ~
o 4* oo ro o> o
— a*
.
>—6
6
-Dp
Up
-^>
•T -^
0 20
R
A. From
^0-"
= 15
= 10.
Oj.
— o-f
Q
3yu.n
i^nas:
•«•— «^
o
i
— o
=o
) 40 80
j
ft
k^
• •
«
e\
7
3^,-fc
^/'
J^
0=
CT
) <
/
o
-fr-
r^>-
40 60
ELATIVE HUMIi
Corn, Reference
u>
4)
C
>»
E
J
u ^
o 6
O
u.
1 4
UJ
I
0
< 2
z
P. c
/
/
--
1)
^
-0-J
31 TY
s 37
/
/8
o
H
>— «
-06.5^1. i
£ iOOp,m
61.
5/im
-ft i5^.m
i
80 100 120
, percent
, Pyrex-Pyrex
r PARTICLE Sl2E=l0.4^im C
FILTER VELOCITY =0.42 n
£
_• }
I
£
~^^
r*--
1
yjf
~i*
/..
-* r
GLASS (50/t
/|
POLYESTER
^j(42Mm
'OLYAMIDE -
.(50am)
) 20 40 60 80 100
RELATIVE HUMIDITY,
percent
B. From L«ffler, Reference 35,
granular quartz on 50 ym
nylon fiber
s RELATIVE HUMIDITY,
percent
C. From LSffler, Reference 36,
granular quartz on indicated
fibers
Figure 118. Effect of particle size and relative humidity on adhesion
for various materials (Reference 1)
302
-------
37 , , 28,35
Corn and others have developed their adhesion theories on the
premise that liquid condensation at the interface between particles or
particles and other surface geometries produces strong capillary forces
that resist separation. Minimum and maximum forces for a particle-to-
particle system and a particle-to-plane system, which differ by a factor
of two, can be estimated by the following relationship;1
2
Fa = 10 dpl dp2/r(dpi + dp2J (Particle to particle)
2
Fa - 10 dp (particle to plane)
Given a 10 ym particle without specification as to physical nature of the
particle or deposition surface, the estimated adhesive force (with dp
expressed in centimeters) is about 0.1 dyne. If a dust cake composed of
10 um particles were in contact with a flat surface, the number of in-
2
dividual particles in contact with a 1 cm surface would be of the order
fi ^9
of 10 and the resultant interfacial force would be 10 dynes/cm . Even
in the event of much greater porosities, the magnitude of the internal
3 2
cohesive forces would probably exceed 10 dynes/cm .
On the other hand, the best estimates of cake adhesion to a fabric based
2
upon the present study indicate that 50 to 150 dynes/cm is the approxi-
mate range of interfacial adhesive forces. The disparity in adhesion
between intracake and interfacial structures is attributed to the greatly
reduced contact area between dust particles and fibers due to the in-
herent openness of the fabric. The above analysis appears to support
the observations that the cake detaches at the dust/fabric interface.
The preceding review provides at best only a qualitative treatment of
the factors that affect adhesion. It does, however, point out that un-
less particle charging is induced by outside means and unless electrical
fields are impressed across the filter media, that electrical phenomena
should not play a major role in determining the performance of woven glass
303
-------
fabrics against cool fly ash. Based upon laboratory tests with Dacron
fabrics, relative humidites ranging from 16 to 42 percent had no dis-
cernible effect on efficiency or resistance to air flow. The good agree-
ment between present field and laboratory studies also suggests that
humidity is not an important factor with fly ash/woven glass systems pro-
vided that filter operation is maintained well above the dew point. It
is also concluded that the only way to estimate cleanability at the pre-
sent is by direct laboratory (or field) measurement. At this time the
cleaning parameters derived from Figure 116 afford the best predicting
capability.
Although the relationship between dust removal (and/or the fraction of
cleaned fabric area exposed after cleaning) and the initial fabric loading
appears to be logically defined by a probability type function, it can
also be described conveniently by the log-log plot shown in Figure 119
if the degree of cleaning is constrained to the approximate range, 5 to
60 percent of the fabric surface. Based upon present field and labora-
tory tests, the above range encompasses most observations of dust removal.
Until further data become available to refine the mathematical description
of the postulated cleaning process, it appears acceptable to use the
simpler relationship indicated below:
a = 1.51 x 10~8 W2'52 (50)
c
in which a is the fraction of filter surface from which the dust cake
c
is dislodged and W is the fabric loading just before dust dislodgement.
Application of Equation (50) is restricted to collapse and reverse flow
cleaning systems. If fabric cleaning is by mechanical shaking, the fol-
lowing relationship should be used:
a = 5.24 x 10~6 (FA)2'52
c A
304
-------
10°
o
o
u
<
cr
LU
or
o
UJ
UJ
_l
O
10 -I
O I NUCLA
O2 SUN BURY
2
I01
3 BOW
?
a =l.5lxIO"8 w2'52
I I L 1 I i
FABRIC LOADING, W-g/m2
Figure 119. Relationship between cleaned area fraction and initial
fabric loading. GCA fly ash and woven glass fabric,
see Figure 116.
305
-------
2
in which F represents the adhesive force (dynes/cm ) that must be over-
A
come by a cleaning force, F , of equivalent magnitude. When the latter
force is induced by mechanical shaking, it is defined as the product of
the fabric loading, W, and the average acceleration, a, imparted to the
bag by the shaking motion (see Equation (49)):
F = F = W "a = Wk47T2f 2A
A c
For the range of shaking frequencies encountered in most commercial appli-
cations, usually less than 6 cps, the cleaned area fraction resulting from
mechanical shaking can then be expressed as:
a = 2.23 x 10~2 (f2AW)2'52 (51)
c
FULL SCALE APPLICATIONS - MODELING CONCEPTS
Equation (50), in conjunction with the several mathematical relationships
discussed earlier in this report, are easily applied to any single bag
filter system. Special considerations are involved, however, when they
are used with multicompartment, sequentially cleaned units. Ordinarily,
it is assumed that all filter bags installed in a given compartment func-
tion in identical fashion although this may not necessarily be true de-
pending upon bag deployment and proximity and gas flow distribution.
In the following paragraphs, the cleaning process is examined for two
typical field situations, the first a filter system in which the cleaning
process is pressure actuated and the second approach wherein cleaning
takes place according to a fixed-time cycle regardless of dust deposition
rate and/or fabric loading.
PRESSURE CONTROLLED CLEANING
The ultimate selection of operating parameters for pressure-controlled
cleaning is based upon the following data inputs:
306
-------
A maximum allowable resistance across the fabric filter
prior to cleaning.
A fixed inlet dust concentration, C., and volume flow rate, Q.,
as determined by fuel burning rate,xfuel composition and excess
air rate.
A fixed average filtration velocity, V±t that has been se-
lected upon the basis of dust penetration properties for
the selected fabric.
Steady-state operation as defined by equilibrium between
dust collection rate and dust removal rate.
Sequential cleaning of all compartments followed by extended
filtration with all compartments on-line until the pressure
limit is again reached.
The first step is to determine what fabric dust loading, W , corresponds
to the maximum allowable operating resistance, P . Since the filter system
W
will operate for lengthy time intervals, ~2 to 3 hours, between cleaning
cycles (roughly 30-minute duration) the dust distribution over the fabric
surface will be nearly uniform just before initiation of cleaning. There-
fore, the term W may be estimated from the characteristic pressure or
drag curve for the specific dust/fabric combination for which the terms
S and K have already been defined; e.g., Figures 43 and 54, Section VII,
E 2
and Table 32. Assuming that the fabric undergoes a conventional collapse
and reverse flow cleaning, the amount of dust dislodged and the fraction
of cleaned fabric area exposed, a^ is readily estimated from Figures 90
and 119 and Equation (50).
The actual dust removal associated with a given ac level is determined
from the relationships:
AW = Wp - WR
and
W= (1 - a)(Wp - WR) +WR (53)
307
-------
Where W is the average residual dust holding after cleaning, W , the fabric
R *-
loading at the limiting pressure, and W the characteristic residual load-
ing of the cleaned fabric surface only. The total dust removed, ZW, fol-
lowing the sequentially cleaning of n compartments expressed in terms of
fabric filtration area then becomes:
ZW = n (W_ - wr) (54)
r K
Since the amount of dust dislodged during the cleaning process must equal
the quantity of dust deposited between successive cleaning and filtration
cycles, the combined operating time, Zt, for the overall cycle is estimated
from the following relationship:
Zt - n (Wp - W^/C^ (55)
and the time interval over which all filters are on-line is computed as :
t 1 . = Zt - nAt (56)
on-line v '
where n is the number of compartments and At the cleaning interval for
each compartment.
Equations (52) through (56) provide a practical estimate of the necessary
cleaning frequency if one is not to exceed the specified operating pressure.
An improved estimate may be obtained, however, by noting that at the in-
ception of the cleaning process only the first compartment has a fabric
loading of Wp. The remaining compartments will acquire succesive incre-
ments of loading as cleaning continues throughout the cycle such that the
last or n compartment to be cleaned will have an increased fabric loading,
EW; i.e.:
ZW = Wp + (n-1) C V^ At (57)
308
-------
During the actual cleaning period, nAt, all filtration must take place
through the (n-1) on-line compartments. Hence, assuming that overall
system volume flow is relatively insensitive to small pressure changes,
the average velocity throughout the on-line compartments must increase
by the ratio n/n-1; i.e.:
V^ = V± (n/n-1)
Equation (57), therefore, is reduced to the form:
ZW = Wp + n C± V± At (58)
One can infer from the above relationship and Figure 119 that more dust
should be removed and, hence, more filtration surface exposed as the
cleaning cycle progresses.
To avoid undue complications in the estimating process and yet take into
account the gradual increase in fabric loading, the original W value
based on pressure limitations has been modified as follows:
wp = wp + n c± v± At/2
where V. is again the average face velocity when all filter compartments
are operating. It is emphasized that use of the procedure described above
assumes that the transient pressure increases associated with the start
of the cleaning process will not reduce the gas handling capacity of the
boiler fans. When the first compartment is taken off line for cleaning,
the system flow must be accommodated by the remaining n-1 compartments
leading to an automatic rise in system resistance.
If the upper pressure criterion is based upon the peak transient values,
the following approach must be used.
309
-------
The increment of dust added to the remaining on-line (n-1) compartments
while cleaning the first compartment can be expressed as:
AW = C V. (n/n-1) At
i i
and the corresponding increase in pressure, AP, over that observed just
before cleaning appears as:
AP = K^ V (n/n-1) AW (60)
Because K has also been shown to be velocity dependent; i.e.:
K = 1.8 (V)'2 (metric units), Equation (60) must be further modified:
AP = 1.8 C± F(V ) (n/n-1)]2'5 At (61)
When the indicated pressure is now specified as a not-to-be-exceeded value,
the pressure used to del
cleaning is defined as:
the pressure used to determine the required fabric loading W just before
PT7 = P - AP (62)
W max > '
Once the term P is determined, the estimate of cleaning frequency is
carried out according to the previously described procedure, Equations (52)
through (56).
It should be noted that in the limiting case, a pressure controlled clean-
ing system with intermittent cleaning cycles will reduce to a continuously
cleaned system with back-to-back cleaning cycles when dust removal during
the cleaning cycle equals that deposited during the same period. If the
deposition rate exceeds the capability of the removal cycle, a new, higher
pressure equilibrium automatically evolves. In the case of cleaning by
bag collapse and reverse flow, the increase in surface loading provides
the added dislodgement force. Where mechanical shaking is employed, it
310
-------
might be possible to avoid a pressure increase altogether by increasing
either the amplitude and/or frequency of shaking.
TIME CYCLE CLEANING
Under conditions of constant flow and constant loading, the behavior and
analysis of filtration systems cleaned on either a pressure or time control
cycle would be the same. During a variable loading process, initiation of
cleaning during periods of low inlet loading may lead to undesirably high
outlet concentrations due to loss of dust cake. To a certain extent,
however, dust removal at lower cloth loadings (and lower operating resis-
tances) is significantly lower with both collapse and mechanically shaking
because the dust layer itself contributes to the separating forces. Hence,
the impact of overcleaning may not be as severe as anticipated.
In the following section, the cleaning versus resistance parameters are
examined with respect to a Sunbury type filtration system in which the
cleaning cycles are repeated sequentially. The analysis of the above sys-
tem is carried out on the premise that C., V. and the collapse and reverse
flow process per se are constant terms. Hence, it can again be stated
that once steady state conditions are established, the total- quantity of
dust dislodged over a complete cleaning cycle (each bag cleaned once) must
equal the amount of dust deposited over the same time interval. The latter
amount, AW, is again determined as:
AW = n C± V± At = C± V± Et
where n is the number of compartments and At the nominal time between
successive cleanings. At the start of any filtration cycle, the fabric
loading for the compartment about to be cleaned can be expressed by Wp,
which, in the present case, is an unknown quantity. The fraction of
cleaned fabric area exposed, a , can be described as indicated earlier
~~ ' ' C
by the relationship:
311
-------
W - AW - W
-
The term a is also definable by the relationship:
c
a = 1.51 x 10 8 W,,2'52 (51)
c "
By combining Equations (63) and (51) , a relationship is obtained that
allows solving for Wp:
W 2'52 (Wp - WR) = 6.62 x 107 AW (64)
If W is significantly greater than W , 10 times or greater, Equation (64)
can be reduced to the simple form:
Wp3'52= 6.62 x 107 AW (65)
Having determined W , the magnitude of the cleaning parameter, a , can be
estimated from Equation (51). Similarly, the equilibrium pressure and
drag associated with all "n" compartments in operation are determined from
the previously established performance data for the dust/fabric combina-
tios of interest.
The maximum pressure level displayed during the cleaning cycle will again
take place when one compartment is taken off-line for cleaning. Although
Equation (61) serves to indicate the increase in resistance, AP, it should
be noted that the fabric dust loading computed by Equations (64) or (65)
applies only to the compartment just taken off-line. The remaining com-
partments through which all flow is diverted have instantaneous fabric
loadings that range from:
312
-------
wp " ci vi
to
wp - ci vi zt
for the next and last compartments to be cleaned in sequence and the time
interval, Zt, cited above represents the total elapsed time for the clean-
ing cycle.
Thus, as a reasonable approximation, the average fabric loading just before
cleaning, W , can be expressed as:
Wp = Wp - C^ V_^ — (n/n-1) (66a)
The resistance corresponding to the W level then becomes:
\ = PE + K2 [ WP - °i Vi T n/n-1] Vi (n/n-1) (66b)
The term P in Equation (66b) , which is defined as the effective resis-
E
tance, is related to the effective drag, S .
EI
If there are many compartments in the system, the maximum or peak resis-
tance, P , occurring when one compartment is undergoing cleaning may nc
max
be much greater than that predicted by Equation (66b).
If there are only a few, approximately five, compartments in the system,
it might be safe to design on the basis of the maximum expected pressure,
P , in the system; i.e.:
max J '
p = P + K W V. (n/n-1) + AP (66c)
max E 2 P i
where W and AP are determined by Equations (66a) and (61), respectively.
P
313
-------
When several compartments are involved as with the Sunbury system, the
difference between maximum and minimum pressures becomes relatively
2
small, approximately 2.5 to 2.75 in. water (550 to 687 N/m ) without the
introduction of reverse air. Reverse air flow with its added volume in-
2
crement further increases the pressure range; i.e., 550 to 750 N/m .
314
-------
SECTION X
PREDICTION OF FABRIC FILTER PENETRATION
In this section, the development of a new model to predict the particle
collection characteristics of woven fabric filters is discussed. The
model is intended to describe the behavior of fabrics in which at least
the fill yarns are spun from staple fibers or are made up of bulked
multifilament yarns. In both cases, many loosened, discrete fibers pro-
trude into the interyarn spaces (or pores) thus forming a convenient sub-
strate for initial dust cake formation. For present purposes, the appli-
cation of the model is directed mainly to woven glass fabric filters
used for the collection of coal fly ash. Thus, the approximate pore
structure shown in Figure 99, Section IX, is the one for which particle
collection characteristics have been modeled.
As indicated in the literature review, most techniques for estimating filter
collection efficiency apply to low porosity, bulk fiber filters or felted
media. They are not intended for use with fabric filter systems in which
particle capture occurs as the dusty gas passes through a parallel array
of pores or channels whose boundaries are defined by the specific weave or
interlacing of the warp and fill yarns. Therefore, syntheses of the type
23
attempted for fabric filters by Fraser et al., the latter based upon a
highly modified single fiber/single particle theory, are not successful
except for describing closely replicated filter systems.
On the other hand, treatment of collection on the basis of particle capture
by obstructed or unobstructed pores (the obstructions consisting of low
porosity, bulk fiber plugs or screens) and by a dust cake composed of the
315
-------
collected particles appear to provide a satisfactory means for com-
pletely describing the particle collection characteristics of a woven
fabric filter.
PARTICLE CAPTURE BY UNOBSTRUCTED PORES
Although it would be highly desirable both from the performance and
analytical viewpoints that (1) all filter pore dimensions be identical
and (2) that any fiber substrate bridging the pores be uniformly dis-
tributed, a real fabric filter may show considerable deviation from the
ideal pattern. In the former (ideal) case, the substrate deposition and
bridging processes will proceed in parallel. Conversely, the imperfec-
tions encountered with real filters will lead to some sequential bridging
of pores although the latter process must occur over a brief time span
if the filter is to provide satisfactory performance. The more serious
deviation or defect is where the loose, interpore fiber substrate fails
to bridge all or part of the pore opening. Depending upon the dimensions
of the unobstructed opening and the pressure gradient across it, the
initial gap must either be bridged during the early stage of filtration
before pore velocities become excessive or fail to be bridged and thus
constitute a permanent opening or pinhole leak in the filter. Within
the context of this report the only distinction made between a pinhole
and a pore is that the pinhole or "see-through" opening is either an
oversized pore or a pore that contains no fiber substrate or plug.
In some cases, it is suspected that a tenuous bridging of the pore open-
ing by the fiber substrate at the onset of filtration may actually rupture
because of the aerodynamic stresses induced by the dust deposit on and
within it. The microscopic examination of several pinholes on heavily
loaded fabrics, Figure 60, Section VII, shows only the bounding yarn sur-
faces and no loose fiber structure whatsoever. Microscopic measurements
of dust accumulation about these pinholes in conjunction with the esti-
mated air volumes passing through these openings also suggest that a
316
-------
pinhole is a very poor fly ash collector, with actual efficiencies in the
10 percent range or lower.
The above observation led to the conclusion that a rigorous analysis of
the dust collection capability of a pore was unjustified insofar as woven
glass fabrics and coal fly ash collection are concerned.
The special tests described in Figure 37 indicated that the rate of dust
accumulation within the pores of a plain weave, mono filament screen, was
very slow. Furthermore, a declining rate of resistance rise coupled with
the fact that previously obstructed pores "blew out," confirming the pin-
24
hole plug releases described by Leith et al., suggested that collection
efficiency would soon fall to negligible levels, Figure 38.
Despite the fact that the openings were slightly larger than the 170 ym,
nominal pore size for the Sunbury fabric, it was very evident from the
5-minute photograph, Figure 37, that very little of the approaching dust
o
load, ~200 g/m , had remained on the fabric. At test completion, the
screen filter retained only 15 percent of the average input loading. The
5-minute photo relates to a free area of roughly 10 percent whereas the
final 30-minute picture shows less than 5 percent free area. Hence the
pore velocity, which was roughly twice as large in the latter case, also
exerted a greater entraining force on the previously deposited dust.
A different approach for predicting pore capture was based upon the rela-
29
tive efficiencies reported by Fuchs for extraction sampling from stagnant
air masses. Figure 120 shows fractional particle size recoveries with a
constant sampling velocity of 6 m/sec at several air-stream velocities.
Those values that relate to pore penetration appear at zero flow field
velocity. They indicate that the sampling probe will capture 95 percent
or more of all particles equal to or less than 15 ym. In the present si-
tuation, the acceleration of the air velocity from 0.61 m/min at the filter
face to roughly 20 m/min at the minimum pore cross section is analogous to
317
-------
29
withdrawing a sample from stagnant air. According to Fuchs, when the
air velocities are low relative to the sampling velocities, the approach-
ing streamlines are either straight or slightly convex with respect to the
axis of flow. Therefore, minimal sampling losses should be expected under
stagnant flow conditions. The maximum losses should occur in the region
where the sampling velocity is roughly twice that of the flow field vel-
ocity. Particle losses are minimized when isokinetic sampling conditions
are attained.
no
u
£ 100
a
u
z 90
UJ
o
t 80
| 70
I-
o
" 60
50
23456
PLOW VELOCITY, m/s«c.
Figure 120. Efficiency of sampling an aerosol from a variable
velocity flow field at a constant sampling velo-
city of 6 m/sec.
Although it is not proposed here that a highly anisokinetic sampling process
is an exact replication of a dust laden gas stream converging to pass through
a fabric pore, the similarity was considered sufficient to justify treating
open pores or pinholes as non- or very-low efficiency collectors.
An extensive series of measurements involving simultaneous gravimetric
sampling of inlet and outlet concentrations and condensation nuclei (CNC)
measurements for filter effluents are given in Tables 19 and 20, Section VII.
318
-------
The CNC data have played two roles in the present study. First, despite
the fact that CNC data do not represent absolute values, they do provide a
relative measure of rapid changes in effluent concentration as filtration
progresses.
More importantly, it was observed that the outlet nuclei concentrations
related directly to the measured outlet mass concentrations as shown in
Figure 86, signifying that all dust particle sizes were collected equally
well by the filter. Since this observation contradicts accepted filtra-
tion theory, which dictates preferential collection of the larger particle
sizes, an explanation was sought for this inconsistency.
First, it was noted that insofar as concurrent upstream and downstream
cascade impactor measurements were concerned, field and laboratory tests
showed no significant differences between the respective mass distributions.
In the case of Nucla measurements, the coarseness of the inlet dust coupled
with the fact that significant gravitational and inertial losses were pos-
sible between the upstream sampling point and the filter face appears to
explain the size reduction in the effluent dust. Although the slough-off
of agglomerated particles from the rear filter face can lead to a coarser
downstream size distribution than expected, test measurements, Section VII,
suggested that agglomerate slough-off can only partially explain the ob-
served downstream size parameters.
However, when the total number of pinhole leaks were actually measured in
conjunction with an estimate of the air volume passing through them with
an assumed 100 percent penetration, the predicted filter efficiency values
were very close to the actual measured values, see Section VII. Volume
flow through the pinholes was based upon the observed dimensions as deter-
mined microscopically, the measured filter pressure loss and the flow
versus resistance parameters developed from special permeability tests,
Figure 21, Section V.
319
-------
The conclusion drawn from the tests cited here was that the downstream
effluent for the fly ash/glass fabric systems, was essentially that which
had passed untreated through pores or pinholes in the filter.
Because of the very low resistance levels for unbridged pores, a very
small fraction of the total filter surface in the form of unbridged pores
will cause a relatively large quantity of gas to pass through them. An
extreme case noted for a Dacron fabric, Section VII, showed that a frac-
-4
tional pore area of approximately 10 resulted in 20 percent fly ash
penetration.
The fraction of the inlet aerosol that actually passes through the bulked
fiber region and the superimposed dust layer is expected to follow the
classical filtration rules as discussed in the following paragraphs. From
a very practical perspective, however, the contributions from the above
sources are usually very small compared to the dust fraction conveyed by
the air penetrating the pores. As discussed in a later section, direct
pore penetration accounts for nearly all the effluent with fly ash/woven
glass filter systems.
PARTICLE CAPTURE BY BULK FIBER SUBSTRATES
The appearance of the pore structure for clean (unused), woven glass fabric
has been presented schematically in Figure 28. Microscopic examination
of the filter surface during the filtration process shows that the initial
dust accumulation is confined almost entirely to the bulked fiber regions
with no buildup upon the multifilament yarns until the interfiber depres-
sions are filled. The filter loading process as viewed by microscope is
shown in a simplified sketch of the fabric surface, Figure 35, Section V.
Based upon the observed pore structure, it is believed that the initial
dust collection process is essentially that of bulk fiber or deep bed fil-
tration. Because the inlet aerosol is highly polydisperse, the preliminary
320
-------
dust deposition will be confined mainly to the upstream region of the sub-
strate such that a distinct and separate dust cake rapidly evolves. The
early dust capture process can be described quite well by classical bulk
fiber filtration theory provided that reasonable estimates can be made for
certain physical and operating parameters.
Dust penetration may be approximated by the relation:
Pn = exp
A p TIL
TT df Pf
(67)
where bed porosity is high, >_ 0.90. The terms p and p refer to bulk and
discrete fiber densities * respectively; L is the bed thickness, d the
fiber diameter and n the single particle-single fiber collection efficiency
for the particle size and particle capture mechanisms(s) of interest. In
the above case the term (1 - p/p ), or (1 - a), which appears in the denom-
30
inator of the exponent form discussed by Dennis, has been deleted. Equa-
tion (67) may also be expressed in the form:
Pn = exp [~-A Lnl (68)
where the product, A L, can be considered as the ratio of total projected
fiber surface to the filter cross-sectional area.
The key fabric properties, operating parameters and the assumptions made
in apply Equation (67) are summarized in Table 44.
The bases for the input parameters listed in Table 44 are as follows:
microscopic observations indicated that roughly 10 percent of the total
yarns, item 4, were dispersed as discrete fibers. Due to the tightness of
the 3x1 twill weave, 25 percent of the pores were effectively blocked in
both the warp and fill directions. Thus, the fabric porosity, item 5, was
reduced from 0.64 to 0.363. It was assumed that the average pore diameter
at the midpoint of the substrate region was roughly 2.5 times smaller than
321
-------
the inlet diameter at the surface of the fabric. This leads to a 6.3 times
velocity increase, item 6, within the bulked fiber region. If the bulk
fiber occupies about 50 percent of the total void volume (0.646), disper-
sion of 10 percent of the fabric weight results in a bulk density value of
3
0.241 g/cm , item 7, for the substrate. The above development is discussed
further in Appendix B.
Table 44. INPUT PARAMETERS FOR ESTIMATING BULK FIBER EFFICIENCY
1. Average fiber diameter 8.5 urn
2
2. Fabric weight 312 g/m
3. Average face velocity 0.61 m/min
4. Only 10 percent of the total fabric fiber content is dispersed as
discrete fibers within the effective pore volume.
5. Because of many contiguous yarns, the effective filter void volume
(i.e., that through which flow takes place) is reduced from 0.646
to 0.363.
6. Average air velocity through the loose fiber occupying the void volume
is increased 6.3 times due to channel shape.
7. The bulk density of the fiber within the effective pores is 0.241
g/cm .
By substitution of numerical values given in Table 44, Equation (67) is
reduced to the functional form:
Pn = exp [ - 5.84 n J
(69)
The term n was then examined with respect to particle collection by inter-
ception and impaction mechanisms which were considered to predominate as
far as mass accumulation was concerned.
The interception efficiency, n , was computed in accordance with proce-
30
dures described by Dennis:
nDI = k' *D ' \ = dp/df and k' = (2 - In Ref) (70)
322
-------
where dp and df refer to particle and fiber diameters, respectively, and
Ref is the fiber Reynolds number.
The impactibn efficiencies, ry were computed in accordance with conven-
tional procedures; i.e.:
C p d
(n ) = c P P
I 18 yf d
(71)
and the classical target efficiency curves given in the literature.7'29
In Equation (71), C is the Cunningham slip factor, p the par-
c P
tide density, v the air velocity and y the fluid viscosity.
The calculated efficiency parameter, n, for capture by either interception
or impaction and the predicted initial filter efficiencies resulting from
impaction alone are listed for several particle sizes in Table 45.
Table 45. COLLECTION PARAMETERS AND INITIAL EFFICIENCY FOR
WOVEN FABRIC FILTERS FOR FIBER PHASE COLLECTION
dp
ym
1
2
3
4
5
10
ni
~0.02
0.10
0.30
0.45
0.60
0.80
nDI
0.003
0.01
0.023
0.041
0.064
0.26
Fractional*
penetration
0.86
0.56
0.17
0.072
0.030
0.004
Fractional3
efficiency
0.14
0.44
0.83
0.93
0.97
0.996
Conservative estimates based on m alone, the larger
of the collection parameters. Note that effective n
should be greater than r\j_.
An estimate of the diffusion parameter, n', for a 0.5 ym particle by the
-L 30
approximate relation, n' = (Pe) = Dg/Vdf, indicated a value of 0.005.
Since n continues to decrease with decreasing diffusion coefficient (and
increasing particle size), it appears that diffusion collection plays a
323
-------
very minor role in the capture of those particles accounting for most of
the projected surface area, > 80 percent.
According to the predicted penetration levels in Table 45, the initial
filter efficiency is low for particles less than 2 ym and greater than 93
percent for all particles greater than 4 ym. In a relatively short time
period, however, sufficient dust will accumulate within the loose fiber
structure to significantly increase its collection capability.
In the following paragraphs, an approximate method is developed for pre-
dicting the increase in efficiency during the first few minutes of filtra-
tion based upon the filtration parameters summarized in Table 46.
Table 46. FILTRATION PARAMETERS FOR COMBINED FIBER-PARTICLE COLLECTION
-
1. Size properties for inlet fly ash aerosol, MMD @ p = 2 g/cm =
6.36 ym, og = 3.28. p
3 2
2. Projected particle surface per gram of dust = 2.366 x 10 cm /g.
3
3. Inlet dust concentration = 3.5 g/m .
2
4. Dust arrival rate at reduced pore cross section = 13.42 g/m /
min = 1.342 x 10~3 g/cm2/min.
2
5. Increase in collector surface area per cm of filter cross
, i j 4. . 3.175 cm2 dust area ,.
section for 1 minute is —~ = AA,
cm filter cross section
If one assumes that the increment of collector surface contributed by the
dust, AA , is as effective as an equal quantity of fiber surface, A ,
Equation (68) can be expressed in the form:
Pn = exp [- (Ap + AAd) n ] (72)
This allows us to calculate the theoretical penetration levels as the dust
loading builds within and upon the fiber substrate as shown in Table 47.
Since these data refer only to the collection of 2 ym particles, it is
necessary to integrate across the particle size distribution with respect
324
-------
to impaction efficiency to determine the overall weight efficiency for the
fly ash aerosol. The simple incremental solution to this problem is sum-
marized in Table 48 for the previously cited GCA fly ash.
Table 47. PENETRATION ESTIMATES FOR A 2 pm PARTICLE AS A FUNC-
TION OF FABRIC LOADING AND INLET CONCENTRATION AT
0.61 m/min FACE VELOCITY, FIBER FILTRATION PHASE
Time
Average0
fabric
loading,
g/m2
v
dimei
• A A )ri
\sionleas
Fractional
Penetration
Efficiency
Inlet loading =3.5 g/m3
0
1
2
3
0.0
2.14
4.28
6.42
(5.84 + 0) 0.1
(5.84 + 3.17) 0.1
(5.84 + 6.34) 0.1
(5.84 + 9.51) 0.1
0.56
0.41
0.29
0.22
0.44
0.59
0.71
0.78
Inlet loading =7.0
0
1
2
3
0.0
4.28
8.56
12.84
(5.84 + 0) 0.1
(5.84 + 6.34) 0.1
(5.84 + 12.68) 0.1
(5.84 + 19.02) 0.1
0.56
0.29
0.16
0.083
0.44
0.71
0.84
0.917
aRefers to dust loading distributed over complete filter
face with 100 percent retention.
The primary reason for exploring the preceding collection concept is to
demonstrate that the proper use of classical theory in conjunction with
some simplifying assumptions provides estimates of early filtration be-
havior which are in good agreement with the actual experimental measure-
ments. For example, the initial fractional penetration values given in
Table 20, Section VII, are generally in the 0.1 range (0.9 fractional
efficiency).
It should be realized that once the fabric undergoes its first cleaning,
2
there will be a permanent residual loading of roughly 50 g/m of which
some 25 percent will reside within the active pore regions. The other 75
percent will be retained in the continuous fill yarns that form blind
pores. If the added particle surface is factored into Equation (72), the
325
-------
Table 48. PARAMETERS FOR, AND ESTIMATION OF, OVERALL WEIGHT COLLECTION FOR FLY ASH
DURING FIBER PHASE FILTRATION
Size range
ym
0.5 - 1.5
1.5 - 2.5
2.5 - 3.5
3.5 - 4.5
4.5 - 5.5
5.5 -> °°
T*/-\ f- r» 1 -f 1 tT
d
lam
1.0
2.0
3.0
4.0
5.0
Percent3
mass
in
range
9.7
10.7
9.0
6.0
6.0
55.0
Timeb
Zero
Fractional0
efficiency
0.14
0.44
0.83
0.93
0.97
0.98
Fractiond
collected
0.0136
0.0471
0.0747
0.0557
0.0582
0.5417
n 7Q1
1 minute
Fractional
efficiency
0.20
0.59
0.933
0.983
0.995
0.999
Fraction
collected
0.0194
0.0631
0.0840
0.0590
0.0597
0.55
n QTR
3 minutes
Fractional
efficiency
0.
0.
0.
0.
32
785
990
999
Fraction
collected
0.031
0.084
0.089
)
J 0.67
)
n Q-?/.
fO
Indicates mass distribution for inlet fly ash aerosol.
'Time from initiation of filtration with unused fabric.
•>
"Efficiency for specified size based on total projected collector surface.
Fraction of inlet aerosol collected.
-------
pentration values for time zero and 1, 2, and 3 minutes, respectively,
would decrease from the initially clean values, Table 47, as follows:
0.56 to 0.088, 0.41 to 0.063, 0.29 to 0.046 and 0.22 to 0.034. Again
these predicted values appear to be in line with the measured results
shown in Table 20. Additionally, it should be expected that the entrap-
ment of residual dust within the fiber matrix should improve particle
collection.
PARTICLE CAPTURE BY DUST CAKE (GRANULAR BED)
Fabric filters, which depend upon the deposited dust layer to provide the
collection capability, would always operate at 100 percent efficiency with
monodisperse aerosols if there were no defects in the supporting sub-
strate and particle contacts with adjacent particles were at a maximum
for the solid geometric array.
Furthermore, the nominal pore openings in a bed of uniformly sized par-
ticles would automatically sieve out any particles greater than about
0.15 times the diameter of the base particle size. Pursuing this analysis
to its logical end indicates that even a polydisperse aerosol will even-
tually generate a dust layer that for all practical purposes is impenetrable.
The only problem is to establish at what point particulate emissions are
no longer detectable.
Unfortunately, most real filter systems fail to demonstrate the postulated
"zero" penetration conditions because the supporting fabric either does
not permit the development of a uniformly structured bed or the gaps in the
supporting fabric may allow low level entrainment of agglomerates from
the clean air face of the filter. These problem areas will be discussed
in detail in a later section. At this point, it is appropriate to examine
the theoretical particle collection efficiency during the early stages of
cake formation to determine how rapidly particle penetration levels will
decrease.
327
-------
In the simple modeling process discussed here, it is assumed that the
supporting substrate enables the development of a dust layer of uniform
density and thickness. The selected penetration expression:
T, F 3 a L n "I ,7,.
Pn = exP I 2d (1 - a) I (73)
L c J
10 29 30
is one that has been discussed extensively in the literature. ' '
30
The term a is the ratio of bed packing density to particle density,
d is the granule (collector) diameter and n again is the single particle-
c
single granule collection efficiency for the system of interest. It
follows from the. definition of a, that the term 1 - a is the bed
porosity, e.
Unfortunately, Equation (73) applies directly only to a highly specialized
system involving a single collector (granule) size and monodisperse aero-
sol. In order to justify its use, for example, with a dust cake produced
by filtering a polydisperse aerosol, certain modifications were necessary.
As a starting point, the situation has been examined wherein the "poly-
disperse" bed is analyzed with respect to its capability to capture
specific particle sizes. Therefore, Equation (73) has been converted to
the form:
Pn =
(74)
where 1 - a refers to the overall cake porosity (which is assumed to be 0.5
based upon present experiments) and ct±, r\± and dc± are the characteristic
porosity, collection efficiency and collector diameter, respectively, for
the i intervals of the size distribution describing the inlet GCA fly
ash aerosol.
The mechanisms of particle collection by diffusion, interception and impac-
tion were then examined in accordance with procedures described for
t. j 30
granular beds
328
-------
Diffusion parameter = n' = 5 (k11)"1^3 Pe'2/3
coefficient.
where Pe is the Peclet number (dc V/D,) and ^ the diffusion
Interception parameter = n = 3 R2 (V'T1
DI D
The term k" is defined by the equation:
k" - (2 - 3 a1'3 + 3 a5/3 - 2 «2>/(l - c,5/3) (77)
and K_ = d /d
T) p c
C p d2 v
Impaction parameter = n = — - — ^ P — (78)
f c
where all terms are described as indicated for Equation (71). In
Table 49, a summary of collection parameters is given for the particle
diameters constituting the polydisperse dust layer (granular bed) and two
inlet particle diameters, 0.25 ym and 1.0 ym.
Inspection of Table 49 shows that the interception parameter predominates
for the indicated particle sizes. Although the combined effect of these
mechanisms when functioning in concert will exceed the largest indicated
value; i.e., n , it will also always be less than the algebraic sum of
the components. For present purposes, a conservative approach was selected
wherein the n values alone were used to describe the collection parameter.
The values for the term, ^ r\±/dc±) were computed for the individual volume
fractions of the fly ash distribution and for two inlet particle sizes as
shown in Table 50. The sum of the terms, E a± r\±/dc±t provides a partial
data input for use in Equation (74). By assuming various bed thicknesses,
L, and a porosity, e, of 0.5, the overall penetration values for 0.25 and
1.0 ym particles were computed, Table 51.
329
-------
OJ
w
o
Table 49. ESTIMATED VALUES FOR DIFFUSION, INTERCEPTION AND IMPACTION PARAMTERS,
GRANULAR BED COLLECTION3
Collector
diameter ,
Vim
0.25
1.0
2.0
5.0
10.0
-
Particle diameter
d = 0.25 urn
Diffusion
"i
0.35
0.17
0.10
0.069
0.035
Interception
nDI
2.27
0.25
0.058
0.018
0.002
Impaction
"I
0.08
0.02
0.01
0.005
0.003
d = 1.0 urn
P
Diffusion
"i
0.32
0.12
0.069
0.037
0.022
Interception
"DI
36.4
4.0
0.94
0.29
"0.038
Impaction
nl
0.5
0.2
0.1
0.05
0.02
a.
Refer to Equations (75), (76) and (78).
-------
Table 50. PARAMETERS USED TO COMPUTE DUST CAKE PARTICLE COLLECTION EFFICIENCY
Particle
size range
Um
5 - 0.5
0.5 - 1.5
1.5 - 2.5
2.5 - 7.5
7.5 - 12.5
> 12.5
Totals
Mean diameter
for range
dc.um
0.25
1.0
2.0
5.0
10.0
Volume of
ith fraction
0.023
0.157
0.140
0.380
0.140
0.160
1.000
<*i
0.0115
0.0785
0.0700
0.1900
0.0700
0.0800
0.5000
k"
1.32
0.75
0.80
0.42
0.80
Particle diameter = 0.25 ym
RD
1.0
0.25
0.125
0.05
0.025
HDI
2.27
0.25
0.058
0.0178
0.002
ai Vdci
x 10~4
0.1044
0.0196
0.0020
0.0006
0.00001
0.1266
Particle diameter = 1.0 ym
RD
4.0
1.0
0.5
0.2
0.1
0.05
HOT
36.4
4.0
0.938
0.286
0.0375
"i ni/
-------
Table 51. ESTIMATED OVERALL WEIGHT COLLECTION
EFFICIENCIES AS A FUNCTION OF CAKE
THICKNESS AND INLET PARTICLE SIZE FOR
FLY ASH
Bed
thickness
ym
10
20
30
100
Areala
density
g/m2
10
20
30
100
Estimated fraction penetration
Particle diameter, urn
0.25
2.2 x 10~2
5.0 x 10"4
1.1 x 10~5
io-17
1.0
io-27
-------
FLY ASH PENETRATION FOR WOVEN GLASS FABRICS
Based upon the preceding analyses, it appears that most dust emissions
from woven glass fabrics of the type commonly used for hot fly ash fil-
tration are the result of direct penetration through pores or pinholes.
It has been pointed out previously that a freshly cleaned filter surface
contains several open pores which, for the most part, become bridged
over during the first part of the filtering cycle. Those pores or pin-
holes that fail to close at any time during the filter cycle are usually
larger, ~150 pm in diameter, than the average pore size or contain no
fiber obstructions. As far as fly ash/woven glass systems are concerned,
the pinholes are the major dust penetration source.
The above statements have been substantiated by the many field and lab-
oratory measurements discussed in Section VII. Additionally, the fact
that measurements of particle size distributions immediately upstream
and downstream of the filters showed no statistically significant dif-
ferences suggests that the filter collects all particles at essentially
the same efficiency.
Penetration Versus Pore Properties ?
If one considers a fabric filter that operates at a constant face velocity
and inlet dust loading, the fractional mass penetration will be directly
proportional to the fraction of the airstream that passes through the pin-
holes. Therefore, the changes in effluent concentration as the average
fabric loading increases shown in Section VII, Figures 75 and 76, in-
dicate that pore area must decrease rapidly once filtration is initiated.
Based upon effluent measurements for face velocities of 0.61 m/min or lower,
the outlet concentrations appears to level out at about 0.3 to 0.5 mg/m .
Unfortunately, with regard to those filters showing the lowest effluent
concentrations, the detection capability of the CNC system did not permit
333
-------
estimation of true minimum values. Depending upon the calibration charac-
teristics of the condensation nuclei counter, the minimum detectable
3
effluent concentrations ranged from an apparent 0.3 to 0.5 mg/m , whereas
the actual levels may have been considerably lower. Thus, the assumption
made in designing the penetration model that effluent concentrations never
3
go below 0.5 mg/m represents a
as predicting system emissions.
3
go below 0.5 mg/m represents a safe or conservative approach in so far
Since particulate emissions are attributed almost entirely to pore or
pinhole penetration, the actual quantity of dust passing the filter
must depend directly on the volume of gas passing through the openings.
In turn, the volume of gas passing through the open pores is determined
by the pore cross sectional area and the pore velocity. Because the
velocity through any pore or pinhole must increase as the pressure gradient
increases across the filter, the fraction of the approaching air passing
through the pores will also increase unless the pore dimensions are re-
duced greatly by effective bridging as filtration progresses. Therefore,
the very rapid decrease in dust penetration observed during the early
stages of filtration must denote a rapid closure of pore openings.
The major problem at this time is to determine precisely what fabric
parameters control its capability to provide an essentially unbroken
substrate for support of the dust cake. It is evident, based upon both
theoretical analyses and experimental observations, that high efficiency
3
filtration of typical inlet concentrations, ~1 to 5 g/m , could always
be attained where there was no problem of pore closure. As this study
progressed, it became apparent that a rigorous study of the basic fabric
variables determining the ultimate performance of the dust/fabric struc-
ture could not be undertaken without detracting from the specific objec-
tive of developing a predictive model for coal fly ash filtration.
Hence, the working parameters proposed for the model described in this
report are based mainly upon practical field and laboratory measurements
relating to coal-fired boiler operations.
334
-------
It has been indicated that certain semiquantitative measurements may
provide key guidelines as to filter performance, clean cloth permeabi-
lity being a prime example. However, the caution has been extended that
the presence or absence of bulk fibers within a pore while having little
influence on clean (unused) cloth permeability may affect dramatically
its dust retention capability.
If one compares the woven glass fabrics commonly used for hot gas fil-
tration with sateen weave cotton, for example, the Frasier perme-
abilities are roughly 60 and 15, respectively (i.e., initial cotton
resistance is four times that for woven glass). Here the permeability
does provides a reliable index of dust collection potential since, as
shown in Table 18, fly ash effluents from sateen weave cotton were
appreciably lower, "10 times. Comparative emission measurements with
atmospheric dust as the test aerosol, using an optical counter and a
condensation nuclei counter, Section VII, have also indicated that sateen
weave cotton is a more efficient fine particle and nuclei collector.
The point that must be stressed is that only the presence of fine, well
dispersed fibers, with separation distances of the order of the fiber
diameters can provide firm supporting substrates for cake development.
It is also emphasized that in conjunction with the reduction of pore size
to enhance pore bridging, it is also important that the number of pores
per unit area be maximized so that the free area is kept as high as prac-
ticable and, conversely, the resistance as low as possible. In the case
of the Sunbury and Nucla twill weave fabrics, the continuous yarns were
responsible for a reduction in the number of active pores, Section V.
However, it is also possible that a looser weave structure (which at
first appears as a possible way to reduce resistance by providing addi-
tional flow would simultaneously lead to a lack of uniformity in pore
dimensions. The latter situation has been demonstrated in this and prior
studies3 to be a primary cause of pinhole leakage.
335
-------
In the modeling relationships proposed in this section for particle pene-
tration, the constants appearing in the working equation apply specifically
to coal fly ash/glass fabric systems. If different fabrics are substi-
tuted or the properties of the dusts differ significantly from the types
used in this study; i.e., coal and lignite fly ashes, and granite dust,
a change in constants should be expected. However, it is again emphasized
that the ease with which bench scale measurements can be performed suggests
that direct measurement, rather than uncertain extrapolation of unproven
theoretical concepts, is the best approach for estimating many basic model-
ing parameters. Methods of performing these measurements have been de-
scribed in Section IV.
Penetration and Inlet Concentration
It has been reported previously ' that effluent concentrations from
fabric filters are not strongly dependent upon the influent concentration,
particularly so in the case of fabrics such as sateen weave cotton that
provide a good support for the overlying dust cake. The above situation
prevails because once the dust cake develops (and in the absence of pin-
hole leaks) the amount of dust penetrating the undisturbed cake is neg-
ligible for cake thicknesses greater than 10 to 20 ym. Therefore, only
in the case of very frequent cleaning wherein a larger fraction of the
gas stream passes through the yet unblocked pores would one expect to
see the effect of inlet concentration changes.
Woven glass fabrics, however, and other similar weaves, often posses
sufficient pinhole leaks to cause a constant low order dust emission.
In those cases where the problem is serious; e.g., penetrations at the
1 percent level or greater, the magnitude of the pinhole leakage will
vary directly with the inlet concentration because the volume of un-
filtered air passing through the pinholes far exceeds that passing
through the dust cake.
336
-------
Low order emissions may arise from two sources, direct pinhole penetra-
tion or the previously discussed rear-face slough-off. Limited gravi-
metric tests during the present program with a full scale (10 ft. x 4 in.)
woven glass (Sunbury type) bag indicated that slough-off contributed about
0.5 mg/m to the total effluent concentration. This (0.5 mg/m3) value
also corresponds to the lowest mass concentration that can be estimated
from CNC measurements. For this reason, use of 0.5 mg/m3 as a constant
background emission rate to be added to that resulting from pinhole pene-
tration appears to be an acceptable procedure.
Penetration Versus Face Velocity
The measurement of dust penetration and effluent concentration at various
face velocities, Section VII, Figures 87 and 88, indicated that velocity
plays a very important role in filtration. The discouraging aspects of
these tests as far as the fly ash/glass fabric combinations are concerned
is that a serious penalty in the form of increased emission levels
(roughly eight times greater) must be accepted if one elects to increase
the air-to-cloth ratio by a factor of 2.5 (0.61 to 1.52 m/min or 2 to
5 ft/min).
The fact that emission rates increase at the higher velocities is con-
sistent with the characteristic pore and pinhole structures noted for
the glass fabrics. Despite the fact that the deposition velocity for
the dust is greater, which should accelerate the pore bridging process,
the higher velocity also causes a greater entraining force to act upon
particles deposited in any partially bridged region. The end result
is that more pores remain unbridged at the higher face velocities.
The most important aspect of the above findings, however, is that any
sequentially cleaned, multicompartment filter is automatically subjected
to a rather broad spectrum of velocities at various points in the system
at any instant depending upon the surface loading distribution. Hence,
337
-------
in computing local dust penetration levels, one must take into account
both the local fabric loadings and velocities. For example, previously
discussed tests showing the resistance versus loading characteristics
for filters from which approximately 50 percent of the dust had been
removed, Table 18, Section VII, showed significant increases in penetra-
tion. The corresponding changes in penetration are summarized in
Table 52. Reference to tests 71 and 96 shows that penetration is about
16 times greater for a 150 percent velocity increase with uniformly
loaded glass fabrics. Partially loaded fabrics, tests 72A-C and 97,
indicate a nine times increase in penetration for a similar, 150 percent,
velocity increase. Most important, when a filter operating at an aver-
age velocity of 1.52 m/min was partially cleaned, its emission levels
were 11 times greater than those for the uniformly loaded fabric when
filtration was resumed.
The data in Table 52 indicate that overcleaning of fabrics as well as
high velocities can lead to undesirably high emission rates. Therefore,
it is very important to determine precisely what contribution is made
by each element of a filter system under parallel flow conditions in
which the filtration velocities over the complete system can easily vary
by a factor of 10 at the initiation of filtration.
Figure 121 shows the velocity versus fabric loading relationship for
the partially cleaned woven glass fiber described in test 72, Table 52.
The maximum velocity is seen to be nearly twice the average velocity
and the initial velocities through the cleaned and uncleaned fractions of
the surface differ by a factor of 10. If the terminal loadings remain
the same and 10 percent, rather than 50 percent of the filter surface
is cleaned, a fivefold increase over the average velocity would be ex-
pected at the resumption of filtration.
338
-------
Table 52. COMPARATIVE PENETRATION CHARACTERISTICS FOR UNIFORMLY LOADED AND PARTIALLY
LOADED FABRICS, GCA.FLY ASH
Test3
71
72-A-C
96
97
Fabric
Used Sunbury
Used Sunbury
New Sunbury
Used Sunbury
Average
face velocity,
m/min
0.61
0.61
1.52
1,52
Dust
loading range,
g/m2
32-660
340-750
0-400
270-390
Dust
distribution
Uniform
Nonuniform
Uniform
Nonuniform
Cleaning
Complete
Partial
Complete
Partial
Fractional
penetration
mass basis
0.0007
0.0135
0.0109
0.1223
Data excerpted from Table 18, Section VII.
-------
3.0
CURVE
FRACTIONAL AREA
CLEANED UNCLEANED
0.515
100 200 300 400
ADDED FABRIC LOADING, gram/m2
500
Figure 121. Filtration velocity through cleaned and uncleaned
areas of filter. GCA fly ash and Sunbury fabric
340
-------
DUST PENETRATION MODEL
Based upon the available field data and the results of the laboratory
testing program, it was decided that the model for predicting coal fly
ash penetration through woven glass fabrics should take into account
the following variables:
• The unique functional relationship between a specific
dust and a specific fabric A
• Inlet dust concentration C.
• Fabric loading \j
• Filtration velocity V
• Residual outlet concentration C
R
• Outlet dust concentration C
o
Thus, in notational form the fractional penetration can be expressed as
Pn = i|> [ , C±, V, CR, C
The term, , should appear as a constant that characterizes the unique
interrelationship between coal fly ash and the Sunbury or Nucla type
glass fabrics. The inlet concentration, C., will appear as an indepen-
dent variable and remain unchanged for a specific set of operating
parameters. As indicated previously, CR, depicts a low-order, relatively
constant emission that is assumed (a) to derive mainly from rear face
slough-off, (b) to be independent of inlet loading, and (c) to be unique
to the fly ash/glass fabric system.
The velocity term, V, refers not to average velocity but to the actual
local face velocity at a specific fabric loading as determined by the
model describing the fabric drag versus fabric loading relationship.
The term, C , is the computed filter effluent concentration associated
with the parameters cited above; i.e.:
341
-------
co - e [ *, c., w, v, CR ]
Working equations for the estimation of outlet concentration and filter
penetration in terms of the previously cited variables are developed in
the following paragraphs. The mathematical relationships indicated in
Figures 87 and 88 and Table 21 provide the data base. The curves show
the effects of both inlet concentration and filtration velocity on out-
let concentration as dust accumulates upon the fabric surface. Because
the overall efficiencies attain the 98 percent or greater level within
a very brief time, the increases in fabric loading per unit time are
equivalent to the quantity of dust approaching the fabric.
Since outlet concentrations showed a dependency on inlet concentration
and since the inlet concentrations varied from test to test, the curves
were normalized prior to the data analysis. This was done by graphing
actual penetrations versus fabric loadings for each of the tests.
Plotting an effective penetration (outlet concentration divided by in-
let concentration) would have dampened the effect of the residual outlet
concentration on total outlet concentration. Thus, the penetration
values used in the analysis were the effective penetration minus the
residual penetration (residual outlet concentration divided by inlet
concentration).
The procedure used in developing an expression for actual penetration
as a function of fabric loading and velocity was based upon conventional
curve fitting methods. An empirical relationship was sought which would
accomplish the following:
• Predict penetration as a function of fabric loading at
constant face velocity
• Predict a penetration level £ 100 percent for zero
fabric loading
• Attain a limiting, steady state penetration once fil-
tration is dominated by cake filtration.
342
-------
The general form selected for the mathematical function was:
Pn = Png + (PnQ - Png) exp (-aW) (79)
where Pn = penetration
Pn = penetration at steady state
s
Pn = initial penetration at W = W
0 R
W = increase in fabric loading above the residual value W
R
a = concentration decay function
Equation (79) reflects both the rapid exponential decay observed for
outlet loadings as well as their ultimate leveling off at a fixed emis-
sion rate as filtration progresses.
The constants Pn , Pn and a_ were evaluated for the 0.39, 0.61 and 1.52
S O
m/min velocity tests along with the steady state value for the 3.35
m/min test. The initial fractional penetration values, which were ob-
tained by extrapolation, ranged from approximately 0.09 to 0.11. A
Pn value of 0.1 was used irrespective of velocity. After the steady
state values and the initial slopes were plotted versus face velocity,
the constants were computed and the working equations developed:
Pn = 1.5 x 10~7 exp j 12.7 [l-exp (-1.03V)]} (80)
a = 3'6 X410 + 0.094 (81)
V
where V is the local face velocity, m/min. The development of these equa-
tions is presented in Appendix C.
Equations (79) through (81) provide the means for predicting penetration as
a function of face velocity and fabric loading. The outlet concentration,
C is found by multiplying the inlet concentration, C±, by the actual
penetration followed by the addition of the residual outlet concentration,
CR:
343
-------
C = Pn C. + C (82)
O IK
The solid curves shown in Figure 122 represent the computed values for
effluent concentrations whereas the symbols depict the actual data
points. Despite the obvious curve fitting problem for the low velocity
f\
(0.39m/min) test in the 20 to 80 g/m fabric loading range, it is em-
phasized that the fit is excellent in the very critical range where the
outlet concentration decreases by at least two orders of magnitude.
The effect of inlet concentration upon effluent concentrations is shown
in Figure 123. Note that the filter effluents tend to follow linearly
the changes in inlet concentration during the early phases of filtra-
tion, approximately up to a fabric loading of 40 g/m2. During this
period the dominant emissions are those from direct penetration through
unbridged pores. However, once significant bridging has taken place,
direct pore penetration may play a minor role with respect to periodic
slough-off of agglomerated particles from the rear (clean) filter face.
Therefore, despite tenfold differences in inlet loadings, the ultimate
emissions after cake stabilization may show relatively small, factor
of 2, differences.
344
-------
t-
o
10"
10
2 4
4 O
TEST
98
AVERAGE
96
97
INLET CONC.(a/m3) FACE V£tOCITY(m/min.)
8.09 0.39
7.01
5.37
4.60
0.61
1.52
3.35
NOTES: SOLID LINES ARE CURVES PREDICTED BY MODEL.
SYMBOLS REPRESENT ACTUAL DATA POINTS.
20
40
60
80
l(
40
FABRIC LOADING (W), g/m'
Figure 122. Predicted and observed outlet concentrations for
bench scale tests. GCA fly ash and Sunbury fabric
345
-------
40 60 80 100
FABRIC LOADING (W),0/m2
120
140
Figure 123. Effect of inlet concentration on predicted outlet concen-
trations at a face velocity of 0.61 m/min. GCA fly ash
and Sunbury fabric
346
-------
SECTION XI
MATHEMATICAL MODEL FOR A FABRIC FILTER SYSTEM
INTRODUCTION
The preceding sections of this report provide the technical background
for the design of the mathematical model describing the filtration of coal
fly ash. In general, it can be said that the literature furnished only
qualitative guidelines and certainly no practical techniques for anal-
yzing or predicting the behavior of large, multicompartmented baghouses.
The sparsity of technical information from both field and laboratory
sources required that a broad based series of laboratory studies be carried
out to provide a rational basis for model development. As the present
study progressed, complexities in experimental measurements coupled with
unexpected performance data, made it clear that the research should be
constrained mainly to the problem of filtering coal fly ash from utility
boiler effluents, if a realistic predictive model were to be developed
within the time frame for this study. By adhering to this resolve it
was possible to identify and define the proper roles of the major variables
entering in to the fly ash filtration process.
It is again pointed out that the predictive model is intended for use
with a coal fly ash/woven glass fabric system in which the collecting
media consists of twill weaves similar to those now commonly employed
by large utilities installations. The performance characteristics of
these glass fabrics (and related weaves of nonmineral composition) reveal
that the particulate effluents consist mainly of dust that passes through
unblocked pores or pinholes with minimal size fractionation and collection
taking place during the ensuing process. Although such penetration
347
-------
occurs mainly during the early filtration phase, it may continue through-
out the filtration cycle if the fabric pore structure is not uniform or
the fabric has been damaged or worn by extended field service. The net
result is that the modeling process for dust penetration is greatly sim-
plified because the overwhelming dust penetration by direct leakage pre-
cludes any reduction in size parameters for the filter effluent. Thus,
despite contrary reports in the literature, the fractional particle size
efficiencies for all dimensions of interest are approximately the same
and equal to that for the overall mass collection efficiency.
If radical changes in pore structure or staple fiber content are intro-
duced, it is expected that the relationships proposed for fly ash/woven
glass fabric systems would require modification. For example, limited
tests with cotton sateen fabric (which have not been elaborated upon in
this report) suggest that both effluent size properties and effluent con-
centrations were appreciably lower than those observed with glass fabrics.
PRINCIPAL MODELING RELATIONSHIPS
A brief summary of those mathematical relationships forming the basis for
the predictive model is given in the following paragraphs. The key equa-
tions used to calculate filter drag and dust penetration behavior, each
identified by its original number, are listed below along with the reasons
for their selection.
Two mathematical functions were developed for describing the nonlinear
drag versus fabric loading curves frequently encountered in industrial
filtration processes. Despite a relatively close adherence to postulated
filter behavior, the first approach, Equation (25), required the evalu-
ation of several constants and the mathematical structure was overly
cumbersome. On the other hand, the second approach, Equation (28),
proved to be a good curve fitting tool despite its purely empirical
structure. Therefore, Equation (28) as shown below is selected to
348
-------
describe all nonlinear drag/fabric loading curves when it is believed
that the nonlinear segments play an important role in determining resis-
tance and/or dust penetration characteristics:
S = SR + K2 W + (KR - K2) W*(l-exp (-W/W*)) (28)
In those instances where the drag/loading relationship is essentially
linear, or the nonlinear segment of the curve can be safely ignored, the
simpler expression, Equation (4), is chosen:
S = SE + K2 W (4)
Because there exist no dependable means to predict the numerical value for
the constants appearing in Equations (28) and (4), it is strongly recom-
mended that the terms S , S , K , K , and W* be determined by experiment.
R E R 2.
Simple and inexpensive laboratory methods to achieve this end are de-
scribed in this report (see Section IV) . The term W* is readily computed
from the relation:
W = CSE-SR+K2WR)/(KR-K2)
(29)
Although the most accurate estimates of K_ values should derive from test
measurements with the dust in question, allowance must also be made for
the impact of increasing face velocity on Y.^ An empirical function,
Equation (20a), applying specifically to coal fly ash/woven glass fabric
systems, satisfies this requirement:
K =1.8 V metric units (20a)
If a rough estimate of ^ is required before confirming tests can be per-
formed, the Carman-Kozeny equations (Equations (31) and (36)) can be modi-
fied as follows: (1) reduction of the constant k from 5.0 to 2.5 and (2)
computing the specific surface parameter, SQ, for the distribution of par-
ticle sizes constituting the dust cake:
349
-------
K0 = 2.5y S (l-e)/p e3 (31)
2. O p
The above modification in conjunction with Equation (20a) provides key data
inputs whose values constantly change during the iteration procedure used
to determine local and average values for velocity, V, drag, S, and fabric
loading, W, throughout the fabric filter system being modeled.
The degree of cleaning attained by collapse and reverse flow is deter-
mined by the empirical relationship developed for coal fly ash/woven
glass fabric systems:
ac =1.51 x 10"8 W2'52 (50)
In the case of mechanical shaking, a modified form of Equation (51) applies
which takes into account shaking frequency f (cps), shaking amplitude A
2
(cm) as well as fabric loading W (g/m ):
-22 2 S?
ac = 2.23 x 10 Z (r AW) (51)
Penetration values and particle effluent concentrations that reflect the
impact of dust inlet loading, face velocity, fabric areal density and the
unique characteristics of the coal fly ash/woven glass fabric system are
determined by the following equations:
Pn = Pn + /Pn - Pn \ exp(- aw) (79)
+ (
where Png = 1.5 x 10~ exp J12.7 f~l-exp(-1.03V) 1 \ (80)
and a = 3'6 * 10— + 0.094 (81)
V
CQ = Pn C. + CR (82)
350
-------
It is again emphasized that coal fly ash undergoes no significant reduction
in size properties following filtration with woven glass media of the
types used at the Sunbury and Nucla power stations. Hence, the fractional
particle size efficiency values are constant, independent of size and, for
all practical purposes, equal to the overall weight collection efficiency.
This anomalous behavior is the result of gross, unfiltered air passage
through unblocked pores or pinholes that far exceeds dust penetration
through the dust cake, per se.
The key to the modeling process for predicting the performance of multi-
compartment systems is the concept that treats a cleaned fabric filter
as two separate elements, one from which no dust has been removed and the
other from which the dust layer has spalled off at the dust/fabric inter-
face. The latter surface has a uniformly distributed drag characteristic
which, for practical purposes, is essentially independent of previous
surface loading, intensity of cleaning and the dust/fabric combination.
System drag values are computed by iterative methods using the several
data inputs noted previously:
S = I > A /S I A (47)
or
(a a •
n a u, u
Er + r1"'!-1
1=1 c u.^ u_^,
where a refers to the fraction of cleaned surface, a^ to the various
uncleaned fractions, and S indicates the associated'drag values.
351
-------
DESIGNED MODEL CAPABILITY
In the previous paragraphs the basic filtration equations and the iter-
ative approach for treating multicompartment filtration systems have been
reviewed for immediate reference. The following discussion is intended
to outline the ground rules with respect to how closely the predictive
model(s) describes the overall fly ash filtration processes for utility
applications. The only major constraint for the model(s) is that (1) the
inlet aerosol should consist of or possess the general physical proper-
ties of a coal fly ash and (2) the fabric characteristics be similar to
the woven glass media of the types used at the Sunbury and Nucla instal-
lations. Aside from the above, the model is sufficiently flexible to
meet the following criteria:
The model is adaptable to either constant flow or constant
pressure conditions. With respect to most large, multi-
compartmented systems, however, the available gas handling
capacity must necessarily be controlled so that excursions
from mean flow rates are minimal.
The model can accommodate to a continuous cleaning regimen;
i.e., the immediate repetition of the cleaning cycle follow-
ing the sequential cleaning of successive individual
compartments, or
The model can also describe the situation where lengthy
filtration intervals are encountered between the cleaning
cycles. In both cases the term cleaning cycle refers to
the uninterrupted cleaning of all compartments in the sys-
tem. No provision is made for the random cleaning of less
than all compartments followed by continuous on-line fil-
tration of all compartments.
The system cleaning characteristics are determined by the fraction of
fabric area cleaned, a , when individual compartments are taken off-line.
With respect to bag collapse systems and/or low energy shaking, the dust
removal parameter, a , is calculated from the fabric loading, W , before
cleaning.
352
-------
• The model can be used with collapse and reverse flow systems,
mechanical shaking systems or combinations of the above. It'
is not intended for use with pulse jet or high velocity reverse
jet cleaning systems.
• The model can be used equally well with pressure or_ time con-
trolled cleaning cycles.
• The model provides estimates of average and point values of
filter drag or resistance for the selected set of operating
parameters and dust/fabric specifications.
• The model provides estimates of average and point values for
penetration and mass effluent concentration for the selected
set of operating parameters and dust/fabric specifications.
In the above instances, it is assumed that the following operating param-
eters are known: inlet concentration (C.), average face velocity (V.),
cleaning parameters (frequency and intensity, energy level, of cleaning
cycles) and the fabric loading before cleaning (W ). In addition, the
related parameters, S , S , K , K0, and W* must also be specified for the
R E R *-
given dust/fabric combination.
The model alternatively provides an estimate of the necessary frequency
of cleaning when the maximum operating resistance P is cited as an
TILcl^x
operating specification along with the expected values of C^ and the
selected value for V..
BASIC MODELING PROCESS
The basic model treats each of the "I" compartments of the filter system
as a separate element. It is also assumed that the inlet dust concen-
trations and the filtration velocities are the same for each bag within
a given compartment. However, the possibilities of both concentration
and velocity gradients are recognized due to the particle size spectrum,
bag proximity and air inlet location.
353
-------
Figure 124 indicates the distribution of volume flow rates for a filter
system consisting of "I" separate compartments. Because of the parallel
arrangement, the resistance P across each compartment is the same just as
the voltage drop would be for the analogous electrical circuit. The volume
flow rate, q, and gas velocity, v, through each compartment vary inversely
with the individual compartment drag.
The distinguishing feature between the new modeling concept introduced in
this study and previously reported efforts is that the surface of each
bag within a given compartment is subdivided into a number of secondary
areas each of which displays its own characteristic fabric loading (W),
drag (S), face velocity (V) and dust penetration (Pn). The fact that the
contributive role of each of these areas with respect to overall system
drag and penetration can be assessed at any time during the cleaning
and/or filtering cycles is a unique feature of the new model. Note again
that since all bags within a given compartment possess identical perfor-
mance characteristics, an "I" compartment system could be described equally
well as an "I" bag system.
The experimental models presented in Equations (40) through (46), Sec-
tion IX, for a two-"element" or two-"surface" system have been expanded
to define the performance of typical field, multisurface systems. In the
former instance, the partial cleaning of a single bag led to an average
residual fabric loading, W , represented by two distinct surface loadings;
K.
the first or cleaned area, a , with its characteristics cleaned residual
loading, W , and the second or uncleaned area, a , whose areal density,
K U
W , was the same as that before initiation of cleaning. By successive
iterations, the temporal and spatial relationships for face velocity,
fabric loading and drag for each surface were developed over varying
periods. With extended filtering times, the areal densities for the dis-
tinct surfaces converged due to the self-equalizing feature of the fil-
tering process.
354
-------
Co
V
Ci
C,2
cu
t n
C22
W
)2
W
U
W
2«
'II
I2
VIJ V2I
22 ----
2J
i, i
VfoWjg
'31
IJ
Figure 124. System breakdown for I bags and J areas per bag
-------
Although the treatment of multicompartment systems follows the same pro-
cedure as that for the single bag unit, it is now necessary to deal with
several randomly distributed areas of varying areal densities for each
bag as well as several compartments, each with its unique variability
pattern. Thus, the following notational system is introduced to describe
the various surface elements in the multicompartment system in which the
subscripts i and j, respectively, designate the i compartment and the
f~Vi
j area subdivision in each compartment. This enables one to identify
the specific element of fabric area; e.g., compartment 2, 1st area sub-
division for which the local face velocity, surface loading and effluent
concentration at a specified time are defined as V , W and C , res-
pectively, Figure 124.
Although the program is designed to accept as many as 10 separate areas
(J=10) per bag, the actual number used in the iteration process (which is
automatically selected by the computer program) depends upon the input
value for a . Given the restriction that the number of subdivisions or
c
areas must always appear as integer values, the program will always select
the number of subareas that comes closest to matching the a input. Thus,
a value of 3 for J will satisfy exactly the requirement that a = 0.333
whereas the same J value will also be selected as the nearest approxima-
tion to the condition that a =0.35. If a is 0.38, the program will
*-* C
select and operate with 8 areas wherein the cleaning of 3 areas provides
a cleaning parameter, a , of 0.375.
c
In Figure 125, a schematic representation of a hypothetical 3 compartment
filter system is shown in which a is assumed to be 0.5. Under these
c
conditions, each compartment need only to be subdivided into two areas to
satisfy the cleaning parameters. The proper interpretation of a in this
c
case is that 50 percent of the fabric surface in each compartment under-
going cleaning is reduced to its residual or W status. The height of
R
the bar representing each filter compartment indicates the relative
fabric loading prior to any cleaning. Hence, the areal density is uniform
356
-------
\--\ 2 3
(A) BEFORE 1st. CLEANING
n
4%.
i 2 3
(B) AFTER 1st. CLEANING
12 3
(C) BEFORE 2nd. CLEANING
i = l 2 3
(0) BEFORE 3rd. CLEANING
I 2 3
(E) BEFORE 4th. CLEANING
I 2 3
(F) BEFORE 3th. CLEANING
i = l 2 3
(G) BEFORE 6th CLEANING
I 2 3
(H) BEFORE 7th. CLEANING
2 3
(I) AFTER 7th. CLEANING
NOTES I. BAR HEIGHT A ROUGH MEASURE OF FABRIC LOADING AND/OR DUST CAKE THICKNESS
2. BEFORE AND AFTER 7th. CLEANING REPRESENTS STEADY STATE OPERATING RANGE
3. THREE COMPARTMENT SYSTEM 11-3) WITH 0C=0.50(J=2)
4. MINIMAL RESIDUAL LOADING FOR ALL FILTERS REPRESENTED BY NARROW BAND AT BASE OF HISTOGRAM
DUST ADDED
AFTER
1st. CLEANING
3nd. CLEANING
3rd. CLEANING
DUST ADDED
AFTER
4th. CLEANING
5th. CLEANING
6th. CLEANING
Figure 125. Schematic representation of approach to steady state
cleaning and fabric loading conditions for a three-
compartment system with 50 percent of each compartment
surface cleaned
357
-------
throughout all compartments. Based upon actual field observations, an input
2
a value of 0.5 corresponds to a surface loading of about 900 g/m .
c
Additionally, with a conservative time allowance of 5 minutes for the
cleaning of each compartment and thus a total elapsed time of 30 minutes
before the first subarea an1 is again ready for cleaning, an increase in
11 2
areal density of roughly 50 g/m might be anticipated for the last sub-
area, a „, to be cleaned.
Therefore, to follow precisely the previously established relationship
for dust removal versus fabric loading, (Equations (51) and (51a),
Section IX), the value of a should increase slightly over the time span
involved in cleaning the a through the a subareas. In the actual
modeling procedure, an average value of a is assumed that is based upon
c
the initial loadings of the first and final compartment to be cleaned.
Figure 125B simulates the distribution of a filter system dust holding
immediately following the collapse cleaning of the 1st compartment in
which 50 percent of the fabric surface is cleaned. The resultant height
represents the characteristic residual dust holding, W for the dust/
R
fabric system under investigation which, in the short term, is always
treated as a system constant.
Although no attempt at exact scaling has been made, the increments of
dust added to the on-line filter compartments and their associated sub-
areas, relate in inverse fashion to the fabric loading at the start of
filtration; i.e., the "just cleaned" or lightly loaded subareas see the
higher face velocities and hence the greater dust deposition rates. As
the successive cleaning steps are traced from Figure 125B through H, it
can be seen that the average filter dust holding has undergone a gradual
decrease as the originally uniformly loaded subareas are reduced to the
partially loaded regions shown in Figure 125H. However, with a conti-
nuation of the cleaning process in which subarea a experiences its
second cleaning, the system will operate at a steady state condition.
358
-------
The filter performance with respect to resistance and particulate emis-
sions will now oscillate within a constant range whose upper and lower
limits are dominated by the fabric loading profiles (without regard to
sequence) shown in Figure 125H and 1251.
In addition to assuming that the dependence of a on point-by-point
changes in fabric loading can be ignored, the impact of successive fabric
collapses (which may weaken adhesive bonds but not necessarily lead to
immediate dust dislodgment) has not been included in the modeling oper-
ations. It is assumed that for a specific cleaning method an equilibrium
adhesion level is reached after 5 to 6 repetitions of the cleaning pro-
cess. The above equilibrium process should not be confused with the rela-
tively long, approximately 2 to 3 weeks, process required for the fiber
dust holdings to reach a steady state. Beyond this point no significant
increase in dust dislodgement can be attained without increasing the in-
tensity of the dislodging force. As far as the modeling procedures for
the fly ash/woven glass fabric systems are concerned, the two simplifying
assumptions discussed above reduce significantly the data processing while
introducing no obvious penalties in predicting filter system performance.
Once the decision is made (by the computer) as to how many subareas will
be used for each compartment (and bag), the calculations proceed by suc-
cessive iterations with the results from the first iteration constituting
the input for the second, and so forth.
The general procedure for calculating all the system parameters at any
time in a cycle is outlined in Figure 126. Individual subareas and
compartment (bag) drags are first calculated so that the total (average)
system values for drag, pressure drop, and flow rate can be determined.
Based on the system pressure drop and individual bag drags, the volume
flow is first partitioned among all the compartments followed by a further
subdivision among the subareas of each bag. Penetration and outlet con-
centration are then computed for each subarea, each compartment (bag) and
359
-------
[DETERMINE FABRIC ORAG.Sp]
-»(j.OOP ON TIME~)
-»(LOOP ON BAGS # \J
-»(LOOP ON AREAS}
CALCULATE FLOW VELOCITY FOR AN AREA ON A BAG, V,j
A
CALCULATE DRAG FOR AN AREA ON A BAG,Sijt
CALCULATE DRAG FOR A BAG,Si
CALCULATE SYSTEM ORAG.S,
I
CALCULATE
SYSTEM
FLOW AND
i
PRESSURE
DROP,
vt.
Pi
•
»(LOOP ON
-»(LOOP ON AREAS)
T
CALCULATE FLOW VELOCITY FOR AN AREA ON A BAG, Vjj
CALCULATE PENETRATION FOR AN AREA ON A BAG.Pnj:
I
CALCULATE NEW FABRIC LOADING FOR AN AREA ON A BAG.Wjj
I
CALCULATE FLOW VELOCITY FORA BAG,Vit
T
[CLEAN A BAG IF NECESSARY)
CALCULATE TOTAL PENETRATION, Pnt
END OF CALCULATIONS
EQUATION
USED
86
83,84,85
87
88
89,90
91
92
91
Figure 126. Baghouse model computational procedure
360
-------
for the total system in the order named. Since the dust deposition rate
is determined by a specified flow velocity and inlet concentration, the
weight of dust added to any area on any bag can be calculated. Thus,
the fabric loadings for all areas can be calculated for succeeding time
increment.
The actual time increments chosen for the iteration represent a compro-
mise between excessive computing and printout operations and the need
to resolve excursions in resistance and/or emissions about their mean
values that may bear upon the adequacy of the control process. The
policy exercised here has been to select time increments that define key
system operating and performance parameters; i.e., flow rates, resistance
and outlet concentration, at the start, middle and end of the filtration
interval during which a compartment has been taken off-line for cleaning.
In the case of the Nucla operation, the overall cleaning period per
filter compartment was 4 minutes. Hence, the selection of a 2-minute
iteration time provides a measure of maximum, minimum and average system
parameters while compartment cleaning is taking place.
PROGRAM DESCRIPTION
A flow diagram for the computer simulation program, which is comprised
of two basic steps, is presented in Figure 127. The main program first
calls the MODEL subroutine in which all calculations are performed and
then transfers control to the SCRIBE subroutine in which the results are
plotted. All data input and output and calculations are performed by the
MODEL subroutine. Data are input to the program via the two subroutines
READIM and READIT.
READIT inputs all operating parameters and constants used in the linear
drag model. Cleaning parameters and constants used in the nonlinear
model are input by READIM. READIM also performs the temperature correc-
tion on viscosity.
361
-------
CALCULATE FABRIC DRA8>
(g) »Q.OOP ON TIMEJ
lOOP ON
.
[LOOP ON AREAS J
x
CALCULATE OBA6]>*-
(END OF AREA LOOP)
-^CALCULATE PENETRATION
( END OF AREA LOoTj
LOOP OX CLEANED AREAS
•( END OF CLEANED AREA
— ( END OF BAG LOOP
<^CALCULATE
<
PRINT a STORE
PENETRATION,
'LOW,
(END OF TIME LOOP")
END SUBROUTINE MODEL
I IUBROUTPNE
1 SCRIBE
Figure 127. Baghouse simulation program flow diagram
362
-------
Headings for graphical output are established by the PLOTIN subroutine.
In performing the calculations for drag and penetration, the program uti
lizes the additional subroutines, CAKDRG and PENET, respectively.
COMPUTATIONAL PROCEDURES
The following paragraphs provide a description of the procedures and
equations used to calculate system performance. A flow diagram for the
entire program is presented in Figure 127 and a diagram of the basic
computations performed is shown in Figure 126. A tabulation of relevant
equations with reference to where they are treated in the report is also
included in Figure 126.
Computation
Cleaned fabric drag is a predetermined input that is not computed by the
program. It is set equal to the effective drag, S , if the linear drag
hj
model is selected and to the residual drag, S , if the nonlinear drag
R
model is used.
Area drag values are computed by the linear or nonlinear drag models
with the subroutine CAKDRG. The choice of subroutines is automatically
*
performed by the program which selects the nonlinear model when W has
any nonzero value. A zero value for W will automatically lead to com-
puter calculations by the linear drag model.
The area drag equations for the linear model are:
S.. = S +K xW (83)
I3t E 2±. ijt
and for the nonlinear:
363
-------
Where S . = the drag for the j area on the i bag at time = t
S = effective drag for cleaned fabric
E
S = residual drag for cleaned fabric
R
v = specific cake resistance for the area
Wf. = absolute fabric loading less the residual fabric loading
K = initial slope of the drag versus loading curve
W* = constant dependent on fabric and dust properties
t = time
The specific cake resistance (K ) is a function of velocity:
K, = K? /V 70.61 (85)
1J
where K° is the specific resistance at 0.61 m/min and the actual gas
temperature. A correction for gas viscosity changes is carried out
within the program's initiation step.
Since the flow velocity for a specified area is not determined until the
system pressure drop and area drag are known, it must be estimated from
the previous system pressure drop and the previous drag on the area:
The total or average drag for a compartment (bag) is calculated for a
parallel resistance network of J equal areas as:
S = J/£ 1/S.. (87)
t =l 12t
364
-------
Similarly, total system drag is calculated for I bags as:
I
St = I/£ i/s (88)
For convenience in data processing, the drag value for any compartment
undergoing cleaning is set equal to 102° in lieu of plus infinity because
the compartment velicity is zero. However, since the parameters describ-
ing overall system performance are based on total fabric area, the value
of I in Equation (88), which designates the total number of system com-
partments, is not changed. Total baghouse flow or pressure drop can,
therefore, be held constant while the average flow velocities for the
individual compartments are permitted to vary.
The total or average system flow and/or pressure drop are calculated
from the total system drag and the specified constant pressure drop
and/or flow. Additionally, when a compartment is being cleaned via re-
verse flow, the reverse flow air is factored into the computed pressure
drop and flow rate.
When the system pressure drop is specified as constant, the average gas
velocity system is calculated by:
v = P /S + v/I (89)
t c t R
and when the system flow is specified as constant, the pressure drop is
calculated by:
P = V S + VD S /I (90)
t C t K t.
where Pr = specified constant system pressure drop
C
V = specified constant system velocity
c
V = reverse flow velocity for a single bag
R
365
-------
It is again pointed out that a constant pressure drop system in most
large field installations would not ordinarily be anticipated. Such a
constraint could lead to unacceptable flow variation in most process
or contaminant control operations.
If no reverse flow is used, V is zero in the above Equations (80) and
R
(90). Once the system pressure drop is known, the calculated flow
velocity through an area can be calculated:
Fabric Penetration
Penetration through a specified subarea is calculated by the subroutine
PENET from the empirical relationships developed in Section X:
C
Pnij = C2 = Pns+ t0-1 - Pns)e~aWiJt + CR/Ci <92>
where Pn. . = penetration through the j area on the i bag
W. . = cloth loading minus residual loading at time t
o
CR = residual concentration, 0.5 mg/m , a system constant
C. = inlet concentration
Pns = 1.5 x ID'7 e^U-e-1'03 V«t> (92a)
a = 3.6 x 10~3/(V )4 + 0.094 (92b)
Jt
and V = face velocity of the jth area on the ith compartment (bag) at
time t.
366
-------
Once the face velocity and penetration have been established for ah area,
the dust deposition rate can be calculated. The fabric loading on the
area used in the calculations for the succeeding time loop is:
W.. = V x (1-Pn ) x At x C. + W.. (93)
Jt + At Jt Jt x 13
Note that when a compartment (bag) is being cleaned, its area velocities
are zero and thus no dust is added to the bag. The average flow velo-
city through a compartment (bag) is calculated in the same manner as
that for an area (Equation (91) except that the total compartment drag
is used.
After the compartment filtering (or on-line) time has progressed to the
point where it is equal to the cleaning cycle time minus the time re-
quired to clean one compartment, the cleaning cycle is initiated. This
entails taking the compartment off line followed by setting its drag at
20
10 to adjust for the zero flow condition.
Total or average system penetration is simply the total mass emitted
divided by the total mass input:
i ! J
P = -±— V V Pn. , V.. (94)
After all calculations for time = t have been completed and the fabric
loading for the next time loop has been calculated, one proceeds to the
next time iteration.
Program Input and Output
A summary of the program input data and their related units are presented
in Table 53.
367
-------
Table 53. SIMULATION PROGRAM INPUT DATA
Item
Number of compartments
Cycle time
Cleaning time
Total area run time
Number of cycles modeled
Number of increments per bag
Average system face velocity (constant)
Inlet concentration
Effective drag
Residual drag
Cake resistance at 0.61 m/min and 25°C
Reverse flow velocity
Residual loading
Initial cake loading
Average system pressure (constant)
Maximum pressure
W*, System constant
Initial S versus W slope
Temperature
Caked area
Print diagnostics
X and Y axis lengths
Units
-
Minutes
Minutes
Minutes
-
-
m/min
g/m3
N-min/m
N-min/m
N-min/g-m
m/min
g/m
2
g/m
N/m2
2
N/m
g/m
N-min/g-m
°K
^ .
True or false
inches
Symbol3
I
t
t
t
-
-
V
c
C.
i
SE
SR
K2
V
WR
P
c
P
max
W*
*R
T
a
u
Symbols used in main body of report.
appear in Appendix A.
Symbols used in program
368
-------
The number of compartments (bags) specified is limited to 100. Cycle
time is the time required to clean all compartments whereas claening
time is the time required to clean only one compartment. Once the
number of compartments and the total and individual cleaning times have
been specified, the time between individual compartment cleanings is
fixed. For example, if the cleaning cycle time is 30 minutes, the in-
dividual cleaning time is 5 minutes and the system has three compart-
ments, then the time between the completion of cleaning in one compart-
ment and the start of cleaning in the next compartment is 5 minutes.
This intermediate 5-minute, on-line period plus the actual 5-minute
cleaning time is considered to be the cleaning cycle time for a single
compartment.
The total area run time is defined as the time between cleaning cycles
when all compartments are filtering. This allows all compartments to
filter for a specified amount of time after the cleaning cycle has been
completed. To determine the time increment used in the calculations,
the number of time increments per compartment between successive
cleanings; i.e., the individual compartment cleaning cycle time, must
be input. Referring to the previous example in which the time between
cleanings is 10 minutes, if five increments were chosen, the incremental
time would be 10 minutes/5 or 2 minutes. The number of time increments
per cleaning cycle is simply the product of the number of increments per
compartment and the number of bags (compartments) in the system. The
time specifications are discussed further in Appendix D.
The air-to-cloth ratio should be specified only if operation at constant
total flow rate is desired. If it is not specified, a constant pressure
drop must be specified. Operation at constant pressure is assumed if
both are nonzero. The actual gas temperature and pressure must be used
in calculating the air-to-cloth ratio, the inlet concentration and the
reverse flow velocity. Reverse flow velocity is the velocity of the
reverse flow air through a single compartment.
369
-------
If the system is operated at a constant volume flow rate, cleaning can
be controlled by pressure. If a maximum pressure is specified, cleaning
will be initiated when the total system pressure exceeds that value.
This will override any total area run time specification. When pres-
sure control is used, total time modeled is simply the number of cycles
modeled times the sum of the cycle time and total area run time.
The residual fabric loading, WR, is the loading which exists on the
cleaned portion of a bag just after cleaning. The absolute cake loading,
W , includes all dust on the fabric, the cleanable function as well as
T
the fixed residual fraction, W_. In the operation of the linear drag
K
model, the absolute fabric loading is used, whereas the nonlinear drag
and the penetration models are based on the cleanable or removal dust
loading W or W - W . Both the linear and nonlinear models (see
T R.
Section IX) can be used to predict drag. The effective drag, S , must
E
be specified in either case since it is used to determine flow velocity
for the calculation of K on a cleaned bag. The specific cake resis-
tance, K?, must also be specified at 0.61 m/min and 25°C for both
models because the laboratory measurements of K were made under the
above conditions. The coefficient K is allowed to vary linearly with
viscosity and with the square root of velocity. Therefore, if a K«
value is available for conditions other than indicated above, it may
be corrected for input to the program. The gas temperature must be
specified for the calculation of viscosity, which is subsequently used
to correct the value of KO input.
The linear model is used to calculate drag if a zero value is entered
for W*, a constant quantity for the fabric and dust under investigation.
A nonzero value for W* indicates that the nonlinear model should be
used. Values for the initial slope of the drag versus loading curve,
Kj^, and the residual drag, SR, must be specified if the nonlinear
model is selected.
370
-------
The fractional area of a bag from which dust is not removed during clean-
ing is input as caked area, a^ The number of areas into which a bag is
subdivided and the number of those areas which will be cleaned is deter-
mined within the program. The fractional area which will be cleaned is
then output. A maximum deviation of 3 percent may arise between the
caked area input and the cleaned area selected by the program since the
total number of areas on a bag is limited to 10. Cleaned area is limited
to the 10 to 100 percent range which appears to embrace the observed
field conditions. The model can be easily adjusted to handle 20 sub-
areas per bag if, for example, the cleaned area were as low as 5 percent.
The type of output desired is controlled by the x and y axis specification
and the print diagnostics. The x axis is limited to a length of 24
inches and the y axis to a length of 12 inches. If no values are speci-
fied, 6 and 5 inches, respectively, are used for the x and y axis
lengths. If step-by-step values of individual' area and bag flows and
drags are desired, print diagnostics should be specified as TRUE.
A description of the card input formats to be used is presented in
Appendix A.
A sample of the program output is shown in Table 54. If the print diag-
nostics have been specified as TRUE, the type of information shown in
the table will be output for every time increment. The total number of
areas per bag (compartment) here is eight, three of which will be
cleaned to achieve a fractional cleaned area of 0.375. Drag, S, for
each area on each bag (compartment) and the entire bag (compartment) is
o
output in metric units, N-min/m . Note that compartment (bag) 5 is off
line, as indicated by a drag of 102° and a velocity of 0.0. Flow velo-
cities are reported in m/min. The next line gives the time, T, min.,
system pressure drop, DELP, N/m2, system flow velocity, DELQ, m/min and
o
outlet concentration, g/m . These terms are summarized along with the
individual bag (compartment) flow rates after all calculations have been
371
-------
Table 54. SAMPLE PROGRAM OUTPUT WITH SUPPLEMENTARY DEFINITION OF TERMS
8AG-DRAG= AREA 1
AREA 2 AREA 3
JL_. L.J7E+Q3.. i.77_E+Q3 1..71£+03
2
3
4
5
6
1.T7E+03
l.T7E+03_
1.76E + 03
I. OOE+20
1.77E+03
_BAG-FLPW=_ .. __JREA I
1
2
3
4
5
6.
.-JTrJl'Q.
T =
CAKE 9
SbAG' 0
Q'UAG 0
8.57E-01
8.SBE-01
8.586-01
8.58E-01
0.0
_ 8. 57 E -01
0
BAG 1
6.00
1.77E+03
I.T2£+fl3
1.76E+03
1«.OQE+20
1.77E+03
_AREA 2
8.57E-01
.. 8..58E-01
S.58E-01
8.58h-01
0.0
8.57E-01
1.77E+03
U7E+03
1.76E+03
L.OOE+20.
1 .77K+03
_A8iA 3.
3.57E-01
SL.SJlfe— OJ
8.58E-01
6.5BE-C1
0.0
8.576-01.
DELP= lilt.
BAG i.
10. 00
.19776+02 9.2291E+02
.1_513E+04 o.
15L8E+04
.1001E+01 0.9972E+00
BAG 3
..14.00
9.2621E+02
O.I524E+04
0.993oE+Ou
AREA 4
1.77E+03
1.77E+03
i»X7E.+0-3
1.76E+03
l.OOE+20
1.77E+03
_ A_RtA 4.
8.57E-01
8.58L-01
8.58t-01
8.5St-01
ij.0
8.57t-01
. . DELQ=
BAG 4
18.00
J.
1
I
1
1
1
8
8
8
8
0
8
ARtA 5
.77E+03
.77E+03
.77E+03
.76E+03
.OOE+20
.77E+03
AREA..... 5
.57E-01
.58E-C1
.58E-01
.58E-01
.O
.57E-01
.8309
BAG
22.00
AREA 6
1.
1.
1.
1.
.1.
1.
/2E+03 I
23E+03 1
246+03 1
25E+03 1
OOE+20 1
216+03 1
AREA 6 _
1.
1.
1.
24E+OC 1
23E+00 1
22t+00 1
1.21E+00 1
0.
1.
5
9.2969E+Oi 9.2874E+Oi
0.1530E+04
C.9697E+00
O.lOOOc+21
0.1514E-16
0 0
25E+00 1
C
BAG
2.00
ARtA 7
. 22E +03
.23E+03
.24E+D3
,25t+03
.OOE+20
.2 IE +03
AREA .7
.24E+30
.23£+00
.22E+00
.21E+CO
.0
.25E+00
L3KiCEtvTRA
6
ARtA 6
1^ i2£*03
1.23E+03
1.24E+03
1.25E+03
l.OOE+^0
1.21E+03
AREA, a
1.24E+00
I.i3h+00
1.22E+00
1.21E+OC
0.0
1.2SE+00
TION= ,«;45:
BAG
~ Tf, =r
S6AG
1.5LE+&3
1.52E+03
1-^52 t+u3
1.53fc+03
1.00E+2Q
1.51t+03
QBAG
l.OOfc+00
9.97t-01
9.94fc-01
9.VOE-01
1.51E-17
l.OuE+00
>t-02 v
\
9. 1676E+02 ••""" "' •""•"'- r^-— ^ A
0.1S08E+04
0.1C04E
+01
OJ
Notes:
BAG DRAG - Areas 1 through 8, N-min/m , drag for individual areas
SBAG - Drag for entire bag, N-min/m^
BAG-FLOW - Areas 1 through 8, m/min, velocities for individual areas
QBAG - Average velocity for entire bag, m/min
T - Test time or operating time after cleaning, min
DELP - System pressure drop, N/nr
DELQ - System (average) velocity, m/min
CONCENTRATION - System outlet concentration, g/m3
CAKE - Indicated bag loading, g/m^
-------
completed. Print diagnostics do not affect the summary table. The
amount of material cleaned from a bag is also output to the right of
concentration. This wieght dumped (dislodged) is reported as grams of
material per unit of bag area (m ). The last four lines are again a
summary of operating times,, individual bag loadings, drags and flow
velocities. The loading is the average fabric loading for the bag. The
indicated time below the bag number is a measure of how long a bag has
been operating after a cleaning. Bag 6 will be the last to be cleaned.
After each cycle is completed, the average system flow, pressure drop
and penetration are printed regardless of the print diagnostics speci-
fications. These are averaged over the time simulated up to that point
rather than over an individual cycle. In addition, after each cleaning
cycle is completed, the average penetration for that cycle is output.
The print diagnostics affect this output.
In addition to the tabular output, the program generates four graphs:
system pressure drop versus time, system flow rate versus time, in-
dividual compartment (bag) flow versus time, and total penetration
versus time. To avoid a cluttered graph, the individual compartment
(bag) flow graph is limited to five compartments (bags).
PREDICTIVE VALIDATION
Introduction
The filtration model described in the preceding section was tested by
introducing measured and calculated input parameters for operating coal-
fired utilities boilers at Sunbury, Pennsylvania and Nucla, Colorado.
It is recognized that the validation procedures cannot be considered as
completely independent checks because certain of the field measurements
of the above plants were used to develop and/or to refine the laboratory
tests that constituted the principal basis for the modeling process. On
373
-------
the other hand, the extent to which the measured and predicted performance
characteristics agree with one another, suggests strongly that the original
concepts introduced in this study relating to cleaning and dust penetration
phenomena represent the correct modeling approach for fly ash/glass fabric
systems.
System Parameters
Dust and fabric properties as well as field data analyses for both Sunbury
and Nucla operations have been presented in Sections IV through VI of this
report. Detailed field measurements at both installations have also been
Q Q
described in separate reports. '
The basic input parameters for modeling the Nucla and Sunbury operations
are summarized in Table 55 under the subheadings Operating Parameters
and Fabric and Flow Parameters. The system parameters were selected from
specific field tests (Sunbury and Nucla) rather than using average operat-
ing conditions. The fabric and dust properties for the Nucla system were
based mainly upon the field data described in Section VI of this report.
On the other hand, the continuous cleaning procedures used at the Sunbury
installation did not allow for direct field determination of some dust/
fabric parameters. However, because the GCA test fly ash was quite similar
to the Sunbury dust, the laboratory measurements with the Sunbury fabric
were considered to provide a good index of field conditions.
Nucla Data Inputs - The cleaning process at Nucla was controlled by fabric
2
pressure loss with a resistance of 1200 N/m (4.8 in. water) actuating the
cleaning cycle. Because the cyclical cleaning of six compartments reduced
2
system resistance to well below 1200 N/m , the system operated with all
compartments on-line for extended, "2 hour periods, prior to again reaching
the pressure level actuating cleaning. The actual cleaning sequence for
each Nucla compartment is summarized in Table 56. It should be noted that
during the 240 seconds (4 min) that each compartment is isolated from the
374
-------
Table 55.
DATA USED FOR MODEL TRIALS WITH THE NUCIA
AND SUNBURY FABRIC FILTER SYSTEMS
Operating parameters
Number of compartments
Cleaning cycle time, min
Compartment cleaning time, min
Face (filtration) velocity, m/min
3
Inlet concentration, g/m
2
Maximum pressure, N/m
Gas temperature, K
Reverse flow velocity, m/min
Fabric and dust parameters
Effective drag, S-, N-min/m
Specific cake resistance, K2, N-min/g-m
3
Residual drag, S , N-min/m
K
Initial slope, K^, N-min/g-m
Residual loading, W , g/m
o
W*, g/m
Nucla
6
24
4
0.824
2.6
1160
412
0.0415
434
0.76a
-
-
50
-
Sunbury
14
32.67
1.4
0.545
5.19
-
442
0.300
352
1.6a
80
7.54
30
46
Pleasured at 25°C, 0.61 m/min
375
-------
Table 56. NORMAL CLEANING SEQUENCE FOR EACH
NUCLA COMPARTMENT3
Event
Settle
Repressure
Settle
Shake
Settle
Repressure
Settle
Interval
Duration,
seconds
54
15
56
10
56
15
34
17
Damper positions
Main damper closed, repressure damper closed
Main damper closed, repressure damper open
Main damper closed, repressure damper closed
Main damper closed, repressure damper closed
Main damper closed, repressure damper closed
Main damper closed, repressure damper open
Main damper closed, repressure damper closed
Main damper open, repressure damper closed
Initiate next compartment cleaning
Table 13 from Reference 8.
376
-------
main system, the bag (or compartment) undergoes two separate cleanings
(collapse and reverse flow) and two separate, low intensity shakings.
Field observations indicated no appreciable difference in performance as
the result of the added shaking.
Since the estimated shaking frequency was 4 cps and the amplitude appeared
to be no greater than 0.5 in., the acceleration introduced by shaking is
less than 1 g. Hence, once equilibrium adhesive levels have been reached
due to multiple perturbations of the fabric surface, the added shaking and
a second collapse are not expected to have a significant effect on dust
removal.
Therefore, the total time involved with the cleaning of a. single compart-
ment has, for filtration purposes, been subdivided into the two intervals
shown in Table 57. The first 30-second period describes the total time
that an additional reverse flow must be accommodated by the 1-1 compart-
ments remaining on-line. The second 210-second interval represents the
period when the on-line 1-1 compartments see only the increased flow due
to reduced fabric area.
Table 57. SIMPLIFIED CLEANING SEQUENCE PER NUCLA COMPARTMENT
USED IN PREDICTIVE MODELING
Event
Repressure
Settle,
shake
Filtration
Duration,
seconds
30
210
240
Operation
Main damper
Main damper
closed
closed, repressure damper open
closed, repressure damper
Total off-line, cleaning, period per
compartment
Rather than treating the reverse flow period as an intermittent function
while cleaning is taking place, the total reverse air volume has been pro-
rated over the complete 240-second cleaning cycle. The net result is that
the average reverse flow velocity is reduced to 0.042 m/nin. The above
377
-------
simplification facilitates the data handling process while still taking
into account the average effect on system resistance and particle pene-
tration. The 17-second "left-over" time interval after 240 seconds was
neglected because its inclusion would have required the use of smaller
time increments in the model.
The input data used for the Nucla modeling is presented in Tables 55 and
58, the latter showing the formal computer printout. Since the actual
filtration time between cleaning cylces is lengthy, about 2 hours, com-
pared to the overall cleaning time, 24 minutes, the bags operate with com-
paratively high fabric loadings for a major portion of their on-line time.
Additionally, the distribution of fabric loading is essentially uniform
over the latter part of the filtration cycle so that a satisfactory field
estimate of the specific resistance coefficient, K , for the dust could
be made. On the other hand, it was not possible to extract sufficient
information from the field data on the other descriptive parameters, K ,
* R
S , S and W , used to define the system drag/fabric loading relationships
R E
nor was it possible to determine directly the total fabric dust holdings
for the loaded Nucla bags at the time of the field survey. Hence, it was
necessary to estimate S in conjunction with the measured K value to de-
£i i
termine the approximately drag versus fabric loading characteristics for
the field system. Therefore, laboratory measurements with Nucla fabric
test panels and the GCA fly ash (which was slightly finer than the field
aerosol) were used to provide the best estimate of S.,.
li
The above step led to the choice of the linear drag model since it involved
only one estimated parameter, S , rather than the three additional values,
ft E
SR, K and W required for the nonlinear model. Because of the extended
filtration periods with all filters on-line, the early filtering phase
with recently cleaned fabric surfaces constituted a relatively small frac-
tion of the total filtering period. Thus, it appeared that any nonlin-
earity in the drag/fabric loading relationship might be ignored in the
Nucla case. It is emphasized here that a few special, but comparatively
378
-------
Table 58
TEST RUN.JL_fl*2£. NUCLA BAGHQUSE SIMULATION-LINEAR
PRINTOUT OF INPUT DATA FOR BAGHOUSE ANALYSIS3
.NUMBER 0_F COMPARTMENTS=
CYCLE TIME=
CLEAN TIME=
TOTAL AREA RUN TIME=
.NUMBER OF .CYCLES MODELED=
NUMBER OF INCREMENTS PER BAG=
Q/A_=YELPCITY =
CONCENTRATION^
SE=EFFECIIVE BAG .DRAG- [
K2=CAKE RESISTANCE AT .61 M/MIN=
24.00000 MINUTES
4.00000 MINUTtS
0.0 MINUTtS
20 CYCLES
2 INCREMENTS
'0.82400 M/MIN
2.6006+00 G/M3
4.340E+OZ N-MIN/M3
7.600E-C1 N-MIN/G-M
-j
*** WR=RESIPUAL LQADING= .... _. _ 50.0 G/M2
INITIAL CAKE LOADING= 806. G/M2
"PRINT ~DIAGNOSTICS= --.—.-.- ^
_ CO]^JANI_f RESSUREf _ _ O.Q N/M^
"MAXIMUM PRESSURE= 1.160E+03 N/MZ
"~M*= .0 G/M2
K0= ~ 0.0 N-MIN/G-M
My^JLJ^!JcQSji'nrr o.2339E-ci CP
SR=RESiOUAL DRAG-~ " ....... "" O.O N-MIN/M3
TEMPERATyRE= 4.1200EtQ2 DEGREES KELVIN
CAKED AREA= 6.2000E-01
CLEANED" ARE A= " o.3?50E+oo
a
, „. All measurements referred to gas temperature of 412°K except for K-.
-------
simple field tests might have been used to establish the drag versus
fabric loading relationship had the need for these measurements been
89
anticipated in the preceding field studies. '
In validating the predictive model with Nucla field measurements, it is
emphasized that the starting point is a given set of field output param-
eters which one attempts to relate to the measured input parameters via
the modeling route.
One of the first terms to define is the fraction of fabric surface that
is cleaned in any given bag compartment when the cleaning process (collapse
or mechanical shaking) is carried out. The above determination is
readily made because once steady-state filtration conditions have been
established, the amount of dust deposited over the period between the
initiation of successive cleaning cycles (which in the case of Nucla op-
erations involves both the cleaning cycle and an extensive filtering
period without cleaning) must equal the amount of dust dislodged during
the cleaning cycle.
Based upon the face velocity and inlet concentration values shown in the
Table 58 and a total cycle time of 150 minutes, the dust deposited over
2
this period, AW, was 321 g/m . The terminal fabric loading, W , just
2 p
before cleaning was estimated to be 850 g/m using the maximum pressure
2
level, Pm , just before cleaning, 1160 N/m , Table 58, and the linear
m
lUclX
drag model in which K- was assumed to be 0.76 N min/g m and S to be
434 N-min/m3.
WP = (Pmax/V - V /K2 (95)
and
W,, - AW - W,,
380
-------
At the present time, the calculation of ^ by Equation (64) is executed
outside the formal computer program, because of the great number of op-
erating modes that may be encountered in the field. Since these calcula-
tions are also easily performed, their exclusion from the program appears
advisable until more experience is attained with the model. With refer-
ence to the numerical values entered in Table 58, Sw and K must be cor-
E 2
rected for gas viscosity and K2 must be further corrected for velocity
as pointed out previously in Section VII, Equation (20a). Therefore, S
E
at field operating temperature (and viscosity) must be expressed as
SE = SE / field \ (96)
field ambient 1 y , . \
\ ambient /
The Kp value at field conditions is calculated as
*2 =K / Afield W Vfield \ '
field ambient! u ,. I V .. I iy/;
\ ambient/\ ambient/
Sunbury Data Inputs - The Sunbury cleaning process consists of back-to-
back cycles with all compartments on-line for brief, ~ 1 min, periods
between each compartment cleaning. The actual cleaning cycle, presented
on a compartment basis, is shown in Table 59. Reduced to its simplest
terms, each compartment is off-line for 83 out of the 140 seconds asso-
ciated with the cleaning of each compartment. For 51 seconds out of the
83 second period when 1-1 (13) compartments remain on line, an additional
reverse flow of 0.49 m/min must be accommodated by the on-line compart-
ments. Again, because this flow persists only for the time fraction
51/83, its average value over the compartment cleaning cycle reduces to
0.30 m/min as indicated in Table 55.
A special feature of the Sunbury system is the air (sweep) flushing of
the reverse flow manifold to minimize dust deposition. Practically speak-
ing, this process, which requires about 125 seconds for every seven com-
partment cleanings, increases the on-line time of all compartments by 250
seconds per cleaning cycle. Thus, in redefining the cleaning cycle for
• 381
-------
Table 59. NORMAL CLEANING SEQUENCE FOR SUNBURY3 COMPARTMENTS
Step
1
2
3
4
5
6
1-4
5,6
Event
Settle
Reverse flow
Settle
Filtering
Collapse duct
sweeping
Filtering
Repea
Duration,
seconds
17
51
15
39
80
45
t Steps 1 t
854
125
1,958
Operation
Main damper closed, repressure damper closed
Main damper closed, repressure damper open
Main damper closed, repressure damper closed
All compartments on line
Sweep valve open, all compartments on line
All compartments on line
irough 6 for second group of seven compartments
Cleaning interval for 7 compartments
Sweeping interval for 7 compartments
Total elapsed time per cycle
No. of compart-
ments cleaned
lb
\
)
7
0
14
U)
CO
Excerpted from Table 7, Reference 9
Steps 1 through 4 repeated seven times and Steps 5 and 6 one time for one-half the cleaning cycle
-------
easier computer treatment, the cycle has been restructured as shown in
Table 60. In the modeling process, the actual time that each compartment
is off-line remains at 83 seconds, but the on-line time associated with
the sweep cleaning is spread uniformly over each compartment cleaning in-
terval. With respect to selecting time increments for the Sunbury op-
erations, a basic time division of 140 seconds was chosen so that the be-
ginning and end of each compartment cleaning interval could be properly
described. The above interval was further subdivided into four increments
so that intermediate resistance and penetration variations could be re-
solved by the program.
i
Table 60. SIMPLIFIED CLEANING SEQUENCE PER SUNBURY COMPARTMENT
Steps
1
2
3
Event
Settle
Reverse flow
Filtering
Duration
32
51
57
140
Operation
Main damper closed, repres-
sure damper closed
Main damper closed, repres-
sure damper open
All compartments on line
Cleaning interval per
compartment
No. of
compartments
cleaned
la
b
aOne compartment off line during Steps 1 and 2
bTotal cleaning cycle = (140 sec/comp)(14 comp) = 1960 seconds
The fabric and dust properties and system operating parameters for the
Sunbury installation have been presented in Table 55. Summary of all data
inputs used in the modeling process are given in Tables 61 and 62, re-
spectively, for the linear and nonlinear drag models. Since continuous
cleaning is used at Sunbury, the fraction of cleaned area, ac, could not
be determined in the same manner as that for the Nucla plant. Instead,
the average fabric loading was first determined by weighing several
loaded Sunbury bags (see Section VI) after removal from the system.
383
-------
TEST RUN # 0422 SUN8URY BAGHOUSb
Table 61
SIMULATION-LINEAR
PRINTOUT OF INPUT
DATA FUK bAGHOUSE ANALYSIS
oo
NUMBER OF COMPARTMENTS=
CYCLE TIME=
CLEAN TIMt =
TOTAL AREA RUM t!ME=
NUMBER OF CYCLfcS MODELED**
NUMBER OF INCREMENTS PER BAG
Q/A=V£LOCITY=
CONCENTRATION=
SH=EFFECTIV£ BAG DRAG=
K2=CAKE RESISTANCE AT .61
REVERSE FLOW VELOCITY=
WR=RfcSIOUAL LOADING^
INITIAL CAKE LOADING^
PRINT DIAGNOSTICS=
CONSTANT PRESSURE=
MAXIMUM PRES"SUI?E =
W*=
K0=
MU=GAS VISCOSITY^
SR=RESIDUAL
TEHPEHATURE=
CAKED AREA=
14
32.67000 MINUTES
1 .^GCOO MINUTES
0.0 MINUTES
14 CYCLES
4 INCREMENTS"
0.54500 M/MIN
5.190E+00 G/M3
3.520E+02 N-MIN/M3
'1.600E-I-60 N-MIN/G-M
0.3000 M/MIN
30.0
30.0
0.0
0.0
G/M2
G/M2
N/M2
N/M2
.0
G/M2
C.7540E+01 N-MIN/G-M
0.2456E-01 CP
8.0000EV01 N-MIN/M3
4.4200E+02 DEGKtES KELVIN
5.5500E-01
CLEANED AREA=
O.I429E+00
-------
Table 62
00
Ui
TEST RUN # 0422 SUN6UKY BAGHOUSE SIMULATION-NON LINEAR
PRINJOUl OF INPUT DATA FOR BAGHOUSE ANALYSIS
NUMBER"OF COMPARTMhNTS- 14
CYCLE TIME= 32.67000 KINL'TtS
CLEAN TIM£= " " 1.4000O MINUTES
TOTAL AREA RUN TIMe= _ 0.0 MINUTEi
NUMBER'"OF'CYCLES MODELED= ~ "]A CYCLES"
NUMBER OF INCREMENTS PbK BAG = 4 INCREMENTS
Q/A=VELOCITY= C.545CO M/MIN
CONCENTRATION= 5.190E+CO G/M3
SE=EFFECTIVE BAG DRAG= 3.520E+02 N-MIN/M3
KZ=CAKE RESISTANCE AT .61 M/MIN= 1.600E+00 N-MIN/G-M
REVERSE FUO"W"TECDCirr= "" 0.3COO M/MIN" "
WR=RESIDUAL LOADING^ 30.0 G/M2
INITIAL CAKE LOADING^ 30.0 G/M2
PRINT DIAGNOSTICS=__ ^ ___ _ F
"T01^STAlTT~TrFCFS*SljR"F=~" ""~""~'~ "" 0.0" N/M2
MAXIMUMPRESSURE^ 0.0 N/M2
W*= _ 46.00 G/H2
K0= 0.7540E+01 N-MIN/G-M
SR=RESIDUAL DRAG= 8.0000E+01 N-MIN/M3
TEMPERATURE^"" " 4.4200E+02 DEGREES KELVIN
CAKED AREA= ti.5500E-01
CLEANED AREA=
0.1429E+OC
-------
The fractional area cleaned, a =0.145, calculated on the basis of the
average fabric loading of the bags (compartments) and the quantity of
dust added to the filter system over the cleaning cycle, the latter de-
fined by the C , V and t values given in Table 55. The values shown in
i i #
Tables 61 and 62 for K2> KR, SR, SE and W were based on laboratory mea-
surements with both used and new Sunbury fabric and GCA fly ash. Since
the size properties for the GCA and the Sunbury fly ash were very similar,
it was considered acceptable to use the laboratory findings to describe
the dust-related parameters involved in the modeling process.
Nucla Plant - Model Validation
Predicted Versus Actual Resistance Characteristics - The actual pressure-
time curve for a typical Nucla field test (Run No. 1) is shown in Fig-
ure 128. These data, which were traced from a field strip chart, also
apply to the operating and dust-fabric parameters summarized in Tables
55 and 58.
The predicted pressure-time curve, Figure 129, developed from the linear
model and the data inputs appearing in Table 58. shows good agreement with
the actual measurements. Peak pressure traces were generally lower
during the cleaning cycle because the reverse flow air was averaged over
the cycle rather than using the transient spike values. The multiple
peaks shown in Figure 128 synchronize quite well with the two brief
"repressuring" operations indicated in Table 56. Note that the extra
pair of pressure spikes per compartment cleaning are not displayed on
the predicted curve because of the averaging process.
Selected reference points for comparing the actual and precicted resis-
tance measurements are outlined in Table 63.
386
-------
00
--J
ro
i
O
-------
u>
oo
00
s. ao
40.00
80.00
120. 00 1 GO. 00
230. OJ ;40.00 280.00
TIME (MINUTES]
320.00 360.00 400.00 440.00 480.00
Figure 129. Test run No. 0422 Nucla baghouse simulation - linear pressure
versus time graph
-------
Table 63.
PREDICTED AND MEASURED RESISTANCE CHARAC-
TERISTICS FOR NUCLA FILTER SYSTEM
Maximum resistance during
cleaning
Initial resistance follow-
ing cleaning
Maximum resistance just
before cleaning3
Time between successive
cleaning cycles
Actual
N/m2
1700
850
1160
in.H20
6.8
3.4
4.7
150 min
Predicted
N/m2
1520
720
1160
in.H20
6.1
2.9
4.7
188 min
Fixed value for predicted conditions.
The main differences between the actual and predicted resistance versus
time curves are (1) the average resistance is slightly lower for the
predicted curve and (2) the range between final and initial resistance
values exclusive of the cleaning intervals is slightly higher for the
predicted curve.
The above results would be expected if the estimated value for the frac-
tion of cleaned area were too large.
If a lower a value were assumed, a smaller reduction in resistance
would take place and the interval between cleaning cycles would also
reduce. In the special case where the dust removed during the cleaning
cycle equals exactly the amount deposited over the same period, the fa-
bric operating resistance can be maintained at the same level. However,
failure to keep up with the deposition rate will automatically drive the
system to a new, higher equilibrium operating pressure. In the extreme
case, lack of fan capacity, bag rupture or other irreversible changes
would necessitate a complete reevaluation of the filter system design.
389
-------
Predicted Velocity Relationships - Total or average system velocity is
shown, as a function of time in Figure 130. The average velocity is based
on a constant volume flow rate, Q, and the total number of compartments
(and fabric) in the system. Therefore, during the 24-minute cleaning
cycle, the average velocity also remains constant except when reverse
flow air is used. The short-term increases in flow velocity shown in
Figure 130 are due to the addition of reverse flow air. Because the re-
verse flow was averaged over the entire cleaning cycle rather than over
the actual transient (~15 second) period, the velocity spikes do not
appear in the computer printout.
Figure 131 is a graph of the individual compartment velocities for compart-
ments 1 through 5 as a function of time. A pressure spike appears when
each compartment is taken on- and off-line. This explains the zero velocity
points which are indicated as each of the six Nucla compartments is succes-
sively isolated over the 24-minute cleaning intervals. After 166 minutes
of filtration with all compartments on-line, it can be seen that the velo-
city range for the "just" and the "first" cleaned compartments falls roughly
within +2.5 percent of the average value. Hence, it is reasonable to assume
that the fabric surface loadings have returned to nearly uniform levels.
Predicted Penetration - Total (overall) system fractional penetration
for the Nucla filter installation is presented as a function of time in
Figure 132. The emission characteristics of the system are best analyzed
by starting at a point of minimum system penetration, roughly 5 x 10 at
188 minutes. The initial penetration increase from 5 to 9 x 10 during
the cleaning of the first compartment is due to the increase in on-line
compartment velocities when one compartment is taken out of service.
According to Equation (96), Section VIII, particle concentration is ex-
pected to vary approximately as 2.2 power of the face velocity. Therefore,
the observed penetration increase appears reasonably consistent with the
fact that average face velocity has been increased by 20 percent.
390
-------
oo
40.00 80.00 120.00 160.00
203.00 240. oa ?ec. oo
TIME (MINUTES)
320.00 362.03 IKS. 00 440.00 480.00
Figure 130. Test run No. 0422 Nucla baghouse simulation - linear flow rate versus
time graph
-------
0BRG « 1
A3RG * 2
+8RG » 3
X3RG » 4
OJ
^O
ro
3
EJ
u. 30
40.08
80.00
120.00 TeoTocT
200.00 24S. OC 283.00
TIME (MINUTES)
320.00 360.00 430.00 440.00 4BO. 00
Figure 131. Test run No. 0422 Nucla baghouse simulation - linear individual
flow rate graph
-------
10
VO
U>
0 00
«0.00
80.00
120.00 160.00
2QQ.00 240.00 2BO. CD
TIKE IHlMUTESi
320.00 360.00 400.00 440.03
Figure 132. Test run No. 0422 Nucla baghouse simulation - linear penetration
versus time graph
-------
When the next compartment is returned to service, its dust loading is
nonuniform with part of the fabric cleaned to its residual loading, a^,
and the remainder having a loading equivalent to that just prior to
cleaning. Since the cleaned areas has a much lower resistance to flow
and, thus, a higher face velocity than that for the uncleaned area, its
efficiency is lower. The above process accounts for the second major
increase in penetration to its maximum level. As more dust is added to
the compartments, penetration decreases significantly to a new minimum
value until the next compartment is returned to service, :at •which point
emissions again rise.
As the cleaning cycle progresses, the availability of partially loaded,
previously cleaned areas tends to reduce the high face velocity through
the most recently cleaned area. Hence, one observes a gradual reduction
in peak emission levels over the time frame of the cleaning cycle. When
the cleaning cycle is completed, penetration initially decreases rapidly
due to a preferential deposition on the most recently cleaned areas. The
velocities and fabric loadings in all compartments then decrease slowly
to an asymptotic value such that penetration is nearbly constant until
another cleaning cycle is begun. The average efficiency for the 190-
minute predicted cycle is 99.81 percent compared to an actual test result
of 99.79 percent. Although the above results suggest excellent agreement
between modeling theory and observed performance; i.e.:
Predicted penentration = 0.19 percent
versus
Observation penetration = 0.21 percent
it is recognized that the above statistic derives from a limited data base.
Sunbury Plant - Model Validation
Predicted Versus Actual Resistance Characteristics - The actual resistance
history for Run No. 1, Sunbury plant, is presented in Figure 133 (see
Table 21, Reference 9). Because of the time scale compression, the
394
-------
11P.M.
Figure 133. Pressure drop history of Sunbury baghouse
(Reference 9)
- run No. 1
395
-------
cyclical pattern for the resistance changes is obscured such that one
can perceive only the nominal maximum and minimum pressure excursions.
However, Figure 43 in Section VI of this report, which shows a greatly
expanded time scale (the latter generated by special high speed chart
tests), indicates clearly the various pressure steps corresponding to
the description of the Sunbury cleaning cycle outlined in Table 59.
The predicted curves for resistance versus time for the linear and non-
linear drag models are given in Figures 134 and 135. Both curves were
developed under conditions where the filtration began with clean fabric
and where the continuously cleaning cycle was initiated immediately.
2
The actual average baghouse resistance was approximately 635 N/m
2
during the test period, with a range of about 150 N/m . After about
5 hours of simulated operation, the average resistance as predicted by
2 2
the linear model leveled off at 550 N/m with a range of 100 N/m . On
2
the other hand, an average resistance of 525 N/m with a range of 125
2
N/m was predicted by the nonlinear model. In both cases the resistance
reached a near steady state value after 4 to 5 hours time indicating how
rapidly the system approaches equilibrium. Limited field data, Sec-
tion VI, suggest, however, that a leveling off in both resistance and
emission characteristics may not be reached until 10 days to 2 weeks
on-line performance.
The discrepancy between observed and predicted resistance characteristics
may also be the outcome (see Nucla resistance analysis) of assuming too
high a value for the cleaning parameter, a . If less dust were removed,
the system would automatically seek a new and higher equilibrium resis-
tance. It is believed that the ratio of the resistance range to the
average value will diminish at the higher operating resistances although
the absolute difference between maximum and minimum pressure excursions
may increase.
396
-------
76-CO 114.00 152.00 190:00 228.00 26600
TINE. (.MINUTES)
304. 00
342.00
380. 00
418. 00
456. 00
Figure 134. Test run No. 0422 Sunbury baghouse simulation - linear pressure versus
time graph
-------
OJ
<£>
00
38. 00
76.00
114.00 152.00
190.00 128.00 266.00
TIME IMINUTES)
304.00 342.03 38C. 00 418. OC 4S5. CG
Figure 135. Test run No. 0422 Sunbury baghouse simulation - nonlinear pressure
versus time graph
-------
The lower average resistance and greater resistance range predicted by
the nonlinear model is due to the lower value assumed for the starting
drag of the cleaned fabric. The nonlinear model uses an S value of
about 80 N-min/m in contrast to an S£ value of 352 N-min/m? for the
linear model. Therefore, when a cleaned compartment is first brought on
line, its drag and that of the system are lower for the nonlinear model.
In both cases the resistance just before a compartment is returned to
2
service is about 600 N/m .
Both models are useful for design purposes. The linear model, which
predicts a safely conservative average resistance, is a good estimator
of power consumption. On the other hand, the nonlinear model provides
a better index of transient pressure changes which might be important
with respect to fan selection. Again, the accuracy of all predictions
depends upon the reliability of the data inputs used in the modeling
process.
Predicted Velocity Relationships - The average compartment velocity for
compartments one through five as a function of time is shown in Fig-
ures 136 and 137 for the linear and nonlinear models, respectively. It
was arbitrarily assumed that the velocities (and hence areal densities)
were the same for all compartments at the initiation of the cleaning
cycle. Once the cleaning cycle begins, however, the sequential compart-
ment cleaning in conjunction with the data inputs given in Tables 61 and
62 will automatically drive the system to its steady state regimen
characterized by the velocity gradients shown in Figure 136.
The minimum or zero velocity excursion occurs when a compartment goes
off-line and the peak value indicated for each compartment represents
the high transient velocity occurring when a cleaned filter is first
returned to service. Reference to the point arrays on both curves shows
that the second highest velocity for each compartment (0.6 minutes later)
is very much lower. The data point dispersion for the nonlinear model
399
-------
K I
c 2
» 3
XBRG e 4
-------
r.M
08RG « 1
ABflG * 2
+BRG * 3
X3RG * 4
« 5
oo
33. 00 76. 00
pWMJ ^W «B^» ^ ^—1 ^B^ 1 <•»•» p-^ ^W *^" «W«
114.00 152.00 130.00 229.00 266.00 304.00 342.00 380.00 418.00 456.00
TIME (MINUTES)
Figure 137. Test run No. 0422 Sunbury baghouse simulation - nonlinear individual
flow rate graph
-------
covers a greater range than that for the linear model for the same reason
given for the resistance models; i.e., lower resistance during the early
filtration phase (or nonlinear region of the drag model) leads to higher
velocities through the just cleaned areas.
Predicted Penetration - The velocity variations described previously have
a direct impact upon penetration behavior as might be expected, Fig-
ures 138 and 139, with the greater range in penetration also associated
with the nonlinear model. In contrast to the Nucla operations in which
there were lengthy time intervals between cleaning, the back-to-back
compartment cleaning cycles leads to a constantly changing effluent con-
centration whose average value at any time is represented approximately
by the midpoint of the envelope curves.
The average steady state penetration values for the nonlinear and linear
models are 0.27 and 0.20 percent, respectively, as compared to an actual
field value of 0.06 for the specific test being modeled. Again, the
difference between the two predicted values (linear and nonlinear models)
is attributed to the difference in local face velocities immediately
after cleaning. Since the local velocities through the just cleaned
areas as predicted by the nonlinear model are higher than those predicted
by the linear model, the higher penetration is expected.
The higher penetration values predicted by the model as compared to the
observed field results are attributed to the following factors:
1. The estimate of the cleaned fractional area, ac, based upon
interpretation of field and laboratory data may be on the
high side.
2. The estimates of dust penetration properties based upon fabric
surfaces cleaned in the laboratory may be on the high side.
Such might be the case if field levels for residual dust hold-
ings, WR, were higher due to increased interstitial deposition
of dust in the bulked fiber region.
402
-------
3.00 38.00 76.0C 114.00 152.02 130.00 228.00 265.00 304.00 342.00 380.00 418.00 456.00
TIME CMINUTES)
Figure 138. Test run No. 0422 Sunbury baghouse simulation - linear penetration
versus time graph
-------
HiyjHrii^iHt^'Hi^HiH^{ii;iii?i^?;^c?:^r.?«H
Thw^iiinwwIeraiiiniOT
II """"" """
0.00 38.00 76.00 114.00 152.00 190.00 228.00 266.00 304.00 342.00 380.00 418.00 45600
TIME (MINUTES)
Figure 139. Test run No. 0422 Sunbury bagtiouse simulation - nonlinear penetration
versus time graph
-------
3. The field data relate to a test with fabric bags that have seen
over 2 years' service.
Despite the fact that data are limited, inspection of Table 11A, Section VI,
indicates that the dust penetration levels for recently installed Sunbury
bags showed significantly higher penetrations than those that had seen
over 2 years' field service. The same trend was also exhibited for par-
ticle concentrations over the same time span, Figure 41, Section VI.
Excerpted data from Table 11A provide an improved picture of the predic-
tive potential of the new model. If one considers that the mathematical
relationships developed within this study for calculating dust penetra-
tion were based upon tests with new fabric panels (generally less than
24 hours total use) the agreement between the linear model predictions
and actual field observations is reasonable and safely conservative with
respect to the nonlinear model, Table 64.
Table 64. COMPARISON OF OBSERVED AND PREDICTED FLY ASH PENE-
TRATION VALUE, SUNBURY INSTALLATION
Runs3
22,23,24
25,26,27
28,29,30,31
1 through 21
Time
period
3/20/75 to
3/22/75
3/23/75 to
3/25/75
3/26/75 to
3/29/75
1/08/75 to
2/14/75
Bag
service
lifeb
1.5 days
4.5 days
7.5 days
>2 years
Percent penetration
Measured
0.15
0.11
0.09
0.07
Predicted
Linear
model
0.20
—
—
Nonlinear
model
0.27
—
—•
aSee Table 16, Section VI, and Table 1, Reference 9.
bAverage values for indicated time frame.
405
-------
SUMMARY OF MODEL HIGHLIGHTS AND DIRECTION FOR FUTURE WORK
The mathematical model developed within this study represents a new and
very effective technique for predicting the average and instantaneous
values for resistance and emission characteristics during the filtration
of coal fly ash with woven glass fabrics.
Two basic concepts used in the model design: (1) the quantitative des-
cription of the filtration properties of partially cleaned fabric surfaces
and (2) the correct description of effluent particle size properties for
fabrics in which direct pore or pinhole penetration constitutes the major
source of emission, have played important roles in structuring the pre-
dict ive equa t ions.
A third key factor in the model development was the formulation of ex-
plicit functions to describe quantitatively the cleaning process in terms
of the method, intensity and frequency of cleaning. By cleaning we refer
specifically to the amount of dust removed during the cleaning of any one
compartment and the effect of its removal on filter resistance and pene-
tration characteristics.
The derivation of two supporting mathematical functions based upon labora-
tory and field experiments provided improved definition of the specific
resistance coefficient, K , for use in the modeling equation. The first
function describes K~ in terms of a specific surface parameter, S , that
relates to the typical polydisperse particle size distributions encoun-
tered in the field. The second relationship takes into account, as
others have also indicated, that K« should be expressed as an increasing
function of face velocity.
Limited information on long-term filter service, ~2 years, suggests that
woven glass fabrics now used for coal fly ash filtration will exhibit a
gradual increase in drag in the range of 125 N-min/m per year (0.15 in.
406
-------
H20/ft/tQin per year). Penetration characteristics under the above con-
ditions show a slight tendency to improve once preliminary fabric condi-
tioning takes place.
The success of the model, based upon limited field confirmations sum-
marized in Table 65, dictates very strongly that it be further evaluated.
In that the required data inputs have been identified, it is believed
that a field oriented program with limited laboratory back-up would sa-
tisfy the final validation needs. Minor changes in existing compliance
type sampling methods and apparatus should provide the key data for re-
sistance fabric loading relationships that are fundamental to the appli-
cation of the model. Additionally, such measurements should help to
confirm the present observation that electrical charge and/or humidity
factors do not play an important role in fly ash filtration with glass
fabrics.
Extending the above program to other dust/fabric combinations will pro-
vide a rational basis for treating heretofore unresolved problems in
many field filtration applications.
407
-------
4s
O
00
Table 65. SUMMARY TABLE SHOWING MEASURED AND PREDICTED VALUE FOR FILTER SYSTEM PENETRATION
AND RESISTANCE, COAL FLY ASH FILTRATION WITH WOVEN GLASS FABRICS
PENETRATION
Data source
Test
case
A
B
C
Description
Nucla, CO
Table 11B,
Run No. 1
Sunbury, PA
Table 11A,
Run Ho. 1
Table 11A,
Runs 22, 23,
24
Testing
period
9/21/74
1/08/75
3/20/75
to
3/22/75
service
life
6 months
2 years
1.5 days
Percent penetration
,a
Average
0.21
0.06
0.15
Predicted
Linear model
Average Cleaning
0.19
0.20
0.20
1.52
-
-
Nonlinear model
Average Cleaning
-
0.27
0.27
-
-
-
RESISTANCE
Test
case
A
B
Measured
Average
1030
635
Maximum
cleaning
1700
710
Maximum
1160d
710e
Minimum
850d
5606
Predicted
Linear model
Average
972
620
Maximum
cleaning
1521
663
Maximum
1160d
663e
Minimum
720d
567e
Predicted
Nonlinear model
Average
.
560
Maximum
cleaning
.
609
Maximum
_
609e
Minimum
_
489e
Based on field measurements. See references 8 and 9.
All values listed as average depict overall system performance (penetration and resistance) for combined cleaning
and filtering cycles.
All values listed under cleaning describe performance parameter during cleaning only.
Maximum-minimum with Nucla tests indicate resistance immediately before and after cleaning.
Maximum-minimum with Sunbury tests indicate values for envelope curve.
-------
REFERENCES
1. Billings, C. E. and J. E. Wilder. Handbook of Fabric Filter Technology
Volume I, Fabric Filter Systems Study: CCA/Technology Division. EPA '
No. APTD 0690 (NTIS No. PB-200-648). December 1970.
2. Symposium on the Use of Fabric Filters for the Control of Submicron
Particulates - April 8-10, 1974, Boston, Massachusetts. Proceedings
published in JAPCA, December 1974. 1140-1197 pp. See also: EPA-650/
2-74-043 (NTIS No. PB237-629/AS). May 1974.
3. Draemel, D. C. Relationship Between Fabric Structure and Filtration
Performance in Dust Filtration. Control Systems Laboratory. U.S.
Environmental Protection Agency, Research Triangle Park, N.C. Report
Number EPA-R2-73-288 (NTIS No. PB222-237). July 1973.
4. Turner, J. H. Performance of Nonwoven Nylong Filter Bags. EPA-600/
2-76-168a (NTIS No. PB 266-271/AS). December 1976.
5. Harris, D. B. and D. C. Drehmel. Fractional Effluency of Metal Fume
Control as Determined by Brink Impactor. (Presented at the 66th Annual
APCA Meeting. Chicago. June 1973).
6. McKenna, J. D. Applying Fabric Filtration to Coal-Fired Industrial
Boilers. Enviro-Systerns and Research, Inc. Roanoke, Va. Control
Systems Laboratory, U.S. Environmental Protection Agency, Research
Triangle Park, N.C. Report No. EPA-650/2-74-058 (NTIS No. PB237-117/AS).
July 1974.
7. Hall, R. R. and R. Dennis. Mobile Fabric Filter System. Design and
Field Test Results. GCA/Technology Division, Bedford, Mass. Control
Systems Laboratory, U.S. Environmental Protection Agency, Research
Triangel Park, N.C. Report No. EPA-650/2-75-059 (NTIS No. PB246-287/AS).
July 1975.
8. Bradway, R. M. and R. W. Cass. Fractional Efficiency of a Utility
Boiler Baghouse - Nucla Generating Plant. GCA/Technology Division,
Bedford, Mass. Control Systems Laboratory, U.S. Environmental Protec-
tion Agency, Research Triangle Park, N.C. Report No. EPA-600/12-75-013a
(NTIS No. PB246-641/AS). August 1975.
9- Cass, R. W. and R. M. Bradway. Fractional Efficiency of a Utility
Boiler Baghouse: Sunbury Steam-Electric Station. GCA/Technology
Division, Bedford, Massachusetts. Control Systems, Laboratory, U.S.
Environmental Protection Agency, Research Triangle Park, N.C. Report
No. EPA-600/2-76-077a (NTIS No. PB253-943/AS). March 1976.
409
-------
10. Dennis, R. and J. E. Wilder. Fabric Filter Cleaning Studies.
GCA/Technology Division, Bedford, Mass. Control Systems Laboratory,
Research Triangle Park, N.C. Report No. EPA-650/2-75-009.
January 1975.
11. Jorgensen, R., (ed.). Fan Engineering, 7th Edition. Buffalo, N.Y.,
Buffalo Forge Co., 1970. 729 p.
12. Snyder, C. A. and R. T. Pring. Design Considerations in Filtration
of Hot Gases. Ind Eng Chem Process Des Dev. 47:960. 1955.
13. Robinson, J. W., R. E. Harrington, and P. W. Spaite. A New Method
for Analysis of Multicompartment Fabric Filtration. Atmos Environ.
1^:499-508, 1967.
14. Spaite, P. W., G. W. Walsh. Effect of Fabric Structure on Filter
Performance. Amer Ind Hyg Assoc J. 24:357-365. 1963.
15. Solbach, W. Derivation of a Computational Method for Multichamber
Cloth Filters on the Basis of Experimental Results. Staub (English).
29(l):28-33, 1969.
16. Noll, K. E., W. T. Davis, S. P. Shelton. New Criteria for the
Selection of Fabric Filters for Industrial Application. (Presented
at 66th Annual APCA Meeting. Chicago. 1973.)
17. Williams, C. E., T. Hatch, L. Greenberg. Determination of Cloth
Area for In-ustrial Air Filters. Heat/Piping/Air Cond. 12:259-263,
April 1940.
18. Durham, J. R. Filtration Characteristics of Fabric Filter Media.
U.S. Department of Health, Education and Welfare, Public Health
Service, National Air Pollution Control Administration. Interim
Report. 1969.
19. Anonymous. Dust Collection With Fabrics of Nomex Aramid Fiber.
DuPont Corp., Textile Fiber Department. Information Memo No. 336.
1974.
20. Noll, K. E., W. T. Davis, P. J. LaRosa. The Generation and Evalu-
ation of Fabric Filter Performance Curves from Pilot Plant Data.
(Presented at 68th Annual APCA Meeting. Boston. 1975.)
21. LaRosa, P. J. Private Communication. 1975.
22. Stinessen, K. 0. The Filter Simulator - A Research Backed Method
for Filter Fabric Evaluation. (Original source uncertain.
Personal Communication, 1976.)
410
-------
23. Fraser, D. and G. J. Foley. A Predictive Performance Model for
Fabric Filter Systems. I. Intermittently Cleaned Single-Compartment
Systems. (Presented at 67th Annual APCA Meeting. Denver. 1974.)
24. Leith, D. and M. W. First. Particle Collection by Pulse-Jet Fabric
Filter. (Presented at 68th Annual APCA Meeting. Boston. 1975.)
25. Dill, R. S. A Test Method for Air Filters. Trans ASHVE. 44:379.
26. Butterworth, G. A. M. and M. Platt. Filter Fabric Selection and
Design: Consideration of Air Permeability and Fiber Characteristics.
Fabric Research Laboratories, Inc. Dedham, Mass. (Presented at
EPA Fabric Filter Symposium. Charleston, S.C. March 1971).
27. Pedersen, G. C. Fluid Flow Through Monofilament Fabrics.
Filtration and Separation. 11(5):586. 1974.
28. Zimon, A. D. Adhesion of Dust and Powder, p. 112. Plenum Press,
New York. 1969.
29. Fuchs, N. A. The Mechanism of Aerosols. The MacMillan Company,
New York. 1964.
30. Dennis, R. (Ed.) Handbook on Aerosols. U.S. Energy and Development
Corporation. TID-26608, NTIS, Springfield, Va. 1976.
31. Dennis, R. Collection Efficiency as a Function of Particle Size,
Shape and Density: Theory and Experience. JAPCA. 24(12)1156.
December 1974.
32. Borgwardt, R. H. and J. F. Durham. Factors Affecting the Performance
of Fabric Filters. (Presented at 60th Annual Meeting of the
American Institute of Chemical Engineers. New York. 1967.)
33. Dalla Valle, J. M. Micromeritics. Second edition. Pitman
Publishing Corp. New York. 1948.
34. LSffler, F. Investigating Adhesive Forces Between Solid Particles
and Fiber Surfaces, Staub, (English Translation), 26, 10. June 1966.
35. LSffler, F. The Adhesion of Dust Particles to Fibrous and Particulate
Surface, Staub, (English Translation), 28^, 32. November 1968.
36. Boehme, G., et al. Adhesion Measurements Involving Small Particles,
Trans Instr Chem Engrs., London. 40, 252. 1962.
37. Corn, M. The Adhesion of Solid Particles to Solid Surfaces, 11.
J Air Pol Cont Assoc., ri, 566. 1961.
38. Corn, M. The Adhesion of Solid Particles to Solid Surfaces, 1. A
Review, J Air Pol Contr Assoc., 11, 523. 1961.
411
-------
APPENDIX A
EFFECT OF SEQUENTIAL PORE CLOSURE ON
SHAPE OF RESISTANCE VERSUS FABRIC LOADING CURVE
Asssume that a sequential pore bridging or closure process follows an ex-
ponential decay pattern in which the rate of pore closure, - dN/dt, at
any time is proportional to the number of remaining open pores, N.
dN/dt = - kN (98)
If the bridging process is instantaneous, the total pore area, at any time,
t, determines the instantaneous pore velocity for a constant volume flow
rate, Q, i.e.;
V = Q/N A
P
where A is the individual pore cross section.
P
Since the number of open pores, N, at any time also determines the total
pore area, the integration of Equation (98) following substitution of N A
P
for N leads to the expression
N A = (N A ) exp (-kt) (99)
p P o
or alternately as
V = Q/(A')o exp (kt)
where A" refers to total pore area.
P
412
-------
If the pore area and depth for the open pores remain unchanged and lami-
nar flow persists, the instantaneous resistance, P, will then depend only
on the instantaneous velocity, i.e.;
P = f(V) = Q/(A') exp (kt) (100)
P o
If both volume flow rate and inlet dust concentration are assumed to be
constant, the dust loading, W, upon the fabric is at all times proportional
to the filter operating time. Hence, Equation (100) in derivative form
appears as
dP/dW = k Q/(A') exp (kW) (101)
in which the slope is always increasing.
413
-------
APPENDIX B
INPUT PARAMETERS FOR ESTIMATING FIBER
EFFICIENCY IN SUBSTRATE LAYER
In the following section, the rationale for the input parameter values
given in Table 44 is presented in more detail.
2
• Given a fabric areal density of 312 g/m and a nominal
fabric thickness of 0.04 cm (400 ym), the fabric bulk
density is 0.78 g/cm-*.
• Assuming the glass fiber density to be 2.2 g/cm, the
porosity of the fabric becomes
6 = 1 - p = 1 - 0.78/2.2 = 0.646
• Because 25 percent of the pores are lost in both the warp
and fill directions due to yarn contact, the effective
pore volume is reduced roughly by a factor of 2.
(0.646)(0.5) = 0.323
• If 10 percent of the total fabric weight is assumed to be
distributed within the effective pore volume, 0.323 cm^
per cm3 of fabric, the following estimate of the fiber
volume fraction in the filter is made:
(0.78 g/cm3) 0.1/0.323 = 0.241 g/cm3 (bulk density)
P = 0.241/2.2 = 0.11
e = 1 - P = 0.89
• Based upon microscopic examination of the fabric structure,
the yarn shape and the fabric thickness, the minimum pore
dimension appeared to be about 100 urn as shown in Figures
28 and 30. Examination of Figure 140 (an excerpted section
of Figure 28 with added dimensional notations) indicates
that the dimension characterizing average pore cross section
at the surface of the fiber substrate is roughly 0.67 times
that of the superficial dimension. In conjunction with the
adjustment for corrected porosity, the average gas velocity
414
-------
within the substrate will be approximately 6.3 times
greater than the superficial value, i.e.;
(1.01/0.363)(1.5)2 = 6.3
FIBER SUBSTRATE
EDGE VIEW
Figure 140. Estimation of pore cross section in fiber
substrate region
According to a previous analysis of fabric structure, Figure 9, the dimen-
sion characterizing the surface of the substrate was also assumed to be
0.67 times that of the superficial layer. In the former instance, it was
shown that the development of a dust layer starting at the substrate sur-
face and continuing until the superficial fabric surface was reached,
provided a rational explanation for the curvalinear filtration range for
drag versus loading curves. Hence, the estimated gas velocity of 6.3
cm/sec at the surface of the fiber substrate appears as a reasonable value.
415
-------
APPENDIX C
DETERMINATION OF CONSTANTS USED IN
DUST PENETRATION MODEL
The reasons for choosing the general form of the model and the constraints
placed upon it have been discussed in Section X. Only the mechanics of
developing the equations and their related constants will be discussed
here.
The general form chosen to model dust penetration was:
Pn = Pn + (Pn - Pn ) exp (-aW) (102)
where Pn = actual penetration
Pn = steady-state penetration
s
Pn = penetration at W = W (residual loading or W = 0)
O R
W = absolute cake loading, W, minus the residual, W
R
-a = initial slope of the penetration versus loading curves
The original outlet concentration versus loading curves obtained from the
bench scale tests were first replotted as penetration versus fabric load-
ing, see Figure 141. Penetration here is defined as the outlet concen-
tration minus
concentration:
3
tration minus the residual concentration, 0.5 mg/m , divided by the inlet
C - C
Pn =
i
Steady-state penetration values were determined at the points where the
curves assumed nearly horizontal paths. Extrapolation of the curves in
416
-------
|
e
10" '
9
e
I0-2
c
a.
1 °
t-
<
te.
t e
UJ
* 10*
IW
9
2
10-"
9
2
to'8
•»
i*
:o
">
•v
„
-
k -
h
-
_
;
r
^
k
1 I 1 1 1 I —
TEST FACE VE! ^CITY (m/mln ) I
V 9t C.59
+ AVERAGE 0.61
O 96 1.58
Q 97 3.38
—
I
-
a a I
o
—
+ I
0
o o -
-
+ _
V -
+
V
+ —
}
-& 40~ «0~" 80 100 120 H
FABRIC LOADING, W'g/m2
Figure 141. Penetration versus loading for bench scale tests
417
-------
the initial decay region to a loading of zero yielded values for the ini-
tial penetration, Pn . Since Pn values were all within the penetration
r o o
range of 0.09 to 0.11 for velocities of 0.39 to 1.52 m/min, a characteris-
tic value of 0.1 was assumed for Pn irrespective of face velocity. The
o
initial slope of the curve, -a, was determined by solving Equation (102)
for a after substituting proper values for Pn , Pn , W", and the penetra-
?
tion corresponding to W. A value of 20 g/nr was chosen for W. Steady-
state penetration, Pn , and the negative of initial slope, a, were then
s
plotted versus velocity in Figures 142 and 143, respectively. A summary
of the data used in the analysis is presented in Table 66.
The choice of the equations used to describe the curves was arbitrary.
The plot of the logarithm of steady-state penetration versus velocity
curve appeared to have the same form as a drag versus loading curve with
one exception. Since steady-state penetration can never exceed a value
of 1, any mathematical relationships must account for this constraint.
The form of the nonlinear drag model is:
S = S + K2 W + (KR - K2)W* exp (1 - exp - W'/W* (28)
If the term K-W is dropped from Equation (28), the curve will actually
level off. Therefore, the form of the equation used to describe the re-
lationship between steady-state penetration and velocity was
In (Pn ) = InX + Y (1 - exp -V/Z ) (104)
S
The constants X, Y, and Z were determined by substituting the actual values
for Pn and V for three velocities, 0.61, 1.52, and 3.35 m/min into Equa-
S
tion (104) and solving the three equations simultaneously. The steady-
state penetration corresponding to a velocity of 0.39 m/min was not used
for determining the constants since its value was essentially zero. The
final equation is:
Png = 1.5 x 10"7 exp J12.7 [l - exp (-1.03 V)][ (105)
418
-------
10"
o
!c
(X
UJ
i
o
<
Ul
in
to
,-4
10
2
,-5
2 -
10
-e
O ACTUAL VALUE
EQUATION 105
1234
LOCAL FACE VELOCITY, V , m/min
Figure 142. Steady state penetration as a function of velocity
419
-------
0.3
o»
o
UJ
>
o:
ID
O
UJ
a.
o
0.2
O.I
©ACTUAL VALUES
— EQUATION 107
0.0
01234
LOCAL FACE VELOCITY , V, m/min
Figure 143. Initial slope of penetration versus loading
curve as a function of velocity
420
-------
Table 66. DATA USED TO DETERMINE CONSTANTS
IN DUST PENETRATION MODEL
Test
Face velocity, V, m/min
3
Inlet concentration, C , g/m
o
Steady-state penetration, Pn
2 s
Pn at W = 20 g/m
a
Pn
o
98
0.4
8.09
0
8.3xlO"4
0.24
0.11
Averagea
0.61
7.01
_«;
5.7x10
8.5xlO"3
0.125
0.087
96
1.52
5.37
_3
3.6x10
1.9xlO"2
0.094
0.098
97
3.35
4.60
_2
3.25x10
-
-
-
Average for test numbers 65, 68, 69, 70, and 99-
421
-------
The equation that describes the relationship between the initial slope,
-a, and velocity was determined in the same manner except that the form
of the equation chosen was:
a = r/V + t (106)
s
The constants r, s, and t were determined by substituting values of a
and V for velocities of 0.39, 0.61, and 1.52 m/min into Equation (106)
and solving the three resultant equations simultaneously. Insufficient
data were available to determine the variation in penetration with loading
at low loadings for the highest velocity and, therefore, a slope was not
determined for that test. The resultant equation for the slope is:
a = 3.59 x 10"3/V4 + 0.094 (g/m2)"1 (107)
where V is in m/min.
Equation 107 and the actual slopes for the three velocities are plotted
in Figure 143.
422
-------
APPENDIX D
BAGHOUSE COMPUTER PROGRAM DESCRIPTION
SPECIFICATION OF OPERATING TIMES FOR BAGHOUSE COMPUTER PROGRAM
The pressure drop versus time curve for a three compartment system shown
in Figure 144 will be used here to illustrate the various times associ-
ated with cleaning and filtering cycles in the program. Vertical (step)
increases and decreases in pressure drop represent compartments being re-
moved from service and returned to service, respectively. A complete
cycle is represented by the cleaning cycle and the period when all com-
partments are filtering, the latter designated as total area run time.
The cleaning cycle in this example is composed of three individual com-
partment cleaning cycles. Each cycle consists of one period where all
compartments are filtering between individual compartment cleanings and
a second period where one compartment is taken off-line for cleaning.
The time increment used in the program is determined from the individual
compartment cleaning cycle time and the number of increments per compart-
ment specified in the input data. Thus, if five increments are specified,
an individual compartment cleaning cycle is split up into five equal time
increments.
Required inputs, regardless of the type of cycle employed, are (1) the
cleaning cycle time, (2) the individual compartment cleaning time, and
(3) the number of time increments desired per individual compartment
cleaning cycle. These three values will define the cleaning cycle.
423
-------
9
Q.
O
QC
Q
liJ
(T
3
V)
tn
UJ
(£.
Q.
to
-F-
TIME INCREMENT
INDIVIDUAL COMPARTMENT CLEANING TIME
INDIVIDUAL COMPARTMENT CLEANING CYCLE TIME
.b
— CLEANING -
CYCLE TIME
TOTAL AREA RUN TIME
ONE COMPLETE CYCLE
TIME
BEGINNING OF FIRST
CLEANING CYCLE
I
BEGINNING OF SECOND
CLEANING CYCLE
NOTES!
0 VERTICAL INCREASES AND DECREASES INDICATE A COMPARTMENT BEING TAKEN OFF
OR PUT ON LINE, RESPECTIVELY.
& ALL COMPARTMENTS ON LINE. FOR TIMED CLEANING CYCLE INITIATION, THIS MUST BE INPUT.
FORA PRESSURE INITIATED CLEANING CYCLE, THIS VALUE IS INDETERMINATE.
FOR CONTINOUS CLEANING CYCLES, THIS VALUE IS ZERO.
c EXAMPLE OF A THREE COMPARTMENT SYSTEM.
Figure 144. Description of time specifications for baghouse computer program
-------
How the cleaning cycle is initiated is determined by the total area run
time and the maximum pressure drop specification, see Table 67. If clean-
ing is to be continuous (i.e., back-to-back cleaning cycles), total area
run time and maximum pressure drop should be specified as zero. If the
cleaning cycles is to be initiated on a time basis, a value should be
input for the total area run time and the maximum pressure drop should be
specified as zero. For pressure-controlled cleaning cycles, the maximum
pressure drop should be specified and a value of zero should be entered
for the total area run time.
Table 67. INPUT SPECIFICATIONS FOR VARIOUS
TYPES OF CLEANING CYCLES
Type of cleaning Maximum Total area
cycle initiation pressure drop run time
Continuous
Time
Pressure drop
zero
zero
Specify value
zero
Specify
value
zero
PROGRAM DESCRIPTION
A listing of the baghouse computer program is presented in Table 68.
The variables and arrays used within the program and their definitions
are given in Table 69. Finally, the format for input data is shown in
Table 70.
425
-------
Table 68. BAGHOUSE SIMULATION PROGRAM LISTING
//* BAGHOUSE PROGRAM IBM J70 MTH CALCO*P PLOTTER
//* 1<>76 GCA TECHNOLOGY ROGER STERN - DOUG COOPER
//* BAGHOUSE SIMULATION PROGRAM- IHM 370- ZETA PLOTTER
//* 1977 GCA TECHNOLOGY DIVISION) HANS KLtMM- RICHARD DENNIS
// EXEC FORTGlCG,AcCT*COST,PARM.GOi'SIZE=175K'
//FORT.SYSIN DO *
CALL MODEL 10
100 DO 500 1=6,15 20
END FILE I 30
500 REWIND I UO
CALL SCRIBE 50
STOP 60
END 70
SUBROUTINE CAKDRG(SZEWQ,*DEL»VnK,wSTAP,ZKZERn,ZK2, VEL.CDRAG) 60
c SUBROUTINE OF BAGHOUSE u/77/HAK-RD GCA TECHNOLOGY DIVISION <»o
C-CALCULAUS CAKE DRAG 100
C-ZK23SPECIF ic CAKE RESISTANCE OF CAKE AT o.bi M/MIN, N-MIN/G-M no
C-WDEL»TOTAL FABRIC LOADING ON AN AREA OF FABRIC, G/M2 120
OwRrRESIDuAL FABRIC LOADING ON AN AREA OF FABRIC, G/«? 130
C-WSTAR= CONSTANT CHARACTERISTIC OF DUST AND FABRICr G/M2 1«0
C-ZKZERO= INITIAL SLOPE OF DRAG VS. LOADING CURVE, N-MI/V/G-M 150
C-CORAG=CAnfc D«AG»S, N-MIN/M3 170
If (wSTAR.GT.l .£-20) GO TO 10 190
C-LINEAR MODEL 2°0
CDRAG=ZK2V*rtOEL ?10
GO TO 20 220
10 WPPIMt*WDEL-wR 230
EXPO=-«VPRIME/WSTAR 2«0
!F(ExPn.LT.-30.) ExPOa-30. 250
C-NON-LINEAR MODEL 260
CDPAG*ZK2V*wpRIME+rzKZERO-ZK2V)*wSTAR*n ."EXP(EXPO)) 270
20 RETURN 280
END 290
SUBROUTINE PENETCCZERO, WEIGHT, VF.L,rtR, PEN) 300
c SUBROUTINE OF BAGHOUSF 4/77/HAK-RD GCA TECHNOLOGY DIVISION iio
C-CALCUL^TES TOTAL PENETRATION j20
C-CZERO»INLET CONCENTRATION, G/M3 JJO
C-WEIGHT=TOTAL FABRIC LOADING ON AN AREA OF FABRIC, G/M2 3ao
C-VELsVELOCITY, M/MIN 3,Q
C-WR=RESIDUAL FABRIC LOADING ON AN AREA OF FABRIC, G/M2 3fcO
CS*0.0005
A=aoo.
IF(VEL.GT.l.E-9) A=0. U 16/( VEL*3.281 ) **
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
SUBROUTINE *ODEL
C SUBROUTINE OF BAdHOUSF 12/1/RwS-DC GCA TECHNOLOGY DIVISION 510
c SUBROUTINE OF BAGHOUSE u/77/HAK-Ro GCA TECHNOLOGY DIVISION sao
C-MAIN DRIVER SUBPROGRAM 530
C-ALL T's ARE TIMES, WIN 5ft0
C-ALL W'S ARE CAKE LOADINGS , G/M2 550
C-ALL S'S ARF DRAGS, N-MIN/M3 560
C-ALL P'S ARE Pf NfcTRATIUNS 570
C-ALL C's ARE CONCENTRATIONS 580
C-A BAG IS A COMPARTMENT 590
C-ZK2rSPECIFIC CAKE RESISTANCE OF CAKE AT 0.61 M/MIN, N-MIN/G-M 600
C-WRsRtSlCUAu FABRIC LOADING ON AN AREA OF FABRIC, G/M2 610
C-WSTAR* CONSTANT CHARACTERISTIC OF DUST AND FABRIC, G/M2 620
C-ZKZEROs INITIAL SLOPE OF DRAG VS. LOADING CURVE, N-MIN/G-M 650
C-$ZERO=RfcSIDUAL DRAG, N-MJN/M3 6UO
C-TEMPK=GAS TEMPERATURE, DEGRESS KELVIN 650
C-ACAKE»CAKED AREA, THAT PORTION OF A BAG WHICH IS NOT CLEANED 660
C-ZK2MU»VISCOSITY CORRECTION FOR 'SPECIF 1C CAKE RESISTANCE 670
C-NsNUMBER OF COMPARTMENTS OR BAGS 660
C-T=CLEANING CYCLE TIME.MIN 690
C-NT=TOTAL NUMBER OF CYCLES TO BE MODELED 700
C»M=NUMBER OF TIME INCREMENTS PER BAG 710
C-SMALQaAVERAGE SYSTEM VELOCITY, IF OPERATING AT CONSTANT TOTAL FLOW, M/M 720
C-CZ£RO=INLET CONCENTRATION, G/M3 730
C-LOIAG»PRINT DIAGNOSTICS 740
C-TLAG»TIMt PERIOD FOR WHICH ALL BAGS ARE ON LINE AFTER ENTIRE CLEANING 750
C-CYCLE 760
C-CONSP=PRESSU«E DROP IF OPERATING AT CONSTANT TOTAL PRESSURE, N/M2 770
C-OPSTOP*PRESSUHE DROP AT WHICH CLEANING IS INITIATED, N/M2 760
C-WS»CAKE LOADING AT ZERO TIME, G/M2 790
C-VRFLOsREVERSE AIR VELOCITY FOR ONE BAG, M/MIN 800
C-SE*EFFECTIVt CAKE DRAG, N-MIN/Mi 810
COMMON/INPUT1/N.T,NT,M,SMALQ,CZFHO,TCLEANFLDIAG,CONSP,TLAG,DPSTOP 820
COMMON/ I NPUT2/ZKZERO.SZERO.TEMPK, ACAKE 630
CQMMON/RESIS/SE,ZK2 8«0
COMMON/ I NPUT3/WR, rtSTAR, WS,VRFLO 850
COMMON ZK2MU 860
DIMENSION IDUM(10),PDP(3),PDQ(3),PT(J),PPS(3),PQ<3,5) 870
DIMENSION TIME(IOO) ,OLDTIM(100),CAKE(100) 880
DIMENSION ftD( 1 0, 100 ) ,SBAG ( 100 ),«BAG( 100), 3(10,1 00 ),QAPE»( 10), PCI 0) 890
LOGICAL LCONP,LDIAG 900
DATA DRAG.BAGlfBAGa/'AREA'j'SBAG^'OBAG'/ 9 I 0
C READ INPUT DATA 920
CALL READIT 930
CALL READIM 940
C-INITIALIZfc DATA
LCONPs. FALSE.
IF(CONSP.GT.l,E-6) LCONPs.TRUE. 980
IRtPTsN/10 + I 990
C-OETERMINE TOTAL NUMBER OF AREAS ON A BAGCIAREA) AND 1000
C-NUMBER TO BE CLE ANEOCNARE A) 1010
ERR=0.01 102°
7 Iel./(l.-ACAKE)*0.5
J=l
IF(£RR.GT.0.06)GO TO 9 1050
DO 8 1=1,10 1060
DO 8 J=1.I 107°
ATESTsFLOAT(J)/FLOAT(I)
IF(ATEST.LE.Cl.-ACAKEtERR).AND.ATEST.GE.(l.-ACAK£-ERH)) GO TO 9 1090
8 CONTINUE 110°
427
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
ERR«ERRt0.01 1110
GO TO 7 1120
9 NAREAsJ 1130
IAREA=I " it«0
AREAsl ./1AREA 1150
CLAREAsAREMNAREA 1160
WRITEC6,210) CLARfA '"" 1170
210 FORMATUX, 'CLEANED AREAa ' ,T«0,E 10.U) 1180
WRITEC6,220) _ 1190
220 FQRMATMH1) 1200
DO 5 Ist.IAREA 1210
QAREA(I)=3MALQ 1220
IF(SMALfJ.EQ.O.)QAREA(I)«DELP/SZERO 1250
IF(wSTAR.EQ.O..AND.SMALQ.EQ.O.)QAREA(I)sDELP/SE 1240
DO 5 1BAG=1»N 1250
OLDTIM(IBAG)s«2 1260
TIME(18AG)*-1 1270
5 WD(I,IBAG)=WS __ 1280
IFBAG«0 " ........ 1290
PAVRsO.O 1300
TCONT«0.0 1310
OTLAST«0.0 1320
PENTOTsO.O 1330
PAVTOT»0.0 13«0
CZEROE=C7ERO 1350
DPAVG*0.0 1360
QAVG=0.0 1370
"~ " ""
IF(TLAG.LT.l.E-9) TCORRaO.O 1390
TMODsTlAGtT l«00
IF(DPSTOP.GT.O.)TMOO»1.E+20 " "" ' 1«TO
IF (OPSTOP.GT.O.) TCORR=0,0 1420
KJ=0 1430
C DETERMINE DRAG THROUGH FABRIC 1«50
SFAB'SZERO 1460
IFfWSTAR.LT.l.E:-20) SFAB=SE U70
C LOOP ON TIME 1480
00 300 JLOQPsliMAXJ 1«90
DELT«T7«7N ..... ~ ~" 1SOD
TTESTSAMOD(TCONT+0.01,TMOD)-0.01 1510
IF(TCpNT.tT.l.E-9.0R.TTEST.LE.-0.01.0R.TTEST.GE.0.01) GO TO 12 1520
OAVGNV(QAVG-QS"VSTM*DTLAST)X2./TCONT - .......... J5JO
PAVMOW=(PAVTOT"PENTOT*DTLA8T)/2./TCONT 15«0
DPAVGN«(OPAV6-DELP*ORAST)/2./TCOMT _ 1550
C -WRITE Av'fRAGe PT«rsJrURE DROT,FLdF AND PiENE'TRA'TItiSrUP' TO TrMEVfCONT '~ " '" 1560
NRITEC6.230) TCONT,PAVNOI*,DPAVGN,OAVGN 1570
12 CONTINUE 1580
IFHTEST.GT.n Go TO 11 1590
C EXTRA PASS FOR CLEANED BAG 1600
C-BAG LOOP 1 1610
DO 13 IBAG*1,N ........ "" •• •—- r62"<
IF(OLDTIM(I8AG).LE.TIMECIBAG)> GO TO 13 1630
IF8AG«IBAG 16^0
TCONT«TCONT+.01 1650
GO TO 1« 1660
13 CONTINUE 1670
C-END Of 8*6" LOOP T "" 1680
11 IFBAG»0 1690
DtLT*T/M/N j700
JTIME«JLOOP-1 ~ --- 17 ft
C. DETERMINE TIME 1720
428
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
1 MO
. . ]?ao
IF (lit ST. Gl .T.ANn.TTt.ST.Gfc.CT + TLAG-OELT)) DELT=DELT+TCORH 1 7SO
SSYS1H=0.0 1760
l.>rLtl=l>bLT 17;0
VKFLLlV. = 0.0 1/80
IKl 01 AG) WRITE (6, 16) (DRAG, I. 1 = 1, I ARE A), BAG1 1790
16 KiPMATMX, 'BAG-DRAGS', tx.H(3X,A«, IX, I?)) 1800
t-BAG LI 'OR £ 1610
OIJ 20 IHAG=1.N lygo
SBAGt IHAG)=0.0 18JO
C-AREA LOOP 1 I6y0
DO 6 I=l,lARt A jfltjQ
C-jf. HAG WAS JUST CLEANED ESTIMATE FLOW VELOCITY f-HOM LIMEAR MODEL I860
IFCSdAPf A,l(jAG).GT.l.f +19) S( I , IBAG ) =SE + WD ( I , IBAG) *ZK2 1870
IFCTCflMT.Gl . l.fc-9)UARFA(l)=DtLP/5(I,IBAG) 1S80
ODETFKMIffc 0«A(i ON EACH AREA 1PQO
CALL CAKDkGCSZERU»wU(I,lB«G).HR,WSTAH,ZKZtRO»ZK2,OAREA(I), 1<>00
* S ( 1 , 1 H A C, ) ) 1910
S( 1 , IhAG)=S( I , IbAlO tSH AB
6 SHAi;(IHAU)=SHAli(IbA
C-ENH Uf A^FA LOU^ I
SHAG( 1RAG) = 1 ,/SBAf, ( IHAG)
C OETEW^I^F Tl^t IN CYCLE 1960
IMTTF.Sl.l'T.CTtO.OOS)) GO TO 21 1970
OK'TII- ( JhAG J = T Il^t ( IB AG) 1980
TI^F(IHA(;)=Ay(.H)(lTFST + 0.01 + IBAG*T/N,T)-0.01 1990
21 1KT TtST.GT .T) Ml Id 19 2000
C-TFST H>K AN (.If-f- LINF BAG 2010
IFITCONT.L T . 1 ,t-9,AN0.1 IMt (IDAG) ,LT. ( T-TCLE AN-. 00 1 ) ) GO TO 19 2020
IF (TlwEdBAlO .LT. (T-TCLEAM-. 001) , AND. TIME(IBAG) ,GT. 0,005) GO TO 19 2030
IKTIMt(lHA(i).LT,(T-TCLtAN-. 001). AND, TTEST.LE, 0,01, AND, TLAG.GT.l.E 20aO
*"9) GO T.i 19 • 20bO
DO 22 1=1 , I A«tA 2060
22 S( I , ]HAG)=1 ,F+20 2070
2090
C-IHITPUT IMFR^tDJATfc RESULTS 2100
19 IF (LDIAfO^RITt Cb, lb)IBAG, ( S ( I . IB AG) , 1 = 1 , I AHfc A ) , SB AG ( IBAG) 2110
IS FHHMATd X, 13, 7X, 1 1 ( IX, 1PF9.2)) 2120
SSYSTM=SSYSTM+1./SBAG(IBAG) 2130
IFdiLDTIfdBAGJ.GT.TIMEdBAO.AND.TTEST.LT.CTtO.OOS)) DELTT = 0,01 21UO
20 CONTIMJt 2 ISO
C-ENO OF Bit; LOOP 2 2160
C-CALCULAlt SYoffM o« At;, PRESSURE DROP AND FLOW VELOCITY 2170
CZFkO = C,?fcRUt 2160
SSYSTM=I ,/SSYSTN 2190
l)EL(L' = SMALQ*SSYSTM*N + \/RFLOW*SSYSTM 2200
IF(LCtiNP) PFl.P = C(iNSP 2210
f j S Y S T N' = S M A L 'J + V R H. ( 1 H / N 2220
1F(ICOI--,P) USYSIM = Cf)NSP/SSYSTM/N 2250
C-CORRFCT I^LfcT CONCENTRATION FOR REVERSE FLOW AIR 22UO
tZtPO = C7tRLiE*CQSYSTM-VRFLOW/N)/OSYSTM 22SO
IF(LDIAG) wRITt(6,30)(ORAG,I»I»l/IARtA),BAG2
30 FOWMATdx, 'BAG-FLOWS'/ IX, lt(3X,A«/lX/ 12))
PF.NTOT=0.0 2280
C-BAG LOOP 3 2300
DO 60 1BAG=1,N 2310
IKTTEST.GT.T) GO TO 26 2320
DELT=OELT1 2330
IFC(TlMt(IBAG)+T/M/N).GT,(T-TCLtAN))OEL,T=T-TCLEAN-TIME(lBAG)
429
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
26 WCOMPsQ.O 2150
CAKE(1BAG)=0.0 2360
C-AREA LOOP 2 _ _ 2370
DO 26 I=l,IA»Eft "" " "" 2380"
QAREACI)*D£LP/S(1,IBAG) 2390
C-DETERM1NE PENETRATION __ 2400
CALL PENET(CZERO,WD(I,iBAG).QAREAO},WR,P*DELT*CZERO 2a20
C«RE(IBAGJ«CAKE{IB*G)+*0(I,_IBAG)*ARE_A __ ____ _ _____ 2030
27 PENTOT«PENTOt+PtI)*Af»EA*QAREA(I)/ 2470
C-OUTPUT INTERMEDIATE RESULTS 2480
_IF(LgiAG)WHlTE(6,15_)lBAG,(OAREA(I),I«l.IAREAL)»OBAG(IlAG) ___ _?«90
IF(TTEST.GT.T) GO TO 60 "" 2500
1F(OLOTIM(IBAG).LE,TIME(IBAG))GO TO 60 2510
C-CLEAN NAREA AREAS ON A BAG IF NECESSARY __ __ J?20
WOUMpsQ.O " " "" ' " " "2530"
DO 36 11=1, NAREA 2540
NCOMPsO.O ___ _2550
C-AREA LOOP "3 " .................... " 2560
DO 35 I»1,IAREA 2570
IF{WO(1,IBAG).LT.WCOMP) GO TO 35 2580
WCOMPSWD(I»IBAG) ~ "" 2590
IFAREAsI 2600
35 CONTINUE 2610
"" -' 3" ..... ' ----------- ......... " ------ '•
WDUMPswDUMP+(WD(IFAREA,IBAG)-WR)*AREA 2630
36 WDCIFAREA, IBAG)«*R 2640
6"0 CONTINUE " " " ' " " " 2650
C-ENO OF BAG LOOP 3 2660
DELT=DELTT 2665
QAVG*OAVG+(OTLAST+DELT)*OSYSTM 2680
PAVTOT=PAVTOT*PENTOT*(DELT+DTLAST) 2690
PAVRsPAVRtPENTOT*'(tJELT+6TL*St ) 2700
DTLASTsDELT 2710
K33K3+1 2720
-"- —
POP(K5)»OtLP 2740
POO(K3)«OSYSTM 2750
PPS(K3)«PENTOT ' ~ 2760
CONTOT=PENTOT*CZERO 2770
LMAX*MINO(5,N) 2780
Od"17JO~"L*l'»i:MAX - •- --------- 2790
100 PQ(K3,L)*QBAGCL) 2800
IF(K3.LT.J) GO TO 120 28J0
K3=0
C 2830
C PUNCH PLOT 2840
110 FORMATC6G10.5) 2860
WRITEC8,110J ((PT(K),POP(K)),Ksl,J) 2670
WRITE(9,110) C(PT(K),POQ(K)),Ksl,3) 2680
DO 115 LM'LMAX
115 WRITt(lUNIT,110) ((PT(K),PQ(K,L)).Krl,3)
WRITEC15,110)(PT(K),PPSCK),K«1,3)
120 IF(,NOT.LDIAG) GO TO 290
C
C PRTNT DIAGNOSTICS
430
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
WRITEC6,130) TCONT,DELP,QSYSTM,CONTOT,WDUMP ^970
130 FORMATOX/' Ti',G10.4,10X,«DELP8',G10.4,10X,'OElQ=',G10.4, 2960
& 10X,'CONCENTRATIONS',G10.4,lOx,'WEIGHT DUMPED'1,610.4) 2990
IDUW(10)*0
00 250 L=1,IR£PT
140 00 ISO K = l,10
MAXKsMINO(K,(N-10*(l-l))) 1050
ISO IDUM(K)rlDUM(10UK 3040
WRITE(6,160) (IDUM(K),K=1,MAXK) 3050
160 FORMATC5X,10(6X.'BAG SI2)) J060
«RITE(6,170) (TIMF_ S080
WRITE<6,180) CCAKE(IDUMCI)),IH,MAXK} 3090
180 FORMAT(' CAKE=',T6,1PE12.4,9E12.4) 3100
WR1TE(6,190) fSBAG(IOUM(tn,I = t,MAXK) - JnO
190FORMATC SBAG' , T6,10E 12.<4) 3120
WRITE(6,200) (OBAG(IOUM(I)),I=l,MAXK) 3130
200 FORMATC QBAG',T6,10E12.«,OPF2.0) "31^0
250 CONTINUE 3150
IF(TTEST.GT.T) GO TO 270 3160
IF(OLDTIM(N).LT,T1ME(N)) GO TO 270 3170
PAVRsPAVR/2./T 3180
MRIT£(6,2bO) PAVR 3190
260 FORMATflX,'AVERAGE PENETRATION*',1PG10.3) 3200
PAVRxO.O 3210
270 CONTINUE 3220
WRITE(6,500) 323T)
500 FORMATC///} 32
-------
60 hRITF UUNITf 70) I
70 FORMATCBAG # Ml)
75 *RJ Tfc.( 15,25) HEAD, AMP
C-PFNtTRATliiM VS HME
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
SMHRdUl INF P|.(UlN 3SKO
c simRdurjNh 10 INITIALIZE PLOTTER ii/n/75/nws-oc ibvo
C SiiHRUUT INt (..IF BAGHUUSt 4/77/HAK-RD GCA TECHNOLOGY DIVISION 3600
/i^T,NT,M,SM6LQ,CZE.RO,TCLEAN,DlAG,CONSP,TLA&,DPSTOP 3610
HtADU9) 3620
DATA Am>/'K'/ 3630
REAO(Sfio) HFAO 3640
JO FORMA I ( IX, 19A4/ TdO, Al) 3650
K»«|1b(6,lb) HtAD 3660
Ib FLIRMA1 (IX, 19A4) 3670
(Hi 20 IUMI T = 8, 10 3680
20 WHITE ( IllMI T,25) HfcAD,AMp 3690
25 FORMAT (i
-------
Table 68. (continued) . BAGHOUSE SIMULATION PROGRAM LISTING
SUBROUTINE KEADlM a200
C SUBROUTINE BAGHOUSE 1 1 /20/75/RhS-DC 6CA 4210
c SUBROUTINE OF SAGHOUSE U/TT/HAK-RD GCA TECHNOLOGY DIVISION «3aO
READ(S,15)ACAKE 1350
15 FORMAT (lOXfFlO. 5) 4360
IF(TEMPK.E0.298.) GO TO 18 ^370
IFCTEMPK.GT.l.) GO TO ISO «380
GO TO 18
150 ZMUE»1 ,i
1« wRITE(6,20)ZKZERO,ZMUE
WRITE(6,30)SZERO, TEMPK, ACAKE
20 FORMAT(
?' KOS',TIO.EIO.«,» N-MIN/G-MV
8' MU=GAS VISCOSITY*', T40,EiOi«,' CP()
30 FQRMATC SR*RESIDUAL DRAG=i, TaO, IPElO.a, ' N-MIN/M3'/ 4«70
1' TEMPERATURE" '. T40,E10.a,' DEGREES KELVIN1/ 4480
2' CAKED AREAS', T40,E10.«/ 4490
7) 4500
ZK2MUSZMUE/1.8E-2 "510
RETURN ~ " " 15^6
END "530
433
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
SUBROUTINE READIT 4540
C SUBROUTINE Of BAGHOUSE 11/24/75/RWS-DC GCA 4550
c SUBROUTINE OF BAGHOUSE 4/77/HAK-RD GCA TECHNOLOGY DIVISION 4560
C READS AND INIT1ALI/ES 4570
C-N*NUMBER OF COMPARTMENTS OR BAGS 4580
C-T=CLEANING CYCLE TIME,HIM 4590
C-NTrTOTAL NUMBER OF CYCLES TO BE MODELED 4600
C-MrNUMBER OF TIM£ INCREMENTS PER SAG 0610
C-SMALQsAVEHAGE SYSTEW VELOCITY,IF OPERATING AT CONSTANT TOTAL FLOW, M/M «620
C-CZERO«INLET CONCENTRATION,G/M3 4630
C-TCLEAN=TIME IT TAKES TO CLEAN ONE BAG 4640
C-TCLEANSTIME IT TAKES TO CLEAN ONE BAG 4650
C- DIAG=PRINT DIAGNOSTICS 4660
C-CONfiPsPRESSURE DROP IF OPERATING AT CONSTANT TOTAL PRESSURE,N/M2 4670
C-TLAGsTIME PERIOD 'FOR WHICH ALL BAGS ARE ON LINE AFTER ENTIRE CLEANING 4680
C-CYCLE «690
C-DPSTOP=PRFSSURE DROP AT WHICH CLEANING IS INITIATED* N/M2 4700
C-WS=CAKE LOADING At ZERO TIME, G/M2 4710
C-VRFLO*REVfcRSE AIR VELOCITY FOR ONE BAG, M/MIN 4720
C-SE*EFFECTIVE CAKE DRAG, N-MIN/M3 4730
C-R2»SPECIf1C RESISTANCE OF CAKE AT 0.61 M/MIN AND 25 C/N-MIN/G-M 4740
C *SET UP COMMON VARIABLE AREAS FOR SUBROUTINES 4750
COMMON/1NPUTl/N.T,NT,M,SMALQ,CZERO.TCLEANED IAG,CDNSP,TLAG,DPSTOP 4760
COMMQN/INPUT3/wR,wSTAR,rtS,VRFLO ' ' 4770
COMMON/RESIS/SE,R2 4760
COMMON EPSLON 4790
LOGICAL tuar; ~ aaoo
C *HEAO INPUT DATA 4810
READ(5,10)N/T.TCLEAN,NT,TLAG,M,SMALQ,CZERO,SE,R2,DIAG,CONSP,wR 4620
10 FORMAT(T15,It.,2(IOX,G10.0),T75,I6/Tll,G10.0/ 4830
2 T1S, l6«3(10XiG10.0)/ 4840
3 TU.G10.5,T3S,L6,2(10X,G10.0)) 4850
READC5,40) WSTAR,DPSTOP,VRFLO " 4860
READ(5,40) WS 4870
c *INITIALIZE PLOTTER 48eo
wRITE(6,13) 4890
13 FORMATC'l') 4900
CALL PLOTIN «9}0
C *WRITE INPUT DATA - - 5930
WRITE(6,15) 4930
15 FORMATC40X,'PRINTOUT OF INPUT DATA FOR BAGHOUSt ANALYSIS1//) 4940
WHITEt"6,20)N,T,TCLEAN,TLAG,NT,M,SMALO,CZERO,SE,R2,VRFLO 4950
20 FORMATCIX,"NUMBER OF COMPARTMENTS*',T4o,i6/ 4950
2 IX,'CYCLE TIME"', T40,OPF10,5,' MINUTES1/ 4970
3 IX,'CLEAN TIMES', T40,F10.5,' MINUTTS'/ 4^0
3 IX,'TOTAL AREA RUN TIM£=',T40.F10.5,'MINUTES " 5070
rtRITE(6,25J «R,«S 5080
25 FORMATC1X,'wR»RESIDUAL LOADINGS',T40,1PG1O.i,' 6/M21/ 5090
1 U,'INITIAL CAKE'LOADING*',140,610.3,' G/W2'/) 5100
WRITE(6,30)DIAG,CONSP,DPSTOP 5HO
30 FORMATdX,'PRINT DIAGNOST ICS*1 , T40,L6/ 5120
1 IX,'CONSTANT PRESSURE*',T40,1PE10.3,' N/Mg'/ $130
2 IX,'MAXIMUM PRESSURES',T40,E10.3,' N/M2'/)
434
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
5150
«0 FOR»AT(«(lOX,GlO.Sn 5160
WRIU(6,50) WSTAR 5170
c,0 FORMATC ***'.Tao.lPG10.«,' G/M2'/) - ^^
RETURN 5190
END
435
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
GCA TECHNOLOGY
SUBROUTINE SCRIBE
C GRAPH LIBRARY 7/16/75/H"S
C VERSION 8/1/70
C CARDS-
C TITLEU-64) OPTIONS: XPOSC65-69) YPOS(70-7«) HEIGHT (75-79) K80)
C XAXIS LABEL(1-64)OPTIONS: 8EGINC69-74) UNITS OR LOGS/INCH(75-80)
C YAXIS (SAM£)
C TYPE (YAXIS-XAXIS)(SEMI,LOG-,PHOB,BAR-)(l-8)
C OPTIONS! LOG-19.J2) FOR A LOGRITHMIC BAR GRAPH
C NEW GRAPH OISTC35-40) OEFAULT=6
C X-AXIS HEIGHTC45-50) D£FAULT=2
C X AXIS L£NGTH(5S-60) DEFAULTS
C Y-AXIS L£fMGTH(65-70) DEFAULTS
c DUUBLE Axisc74) i FOR x, 2 FOR Y, 3 FOP BOTH
C SYMBOL(75-80) POINTS BETWEEN PLOT SYMBOLS
C NEGATIVE FOR SYMBOLS BUT NO LINES
C DATA X(l-10) YU1-20)
C OPTIQNt X(2i-30) Y(31-40) X(41-50) Y(51-60)
C OPTION (END,NEW,SAME)(75-78) (NEW MAKES NEW GRAPH-REPEAT ALL CARDS)
C (SAME PLOTS ON OLD GRAPH-NO X-Y AXIS)
C (79-80) (CHANGE 'SYMBOL' FOR NEXT PLOT)
DIMENSION IBUF(4000),XAR(1002),YAR(1002),PRN(50),PRBUOO)
DIMENSION XPLAB(26),YPLAB(26),XPROB(38),YPLA5(26)
REAL LOG,NEW,NEXT,NEX
R£AL*8 TAR(fc),XLAB(8),YLAB(8),SPLAB(12),SPLAT(12)
DATA XPLAB/.00,.30,.46,.65,.91,1.10,1.32,1.65,1,95,2.30,2.56,2.76,
&3.00,3.22,3.««,3.70,4.05,4.35,fl,68,4.90,5.09,5.15,5.52,6.00,0.,V./
DATA YPLAB/25*0,0,1,/
DATA YPLA5/2fl*5.0,.0,l./
DATA SPLAB/'.Ol .05 '
2 ' 40 50 6'
3 ' 99.9 9'
DATA SPLAT/' 99.99 9"
2 '0 60 50 '
3 '.2 .1 '
DATA XPROB/.O, .16, .45
'.1.2 .5',' 1 2
'0 70 80',' 90
'9.99 '/
•9.9 99.' , "5 99 98
' 40 30 ','20 10
'.01 '/
.65, .91, 1,10, 1.42,1.
',' 5 10 ',
','95 98 ',
>,' 95 90",
',' 5 2 ',
68,1.94,2.17,
' 20 30',
'99 99.5 ',
i 80 7",
' 1 .5 ',
2.43,2.62,
S3.04, 3,33, 3,63,3.88,4.24,4.53,4.92,5.21,5.57,5,89,6.28,6.63,
& 7.06,7.51,7.90,8.32,8,84,9,29,9.81,10.36,10.91,11.55,12.27,
413.09,13,95,15.OO/
DATA BLA,SEMI,LOG,P80B,BAR/' ','SEMI',»LOG-','PROS','BAR.'/
DATA SAME,NE«,ENDD/»SAME«,'NEW ','END '/
CALL PLOTSCIBUF.4000)
INUNIT»3
IOUTUN«4
NEXTsNEW ' ' "-•
10 ISYMSO
CALL PLOT(0.,-36,,-3)
CALL PLOT(0.0,2.,-3)
IPOS»0
BARXsQ.
BARYsO.
PROBXsO.
LTYP«0
20 ISYM»ISYM»1
NEX«NEXT
IF(NEXT.NE.SAME) GO TO 30
YBEG»YAR(IMAX+1)
YlNC*YAR(IMAx»2)
TITLE
5200
5210
5220
5230
5240
5250
5260
5270
5280
5290
5300
5>310
5320
5330
5310
5350
5360
5370
5360
5390
5400
5410
5420
5430
5«40
5450
5460
5470
5480
5490
5500
5510
5520
5530
5540
5550
5560
5570
5580
5590
5600
5610
5620
5630
5640
5650
5660
5670
5680
5690
5700
5710
5720
5730
5740
5750
1760
5770
5780
5790
5800
436
-------
Table 68. (continued). BAGHOUSE SIMULATION PROGRAM LISTING
30 REAO(INUNIT,40,£NDslOOO) TAB, XPOS, YPOS, CHIT, CONT 5610
40 FOR*AT(BA8,4G5.2,A1) S820
IFCABSCXPOS).LT.1.E-20) XPOS».5 5630
IFCABS(YPOS).Lf,l.E-20) YPOSs8.0-(.25*IPOS) ~ 5840
IF(CHIT,LT.1,E-20.AND.ISYM.EQ.l.AND.CONT.NE.BLA) CHlTs.21 5650
IF(CHIT.LT.1.E-20.AND.CONT.EQ.BLA) CHIT».l4 5860
«RITE(IOUTUN,4l) TAR,XPOS,YPOS,CHIT,CONT 5870
41 FORMAT(lX,6A8,3X,'XPOS*',F7.3,3X,'YPOS»',F7.3,3X,'HEIGHTa',F7.3, 5680
& 3X,'CONTB»,A1) 5890
IFCCONT.NE.BLA.OR.iPOS.EQ.O) GO TO 45 ~ ~ " 5900
XPOS*XPOS+.2 5910
DO 02 I«l,7 5920
IF(TAR(8).£Q.TAR(I)) GO TO 42 5930
CALL SYM80LUPOS-.1,YPOS,CHIT,ISYM,0.,-1) 5940
GO TO 45 _ 5950
42 CONTINUE "" ~5«60
45 CALL SYMBOL(XPOS,YPOS,CHIT.TAR,0.,64) 5970
IPOS»JPOS+1 5980
IFCCONT.NEtBLA) GO TO 30 5990
C LABELS 6000
IFCISYM.GT.l.AND.NfXT.EQ.SAME) GO TO 70 6010
READ(INUNIT,50) XLAB,XBEG,X INC " "6020
50 FORMAT(8A8,T69,aG6.2) 6030
WRITE(IOUTUN,55) XLAB,XBEG,XINC 6040
55 FORMAT(1X,8A8,3X,'XBEG=',G10.3,3X,'XINCa',G10.3) 6050
READ(INUNIT,50) YLAB,YBEG,YINC 6060
WRITE(IOUTUN,58) YLA8,YBEG,YINC _ 6070
58 FORMAT(1X,8A8,3X,'YBEG*1,610.3,3X,'YINC=«,G10.S) 6060
C TYPt 6090
READ(I NUNIT,60) YTYP,XTYP,ZTYP,XOVER,YUP,XAXL,YAXL,IDOUB.LTYP 6100
60 FORMAT(3A4,T31,4(4X,G6,2),T74,I1,I6) 6110
IF(XTYP,EQ,8LA) XTYPaSEMI 6120
IF(YTYP.EQ.BLA) YTYPsSEMI 6130
IFCYUP.LT.l.E-5) YUP*2. " 6140
YUPiYUP-2. 6150
CALL PLOT(0.,YUP,"3) 6160
IF(XAXL.LT..5> XAXL=6. 6170
IFCABSCXOVER).LT.1.E-20) XOVER=6. 6180
PMUVE»XAXL+XOVER 6190
IFCYAXL.LT.,5) YAXL«5. 6200
WRITE(IOUTUN,65) YTYP,XTYP,ZTYP,XOVER,YUP,XA XL,YAXL,1DOUB,LTYP 6210
65 FORMAT(1X,3A4,3X,«XOVER=',F6.2,3X, 6220
2 'XAXlS HT»',F6.2,5X,'XAXIS L=',F6.2,5X, 6230
J 'YAXIS L=',F6.2,5X, 6240
a tWAXISc',II,JOX,'POINTS PER TICKS',16) 6250
IF(2TYP.E
-------
Table 68 (continued). BAGHQUSE SIMULATION PROGRAM LISTING
IF(J.GT.IOOO) GO TO 90
IF(MEXT.EO.BLA) GO TO 100
IF(XAR(J-!>.LT,1.E-20.AM>,YARO1).IT.1.E-20) J = J-1
IF(NEKSYM.NE.O) LTYP=NEwSYM
90 IMAXsJ-l
iF(Nt XT.EQ.BLA) NEXT=ENDD
GO TO 102
100 CONTINUE
C SCALES AND AXIS
102 XARUMAX + 1 )=XBEG
XAR(IMAXt2)*XINC
YARU^AXtl )=YBEG
YAR(IMAX+2)=Y1NC
C CUT OFF VALUES OUT OF RANGE
IF(A6S(XINC) .LT.l .E-20) GO TO 106
IF(XTYP.EQ.PROB) GO TO 106
xBYG=xBEGf XINC*XAXL
IF (XTYP.EO.LOG) XBYG«XBEG*10**(XINC*XAXL)
DO 104 IMLOOP»1,1MAX
IF (XBYG.GT.XBEG, AND. XAR ( IMLOOP). GT.XBYG) XAR (IMLQOP) =XBYG
IF(XBYG.GT.X8EG.AND.XAR(IMLOOP).LT.XBEG) XAR ( IMLOOP) =XBEG
IF (XBYG.LT.X9EG, AND. XAR( IMLOOP), LT.XBYG) XAR( 1*LOOP) = XBY6
IF (XBYG.L.T.XBEG. AND. XAR (IMLOOP) .GT.XBEG) X AR ( IMLOOP) =XBEG
104 CONTINUE
106 IF(ABS(YINC). LT.l. E-20) GO TO 110
YBYG=YBEG+YINC*YAXL
IF (YTYP.EQ.LOG) YBYG*YBEG*1 0** ( YINC*YAXL )
DO 108 IMLOOP*!, IMAX
IF ( YBYG.GT.YBEG. AND. YAR( IMLOOP ).GT.YBYG) YAR( IMLOOP) =YBYG
IF(YBYG.GT.YBEG.AND.YARUMUOOP).LT.YBEG) YAR ( IMLOOP) =YBEG
IF ( YBYG.LT.YBEG, AND. YAR( IMLOOP). LT.Y8YG) YAR(IMLOOP)=YBYG
IF (YBYG.LT.YBEG.AND.YAR{IMLOOP).GT.Y8tG) Y AR ( IMLOOP ) a YBEG
108 CONTINUE
C CUT OFF LOW VALUES
110 XBYG«1.E-20
YBYGsl,E-20
IF(XTYP.NE.LOG) GO TO 113
DO 112 IMLOOPalr IMAX
IF(XARCIMLOOP).LT.XBYG) X AR ( IMLOOP JsXBYG
112 CONTINUE
11J IF(YTYP.NE.uOG) GO TO 115
DO IM IMUJOPal, IMAX
IF(YARnMLOOP).LT.YBYG) Y AR ( IMLOOP]*YBYG
lia CONTINUE
lib IF(NEX.EO.SAME) GO TO 117
IF(XTYP.EQ.BAR.OR.YTYP.EO.BAR) GO TO 200 -------
IF(XTYP.NE.SEMI) GO TO 120
IFIXINC. LT.l. E-20) CALL SC ALE ( XAR, XAXL, IMAX , 1 )
116 CALL AXIS(0.0,0.0,XLAB,-6U,XAXL/0.0,XAR(IMAX+l),XAR(IMAXl-2))
IF(IDOU8.EQ.1.0R.IDOUB.EO. J)
&CALL AXIS(0.0,YAXL,XLAB,+6a,XAXL,0.0,XAR(IMAX+l),XAR(IMAX+2))
120 IF(YTYP.NE.SEMI) GO TO 130
IFCYINC. LT.l. E-20) CALL SCALE( YAR, YAXL» IMAX, 1 )
126 CALL AXIS(0.0,0.0,YLAB,6«,YAXL,<>0.0>YAR(IMAX-M).YAR(IMAX+2))
IF(IDOUB.GE.2)
&CALL AXlS(XAXL.O.OfYLAB,-6a,YAXL,90.0,YARUMAX*l),YARUMAXt2))
130 IF(XTYP.NE.LOG) GO TO HO
IFCXTNC. LT.l. E-20) GO TO 135
IF(XBEG.GT.1.E-20) GO TO 133
XBEG=1.
6440
6450
6460
6U70
6460
6490
6500
6510
6520
6530
6540
6550
6560
6570
6580
6590
6600
6610
6620
6630
6640
6650
6660
66TO
6680
6690
133 CONTINUE
6710
6720
6730
6740
6750
6760
6770
6760
6790
6600
6610
6820
6630
6640
6650
6660
6870
6860
6890
6900
6910
6920
6930
6940
6950
6960
6970
6980
6990
7000
7010
7020
7030
7040
438
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
GO TO 1J6 7050
135 CALL SCALGUAR.XAXL»IMAX,1) 7060
136 CALL LGAXS(0.0.0.0,XLAB,-64.XAXL»O.OrXAR(IMAX + l),XAR(IMAX + 2)) 7070
"""" IF(IOOUB.£Q.1.0R,IDOUB.EQ.3) " "" """ ' 7080
1CALL LGAXS(0.0,5,0»XLAB,64,XAXL,0,0,XAR(lMAX+U»XAR(IMAX+2)) 7090
140 IF(YTYP.NE.LOG) GO TO 147 _ 7100
~ IF.(YIMC,LT.1.E"20) GO TO 145 7110
IF(YBEG.GT.l.E-20) GO TO 143 7120
YBEG=1 . _
YAR(1MAX+1)S1.
143 CONTINUE
GO TO 146
145 CALL SCALG(YAR,YAXL,IMAX,1) ,6a,YAXL.90.0,YAR(IMAX+l),YAR(lMAX*2)) 7180
.. .._.
iCALL"LGAXS(6.0,0.0,YLAB,«6U,YAXL,"=XARCIMAX) 7570
00 230 I»1»IMAX 7580
jsjMAX-I+1 7590
YAR(2*J)«YAR(J) 7600
YARC2*J»DsirAR(J) 74,10
XAR(2*J)*XAR{J»1) 7fc2'0
230 XAR(2*J-1)S>(ARCJ) 7630
IMAXs2*IMAX 7640
BARYsl. 7650
7660
)=XHtG
439
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
XAR(IM»x+2)=XINC
YARCIMAX+1)=YBEG
GO TO 110
C PROS GRAPH
250 IF(XTYP.NE.PR06) GO TO 300
" IF(NEX.EQ.SAME) GO TO 255
XPLAB(26)=6.0/XAxL
CALL LINE(XPLAB,YPLAB,24,1,1,13)
CHXP»XAXL/6.*.0681
PSYMSs-CHxP
PSYT=-.17*(XAXL/6.J
CALL SYMBOL(PSYMS,PSYT,CHXP,SPLABiO.,96)
CALL SYMBOLCO.»»,3S,.14,XLAB,0.,69)
IF(IDOU6.NE.1.AND,IDOU6,NE.3) 60 TO 255
00 251 IDUMIi=I,24
251 YPLA5(IDUMII)»YAXL
CALL LINE(XPLAB,YPLA5,24,1,1,13)
PSYMS=2.*("CHXP)
PSYT=YAXL-PSYT
CALL SYMBOL(PSYMS,PSYT,CHXP,SPLAT,0.,96)
CALL SYMBOLCO.,5,35,.14,XLAB,0.,69) "
255 DO 270 I«l,IMAX
LEFTsJ
IF(XAR(I).LT..01) XAR(I)x.01
IF(XAR(l).LT.50.) GO TO 260
LEFTSO
IF(XAR(I).GT,99.99) XAR(I)=99,99
XAR(I)slOO.-XARd)
260 RLP*ALOG10(XAR(I)*100.)*10.+1.
IFCRLP.LT.l.) RLPal.
IFCRLP.GT.36.) RLP»38.
LP=IFIX(RLP)
XAR(I)=(XPROBCLP)+CRLP-LP)*(XPROB(LP+l)-XPROBCLP)))/5.
IF(LEFT.EQ.O) XAR(I)=6.»XAR(I)
270 CONTINUE
XAR(IMAXM)=0.
XAR(IMAx+2)sXAXL/6,
PROBX«1.
XTYPsSEMI
GO TO 147
300 IFCYTYR.NE.PROB) GO TO 450
450 WR1TEC6.460) XTYP,YTYP
460 FORMATc NO SUCH GRAPH TYPE AS ',2A4j
GO TO 1000
: AGAIN
500 CONTINUE
IM8ARX.GT. .5) XTYPsBAR
IF(BARY.GT..5) YTYPsBAR
IF(PR08X.GT..5) XTYPoPROB
IF(NfcXT.NE.NEW) GO TO 510
CALL PLnT(PMQVE,0.,-3)
GO TO 10
510 IF(NEXT.EQ.SAME) GO TO 20
1000 WRITE(IOUTUN.IOIO) NEXT
1010 FORW4T(' END NEXT« ',A4)
CALL PLOT(PMOVE,0.,999)
RETURN
END
7670
7660
7690
7700
7710
7720
7730
7740
7750
7760
7770
7780
7790
7600
7810
7620
7630
7840
7650
7860
7870
7680
7690
7900
7910
7920
7930
7940
7950
7960
7970
7<>80
7990
"8000
8010
8020
6030
6040
8050
8060
8070
8060
8090
8100
8110
8120
8130
8140
6150
8160
6170
8180
8190
8200
8210
8220
8230
8240
8250
//*
INSERT SOURCE DECK MODIFICATIONS HERE
440
-------
Table 68 (continued). BAGHOUSE SIMULATION PROGRAM LISTING
//GO.SYSUN DO
// 00 *
//* INSERT OBJECT DECKS HERE TAKF HUT SOURCE DECKS ABOVE
//GO.SYSUB OD DISP=SHP
// 00 DSNeSYSI•CACCOMP,DISPsSHR
//GO.FT08F001 00 UNITsSYSOA,DISPs(NF*,PASS),DSM=&&bAGl,
// OCB=(RfrCF* = FB,LRECLsftO,BLKSIZE«1600),SPACE*(TRK,<1,1), RISE)
//GO.FT09F001 DO UMTsSYSDA,DISP= (NE»,PASS) ,DSNsSS8AG2,
// DCBs(RECFM=FB,LRECL=80,BLKSIZE=1600),SPACE*(TRK, (1,1}, RISE)
//GO.FT10FOOI OD UNI TsSYSDA,OISP=(N£w,PASS) , DSNs&8,eAG3,
// DC8=(RECFMsFB,LRECL360,8LKSIZt»1600),SPACE*(TRK,(1,1),RISE)
//GO.FT11F001 DO UNIT3SYSDA,DISP=(NEw,PASS).03NB&iBAG4,
// DCB*(RECFM»FB,LI>ECL»eO,8LKSI2F»1600),SPACe»(TRK,(l,l),»LSE)
//GO.FT12F001 DO UNITsSYSDA,DISP«lNEh,PASS),OSNs&g,BAG5,
// DCBs(RECFM*F6,l.RECLa90,BtKSIZE=1600),SPACE"(TRK,(1,1),RLSE)
//GO.FT13F001 DO UNIT3SYSDA,0!SP=(NEW,PASS),DSNsftg,BAG6,
// t>CB=(RECFMsFB,LRECL=80,BLKSIZE=1600),SPACE3(TRK,(1,1),RLSE)
//GO.FTiypOOl 00 UNTT*SYSOA,DISPa(NEw,PASS),DSNs&&BAG7,
// OCB=(RECFMxF6,LRECL=80,BLKSlZE*1600),SPACE*(TRK,(1,1),RLSE)
//GO.FT15F001 DO UN1T=SYSDA,OISP=(NEW,PASS),DSNa&8,BAG6,
// DCB*(RECFMsF6,LRECL=80,BLKS17E«1600),SPACEs(TRK*(1,1),RUSE)
//GO.FT03F001 00 OISP«(OLD,PASS),DSN=&&BAG1,UNIT=SYSOA,
// VOL=REF»*.FT08F001
// DO DISPs(OLD,PASS)»DSN = &&tiAG2,UNITaSYSDA,VOL=REF = *.FTOKEEP),UMT = CTAPE7,,DEFER),
// OCBaDENsl,LABEL«(,NL),
// VOL»SER=PLOO««
//GO.SYSIN 00 *
//» INSERT INPUT DATA HERE
OF BAGSs **CYCUE T= **C'.,EAN Ts *****# CYCCES=
2****TLAG« ***
3NUMB OF INCS*" ******Q/Aa ***CQNCEN« **MIN RSE=
«*****»K?» **DIAGNOS1IC3= ***CONS Pa **»***»rtRs
5***WSTAR» ***DPSTOP* **R£VFl.O«*« ********************
6**WSTART= ***
7TEST RUN * BAGHOUSE SIMULATION ***
ft***XAxISx ****YAXJS* **
<>******KRs *******SR* ****TEMpKs ********************
**
441
-------
Table 69. VARIABLES AND ARRAYS USED IN BAGHOUSE SIMULATION PROGRAM
VARIABLES
ACAKE - fractional area on a bag that is not cleaned, input.
AREA - fractional area on a bag. The product of AREA and the number
of areas cleaned gives the fractional area cleaned.
ATEST - intermediate calculation in determining AREA.
BAG1 - heading, 'SBAG'.
BAG2 - heading, 'QBAG'.
CLAREA - fractional area cleaned on a bag, calculated.
CONSP - system pressure, if the system operates at constant pressure,
N/m2.
3
CONTOT - total outlet concentration from the system, g/m .
3
CZERO - inlet concentration, calculated, g/m .
CZEROE - inlet concentration, input, g/m .
2
DELP - system pressure drop, N/m .
DELT - time increment, min.
DELTT - intermediate in determining time increment, min.
2
DPAVG - intermediate in calculating average pressure drop, N/m .
2
DPAVGN - average pressure drop at the end of a cycle, N/m .
2
DPSTOP - maximum system pressure, if exceeded cleaning begins, N/m .
DRAG - heading, 'AREA'.
DTLAST - time increment of last loop, min.
ERR - error used in determining cleaned area.
I - index.
IAREA - number of areas on a bag.
IBAG - bag index.
IFAREA - number of the area to be cleaned.
IFBAG - number of the bag just cleaned.
II - index.
IREPT - line counter for output of intermediate calculations.
IUNIT - output file number.
J - index.
JLOOP - index in time loop.
442
-------
Table 69 (continued). VARIABLES AND ARRAYS USED IN BAGHOUSE SIMULATION
PROGRAM
JTIME - JLOOP - 1.
K - index.
K3 - index in determining when to write on a file, data points
for graphs are written three at a time.
L - index.
LCONP - constant pressure diagnostics; if true, operation is at con-
stant pressure.
LDIAG - print diagnostics; if true, intermediate calculations are
output, input.
LMAX - maximum number of individual flow rate graphs, limit = 5.
M - number of increments per bag, input.
MAXJ - total number of increments used in time loop-
MAXK - maximum number of bags for which calculations are output
per line.
N - number of bags (compartments), input.
NAREA - number of areas to be cleaned.
NT - number of cycles modeled, input.
PAVNOW - average penetration of the end of a cycle, referenced to
time = 0.
PAVR - average penetration at the end of a cleaning cycle.
PAVTOT - intermediate in calculating average penetration.
PENTOT - total system penetration at any time.
QAVG - intermediate in calculating average system flow, m/min.
QAVGN - average system flow at the end of a cycle, m/min.
QSYSTM - total system flow, m/min.
3
SE - effective drag, input, N-min/m .
SFAB - fabric drag, N-min/m3.
SMALQ - specified constant total flow, input, m/min.
3
SSYSTM - total system drag, N-min/m .
3
SZERO - residual drag, S^, input, N-min/m .
K.
T - cleaning cycle time, input, min.
TCLEAN - single bag cleaning time, input, min.
TCONT - actual simulated time, min.
443
-------
Table 69 (continued). VARIABLES AND ARRAYS USED IN BAGHOUSE SIMULATION
PROGRAM
TCORR
TEMPK
TLAG
TMOD
TTEST
VRFLO
VRFLOW -
WAREA
WCOMP
WR
WS
WSTAR
ZK2
ZK2MU
ZKZERO -
correction for time interval splitting at the end of a
cycle, min.
gas temperature, input, °K.
total area run time, input, min.
total cycle time = T+TLAG, reference time for cleaning
cycle, min.
TCONT in a modulo TMOD system, it is normally the time since
cleaning cycle started, min.
reverse flow velocity based on a single compartment, input,
m/min.
reverse flow used in calculations; zero if not cleaning,
VRFLO if cleaning, m/min.
weight permit area added to an area in one time increment,
g/m •
2
intermediate in determining areas of highest loading, g/m .
2
residual fabric loading, input, g/m .
2
absolute fabric loading at time zero, input, g/m .
2
constant for nonlinear drag model, input, g/m .
specific cake resistance, K , input, N-min/g-m.1
viscosity correction for K .
initial slope of drag versus loading curve, 1C, input,
N-min/g-m.
ARRAYS
CAKE(IBAG)
IDUM(I)
OLDTIM(IBAG)
P(IAREA)
PDP(K3)a
PDQ(K3)a
PPS(K3)a
PQ(K3,LMAX)a
PT(K3)a
- average fabric loading on bag # IBAG, g/m .
- variable array index for output of intermediate
results.
- previous time for bag # IBAG, min.
- penetration for area # IAREA.
2
system pressure drop, N/m .
- system flow, m/min.
- system penetration.
- individual compartment flow, m/min.
simulated time, min.
444
-------
Table 69 (continued). VARIABLES AND ARRAYS USED IN BAGHOUSE SIMULATION
PROGRAM
QAREA(IAREA)
QBAG(IBAG)
S(IAREA,IBAG)
SBAG(IBAG)
TIME(IBAG)
UD(IAREA,1BAG)
face velocity on area // IAREA, m/min.
average face velocity for bag # IBAG, m/min.
drag of area # IAREA on bag # IBAG.
total drag of bag # IBAG.
time after cleaning for bag # IBAG.
dust cake loading on ares # IAREA on bag # IBAG.
These arrays contain only 3 entires. When data is output for subsequent
processing by the plot routine SCRIBE, they are output in groups of 3.
445
-------
Table 70. DATA INPUT FORMAT
Parameter
Number of bags
Cleaning cycle time
Single bag cleaning time
Number of cycles modeled
Total area run time
Number of increments per bag
Constant flow velocity
Inlet concentration
Effective drag, Sg
Specific cake resistance, K~
Print diagnostics
Constant pressure drop
Residual fabric loading, W
K.
W*
Maximum pressure drop
Reverse flow velocity
Initial cake loading
TITLE
X-axis length
Y-axis length
Initial slope, K
Residual drag, S
K
Gas temperature
Caked area
Units
-
Minutes
Minutes
-
Minutes
-
m/min
g/m
3
N-min/m
N-min/g-m
-
N/m2
g/m
2
g/m
N/m2
m/min
g/m
-
inches
inches
N-min/g-m
N-min/m3
°K
-
Record
1
1
1
1
2
3
3
3
3
4
4
4
4
5
5
5
6
7
8
8
9
9
9
10
Columns
15-20
31-40
51-60
75-80
11-20
15-20
31-40
51-60
71-80
11-20
35-40
51-60
71-80
11-20
31-40
51-60
11-20
2-77
11-20
31-40
11-20
31-40
51-60
11-20
Format
16
G10.0
G10.0
16
G10.0
16
G10.0
G10.0
G10.0
G10.5
L6
G10.0
G10.0
G10.5
G10.5
G10.5
G10.5
19A4
F10.7
F10.7
F10.5
F10.5
F10.5
F10.5
446
-------
INDEX
Adhesion, dust cake
fabrics tested, 296
general discussion, 300-305
Adhesive forces , interfacial
cleaned fabric surface (a ),
relation to, 295-299 C
range of, 300, 302
Aerosol size properties
(See particle size properties)
Air flow
pore structure, 77-81
Atmospheric dust
fabric rating tests, 199, 203-205
Atmospheric dust concentration
optical versus gravimetric
measurements, 50, 51, 190
sateen weave cotton, 204, 205
woven glass fabrics, 204, 205
Bench scale tests
apparatus, 31-33, 38
data summaries, 169-172
Capillary flow
see Hagen Poiseuille flow, 77-79
Carman-Kozeny equations
K determinations, 164
Computer printouts
input data for Nucla and Sunbury
modeling, 379, 384, 385
Condensation nuclei counter
detection sensitivity, 177, 202
Condensation nuclei measurements,
175, 176
Dacron fabrics
fly ash collection, 171
humidity effects, 166
physical properties, 63
Diffusion parameter (n' )
(See particle collection)
Direct interception parameter (^-r)
(See particle collection)
Drag (resistance) model
bilinear, single bag, 288, 289
linear, critique, 230
linear, single bag
applications, 285-287
critique, 230
nonlinear, empirical, 236-242
nonlinear, predicted versus
experimental results, 244
nonlinear, theoretical, 233, 234, 236
Dust cake density
fly ash, 145, 146, 265, 266, 270
Dust dislodgment
(See also fabric cleaning)
447
-------
INDEX (Continued)
adhesion, 272
appearance of cleaned fabric, 282
bag collapse, 153, 155
bag tensioning effect, 218, 219
cleaned area, 290-306
fabric loading effect, 153, 209,
272
interfacial adhesive forces,
distribution, 209, 213, 214
interfacial separation, 153, 155,
230
number of cleaning cycles to
attain maximum removal, 218, 219
repeated cleaning and filtering
cycles, 214, 215, 218, 219
shearing forces, 272
tensile forces, 153, 272
Dusts/fabrics
test combinations, 130
Dust/fabric photomicrographs
before and after cleaning, 148
cake cracking by flexing, 146
cleaned and uncleaned areas, 151,
154
discussion, 130-167
dust cake at 2OX magnification, 147
pinhole leaks, 137, 138
Dust generator
NBS design, 37
Dust removal
(See dust dislodgment)
Effluent concentrations
fabric loading effect, 202
face velocity effect, 199, 201
336-339
Fabric acceleration
calculation for mechanical
shaking, 295
Fabric cleaning
(See also dust dislodgment)
acceleration, 272
average residual dust holding
versus number of shakes,
293, 294
cleaned area (a ) estimation,
290, 291 °
cleaned area (ac) versus dust
separation force, 292
cleaning force calculation,
mechanical shaking, 306
dust dislodgment forces, 272
dust spallation, 230, 272
filter performance, 271-289
partially cleaned filter
photograph, 154
pressure controlled, 307-310
pulse jet systems, 272
sequential cleaning schematic,
357
surface loadings, 271
time cycle control, 311-314
Fabric collection
pore cross section, effect of,
332
Fabric drag (resistance)
clean (SQ), 231, 233, 245, 253
clean (So) versus effective (S ),
247-249 E
effective (S£), 245, 247
experimental values, 249
448
-------
INDEX (Continued)
fabric loading effect, bag tests,
128, 129, 131, 132, 220, 222
partial cleaning, 273-275, 279
pore plugging, blinding, 127
pore velocity, clean Sunbury
fabric, 64
previous dust loading, effect of,
250, 251
residual (SR) for miscellaneous
dust/fabric combinations, 155-157
structure effects, 247
tensioning, effect of at constant
velocity, 88, 91, 92
Fabric loading, W,
residual (WR) for various dust/
fabric combinations, 156, 157
Fabric permeability
new and cleaned woven glass fabrics,
118-124
Fabric photomicrographs
Nucla fabric, unused, 69
Sunbury fabric unused, 68, 75
yarn appearance, warp and fill,
70, 71
Fabric properties
acrylic, spun 2/2 twill, 23
ASTM ratings, 82
cotton, sateen weave, 63
Dacron, crowfoot, 63
general, 81-94
Nomex fabrics, 23
Nucla (W. W. Criswell), 60, 61
rigidity and flexing, 84
Sunbury (Menardi Southern), 60, 61
tensile modulus, 84-90
Fabric weave
pore cross section, schematic,
76
pore density versus yarn
proximity, 72, 74
Sunbury, textile schematic, 65
yarn and pore structure, schematic,
74
Fabrics, woven glass
manufacturers, 63
Filter capacity
number of shakes, 277
shaker acceleration, 277
Fly ash collection
cotton fabric, sateen weave, 171
Dacron fabric (crowfoot weave),
171
glass fabric, 3/1 twill, 169-172
partially cleaned filters,
158-160, 169, 171, 172
Frasier permeability
(See fabric permeability)
clean fabrics, 63
Glass bag
partially cleaned, photomicrograph,
282
Glass fabrics
fiber size, 320
field performance, 101-117
fly ash deposition, initial,
94-96
thickness measurements, 93, 94
Hagen-Poiseuille flow
pressure loss, pore, 77-79
449
-------
INDEX (Continued)
Humidity effects
Dacron fabric/fly ash, 166
Impaction parameter (ru)
(See particle collection)
K , nonlinear modeling parameter
drag/loading relationships,
236-242
experimentally derived values,
243, 249
Lignite fly ash collection
glass fabric, 3/1 twill, 170, 171
Models
(See drag (resistance) and
penetration)
Model, filter system
basic drag equation, 351
basic modeling process, discussion,
352-361
capability, .352, 353
computational procedure for
baghouse, 360
data inputs required, 353
drag computations, 363-366
input data summary, 368-371
Nucla cleaning schedule,
simplified, 377
penetration calculations,
366, 367
program (computer) description,
361-363
program flow diagram, 362
program output, sample, 372
Summary, design highlights, 406
Sunbury cleaning schedule,
simplified, 383
validation, 373-405
validation, Nucla installation,
386-394
validation, Nucla and Sunbury
data inputs, 375
validation, summaries of pre-
dicted and observed perfor-
mance, 406-408
validation, Sunbury installation,
394-405
working equations and relation-
ships, 348-351
Modeling concepts
full scale applications,
resistance, 306-315
Modeling, general
variables controlling performance,
5, 6
Modeling (Historical)
Fraser and Foley, 25
Leith and First, 25-27
Noll, Davis and LaRosa, 23, 24
Robinson, Harrington and Spaite,
14-17
Solbach, 18-20
Stinessen, 24
Monofilament screens
fly ash deposition, 97, 99
pore bridging, 100
Nucla fabric
field performance, 101-117
Nucla field tests
data summaries, 112, 114, 117
450
-------
INDEX (Continued)
normal cleaning procedure, 376
Nuclei concentrations (effluent)
fly ash/glass fabrics, 176-182
instrumentation for measurement,
49
mass concentration, relation to,
191, 194-199
partially loaded glass fabrics,
182
pinhole effect, severe leakage,
176
velocity and fabric loading
effect, 186
Nuclei concentrations (influent)
estimation from effluent concentra-
tion and filter penetration, 186
Nuclei concentration measurements
optical (B&L) measurement,
comparison, 175-193
Nuclei versus mass concentrations
calibration curve, 195
discussion, 191, 194-199
summary of bench scale measurements,
197, 198
fiber substrate with dust
deposits, 322-325
impaction and direct interception,
321
pore bridging, 315
pore capture, discussion, 315-319
pore penetration, estimation,
318, 319, 331
Particle size properties
atmospheric (laboratory) dust, 51
dust slough-off from clean side
of filter, 227, 228
fly ash, GCA, 34, 42, 44
lignite fly ash, 48
logarithmic-normal, surface and
volume mean diameters, 165
Rhyolite (granite), 47
Sunbury fly ash, 44, 45
Peclet number
calculation of diffusion
parameter, 328, 329
Penetration
(See also effluent concentrations)
inlet dust concentration, 334, 335
Optical counter measurements (effluent) model, single bag, 338-343
coal fly ash, size and concentration pinhole leaks, 136, 138-144
versus fabric loading, 187-189, 193
pore structure effects, 174, 175
lignite fly ash, size and concentra- , , ... . _ ,..- ,_
tion versus fabric loading, 183, 192 rear face,f ^f £
agglomerates, 224-228
Particle collection
bulked fiber substrate theory,
319, 324
dust cake, collection theory, 327
dust cake, granular bed, 325-331
residual outlet concentration,
338, 340, 342
surface dust load distribution, 337
variables controlling, single
bag, 338
velocity effects, 220, 224, 335,
336
451
-------
INDEX (Continued)
Photomicrographs
(See fabric, dust)
Pilot plant baghouse
bag illumination, 207
schematic drawing, 40
test facility, schematic, 36-40
testing procedures, 206, 207
vibration problems, 207
Pinhole velocity, 142
Pinhole leaks,
leak velocity estimates,
139, 142
pinhole area estimates, 139
Pore dimensions
equivalent circular diamter, 78
hydraulic radius, 77-80
tabular summary, 78
Pore structure (type)
Sunbury fabric, schematic, 76, 78
Pore type
cross section versus type,
73, 74, 76
hydraulic radius, 78
Pore velocity
maximum, 80
Porosity (dust cake)
bulk density (dust cake), 262
effect on KZ, 254, 263
Program (computer)
(See model, filter system)
Resistance (drag) models
see drag (resistance) models
Rhyolite (granite) collection
glass fabric, 3/1 twill, 170
Sateen weave cotton
fly ash collection, 191, 193
Specific resistance coefficient (K_)
dust cake porosity, 254-263
experimentally derived values, 249
fabric permeability, 256
fabric surface effects, 256
face velocity effect, 161-163,
259-261
particle shape effect, 256
particle size effect, 164, 165
predicted and/or measured values,
220, 222, 266-269
specific surface parameter,
164, 165, 261-271
viscosity effects, 254, 266-269
Specific surface parameter (SQ)
calculation for polydisperse
distribution, 261, 262
coal fly ash, 267
granite dust, 268
lignite fly ash, 267
Nucla fly ash, 264
talc dust, 268
Sunbury fabric
field performance, 101-117
452
-------
INDEX (Continued)
Sunbury field tests
data summaries, 102, 103, 105-107,
109, 111, 113, 115
fabric loading (average) after
cleaning, 102, 103
normal cleaning procedure, 382
Tensile modulus, 84-88
stress/strain factors, 88
Tensile properties
apparatus for measurement, 56
Sunbury fabric, 85
Test aerosols
discussion, 43-46
Test aerosol size properties
(See particle size properties)
Yarn shape (dimensions)
schematic drawing, 61, 74
Yarn shape
photomicrographs, 70, 71
453
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
REPORT NO.
EPA-600/7-77-084
2.
3. RECIPIENT'S ACCESSION-NO.
riTLE ANDSUBTITLE
Filtration Model for Coal Fly Ash with Glass
Fabrics
5. REPORT DATE
August 1977
6. PERFORMING ORGANIZATION CODE
.AUTHORIS) Richard Dennis, R.W.Cass, D.W.Cooper,
R.R.Hall, Vladimir Hampl, H.A.Klemm,
.T.R-T,ang1<=.yf and " " "'
i DC D CrtB tk/tl Mt~l n do A M I "7 ATI /
8. PERFORMING ORGANIZATION REPORT NO.
GCA-TR-75-17-G
. PERFORMING ORGANIZATION NAME AND ADDRESS
GCA Corporation
GCA/Technology Division
Bedford, Massachusetts 01730
10. PROGRAM ELEMENT NO.
EHE624
11. CONTRACT/GRANT NO.
68-02-1438, Task 5
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Task Final; 6/74-6/77
14. SPONSORING AGENCY CODE
EPA/600/13
i5. SUPPLEMENTARY NOTES IERL_RTp project officer for this report is James H.
Turner, Mail Drop 61, 919/541-2925.
16. ABSTRACT
repOrt describes a new mathematical model for predicting
woven glass filter performance with coal fly ash aerosols from utility
boilers. Its data base included: an extensive bench- and pilot-scale
laboratory investigation of several dust/fabric combinations; field data
from three prior GCA. studies involving coal fly ash filtration with glass
fabrics; past GCA studies of fabric filter cleaning mechanisms; and a
broad based literature survey. Trial applications of the model to field
filter systems at Sunbury (PA) and Nucla (CO) indicate excellent agree-
ment between theory and practice for both penetration and resistance.
The introduction and experimental confirmation of two basic concepts
were instrumenta.1 in model design: one relates to the way dust dislodges
from a fabric and its subsequent impact upon resistance and penetration
in a multichambered system; the other, to the relatively large fly ash
fractions that pass with minimal collection through temporarily or per-
manently unblocked pores or pinholes such that observed particle pene-
trations are essentially independent of size. Cleaning parameters were
quantified, and estimates of specific resistance coefficient, K2 , were
improved . _ _______ _ __ _
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air Pollution
Mathematical Models
Filtration
Fly Ash
Coal
Woven Fabrics
Glass Fibers
Aerosols
Dust
Utilities
Boilers
Air Pollution Control
Stationary Sources
Fabric Filters
Particulate
13B
12A
07D
21B
21D
HE
11B
11G
13A
8. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (ThisReport)'
Unclassified
21. NO. OF PAGES
491
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
455
-------