Environmental Protection Technology Series
BACKWASH OF GRANULAR FILTERS USED
IN WASTEWATER FILTRATION
Municipal Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, Ohio 45268
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine 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
8 "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL PROTECTION TECH-
NOLOGY series. This series describes research perlormed to develop and dem-
onstrate instrumentation, equipment, and methodology to repair or prevent en-
vironmental degradation from point and non-point sources of pollution. This work
provides the new or improved technology required for the control and treatment
of pollution sources to meet environmental quality standards
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-77-016
April 1977
BACKWASH OF GRANULAR FILTERS USED IN
WASTEWATER FILTRATION
by
J. L. Cleasby and E. R. Baumann
Iowa State University
Ames, Iowa 50011
Grant No. R802140
Project Officer
S. A. Hannah
Wastewater Research Division
Municipal Environmental Research Laboratory
Cincinnati, Ohio 45268
MUNICIPAL ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
-------
DISCLAIMER
This report has been reviewed by the Municipal Environmental Re-
search Laboratory, U.S. Environmental Protection Agency, and ap-
proved for publication. Approval does not signify that the con-
tents necessarily reflect the views and policies of the U.S. En-
vironmental Protection Agency, nor does mention of trade names or
commercial products constitute endorsement or recommendation for
use.
ii
-------
FOREWORD
The Environmental Protection Agency was created because of increasing
public and government concern about the dangers of pollution to the
health and welfare of the American people. Noxious air, foul water,
and spoiled land are tragic testimony to the deterioration of our
natural environment. The complexity of that environment and the
interplay between its components require a concentrated and integrated
attack on the problem.
Research and development is that necessary first step in problem solu-
tion and it involves defining the problem, measuring its impact, and
searching for solutions. The Municipal Environmental Research Laboratory
develops new and improved technology and systems for the prevention,
treatment, and management of wastewater and solid and hazardous waste
pollutant discharges from municipal and community sources, for the
preservation and treatment of public drinking water supplies, and to
minimize the adverse economic, social, health, and aesthetic effects
of pollution. This publication is one of the products of that research;
a most vital communications link between the researcher and the user
community.
The conventional methods for treatment of municipal wastewaters fre-
quently produce effluents that will not meet local discharge require-
ments. Granular media filters are being installed to provide tertiary
treatment for increased removals of suspended solids and particulate
BOD. This report provides valuable information on criteria for selec-
tion of media for wastewater filters and design considerations to pro-
vide adequate cleaning of the media during backwash.
Francis T. Mayo
Director
Municipal Environmental Research
Laboratory
iii
-------
ABSTRACT
The use of deep granular filters in waste treatment is of growing
importance. The key to long-term operating success of such filters
is proper bed design and adequate bed cleaning during backwashing.
A number of questions related to adequate backwashing of granular
filters were investigated and the study results lead to the follow-
ing conclusions:
Cleaning granular filters by water backwash alone to fluidize
the filter bed is inherently a weak cleaning method because
particle collisions do not occur in a fluidized bed and thus
abrasion between the filter grains is negligible.
Due to the inherent weakness of water backwashing cited above,
auxiliary means of improving filter bed cleaning are essential
for wastewater filters. Three auxiliary methods were compared
in a wastewater pilot filtration study. The most effective
backwash was provided by air scour and water backwash simul-
taneously at subfluidization velocities. The other two aux-
iliary methods, surface and subsurface wash and air scour
prior to water fluidization wash were about comparable in
effectiveness.
The performance of coarse sand, dual-, and triple-media fil-
ters was compared, and the backwashing routines appropriate
for each media are discussed. A number of investigations con-
cerning the design and backwashing of dual-media filters are
also included.
This report was submitted in fulfillment of Grant No. R802140 by Iowa
State University under the partial sponsorship of the U.S. Environ-
mental Protection Agency. This report covers a period from September 1,
1971 to May 31, 1976, and work was completed as of December 31, 1975.
iv
-------
TABLE OF CONTENTS
Page
ABSTRACT iv
LIST OF TABLES vii
LIST OF FIGURES xi
NOTATIONS, ABBREVIATIONS AND CONVERSION'FACTORS xviii
ACKNOWLEDGMENTS xxiv
I. INTRODUCTION 1
II. CONCLUSIONS 4
Conclusions Regarding Backwash Effectiveness 4
Conclusions Comparing Filter Performance 5
Conclusions Regarding Expansion, Intermixing, 5
and Dual Media
Conclusions Regarding Wastewater Filter Design 7
III. RECOMMENDATIONS 10
IV. BACKWASHING-POTABLE WATER EXPERIENCE 11
V. BACKWASHING WITH FLUIDIZATION AND EXPANSION 16
Some Fluidization Fundamentals 16
Predominance of Hydrodynamic Forces in Cleaning 23
by Water Fluidization Alone
VI. OPTIMUM CLEANING BY WATER BACKWASH ALONE 26
Particulate Fluidization and Optimum Turbulence - 26
Evidence from the Literature
A New Theory of Optimum Backwashing by Water Fluidization 31
Only
Experimental Support for the Optimum Theory 39
VII. WASTEWATER FILTRATION AND BACKWASHING - 68
LITERATURE REVIEW
Types of Wastewater Filters and Cleaning Techniques 6.8
Case Histories 76
VIII. EXPERIMENTAL COMPARISON OF BACKWASH METHODS IN 89
WASTEWATER FILTRATION
Pilot-Plant Equipment g
-------
Chronology of Experiments 91
Equipment Details 91
Analysis of Samples 102
Operation and Results - Phase I Dual-Media 106
Filtration of Alum-Treated Secondary Effluent
Operation and Results - Phase II Dual-Media 126
Filtration of Secondary Effluent
Operation and Results - Phase III, IV, and V Single-, 157
Dual-, and Triple-Media Filtration of Secondary
Effluent
Operation and Results - Phase VI Coarse Sand Filtration 194
of Secondary Effluent
IX. EXPANSION AND INTERMIXING OF MULTI-MEDIA FILTERS 204
Introduction 204
Dual-Media and Multi-Media Filtration Literature 205
Backwashing of Granular Filters 207
Bed Expansion Correlations from Fluidization Literature 211
Prediction of Settling Velocities 224
Existing Models for Predicting the Expansion of Fluidized Beds 230
X. EXPANSION AND INTERMIXING EXPERIMENTAL INVESTIGATION 233
Experimental Apparatus 233
Experimental Procedures 239
Illustrative Calculations 242
Results and Analysis - Summary 247
Results - Media Characteristics 247
Fixed Bed Hydraulic Profiles in Dual-Media 256
Filters - Coal and Sand
Expansion - Flow Rate Observations 271
Intermixing Observations
IX. EFFECT OF MEDIA INTERMIXING ON DUAL-MEDIA FILTRATION 315
Introduction . 315
Objectives and Scope of this Study 315
Experimental Investigation 3^
Results 323
Discussion 336
Conclusions 340
XII. ABRASIVE LOSS OF COAL DURING AIR SCOUR 341
Experimental Procedure 341
Results 342
Conclusions 343
XIII. REFERENCES 345
XIV. APPENDIX - Sieve Analyses of Uniform Media 355
vi
-------
LIST OF TABLES
Page
Table 1. Analysis of University tap water. 43
Table 2. Experimental design for series 2. 47
Table 3. Backwash procedures for graded sand. 48
Table 4. Effluent quality index and porosity for series 2. 54
Table 5a. Manual backwashing procedure used on full-scale rapid 83
sand filters at Luton [86].
Table 5b. Backwashing procedure in full-scale rapid sand filters 83
at Luton in November 1975.
Table 6a. Automatic backwashing sequence used on pilot scale 84
Immedium upflow filter at Luton [86].
Table 6b. Backwash procedure for full-scale upflow filters at 84
Luton in November 1975.
Table 7. Summary of experimental phases for wastewater filtration 92
backwashing study.
Table 8. Filter media details for wastewater filtration pilot 99
studies.
Table 9. Initial filter head losses during various portions of 113
the study.
Table 10. Backwash rates required to achieve 38 to 40% bed 120
expansion.
Table 11. Results of analyses during alum treatment (Phase I) 121
for samples from May 17 to July 11, 1973, when both
filters were washed by water fluidization only. (All
results from composite samples except solids contact
influent.)
Table 12. Results of analyses during alum treatment series 122
(Phase I) for samples from July 11 to August 20, 1973,
when air scour was being used on the south filter.
(All results from composite samples except solids
contact influent.)
Table 13. Suspended solids released from filters in special 125
backwashes after run 78.
Table 14. Summary of head loss development during observation 150
runs of Phase II, during direct filtration of
secondary effluent.
vii
-------
Table 15. Results of analyses during direct filtration of 151
secondary effluent (Phase II) from August 30 to
October 30, 1973. (All composite samples except
as noted.)
Table 16. Data summary of clean-up operation at the end of 153
Phase II.
Table 17. Solids capture per unit head loss results for direct 183
filtration of trickling filter effluent, 1974.
Table 18. Summary of analytical test results for Phase III. 185
Table 19. Summary of analytical test results for Phase IV. 186
Table 20. Summary of analytical test results for Phase V. 187
Table 21. Data summary for clean-up operation at the end of 189
the operation period in 1974.
Table 22. Action of simultaneous air and water backwash on 197
coarse sand at subfluidization velocities.
Table 23. Results of analyses during direct filtration of 199
secondary effluent (Phase VI) from June 24 through
August 2, 1975, using coarse sand filters of
different depths.
Table 24. Mean total head loss during filtration of secondary 200
effluent on coarse sand filters during Phase VI.
Table 25. Average initial head loss for three coarse sand filters 200
before and after run 35 in Phase VI when increase of
backwash rate was adopted.
Table 26. Height that sand is thrown by simultaneous air and 202
water backwash.
Table 27. Comparison of n slopes of sea sands (using Jottrand's 217
analysis [69] and Richardson and Zaki's equations).
Table 28. Comparison of n slope of crushed coal. 218
Table 29. Size and source of raw graded silica sands and 238
coal studies.
Table 30. Sieve analysis of garnet sand media (-14+16). 243
viii
-------
Page
Table 31. Upflow and/or downflow experimental runs with the 248
various single media in the 6-in. column.
Table 32. Upflow and downflow experimental runs with dual- 249
media filters in the 6-in. column.
Table 33. Upflow experimental runs with various uniform media 250
and uniform dual media in 2-in. column, 25 °C.
Table 34. Average diameter of uniform media by two methods, mean 254
of adjacent sieve sizes and mean equivalent spherical
diameter by the count and weigh method [Eq. (40)].
Table 35. Summary of the average diameters of the media - d 255
(by several methods).
Table 36. Densities of media - p 256
Table 37. Fixed-bed porosities of the three media determined 257
by the two techniques (e ) and two investigators.
Table 38. Settling velocities of uniform garnet and sand media. 258
Table 39. Expansion - flow rate data of run 1, Series A-13 272
(-14+16 garnet and sand media).
Table 40. Summary of minimum fluidization velocities of garnet 285
sand media - V c.
mf
Table 41. Results of the log V vs log e relationship for 289
garnet sand media.
Table 42. Values of Reynold's numbers and Galileo's number for 291
the garnet sand media.
Table 43. V., n, e f to be used in author's expansion models. 292
Table 44. Results of log V vs log e regression analyses for 293
uniform sized silica sands and coals.
Table 45. Predicted values of m, VA with errors of prediction 295
and maximum error of prediction of expanded bed
depth for uniform sands.
Table 46. Predicted values of n, V^ with errors of prediction 296
and maximum error of prediction of expanded bed
height for uniform coals.
ix
-------
Table 47. Prediction of expanded bed depths for graded
sand using models developed for uniform sands
and average diameter based on the inverse
definition [Eq. (38)].
Table 48. Bulk density difference, lb/ cu ft (garnet-silica
sand).
Table 49. Identification of experimental series.
Table 50. Influent and effluent suspended solids data, average
and range, for wastewater series.
Table 51. Average values of background (bg) turbidity for the
treated wastewaters.
Table 52. Summary of results comparing filter performance for
sharp and mixed interface media.
Table 53. Changes in coal bed over a two-week period of air-
scour exposure equivalent to about 20 years of
normal service.
Page
298
307
324
335
336
336
343
-------
LIST OF FIGURES
Page
Fig. 1. Characteristics of fluidized beds. 17
Fig. 2. Fundamental behavior patterns. 19
Fig. 3. Shear forces on an elemental volume of fluid. 33
Fig. 4. Schematic layout of experimental apparatus. 40
Fig. 5. Head loss curves for run 3B, top 6 in. of filter media. 51
Fig. 6. Variation of the ratio of effluent to influent iron 52
with time.
Fig. 7. Cumulative differential effluent iron vs time, run 8. 53
Fig. 8. Cumulative effluent quality index vs porosity, series 2, 55
12-in. depth.
Fig. 9. Cumulative effluent quality index vs porosity, series 2, 56
all depths.
Fig. 10. Cumulative effluent quality index vs porosity, series 1, 56
* 12-in. depth.
Fig. 11. Cumulative effluent quality index vs expansion, series 3, 57
18-in. depth.
Fig. 12. Cumulative effluent quality index vs expansion, series 3, 58
all depths.
Fig. 13. Backwash water quality vs washwater volume, series 1. 59
Fig. 14. Backwash water quality vs washwater volume, series 1. 60
Fig. 15. Backwash water quality vs washwater volume, series 1. 61
Fig. 16. Terminal backwash water quality vs porosity, series 1. 62
Fig. 17. Backwash water volume vs porosity, series 1. 64
Fig. 18. Iron removable by physical abrasion test vs expansion, 65
runs 20 and 21.
Fig. 19. Pilot-scale Immedium filter used at West Hertfordshire, 71
England [142].
Fig. 20. Immedium filter arrangements for full-scale installations, 72
open and pressure [13].
xi
-------
Fig. 21. Environmental Elements Corp. (Koppers) full-scale 75
automatic backwash filter (from manufacturer's brochure).
Fig. 22. Pilot-scale "Simater" radial-flow, moving-bed sand 77
filter by Simonacco Ltd. of Carlisle, England [68].
Fig. 23. Schematic representation of pilot-scale filter plant 90
used in experimental investigation.
Fig. 24. Pilot plant solids contact unit. 94
Fig. 25. Details of filter boxes. 96
Fig. 26. Abrasion test when filtering secondary effluent treated 112
with alum for phosphorus reduction in Phase I.
Fig. 27. Head loss vs time at various media depths, run 27. 115
Fig. 28. Head loss vs time at various media depths, run 42. 116
Fig. 29. Head loss vs time at various media depths, run 59. 117
Fig. 30. Head loss vs time at various media depths, run 63. 118
Fig. 31. Head loss vs time at various media depths, run 71. 119
Fig. 32. Suspended solids concentration of backwash water vs 124
quantity of backwash water used, run 27, second back-
wash of the south filter immediately following the
first application of air scour.
Fig. 33. Standardized abrasion test results (Phase II) during 136
direct filtration of secondary effluent.
Fig. 34. Initial head loss data for north, south, and west 138
filters for entire Phase II study during direct
filtration of secondary effluent.
Fig. 35. Frequency plot of initial head loss data, phase II. 139
Fig. 36. Chronological head loss development at various media 141
depths, run 2.
Fig. 37. Chronological head loss development at various media 142
depths, run 14.
Fig. 38. Chronological head loss development at various media 143
depths, run 22.
Fig. 39. Chronological head loss development at various media 144
depths, run 29.
xii
-------
Fig. 40. Chronological head loss development at various media 145
depths, run 36.
Fig. 41. Chronological head loss development at various media 146
depths, run 43.
Fig. 42. Chronological head loss development at various media 147
depths, run 54.
Fig. 43. Chronological head loss development at various media 148
depths, run 64.
Fig. 44. Results of special backwash, day no. 242.
Fig. 45. Initial head loss in bottom 16 in. of coarse media 168
filter, observation runs only.
Fig. 46. Standard abrasion test results for Phase III. 17°
Fig. 47. Standard abrasion test results for Phase IV and V. 171
Fig. 48. Initial head loss data for each filter for entire study. 173
Fig. 49. Chronological head loss development at various media 176
depths, day no. 155, Phase III.
Fig. 50. Chronological head loss development at various media 177
depths, day no. 183, Phase III.
Fig. 51. Chronological head loss development at various media 178
depths, day no. 239, Phase IV.
Fig. 52. Chronological head loss development at various media 179
depths, day no. 240, Phase V.
Fig. 53. Chronological head loss development at various media 180
depths, day no. 267, Phase V.
Fig. 54. Chronological head loss development at various media 181
depths, day no. 269, Phase V.
Fig. 55. Chronological head loss development at various media 182
depths, day no. 281, Phase V.
Fig. 56. Relationship between superficial velocity - porosity. 212
Fig. 57. Relationship between n slope and Reynolds number Re . 214
o
Fig. 58. Schematic layout of 6-in. fluidization column. 234
Fig. 59. Schematic layout of 2-in. fluidization column. 236
xiii
-------
Fig. 60. Sieve analysis of graded sand media. 251
Fig. 61. Sieve analysis of graded coal media. 251
Fig. 62. Sieve analysis of garnet sand media. 252
Fig. 63. Fixed-bed head loss for graded sand A at 9 gpm/sq ft 259
for various temperatures.
Fig. 64. Fixed-bed head loss for graded sand A at 26.5 °C for 260
various flow rates.
Fig. 65. Head loss for individual graded media in 1-1/2 in. unit 262
filter sections.
Fig. 66. Head loss for individual graded media in 1-1/2 in. unit 263
filter sections.
Fig. 67. Fixed-bed head loss of dual media AA at various tempera- 264
tures and flow rates.
Fig. 68. Head loss per 1-1/2-in. unit depth in dual media AA and 265
head loss for the two-component media if unmixed.
Fig. 69. Head loss per 1-1/2-in. unit depth in dual media AC and 266
head loss for the two-component media if unmixed.
Fig. 70. Head loss per 1-1/2-in. unit depth in dual media AD and 267
head loss for the two-component media if unmixed.
Fig. 71. Head loss per 1-1/2-in. unit depth in dual media A2E 268
and head loss for the two-component media if unmixed.
Fig. 72. Head loss per 1-1/2-in. unit depth in dual media AF 269
and head loss for the two-component media if unmixed.
Fig. 73. Pressure loss - flow rate diagram for garnet sand media 273
(-14+16).
Fig. 74. Pressure loss - flow rate diagram for garnet sand media 274
(M-60-80).
Fig. 75. Expansion - flow rate characteristics (garnet sand, run 275
1).
Fig. 76. Expansion - flow rate characteristics (garnet sand, run 276
2).
Fig. 77. Expansion - flow rate characteristics (garnet sand, run 277
3).
xiv
-------
Fig. 78. Expansion - flow rate characteristics (garnet sand, run 278
4).
Fig. 79. Expansion - flow rate characteristics (garnet sand, run 279
5).
Fig. 80. Expansion - flow rate characteristics (garnet sand, run 280
6).
Fig. 81. Expansion - flow rate characteristics (garnet sand, run 281
7).
Fig. 82. Expansion - flow rate characteristics (garnet sand, run 282
8).
Fig. 83. Expansion - flow rate characteristics (garnet sand, run 283
9).
Fig. 84. Expansion - flow rate characteristics (garnet sand, run 284
10).
Fig. 85. Log plot of V vs e for garnet sand media (-14+16) (run 286
1, Series A-13).
Fig. 86. Log plot of n slope vs Reynold's number - Re^ (for 287
garnet sand media, runs 1 through 10, Series A-13
through A-17).
Fig. 87. Log plot of Reynold's number, Re^, vs Galileo number, Ga 290
(for garnet sand media, runs 1 through 10, Series A-13
through A-17).
Fig. 88. Minimum fluidization velocity, Vmf, to achieve 10% bed 300
expansion at 25 °C.
Fig. 89. Effect of temperature on Vmf for sand and coal and on 300
absolute viscosity of water.
Fig. 90. Bulk density vs flow rate for garnet sand and silica 302
sands.
Fig. 91. Intermixing of -50+60 garnet sand and -20+25 silica 303
sand.
Fig. 92. Intermixing of -50+60 garnet sand and -30+35 silica 303
sand.
Fig. 93. Intermixing of -50+60 garnet sand and -35+40 silica 304
sand.
Fig. 94. Intermixing of -50+60 garnet sand and -40+45 silica 304
sand.
xv
-------
Fig. 95. Intermixing of silica sand and coal according to the 309
model of Camp et al. [26].
Fig. 96. Bulk density vs flow rate for coal and silica sands 310
(data points not shown on all curves for drafting
convenience).
Fig. 97. Expansion vs flow rate of dual media AA and the two- 312
component media at 22 °C.
Fig. 98. Expansion vs flow rate of dual media A2C2 and the 314
two-component media at 22 °C.
Fig. 99. Schematic diagram of apparatus. 317
Fig. 100. Head loss and filtrate quality vs volume of filtrate, 325
Series IS, run 2, sharp interface, filtration of iron
with CQ = 8.68 to 9.64 mg/1 Fe, Avg 9.07 mg/1.
Fig. 101. Head loss and filtrate quality vs volume of filtrate, 326
Series IM, run 11, mixed interface, filtration of iron
with CQ = 9.2 to 9.7 mg/1 Fe, Avg 9.42 mg/1.
Fig. 102. Head loss and filtrate quality vs volume of filtrate, 327
Series II S, run 1, sharp interface, filtration of
activated sludge effluent with C = 4.5 to 8.5 FTU,
Avg 6.16 FTU. °
Fig. 103. Head loss and filtrate quality vs volume of filtrate, 328
Series II M, run 1, mixed interface filtration of
activated sludge effluent with C = 2.6 to 8.6 FTU,
Avg 4.15 FTU. °
Fig. 104. Head loss and filtrate quality vs volume of filtrate, 329
Series III S, run 3, sharp interface, filtration of
alum coagulated trickling filter effluent C - 3.8 to
10 FTU, Avg 5.21 FTU. °
Fig. 105. Head loss and filtrate quality vs volume of filtrate, 330
Series III M, run 5, mixed interface, filtration of
alum coagulated trickling filter effluent with CQ = 1.7
to 6.2 FTU, Avg 2.95 FTU.
Fig. 106. Head loss and filtrate quality vs volume of filtrate, 331
Series IV S, run 3, sharp interface, filtration of
trickling filter effluent with CQ = 4.7 to 8.5 FTU,
Avg 6.29 FTU.
Fig. 107. Head loss and filtrate quality vs volume of filtrate, 332
Series IV M, run 3, mixed interface, filtration of
trickling filter effluent with CQ = 12 to 15 FTU,
Avg 12.6 FTU.
xvi
-------
Fig. 108. Headless and filtrate quality vs volume of filtrate, 333
Series V S, run 4, sharp interface, filtration of
limesoda ash softening precipitate with C = 5.8 to
6.5 FTU, Avg 6.14 FTU.
Fig. 109. Head loss and filtrate quality vs volume of filtrate, 334
Series V M, run 3, mixed interface, filtration of lime-
soda ash softening precipitate with C = 6.8 to 8.3
FTU, Avg 7.5 FTU. °
Fig. 110. Sieve analysis of coal before and after 14 days of 344
continuous air scour.
xvii
-------
NOTATION, ABBREVIATIONS AND CONVERSION FACTORS
B = Camp's backwash ing number
C - volume
C • influent concentration
o
C = effluent concentration
C/C = effluent concentration/ influent concentration
o
Cn = drag coefficient
d = particle diameter L
D = diameter of column or bed L
d. = diameter of particle in 'i'th layer L
d = equivalent diameter of spherical particle L
d607 finer= 60^° finer of a particle from a probability plot L
d. = l/£(w./d.) = average diameter of particle by
inverse definition L
arithmetic mean diameter L
_2
F. = buoyant force ML T
_2
F£ = impelling force ML T
-2
g * acceleration due to gravity LT
dV1 -1
G = -: — = mean velocity or shear gradient in pores T
CLZ
3 2
Ga = d p(p - p)g/fa Galileo number
s
Gf = superficial fluid mass velocity, Ib(mass)/ „ -
hr sq ft ML T
G , = superficial fluid mass velocity at minimum
m fluidization, Ib (mass)/hr sq ft ML T
h- or HL = head loss in flow through granular bed L
i = subscript denoting the 'i'th layer of the bed
xviii
-------
K = f(Vs, \|f, d/Dt) = constant for a particular LT~
fluidized system
A = height of bed L
SL = height of static bed L
o
m = index of fluid regime
mf = subscript denoting condition at minimum
fluidization
n = slope of log V vs log e plot
N = number of particles
-1 -2
p = pressure intensity ML T
2 -3
P = power dissipated ML T
r = d /d ratio of the particle diameters of
x y
x and y components
R = resistance force per unit projected area of ML T
the particle
Re = pV d/M, = Reynold's number based on the
superficial velocity of the fluid above the bed
Re. = Reynold's number based on the velocity V^
intercept at porosity equals one of the log V vs log
e plot
Re = pVgd/n = Reynold's number based on unhindered
0 settling velocity of particle
-1 -2
S = mean shear stress ML T
t = time T
3
v = volume of particle L
V = superficial velocity of the fluid LT~
above the bed
V. • velocity intercept at a porosity ratio of
one of the log V vs e plot, equal to the
settling velocity of a discrete spherical ,
particle (Vs) when d/D is negligible LT
xix
-------
V f = minimum fluidization or critical fluid
velocity expressed as a superficial _^
velocity LT
V = unhindered settling velocity of a _j_
discrete particle LT
V' = V/e = average fluid velocity within ,
pores of filter LT
W - total weight of particles
W. = weight fraction of 'i'th layer
= Cartesian coordinates L
Greek Symbols
I In
a = [MgK(pg-p>] ' = constant for a particular _j_ _2
fluidized system in optimum backwashing theory ML T
-2 -2
Y = specific weight of fluid ML T
\ - (Ps - P)/
x y -2-2
Y0 = specific weight of particle ML T
a
6 = x/d = dimensionless spacing of particles
in fluidized state
A = prefix signifying increment
e = porosity ratio
ee = porosity ratio of expanded bed
e £ = porosity ratio at minimum fluidization
velocity
e = porosity ratio of fixed bed
,,-lnT1
|a = viscosity of fluid in centipoise ML T
2 -1
V = H/P » kinematic viscosity L T
-3
p = fluid density ML
-3
p, = (1 - e)p + pe = bulk density of mixture MT
.3
p = particle density ML
s
xx
-------
sphericity ratio of the surface area of
an equivalent volume sphere to the actual
surface area of the particle
Abbreviat ions
American Society for Testing Materials ASTM
biochemical oxygen demand BOD
centimeter cm
chemical oxygen demand COD
cubic centimeter cc
cubic feet cu ft
cubic feet per minute cfm
cubic feet per second cfs
degree(s) Celsius C
effective size ES
degree(s) fahrenheit F
feet ft
feet per second fps
Formazin turbidity units FTU
gallon(s) gal.
gallon(s) per minute gpm
gram(s) g
horsepower HP
hour(s) hr
inch(es) in.
inches per minute ipm
xxi
-------
inside diameter ID
Jackson turbidity units JTU
mercury Hg
micrometer(s) pro
microliter(s) Ml
million gallon(s) MG
million gallons per day MGD
milligram(s) per litre mg/1
milliliter(s) ml
millimeter(s) nnn
mimute(s) min
nanometer nm
number no •
outside diameter OD
parts per million PP™
parts per trillion PPfc
percent %
pound(s) 1°
pounds per square inch, absolute psia
pounds per square inch, gage psig
revolutions per minute rPm
root mean square rms
second(s) sec
square centimeter S<1 cm
square feet sq ft
suspended solids ss
xxii
-------
trickling filter
uniformity coefficient
versus
volume
weight
Conversion Factors
TF
UC
vs
vol
wt
The following factors convert the units used in this report to the
more common SI metric unit in popular usage in water engineering
practice.
English Unit x
cfm(cu ft/min)
cfm/sq ft
ft
fps (ft/s)
gal
gal/sq ft
gpm (gal/min)
it it
gpm/sq ft
g/sq ft (gram/sq ft)
hp
in.
in./min
Ib
Ib/cu ft
Ib/hr ft
Ib/hr sq ft
mgd (million gal/day)
psi
sq ft
Multiplier - Metric Unit (SI)
1.68 m3/h
18.288 m3/m2h
(i.e., m/h superficial velocity)
0.3048 m
0.3048 m/s
.003785 m3
.0407 m3/m2
0.2272 m3/h
0.0631 1/s
2.442 m3/m2 h
(i.e., m/h superficial velocity)
10.750 g/m2
0.7457 kw
0.0254 m
1.524 m/h
0.454 kg
16 kg/m3
1.489 kg/h m
4.88 kg/h m2
3785 m3/d
6.9 kN/m2
0.0929 m2
xxiii
-------
ACKNOWLEDGMENTS
The research reported herein was supported by the Engineering Research
Institute at Iowa State University, Ames, Iowa, in part through funds
made available by research grant No. R802140 (formerly 17030 DKG)
from the Office of Research and Monitoring of the U.S. Environmental
Protection Agency.
The report incorporates the work of several graduate student theses
in sanitary engineering at Iowa State. The students were A.
Amirtharajah, R. R. Boss, W. J. Carvalho, J. C. Lorence, A. M. Malik,
G. A. Rice, G. D. Sejkora, E. W. Stangl, and C. F. Woods. Their
contribution is gratefully acknowledged. The report was compiled by
the principal investigator, John L. Cleasby. The assistance of
Oliver Hao in the statistical analysis of the data is also
acknowledged.
xxiv
-------
I. INTRODUCTION
The use of deep granular filters in wastewater treatment is of
growing importance. Filters are an essential unit operation in
many tertiary wastewater treatment flow schemes and in all physical-
chemical wastewater treatment flow schemes.
The deep granular filter used in wastewater treatment is subjected
to more severe operating conditions than it faced in potable water
treatment. The wastewater filter receives heavier and more varia-
ble influent suspended-solids loads, and the solids tend to stick
more tenaciously to the filter media.
The key to long-term operating success of deep granular filters is
proper bed design and adequate bed cleaning during backwashing.
Due to the heavier burden received by such filters, a coarser media
at the entering surface is essential to achieve reasonable filter
run length. Dual- and triple- (multl-) media are commonly being
used to achieve the coarser surface media in the United States.
Coarser coal sizes are advocated to encourage better penetration of
suspended solids (and thus longer filter runs), preventing the
ready transfer of prior experience from the water filtration field
to wastewater filtration. Other approaches, such as deep beds of
coarse sand backwashed without bed expansion and shallow beds of
fine sand backwashed automatically at frequent intervals are also
being used.
The research conducted on this grant was designed to answer a num-
ber of important questions related to the design and operation of
wastewater filters. In general, the original questions proposed
for study were how to select the appropriate media sizes in dual-
and triple-media filters, how to predict the minimum and optimum
backwash flow rate for the media selected, and how to demonstrate
the value of air scour or surface wash as auxiliaries to water
backwashing.
Recent developments have added even greater importance to the proj-
ect, as discussed in the following paragraphs.
The importance of wastewater filters is emphasized by the require-
ment of secondary treatment as the minimum acceptable treatment and
by the recent EPA definition of secondary treatment which requires
an effluent quality of 30 mg/1 BODs and 30 mg/1 suspended solids,
both for 30 consecutive day averages. Many existing municipal
plants cannot meet the 30 BOD, 30 suspended solids goal, especially
trickling filter plants during winter seasons. Tertiary filtration
of such effluents appears to be one of the most attractive alterna-
tives for upgrading such plants to meet the new standards. This
attractiveness exists because the technology is well known from
-------
long use in potable water treatment, and because the operating
costs and energy requirements are low.
Tertiary wastewater filtration is also attractive for plants which
must meet more stringent effluent standards. In some locations,
effluent BOD5 and suspended solids limits of 5 mg/1 each have been
established. Direct filtration of a good secondary effluent will
come close to meeting this goal. Chemical coagulation prior to
filtration will be needed in most cases.
Thus, it appears that a growing number of tertiary filtration
plants will be installed in the next few years. It is vitally im-
portant that tertiary filters be designed and built so they do not
fail to meet the desired standards over their service life. A
vital element of proper design is the provision of adequate back-
washing facilities.
The growing interest in wastewater filtration is evidenced by a.
growing number of equipment companies marketing such equipment.
Some of these companies know very little about filtration, espe-
cially wastewater filtration. The interest is also evidenced by
requests from EPA regional offices for Technology Transfer Seminars
on upgrading secondary plants by tertiary filtration.
During the third year of the current project, the crucial impor-
tance of the backwashing provisions for tertiary wastewater filters
was demonstrated and is reported herein. Water fluidization alone
as a means of backwashing was found to be totally inadequate to
maintain the filter media in acceptable condition. Both the use of
an air-scour auxiliary before the water backwash and the use of a
surface wash auxiliary before and during the water backwash made
substantial improvement in the cleaning effectiveness. However,
some evidence of inadequate backwash remained, even with these
auxiliary cleaning schemes.
Specific Aims
The specific aims of the project from the original research propos-
al are repeated here for reference, with modifications as adopted
in the continuation proposals.
A-l To attempt to extend the successful prediction model of
Amirtharajah for sandexpansion to anthracite coal and garnet
sand filter media. If unsuccessful, attempt to develop a
satisfactory model.
B-l To test the validity of the equal bulk density approach to
prediction of degree of intermixing of two filter media of
different size and specific gravity.
-------
C-l To evaluate the hydraulic behavior of dual- and multi-media
beds with different degrees of interfacial intermixing and
develop a tentative bed design approach.
C-2 To compare the filtering efficiency of a partially intermixed
filter bed with the same media without intermixing. Also, to
compare the ease of backwashing of the two types of filter bed.
C-3 To compare the effectiveness of backwash at various rates pro-
viding different degrees of expansion, and to attempt to veri-
fy or refute the theoretical optimum expected at porosities
of 0.65 to 0.70. Backwash effectiveness will be measured
first by the bed cleanliness resulting from the wash and later
by the quality of the water in the next filter run.
C-4 To compare the effectiveness of a programmed backwash covering
a range of wash rates with a constant rate backwash, and to
recommend an optimum water backwashing procedure. This objec-
tive was dropped in the second year continuation proposal.
C-5 To compare the effectiveness of various air-water backwashing
schemes with the optimum water wash procedure alone. This ob-
jective was expanded to include surface wash auxiliary in the
second year of the study.
C-6 To reaffirm or revise the tentative design of dual- and multi-
media filters suggested in (C-l).
Additional objectives were added in the third year continuation
proposal as follows:
C-7 To determine if coarse, single-media, sand filters can be
cleaned successfully by using air and water simultaneously,
with the water backwash rates well below minimum fluidization
velocity for the media.
C-8 To evaluate the acceptability of filter designs which permit
backwashing with various qualities of backwash water, including
the feed water (secondary effluent).
The original objectives were heavily oriented to dual- and triple-
media filters because that is the common approach in the United
States. During the conduct of the project over a four-year time
period, it became evident that other types of granular filters may
have merit, and some new objectives were added dealing with the
deep-bed coarse sand filter. No work has been included in this
study on proprietary filters such as the ABW filter, the Hydroclear
filter, the Immedium upflow filter, etc. The absence of such work
should not be taken as either a favorable or unfavorable view
toward such filters. Rather, their omission from the study only
reflects that limited objectives had to be selected to fit within
the scope of the project.
-------
II. CONCLUSIONS
A four-year study of wastewater filtration and filter backwashing
is reported herein. Various granular media filters were studied
including those using single-, dual-, and triple-media. Various
methods of backwashing were compared including: (1) water fluidi-
zation only, (2) air scour followed by water fluidization, (3) sur-
face wash and subsurface wash before and during water fluidization
backwash, and (4) simultaneous air scour and subfluidization water
backwash.
Conclusions Regarding Backwash Effectiveness
1. The cleaning of granular media filters by water backwash alone
to fluidize the filter bed is an inherently weak cleaning
method because particle collisions do not occur in a fluidized
bed and thus abrasion between the filter grains is negligible.
2. The cleaning which results in a water fluidized bed is due to
the hydrodynamic shear at the water-media grain interfaces. A
simple mathematical model was developed to calculate the poros-
ity of maximum hydrodynamic shear in a fluidized bed. Maximum
hydrodynamic shear in a fluidized bed occurs at a porosity of
0.68 to 0.71 for sand sizes normally used in filtration. Opti-
mum cleaning of the filter media at this porosity was demon-
strated experimentally.
3. When backwashing by water fluidization alone, a slight economy
in total washwater used is achieved by expanding the bed to
the optimum porosity outlined in conclusion 2 above. Lower
wash rates (anywhere above the rate for minimum fluidization)
will result in nearly the same terminal washwater turbidity,
but proportionately longer backwash times will be required.
Therefore no economy of water use is achieved by use of low
backwash rates.
4. The weakness of water fluidization backwash alone was clearly
demonstrated during wastewater filtration studies where a
dual-media filter which was washed by water fluidization alone
developed serious dirty filter problems such as floating mud
balls, agglomerates at the walls, and media surface cracks.
These problems were observed when filtering either secondary
effluent or secondary effluent which had been treated with
alum for phosphorous reduction.
5. The heavy mud ball and agglomerate accumulations caused higher
initial head losses and shorter filtration cycles. They may
also cause poorer filtrate quality in some cases, although
such detriment was not demonstrated in this study.
-------
6. Simultaneous air scour and subfluidization backwash of coarse
sand filters proved to be the most effective method of backwash.
However, this method should not be used for finer filter media
such as the coals and sands of the typical sizes used in dual-
and triple-media filters because loss of media will occur during
backwash overflow. The choice of the simultaneous air and water
flow rates must be appropriate for the size of sand being used,
and should result in some circulation of the sand for effective
backwashing.
7. The other two methods of improving backwashing, namely air
scour followed by water fluidization backwash, and surface
(and subsurface) wash before and during water fluidization
backwash, proved to be comparable methods of backwash which
can be applied to single-, dual- and triple-media filters.
These two methods did not completely eliminate all dirty fil-
ter problems, but both auxiliaries reduced the problems to
acceptable levels so that filter performance was not impaired.
Conclusions Comparing Filter Performance
8. In a comparison of a coarse sand filter (2 to 3.6-mm sand,
46 in. deep)* with a dual-media filter (1 to 2-mm coal, 15 in.
deep; 0.5 to 0.8-mm sand, 9 in. deep) and a triple-media fil-
ter (same as dual except for the addition of 3 in. of 0.27 to
0.54-mm garnet sand) while filtering secondary effluent, the
coarse sand filter produced a filtrate slightly poorer in
quality, but produced substantially more filtrate to a common
terminal head loss. The performance of the dual- and triple-
media filters was comparable.
9. In a comparison of three coarse media sand filters (2.5 to
3.7-mm sand) of different depths (24, 47, 60 in.) while fil-
tering secondary effluent, there was no apparent difference in
filtrate quality or in rate of head loss development.
Conclusions Regarding Expansion,
Intermixing, and Dual Media
10. Methods for prediction of filter bed expansion are desirable
for rational design of filters and filter backwashing provi-
sions. Existing models for prediction of filter bed expansion
are not adequate for the three filter media in prominent use
today; coal, silica sand, and garnet sand. New unified
*
This report contains common English units since part of the data are
derived from progress reports prepared before the requirement for
reporting in metric units. Metric equivalents will be found on page
xxvi
-------
empirical models for prediction of bed expansion for all
three common media are presented.
11. The expansion prediction models provide acceptable prediction
accuracy for garnet and silica sand, but the coal model is not
sufficiently accurate without further refinement. Some mea-
sure of sphericity for coal seems essential to improvement of
the model. In the expansion models, prediction is made from
measurable properties of the media and the fluid including
media density, fixed bed porosity, sieve analysis, fluid den-
sity and fluid viscosity. The models permit calculation of
expanded bed porosity, bulk density, and bed depth as a func-
tion of backwash flow rate.
12. Complete fluidization of the filter bed during backwashing is
common practice in the United States. Data are presented for
the minimum fluidization velocity of all three media of vari-
ous common sizes. The data will be useful in selecting mini-
mum backwash rates for specific media and water temperatures.
13. The prediction of intermixing tendency between different fil-
ter media in use today is important to rational design of dual
and triple media filters. Lack of such capability may lead to
filters with the media excessively intermixed, or with the
fine media located on the top of the filter bed where it is
not wanted. Two existing models for prediction of intermixing
between media of different size and specific gravity .were
tested against experimental data. Both were found to be lack-
ing in sensitivity.
14. The degree of intermixing is a function of the backwash flow
rate and the manner in which the backwash valve is shut off at
the end of the filter backwash.
15. There is an intermixing tendency between garnet sand and silica
sand, and between silica sand and coal; intermixing in both
tends to increase at higher backwash rates. This is because
the bulk density of the heavier media, in each case, decreases
more rapidly than that of the lighter media as flow rate is
increased.
16. Under some circumstances, the smaller but higher specific
gravity media can be on the bottom of the bed at lower back-
wash rates but move to the top of the bed at higher backwash
rates. Recommended guidelines for differences in bulk density
and ratios of diameter between garnet and silica sand to pre-
vent this occurrence are presented.
-------
17. Rapid shut off of the backwash valve tends to trap the differ-
ent media in the general position associated with the maximum
backwash rate. Very slow closing over several minutes would
be required for restratification of the bed in new positions
associated with lower backwash flow rates.
18. Intermixing at the interface of typical dual-media filters
comprised of coal and silica sand decreases the permeability
of the lower coal layers and increases the permeability of the
upper sand layers. Thus, the intermixed zone has pore diam-
eter, permeability, and fixed-bed hydraulic gradient values
during flow which lie between the values for those parameters
for the individual media. This tends in the desired direction
of providing coarse to fine filtration.
19. The degree of intermixing at the interface of coal and silica
sand filters is determined more by the amount and size of the
finer sands present than by the coal.
20. Interfacial intermixing of dual-media filters does not in it-
self affect filter performance as measured by both head loss
development and effluent quality.
21. Intermixing at the interface of dual-media filters is an un-
avoidable phenomenon which results when United States anthra-
cite coal (sp gr about 1.7) and sand are used and the sizes
are selected to achieve coarse to fine filter media in the
direction of flow.
22. The possible abrasive loss of anthracite coal media due to air
scour was evaluated in a speedup test to simulate the total
period of abrasive exposure which the coal might experience in
a 20-year life. The abrasive loss was found to be negligible.
Conclusions Regarding Wastewater Filter Design
23. The use of some form of air-scour auxiliary or some form of
surface wash auxiliary is essential to the satisfactory func-
tioning of wastewater filters comprised of deep beds (2 to
5 ft) of granular material which are backwashed after several
feet of head loss development. The auxiliary and the backwash
routine must be appropriate to the filter media. For example,
subfluidization wash is limited to single-media filters be-
cause stratification is not essential (or even desired) for
such filters. Fluidization capability is essential for dual-
or triple-media filters to permit restratification of the
layers in their desired positions at the end of the backwash.
Air scour and water backwash simultaneously during overflow is
primarily useful on coarse sand filters because finer media
-------
will be lost due to the violence of the combined air and water
action. However, the simultaneous use of air and water can be
useful on dual- and triple-media prior to the onset of backwash
overflow.
The above conclusion is not intended to apply to all types of
wastewater filters such as the various proprietary filters
with their special backwashing provisions. Such filters and
provisions were not studied in this research.
24. The use of graded gravel to support the filter media without
special provisions to prevent gravel upset is not recommended
where the simultaneous flow of air scour and backwash water
can pass through the gravel by intention, or by accident, due
to the danger of moving the gravel and thus upsetting the de-
sired size stratification of the gravel.
25. Media-retaining underdrain strainers with openings of less
than 1 mm are not recommended for wastewater filters, due to
the danger of progressive clogging.
26. The filter influent feedwater (e.g., secondary effluent) is
not recommended as a backwash water source because of the
danger of progressive clogging of underdrain strainers and/or
gravel. The advantages of using feedwater do not justify the
risks that result therefrom.
27. Air scour is compatible with dual- or triple-media filters
from the standpoint of minimal abrasive loss of the coal
media. However, the backwash routine must be concluded with a
period of fluidization and bed expansion to restratify the
media layers after the air scour.
28. The coal and sand sizes for dual-media filters should be se-
lected to encourage some intermixing at the interface to
achieve improved filtrate quality. However, to prevent some
of the fine sand from reaching the top surface of the coal,
the fines should be skimmed from the sand after installation
of the sand in the filter and hydraulic grading by backwashing.
This skimming is a desirable construction specification even
though it may be inconvenient in the construction scheduling.
29. In the filtration of secondary effluent, a coarser surface
filter media is favored to achieve longer filter run length
and greater solids capture per unit of head loss development.
Substantial difference was observed when comparing a strati-
fied dual-media filter having a coal size of 1 to 2-mm with an
unstratified coarse sand filter of 2 to 3.6-mm sand. The 1 to
2-mm coal gave much shorter run length. As coarser coal sizes
are considered, backwash rates required for fluidization may
-------
become excessive (uneconomical), and the advantage of sub-
fluidization backwash becomes increasingly attractive. How-
ever, the choice of subfluidization backwash dictates the use
of a single-media filter and an air-scour auxiliary rather
than a surface-subsurface wash auxiliary.
30. The use of coarser filter media is also favored by the desire
to eliminate supporting gravel and fine slot strainers in the
underdrain design.
-------
III. RECOMMENDATIONS
The objectives of this research project have been largely fulfilled.
Some require additional work, and, as is the usual case, the re-
search conducted has led to the knowledge of additional work needed.
The following recommendations for additional work are offered;
1. The work reported herein on the stability of double reverse
graded gravel as a media support should be repeated using fil-
tered water as a backwash source.
2. The work reported herein on prediction of the expansion of coal
filter media should be refined to improve its accuracy, perhaps
by incorporation of a sphericity measure in the expansion model,
3. The problem of media loss when backwashing with air and water
simultaneously during overflow should be studied in a system-
atic manner which includes the study of the prominent media ma-
terials and sizes as well as a range of air and water flow
rates.
4. A comprehensive field investigation of existing full-scale
wastewater filters should be made to see if the backwash rou-
tines in use are maintaining the filters in good condition.
The study should include the prominent proprietary filters cur-
rently in use because of the unusual backwash routines employed
in some of these filters.
10
-------
IV. BACKWASHING-POTABLE WATER EXPERIENCE
Rapid filters are washed to restore their capacity when the efflu-
ent quality becomes unacceptable, or when the pressure drop through
the filter reaches a predetermined value. For gravity filters, the
terminal head loss selected is usually the actual head available;
thus operation beyond this time results in an inability to maintain
the desired filtration rate. In some cases, filters are backwashed
on a regular time cycle based on experience with the two criteria
above, or at more frequent intervals dictated by past maintenance
problems created by excessively long filter runs. Filter runs in
different plants may vary from 12 hr to several days, one day being
considered an acceptable average value.
The filter is usually washed by reversing the flow of water through
the filter. In typical United States practice today (1975), the
rate is adequate to lift the grains of filter medium into suspen-
sion, that is, the rising water causes expansion of the filter me-
dium. The deposited material is thus flushed up through the ex-
panded bed and to waste through the washwater gutters.
If the backwash is not effective, dirty filter problems such as
filter cracks and mud balls may occur. Inadequate cleaning leaves
a thin layer of compressible dirt or floe around each grain of the
medium. As pressure drop across the filter medium increases during
the subsequent filter run, the grains are squeezed together and
cracks form in the surface of the medium, usually along the walls
first.
The heavier deposits of solids near the surface of the media break
into pieces during the backwash. These pieces, called mud balls,
may not disintegrate during the backwash. If small enough and of
low enough density, they float on the surface of the fluidized
media. If larger or heavier they may sink into the filter, to the
bottom, or to the sand-coal interface in dual media filters. Ul-
timately they must be broken up or removed from the filter or they
reduce filtration effectiveness or cause shorter filter runs by
dissipating available head loss.
Early sand filters of the late nineteenth century in the United
States were provided with low washwater rates, 8 to 15 in. rise per
minute (in./min), which did not expand the bed, or expanded it only
slightly. Auxiliary methods of agitation were provided in an at-
tempt to clean a filter. These methods included mechanical rakes
that stirred the sand as they rotated in the filter sand and air
agitation systems in which air was introduced into the bottom of
the sand bed before or during the water backwash. Neither method
was completely successful, due to inadequate washwater rates.
11
-------
United States potable water practice has abandoned these methods in
favor of a high velocity backwash (commonly 24 to 30 in./min for 5
to 10 min duration). This wash rate results in about 15 to 30%
expansion of the sands currently specified, depending upon water
temperature, media gradation, and specific gravity. Early work
[62] on the high velocity wash indicated that with the fine sands
then in use, 50% sand expansion gave the best washing results.
However, the trend to the use of coarser sands results in exces-
sively high wash rates if 50% expansion is to be achieved, so it is
seldom attempted in present practice.
The high velocity wash commonly employed in the United States did
not solve all problems with dirty filters, and it has created prob-
lems with the shifting of finer supporting gravel layers when they
are used. The provision of a surface wash system which introduces
high velocity water jets before and during the backwash has largely
solved the problem of dirty filter medium for potable water filters,
but has not solved the problem of shifting gravel.
Evidence of the benefits of surface wash led to its wide adoption
for potable water filters in the United States [60,10]. Surface
wash is introduced at pressures of 45 to 75 psig through orifices
on a fixed piping grid or on a rotating arm, located 1 to 2 in.
above the fixed bed. Surface wash flow rates are about 1 in./min
for the rotary type, and 3 to 6 in./min for the fixed nozzle type [3]
The desired operating sequence involves draining the filter to the
wash trough level or below, applying the surface wash flow with no
concurrent backwash flow for 1 to 2 min to break up surface layers
on top of the media, then continuing the surface wash with concur-
rent backwash flow for several minutes until the backwash water be-
gins to clear up. The concurrent application may be at two rates,
a low rate to barely immerse the surface wash jets in the media
followed by a period with normal bed expansion. The surface wash
is then terminated and water fluidization backwash alone follows
for 1 to 2 min to stratify the bed, which is only important in dual-
media filters.
British and European continental backwashing practice continues to
use low rate backwash with little or no bed expansion assisted by
air scour. This continued use has sparked a, renewed interest in
and use of such practice in the United States since about 1965.
The interest in air scour has also been stimulated by the problem
of shifting gravel and the more difficult backwashing of wastewater
filters. There is also interest in the use of underdrains with
fine strainers that do not require supporting gravel, a system
which was abandoned in the early twentieth century due to clogging
and corrosion problems.
Air scour consists of the distribution of air over the entire fil-
ter area at the bottom of the filter media so that it flows upward
12
-------
through the media. It is used in a number of fashions to improve
the effectiveness of backwashing, and/or to permit the use of lower
backwash water flow rates. The air may be used prior to the water
backwash or concurrently with the water backwash. When used con-
currently during backwash overflow, there is legitimate concern
over potential loss of filter media to the overflow due to the vio-
lent agitation created by the air scour. When air is used alone,
the water level is lowered a few inches below the overflow level to
prevent loss of filter media during the air scour.
Air scour may be introduced to the filter through a pipe system
which is completely separate from the backwash water system, or it
may be through the use of a common system of nozzles (strainers)
which distribute both the air and water, either sequentially or
simultaneously. In either method of distribution, if the air is
introduced below graded gravel supporting the filter media, there
is concern over the movement of the finer gravel by the air, or
especially by air and water used concurrently, by intention or
accident. This concern has lead to the use in some filters of
media-retaining strainers which eliminate the need for graded sup-
port gravel in the filter. However, these strainers may clog with
time, causing decreased backwash flow capability or, possibly,
structural failure of the underdrain system.
In view of the concerns expressed above and the renewed interest in
air scour in the United States, a summary of European air-scour
practice is appropriate because it has been used there since the
beginning of rapid filtration.
British potable water practice has included the use of air scour
for many years. Air scour is used alone first, followed by water
backwash. Plastic strainer nozzles with 3-mm slots are commonly
used in the underdrains and are generally covered by layers of
graded gravel to support the media. Single-media sand filters have
been the most common, but dual-media filters are becoming more com-
mon since about 1970. In the single-media sand filters, the wash
rate is only intended to reach minimum fluidization velocity with
only 1 to 27, expansion of the bed. The sand grain size used in
Britain for potable water filtration is about the same size as in
the United States, although they specify the range of size of the
sand (e.g., 0.5 to 1 mm, meaning that the sand would pass a 1-mm
sieve and be retained on a 0.5-mm sieve) rather than the effective
size. In the backwash operation, the air scour is intended to
loosen the dirt and is followed by the water backwash to flush away
the dirt. Air is introduced through the gravel at rates of 1 to
1.5 scfm/sq ft, sometimes up to 2 scfm/sq ft, followed by water at
8 to 12 in./min [118]. The upper rate is common in current prac-
tice. These British water rates of flow are substantially lower
than the United States practice and are not sufficient for bed ex-
pansion. The apparent success of such low rates must, therefore,
be attributed to the prior air scour and the use of single media
13
-------
which do not require restratification. Problems with gravel move-
ment have occurred but only in a few cases [65]. The absence of
such problems must be attributed to the low water and air flow
rates which are presumably not sufficient to move the fine gravel,
and the fact that air and water are not used simultaneously. Prob-
lems with clogged strainers have been controlled by using strainers
with 3-mm slots after prior clogging problems were experienced
using strainers with 0.5-mm slots.
In the British practice, the water level is lowered to 1 to 2 in.
(5.75 cm) above the fixed bed surface, a little below the edge of
the washwater overflow which is located about 4 in. (10 to 15 cm)
above the surface of the sand. Then air is applied through mush-
room type strainers that are used both for air and water distribu-
tion. The air scour is applied for 3 to 5 min and then is shut
off. Backwash water is then injected immediately through the same
nozzles.
The backwashing practice on the European continent as described
below is obtained from four sources [61,37,70,108] and therefore
probably does not reflect the diversity of the continental practice.
The sources describe several points about continental practice
which differ substantially from both United States and British
practice.
a. Deeper beds of coarser sand are used and the backwash
of these sands is at low rates, with little or no expansion
of the bed (<10%)
b. Backwash is with air and water simultaneously at low water
rates followed by water alone to flush the solids out of the
bed and to the overflow. The air rate is 2 to 4 cfm/sq ft and
the water rate is 10 in./min for the smaller sand sizes (1 to
2-mm size range); for the coarser sizes such as 2 to 3 mm or 2
to 4 mm, the rates are 6 to 8 cfm/sq ft air and 10 to 12 in./
min water. In fact, the use of air and water simultaneously
is considered absolutely essential because air alone compacts
the bed and causes solids to be pushed deeper into the bed
between the rising columns of air bubbles. The potential
danger of media loss is acknowledged if the simultaneous air-
water backwash is continued during overflow. It is suggested
that if loss is observed that the backwash water rate be re-
duced during the simultaneous air-water backwash [37]. Mud
balls are unknown in Europe using this type of bed design and
backwash system.
c. Supporting gravel is sometimes used but it is of the double
reverse graded gravel arrangement, i.e., coarse to fine to
coarse in gradation [108]. Otherwise, media-retaining strain-
ers are used but the dangers of clogging and underdrain
failure are acknowledged [61],
14
-------
The renewed use of air scour in the United States has been pat-
terned more after the British practice of using air scour alone
first, followed by water backwash. United States air rates have
been typically 3 to 5 scfm/sq ft and the subsequent water wash is
above fluidization velocity to expand and restratify the dual-media
bed, typically 24 to 36 in./min. The filters are usually equipped
with media-retaining underdrain strainers without graded gravel
support for the media. Because of the fine sand media used in
potable water filters (0.5 mm effective size), the strainer open-
ings are very small (0.25 to 0.5 mm), and strainer clogging prob-
lems and some underdrain failures have occurred therefrom. Because
of these problems, a reconsideration of the United States air-scour
design practice may be appropriate.
15
-------
V. BACKWASHING WITH FLUIDIZATION AND EXPANSION
In view of the common practice in the United States of backwashing at
high velocity with expansion and fluidization of the filter bed,
studies were conducted on this method of backwashing. Extensive lit-
erature reviews have been completed of both the chemical engineering
fluidization literature and the sanitary engineering backwashing lit-
erature [4,5,28,80]. That literature will not be repeated in total
in this Final Report, but appropriate portions are included to sup-
port the stated conclusions.
Some Fluidization Fundamentals
The phenomenon of fluidization can best be visualized by passing a
fluid (gas or liquid) upward through a bed of solid particles in
which it encounters a resistance to flow and a resultant pressure
drop Ap. As the flow rate V is increased there is a linear relation-
ship between Ap and V. As V is further increased a point is reached
at which the pressure drop is sufficient to bear the weight of the
solid particles. Any further increase in flow rate causes the bed to
expand and accommodate the increased flow while maintaining the pres-
sure drop Ap effectively the same. The fluidized bed thus formed
closely resembles that of a liquid [50]. The feature which distin-
guishes the fluidized bed from other processes (fixed bed, filtration,
etc.) is the motion of the particles within the bed. The character-
istics of an ideal fluidized bed and the distortions due to real con-
ditions are indicated in Fig. 1.
Within the last two decades a flowering of thought has occurred in
the field of fluidization, fertilized by the necessity of its use in
the catalytic cracking of heavy hydrocarbons into petroleum products.
This has given rise to five books in English [35,76,91,144,145] , six
symposia, and countless papers in the literature. The above books
and the reviews of recent work by Coulson and Richardson [33] and
Botterill [15,16] form excellent general references for this study.
Point of Incipient Fluidization or Minimum Fluidizing Velocity - Vfflf
This is the fluid velocity required for the onset of fluidization. It
could be defined exactly as point A in Fig. I for an ideal fluidized
bed. For a real graded bed it is defined by some as the intersection
of the two linear sections of the curve [35,76] while alternate def-
initions are presented by others [110].
The bed is completely fluidized when the friction drag or pressure
drop across the bed is just enough to support the weight of the
filter media [35]. Mathematically, this relationship is given by
hpS = 4(ps - p)gd - e) (1)
16
-------
BED HEIGHT
REAL GRADED BED
IDEAL UNISIZED BED
FIXED BED
LOG
(PRESSURE DROP)
Ap
'HYSTERESIS1
EFFECT
FLU ID I ZED BED
DUE TO NONUNIFORM
FLUID DISTRIBUTION
(CHANNELING)
IF PURE FLUID SECTIONS
PPAI rpAOFn RFn PASS THROUGH BED
REAL GRADED BED (SLUGGING)
IDEAL' UNISIZED BED
JVmf = MINIMUM FLUIDIZING VELOCITY
LOG (SUPERFICIAL FLUID VELOCITY)
Fig. 1. Characteristics of fluidized beds.
17
-------
where
h = pressure drop across the fluidized bed
H = height of expanded bed
e = porosity of expanded bed
g = acceleration due to gravity
p = particle mass density, and
p = fluid mass density.
The simplest bed expansion can be worked out by considering a bed
which is fluidized from initial porosity e0 at height SLO to a poros-
ity e and a height i,. Since the volume of solids within the bed re-
mains constant, then for a bed of constant cross section we have
V1 " so) = A(1 " e)' (2)
Particulate or Homogeneous Fluidization
In most liquid fluidized beds there is a uniform increase of bed
height for velocities greater than Vmf, and the liquid passes smoothly
and appears uniformly distributed within the interstices of the solid
particles as shown in Fig. 2. This type of fluidization with a uni-
form distribution of particles was termed "particulate" by Wilhelm
and Kwauk [141]. The condition of a filter bed while being washed by
water is particulate.
Aggregative or Nonhomogeneous Fluidization
For most gas-fluidized systems, part of the gas breaks through the
bed in the form of bubbles which are considered shaped as a spherical
cap or a sphere with a collection of solid particles at the bottom
as shown in Fig. 2. The two-phase theory of aggregative fluidization
postulates that all gas in excess of minimum fluidization passes
through the bed as bubbles. The bubbles increase in size as they pass
through the bed and burst at the surface of the fluidized bed with a
light scattering of the surface solid particles and those carried by
the bubbles. This is termed "aggregative" fluidization. At higher
rates of gas flow, the frontal diameters of the bubbles build to the
diameter of the containing apparatus, and a condition of the bed
called "slugging" is developed. It cannot be overemphasized that all
discussion in this review referring to aggregative fluidization is a
description of a two-phase system, composed of solids and gas. Air
scouring in backwashing of filters is a three-phase system containing
the media, water, and air, as will be discussed later.
In recent years it has been shown theoretically [7,57,66] as well as
experimentally [35,57,98,113] that both particulate and aggregative
fluidization are the ends of a continually changing spectrum of an
intrinsically unstable system. There are bubbles of liquid in a par-
ticulately fluidized bed, but their magnitude is of the order of the
18
-------
" " ».*•*" ~ ^ ."
***-*• '-'•l* *"+**'
* ',"\^»\\ ^r*%V»'
fTT'
V = V e
"-c^X^-
"> x.'"\'"- t i,
^ _ ,** x^»
M'f
V > V ,
V * >
*» ^
\ , *
X
** *
^ %•
*
* "" ^
•*• s "*k
s. %
w •«-
- ./
N ^
•**»**-*
\ * r
c-»l
DILUTE PHASE
PARTICULATE FLUIDIZATION
V, V>»V, c-*l
IT1T IDT ITlf
BUBBLING SLUGGING DILUTE PHA
AGGREGATIVE FLUIDIZATION
(MOST GAS SYSTEMS)
Fig. 2. Fundamental behavior patterns.
19
-------
solid particles; hence, they are not discernible for a bed of small
height. However, their effects are noticeable in beds of large depth
or when the ratio of the density of solid to liquid increases [66],
Thus steel or lead balls fluidized in water behave aggregatively
while hollow resin or FCC catalysts fluidized by air behave particu-
lately.
It should be noted that an aggregatively fluidized condition cannot
be maintained at large average porosities of the bed due to an in-
adequacy of available solid material to form the shells of the bub-
bles. Thus a reappearance of homogeneous fluidization can be ex-
pected at very high mean bed voidages. Several workers have pre-
sented relationships for predicting whether particulate or aggrega-
tive fluidization will occur [35,104,141,145].
Channeling
Channeling [76,145] is a condition in which the fluidizing medium
passes through a bed of particles along a preferred path. Channeling
can occur in both particulate and aggregatively fluidized beds but
is more common and pronounced in gas-fluidized beds. In an aggrega-
tively gas-fluidized bed, channeling can be described as follows.
The majority of the gas passes through in the form of bubbles ran-
domly distributed throughout the bed. When these bubbles are not
randomly distributed but tend to rise through the bed along a pre-
ferred path, then the bed is approaching a channeling condition.
Once this trouble has started, it will lead to a greater and greater
degree of channeling. Two typical cases of channeling have been
mentioned, "through channeling," when the flow paths extend through
the entire bed, and "intermediate channeling," when only a portion of
the bed is subject to irregularity. The design of the gas inlet de-
vice at the bottom of the bed has an important effect on channeling.
Channeling is also affected by the characteristics of the solid phase
such as size, shape, and density. It is more pronounced for finer
particles which tend to agglomerate. Channeling tendencies are al-
ways smaller with porous plate distributors than with multi-orifice
distributors.
Spouting
Spouting [12,76] is a technique for agitating a bed of particles too
coarse to fluidize well. It is in a sense a combination of a dilute
fluidized phase in the form of a rising spout surrounded by a down-
ward moving fixed bed of solids. Becker [12] observed that there is
a maximum bed depth above which spouting cannot occur. He observed
that the minimum spouting velocity, expressed as a superficial veloc-
ity, required to fluidize the maximum spoutable bed depth is identi-
cal with the minimum superficial gas velocity causing onset of aggre-
gative fluidization for the same fluid and solid system. The author
differentiated between real spouting and pseudospouting. According to
20
-------
him, pseudospouting is a case in which an internal channel is formed
under any condition. Pseudospouting is described as a feeble form of
agitation. It is nothing more than an empty vent containing essen-
tially no rising solids and owing its stability to the bridging tend-
encies of the particles. In real spouting, agitation is caused by an
intense, continuous jet of gas piercing a quiescent bed and yielding
its energy to generate a well defined stream of particles, the spout,
whose upward motion is balanced by the downward motion of the sur-
rounding annulus.
Gas-Liquid Fluidization
Some research workers [39,89,119] have studied the use of gas-liquid
fluidized beds. In gas-liquid fluidization, the liquid flows upwards
through a bed of solid particles which is fluidized by the flowing
liquid, while the gaseous phase moves as discrete bubbles through the
liquid fluidized bed. Gas-liquid-particle operations, also called
three-phase systems, are of comparatively complicated physical nature.
Three phases are present, the flow patterns are extremely complex,
and exact mathematical models of the fluid flows and mass transport
in these operations probably cannot be developed at the present time.
Description of these systems will be based upon simplified concepts.
These processes are, however, distinguished by their high rate for
various purposes (e.g., heat transfer), good phase contact, and wide
ranges over which the process can be varied.
For air scouring to be identical to aggregative fluidization, the
water from the filter must be completely drained and sufficient air
supplied to fluidize the sand bed. The rate of air supply required
to fluidize a sand bed of average size (0.8 mm) would be 80-100
cfm/sq ft [4] as compared to the 3 to 5 cfm/sq ft commonly used for air
scour in backwashing. It is clear that a true case of aggregative
fluidization is not possible in a normal air-scour operation in fil-
tration. But there are similarities between three-phase fluidization
and aggregative fluidization. Three-phase fluidization in filter
backwashing exists when air is applied to a bed which is fluidized
with water. A three-phase system of backwashing rapid sand filters
such as this is discussed by Camp [26] and Simmonds [112]. Note
that there is a danger of loss of filter media in this operation un-
less the air scour is stopped before the water level overflows into
the wash troughs. The more common method of air scouring is a special
case of a three-phase system where the water level is lowered to just
above the top surface of the sand and then the bed is air scoured at
the rate of 3 to 5 cfm/sq ft without the concurrent upward flow of water
and without fluidization of the bed.
It has been observed [89,119] that an increase of the gas flow rate
often causes a decrease of bed expansion for three-phase fluidization,
whereas bed expansion increases as the flow rate of the fluid medium
is increased for two-phase fluidization. It was reported [89] that
this reduction in bed height or porosity is more marked in beds of
21
-------
small particles than in beds of large particles. A contraction of
48% has been observed in a highly expanded bed of 0.28-nnn ballotini.
The reduction in bed height was explained by the hypothesis that a
portion of the liquid moved upward in the wakes of the gas bubbles at
a velocity much higher than the average liquid velocity. The liquid
velocity in the rest of the bed was consequently reduced below the
average liquid velocity, and the expansion of the bed is reduced cor-
respondingly. This reduction in bed height is more than the gas vol-
ume present at any instant, thus causing a net reduction of total bed
volume.
Rigby et al. [101] have investigated bubble properties in three-phase
systems. They showed that gas bubbles have a greater tendency to rise
in the center of the bed than near the walls. This tendency was
observed even when bubbling at the base of the column was quite
uniform. As result of this tendency, these authors concluded that
the most favorable gas distributor may not be one which distributes
gas evenly over the bed cross section, as has generally been assumed,
but one which introduces relatively more gas near the walls to
counteract the natural tendency for bubbles to rise in the center
of the bed.
Ostergaard [90] measured the rate of growth of gas bubbles formed in
a liquid-fluidized bed at a single orifice of 3.00-mm diameter for gas
flow rates varying from 9 to 63 cc/sec. The experiments were carried out
with tap water, atmospheric air, and sand particles having an average
equivalent diameter of 0.64 mm. The equivalent diameter was defined
as the diameter of a sphere with the same average particle volume.
The bubble frequency at the orifice was measured by an electrical re-
sistance probe connected to an oscilloscope. The bubble frequency at
the bed surface was calculated from cinephotographs. The measured
rate of coalescence was markedly dependent on bed porosity, having a
relatively high value near the point of incipient fluidization and de-
creasing with increasing liquid velocity and bed porosity. This is in
general agreement with the results of Rigby et al. [101], as is the
observation that the main change in bubble frequency occurs within a
relatively short distance from the orifice. The rate of coalescence
did not vary significantly with gas flow rate. Observations of bubbles
emerging through the bed surface show that bubble shape is markedly
dependent on liquid velocity. A bed near incipient fluidization is
characterized by a high viscosity, and an emerging bubble is of nearly
spherical shape, whereas a fluidized bed of high porosity is charac-
terized by a viscosity not very much higher than that of water, so
that an emerging bubble is of spherical cap shape. The author [90]
also observed that no individual bubbles were observed at the orifice
at zero liquid velocity. This was probably due to the formation of
gas channels in the fixed bed.
22
-------
Predominance of Hydrod.yn.amic Forces
in Cleaning by Water Fluidizatlon Alone
The basic hydrodynamic behavior of a fluidized bed is frequently for-
gotten in trying to explain the removal of impurities from the grain
of a filter sand during backwashing. Fair and Geyer [45] mention
that "substances adhering to the filter grains are dislodged ... by
the rubbing together of the suspended grains." Babbitt et al. [8]
state that the "purpose of such expansion is to cause the sand grains
to rub against one another."
From purely theoretical grounds the suspension of particles in a ris-
ing stream of fluid is expected to require a field of flow around each
particle, thus negating the concept of a number of particles rubbing
together when fluidized. Considerable direct evidence [105,106,135]
as well as most correlations [4,76,91] are based on the assumption that
the particles are uniformly distributed within the beds. Thus, exper-
imental verification of these correlations implies that the assumption
of uniform distribution is reasonably valid. Further qualitative sup-
port of the above assumption is the fact that particle attrition [145]
is negligibly small in fluidized beds and also the fact that however
well filters are backwashed, sand growth by layers of deposited mate-
rial frequently occurs in significant amounts. Johnson and Cleasby
noted a growth from 0.43 mm to 0.65 mm in 14 years at the Ames plant
[67].
The most significant work which removed the above from the realm of
postulates to that of fact is Rowe's studies of "Drag Forces in Hy-
draulic Models of Fluidized Beds I, II" [105,106]. He showed in a
fundamental study that the drag forces on spheres arranged in regular
arrays is extremely sensitive to the separation between the particles.
The required modification to the drag coefficient for a single parti-
cle CD> due to neighboring particles, was effectively to multiply it
by [1 + (0.68/6)], where 6 = x/d, the dimensionless spacing of the
particles based on the particle diameter, d, and the clear distance
between particles, x. Particles in rhombohedral packing were found
to be subjected to a drag about 70 times greater than that on an iso-
lated particle for the same superficial velocity of the fluid [106].
The value of the drag coefficient given by the above expression when
6 = 0.01 is 69 and was considered to refer to the maximum condition
[105].
Rowe's studies showed that small local changes of particle concentra-
tion were unstable because they required a very large change in the
velocity distribution. A local decrease in particle concentration of
37,, required the velocities to be doubled. It can be seen that as 6"*0,
the drag coefficient -» »; however, the expression does not apply for
6=0. The studies indicate the existence of lateral repulsive hydro-
dynamic forces between particles; these became extremely large as the
spacing between particles was reduced. Thus physical contacts between
particles in fluidized beds were extremely limited, and the particles
23
-------
were uniformly distributed in the fluid field. Rowe clearly indicates
that his development does not eliminate the existence of particle con-
tacts and concentration effects, but only shows that they cannot per-
sist and that their effect is negligible. Adler and Happel [1] have
also indicated that solid-solid frictional effects in the low porosity
range of fluidized beds where they should be most significant were
inconsequential.
In a study of the stability of particulate fluidization, Jackson de-
veloped general equations of motion for a fluidized assembly of iden-
tical particles [66] . In discussing the equations of motion, he ne-
glected terms due to the direct interaction by collision and justified
it in two ways. First, if collisions between particles were of com-
parable importance to drag forces then the number of particles per
unit volume would be expected to decrease with height above the bed
support, in the same way as the pressure of the atmosphere decreases
with height above the earth's surface. Second, a posteriori justi-
fication was provided by the fact that the equations of motion ob-
tained by neglecting collisions gave a good qualitative account of the
main phenomena of fluidization.
Murray [85] summarized Jackson's arguments and also gave some of the
other reasons for the fact that negligible particle collisions occur
in fluidized beds. He stated that
Collision forces, which are a form of particle pressure,
are also small, since, if such a term were important it
would probably increase with rip (ru, = number density of
the particles). This would result in a gradation in n«
from the surface into the bed from zero to a finite val-
ue, this is not observed. The surface appears to be a
discontinuity. Furthermore, observation of particle
flow round a bubble by X-ray techniques seems to show
little or no contact interference between neighboring
particles. Also, if collisions were frequent, the noise
would be noticeable, which is not the case.
Though Murray was chiefly concerned with aggregative fluidization, it
is a well known fact [4,35,85,145] that the fluidized section of the
bed outside the bubbles is very similar to a particulately fluidized
bed at minimum fluidization. The hydrodynamical behavior of the con-
tinuous phase is hence similar to that of a particulately fluidized
bed.
In a recent two-dimensional study (which may not effectively extra-
polate to three dimensions) using a monolayer of fluidized particles,
Volpicelli et al. [135] indicated the presence of inhomogeneities in
particle distribution within the bed. They presented photographic
stills from their motion picture, study of the fluidized beds. A
study of these photographs by the author of this report showed that
particle contacts are rare, which confirms Rowe's studies. Their
24
-------
studies must be interpreted with caution since they used steel balls
and water as one of their systems; this system will exhibit marked
aggregative tendencies. This study also indicated that particle flow
patterns switch from a circulation regime to a regime of quasirandom
motion as the voidage increases. This characteristic, confirmed by
other workers too, has important implications in the development of
criteria for backwashing at optimum rates.
In a recent Russian theoretical study, on the pseudoturbulent diffusion
of particles in homogeneous suspensions using tensor analysis, turbu-
lence equations were developed for a two-phase system by Buevich and
Markov [20], In a discussion on the .collisional dissipation of energy
the authors stated that particles undergoing collision have step
changes in velocity which will always be small for dilute suspensions.
They also said that the collisions of particles suspended in a liquid
are characteristically very gradual, and there are no step-wise
changes in the particle velocities. The latter is due to significant
increases in the pressure in the liquid layer between the particles
as they approach each other and to the need for "squeezing out" this
layer before direct contact of the particles occurs. An analogous
effect also occurred as particles approach a solid wall and in lubri-
cation processes, when the lubricating liquid in the gap between bear-
ing and slider played the part of this liquid layer. Hence, they con-
cluded that any model of energy dissipation based on elastic colli-
sions will be in error by at least an order of magnitude.
Ruckenstein [107] developed a physical model for a homogeneous (par-
ticulate) fluidized bed, using the equation of motion of one particle
which is part of an ensemble of particles in interaction with a fluid.
The equation is established by neglecting the interaction by collision
of the particles of the ensemble.
All the above evidence in the fluidization literature pinpoints one
single fact: the effect of collisional interactions between particles
in the fluidized state is comparatively insignificant. This fact, now
becoming more and more accepted in the sanitary engineering literature
[4,6,26,132], indicates that the age-old argument whether abrasions
between particles or the hydrodynamic shear forces are the predominant
cleaning mechanism, should finally be laid to rest. This conclusion
is also one of the basic assumptions of the theory of optimum back-
washing developed in this chapter.
25
-------
VI. OPTIMUM CLEANING BY WATER BACKWASH ALONE
Particulate Fluidization and Optimum Turbulence —
Evidence From the Literature
Analyzing a filter being backwashed, qualitatively, will indicate that
at minimum fluidization individual particles have no motion and that
frequently the fluid motion is streamline, and hence cleaning of the
media will be negligible; at the other extreme overexpansion of the
bed will also reduce the cleaning action due to large separations be-
tween particles. This macroscopic analysis indicates that somewhere
between the two extremes lies an optimum condition which we seek.
The review of the literature quoted in this section indicates a strik-
ing phenomenon discovered in particulate fluidization research: the
existence of a maximum value for most turbulence parameters at a
porosity of 0.65 to 0.70. This was the fact that originally led
Amirtharajah toward inferring that the elusive condition of optimum
backwash would probably be centered around this porosity [4]. This
section summarizes his theoretical and experimental proof of this
hypothesis [5].
Considerable evidence was collated in a previous section to show that
particle abrasions or collision are inconsequential in a fluidized
bed. This fact leads immediately to the deduction of two very impor-
tant hypotheses: (1) the present United States mode of cleaning
filters by fluidization has an intrinsic weakness in the process it-
self . and (2) the most that can be achieved from the process is to
backwash at flow rates which will produce the maximum turbulence and
the maximum shear in the fluid-par tide field, for this is the prin-
cipal mode of cleaning. The first weakness is being remedied by the
use of backwash auxiliaries.such as air scour or surface washers.
The second problem is far more tractable both theoretically as well
as experimentally with the systems we have at present and the know-
ledge we have in fluidization.
Hanratty, Latinen and Wilhelm [56] were the first to use Taylor's tur-
bulence equations to describe the diffusion of a tracer dye in par-
ticulately fluidized beds. They established the mixing parameters —
eddy diffusivity and the scales and intensities of turbulence. The
experimental studies were made in a 5.40-cm Lucite tube, by admitting
methylene blue dye from a central location to the bed of fluidized
particles. Four different systems consisting of glass spheres of dia-
meter 0.47, 0.93, and 3 mm, as well as silica spheres of 1.84 mm were
used for the solids in the bed. In all runs except two, a constant
expanded bed height of 20.3 cm was maintained, and the porosity was
adjusted by changing the amount of solids in the bed. Two types of
sampling traverses — radial and centerline — were used to measure the
spreading of the dye.
26
-------
The results indicated that the theoretical equations of Taylor's
theory of turbulence were applicable for diffusion in a particulately
fluidized bed. A minimum Peclet number, i.e., maximum eddy diffu-
sivity, was found for all particle sizes and corresponded to a poros-
ity of 0.70. This maximum eddy diffusivity was directly related to
the maximum in the length scale of turbulence, which also occurred at
this porosity. The intensity of turbulence increased continuously for
all porosities. For the bed of 3-mm spheres the maximum scale of
turbulence was approximately 4 mm.
Hanratty et al. did not attempt to provide a quantitative explanation
for the minimum Peclet number at the porosity of 0.70. Qualitatively
explaining the observed phenomenon in terms of the random-walk model
they stated
Mixing in a packed bed was found to be explainable in terms
of a random-walk model, and it is suggested that a dense
fluidized bed retains elements of this mechanism. The dis-
tance a fluid element must side step in order to pass around
a particle decreases as the bed is expanded. Eventually, at
a fraction void of 0.70, a fluid element may begin to flow
past solid particles without the necessity at each level of
flowing laterally in order to evade a particle. Beyond the
critical fraction void, in dilute beds, the turbulence is
particle generated, and the eddy diffusivity is a direct
function of particle population, leading to an increase in
the Peclet number as the velocity is increased.
Cairns and Prausnitz [21,22] studied macroscopic mixing and longitu-
dinal mixing in solid liquid fluidized beds. The studies in longitu-
dinal mixing were made by determining the electrical conductance
break-through curves using very small electrical conductivity probes
with a step-function input of salt-solution tracer. The principal
advantage of the conductivity method is that it enables continuous
monitoring, and hence the tracing, of transient velocities in the
system. Longitudinal eddy diffusivities were determined for 1.3- and
3.0-mm lead spheres and 3.2-mm glass spheres in 2- and 4-in. diameter
beds at a distance of five bed diameters from the injection point.
The analysis of the data was based on a statistical model developed by
H. A. Einstein in connection with the motion of pebbles in a water
stream. The model gives an easy and rapid method of determining the
Peclet groups from the experimental data. The longitudinal eddy dif-
fusivities were determined for various solids-to-column-diameter
ratios, various radial positions and various void fractions. The re-
sults were consistent with the fact that the fluidized bed was con-
sidered as a transition between a packed bed and an open tube. It
was found that the ratio of longitudinal to radial eddy diffusivity
was approximately 20 to 30. Thus, the rate of longitudinal mass
transfer was very much greater than the radial transfer.
27
-------
In all cases a maximum of the longitudinal eddy diffusivity occurred
at a porosity of 0.65 to 0.70. This maximum was much more pronounced
for the lead sphere system than for the glass sphere system. On
defining the Peclet group in terms of the diameter of the solids in
the system, a minimum Peclet number occurred at a porosity of 0.7 in
all cases. However, for the glass sphere system an asymptotic mini-
mum value was reached for all porosities greater than or equal to
a porosity of 0.7. The authors concluded that the eddy diffusivity
was strongly affected by the density and concentration of particles in
fluidized beds and a maximum in the mixing properties occurred at a
fraction voids of 0.7.
In the study reported later the authors investigated macroscopic mix-
ing [22] with a similar experimental setup. They measured the fre-
quency distribution of fluctuations and the correlation coefficients
at two points separated by a known distance. From the mixing data,
radial eddy diffusivities, scales of turbulence, and intensities of
turbulence were measured.
Since mixing-length theory suggested that the mixing process in
fluidized beds was analogous to molecular diffusion, a similar ana-
lytical model for a steady-state system in cylindrical coordinates was
used for analysis of the data. The results showed that the scale of
turbulence maximized at a porosity of 0.70. A plot of the Peclet num-
ber indicated a minimum corresponding to this same porosity. Detailed
visual records of the behavior of the solids and the fluid were also
noted. The most active particle motion was found to occur in the
range of e = 0.70; the motion of the particles changed from a circu-
lation pattern to that of random motion. This change in flow pattern
was also noted by Volpicelli et al. [135], as recorded before, in
their studies of a monolayer of particles in the vertical plane.
Lemlich and Caldas [75] studied the heat transfer characteristics from
the wall to the fluid within a particulately fluidized bed. They
used a bed of glass spheres fluidized by water and found that the heat
transfer coefficient maximized at a transition between two regimes of
flow. The lower regime indicated limited axial mixing, while con-
siderable mixing was evident in the temperature profiles at the higher
rates of flow. The maximum heat transfer coefficient occurred at
porosities of 0.66 and 0.81 for solid particles having diameters of
0.50 and 0.29 nun, respectively.
Handley et al. [55] studied the mechanics of fluid and particles in a
particulately fluidized bed using unusual experimental techniques.
They obtained a transparent solid-liquid system using soda glass
(density = 2.50, refractive index = 1.52) and methyl benzoate (density
= 1.08; refractive index - 1.52). An opaque white glass tracer parti-
cle having the same properties as the transparent soda glass was used,
and the motion of this particle was studied by cine photography. A
similar technique was used in Ulug1s studies of flow visualization
[132J.
28
-------
Fifty histograms of displacement vectors measured from projected
stills showed a mean zero solids velocity and showed that standard
deviations were independent of radial or vertical positions in the
bed. The mean zero solids velocity means that a single particle over
a sufficient length of time does not have a velocity, which is to be
expected because solids in a fluidized bed do not have a finite ve-
locity over a long period. Thus, particle motion was random, and ho-
mogeneous although isotropic conditions were not obtained. They then
applied Taylor's random-walk statistical analysis [49,125,126] to the
motion of the particles as well as the fluid regimes. The results
showed that the root mean square (rms) of the turbulent fluid velocity,
passed through a maximum at some voidage between 0.44 and 0.75; the
extrapolated graph indicated a value near 0.70. The rms of the tur-
bulent particle velocity similarly passed through a maximum at a void-
age between 0.44 and 0.75. Using fluid dynamic pressure measurements
(made with a pitot meter), they also found that the vertical turbulent
fluid velocity component passed through a maximum at e = 0.68 to 0.70.
Unfortunately, due to the experimental limitation requiring trans-
parency, the authors did not investigate sufficient systems expanded
to porosities less than 0.65.
In one of the pioneering studies in fluidization, McCune and Wilhelm
[81] determined mass transfer characteristics by measuring the partial
dissolution of spherical and flake-shaped napthol particles in a flu-
idizing stream of water. They found that the mass transfer character-
istics for 1/8-in. pellets maximized around a porosity of 0.70 to 0.75.
In a recent study, Galloway and Sage [51] , using an instrumented cop-
per sphere within a packed bed of spheres, studied the local thermal
transfer from the instrumented sphere. Using the data of McCune and
Wilhelm [81] and Rowe [105,106], they established a boundary layer
model based on the behavior of thermal and material transfer from sin-
gle spheres and cylinders in turbulent fluid streams. The studies
showed that a maximum mass transfer occurred at fraction voids of 0.70
for fixed beds. Extending the use of the model for fluidized beds with
literature data, they showed that the maximum mass transfer correlated
with the maximum turbulence at the expanded porosity of 0.70.
Except for the qualitative observation of changing flow fields, the re-
searchers have not sought to explain why this maximum occurred at this
porosity. Probably the answer to this lies in Jackson's studies [66].
Jackson probed the fundamental mechanics of the fluidized bed and con-
cluded that particulate and aggregative fluidization were both mechan-
istically unstable systems. The instability manifested itself as
traveling waves of increasing amplitude. His theory predicted that
disturbances similar to those of bubbles in aggregative systems would
also develop in particulately fluidized beds. However, for particulate
systems these disturbances are of the same order of magnitude as the
size of the solids in the system and do not develop to a noticeable
extent, except in very deep beds. The visual observations of Cairns
and Prausnitz [22] previously described were mentioned by Jackson as
29
-------
evidence for the results he deduced from his stability study. As ad-
ditional evidence he referred to the papers by Slis et al. [115] and
Kramers et al. [73].
Kramers, who was a coauthor of both papers [73,115], studied in the
second paper [73] the longitudinal dispersion of liquid in a fluidized
bed. The study used an experimental setup similar to that used by
Cairns and Prausnitz [21], except that Kramers et al. used very deep
beds (12 and 6 m) and fluidized glass spheres of diameters 0.50 and
1.0 mm [73]. In order to avoid any external influences they took
great care to eliminate as far as possible all visible systematic
eddies. In fact the tubes used for the bed were purchased as a single
piece and had no connections or protuberances.
The results indicated [73] a hump in the longitudinal diffusivity at
a porosity of 0.7 for the 0.50-mm particles. However, at porosities
greater than 0.75, the value of the diffusivity continued to increase.
For the 1.0-mm system the hump at the porosity of 0.7 was barely per-
ceptible. Analyzing these results in terms of the reported maxima of
Cairns and Prausnitz's studies [21,22], the authors concluded that the
eddy diffusivity was composed of two parts. One part was supposed to
be due to the eddies produced by individual particles, and this con-
tribution passed through a maximum at the porosity e =* 0.70. The
other part, which strongly increased at higher values of porosity, was
thought to be connected with the presence of local porosity flucta-
tions which were seen to travel upwards through the bed.
All the above studies have considered a solid phase and a pure fluid-
izing liquid. In order for the above results to be directly appli-
cable to backwashing, it is necessary to assess whether the presence
of a number of small floe particles in the liquid phase will affect
the turbulence parameters. Precisely this question, the effects of
solids on turbulence in a fluid, was studied by Kada and Hanratty [71].
Using the same technique as that used by Hanratty et al. [56] to study
turbulent diffusion in fluidized beds, mentioned previously in this
report, Kada and Hanratty studied the effect of solids concentrations
of 0.13 to 2.5% by volume on the turbulent dispersion characteristics
of a pure fluid. They found that one of the chief variables affecting
the system was the slip velocity, which was the difference between
the particle and fluid velocities in the direction of flow. For the
systems studied the slip velocity was equal to that of the free fall
velocity. It was found that glass particles having diameters of 0.10
and 0.38 mm and concentrations of 1.5 and 1.7% by volume had no ef-
fect at all on the turbulent dispersion characteristics of the fluid.
One of the principal conclusions of the study was that for systems
with small slip velocities the effect of solid concentrations up to
1.5% had no effect at all on the dispersion characteristics. This
study provides the final argument for applying the results of the
studies reviewed in this chapter to the backwashing of filters.
30
-------
The evidence recorded above is impressive, since several studies have
confirmed the central result. The impressiveness lies in the fact
that totally different experimental techniques used to study differ-
ent, but hydrodynamically related characteristics, namely, scales of
turbulence, eddy diffusivities, Peclet numbers, particle and fluid
motions, mass transfer and heat transfer effects, all yielded the
surprising fact of maximization at a porosity of approximately 0.65
to 0.70. The very few studies which have sought to explain the reason
for this maximization of the turbulence parameters indicate that the
diffusivity is probably due to two factors: the eddies associated
with the particles and the porosity fluctuations which travel up the
bed.
The essential conclusions of the preceding literature review can be
summarized as follows:
1. Considerable evidence exists in the fluidization literature that
particle collisions in the fluidized state are of negligible
consequence compared to the hydrodynamic effects. This fact is
also being realized in the sanitary engineering field and as a
corollary it implies that the use of water fluidization alone to
clean a filter is inherently unsatisfactory, and various auxili-
ary scour systems can be used to overcome this weakness.
2. The fluidization literature abounds with evidence that the fluid
and particle fields in particulate fluidization can be described
by the statistical turbulence theories, even though the actual
fluid velocities are not in the turbulent regime. It has been
found that most turbulence parameters have a maximum at an ex-
panded porosity of 0.65 to 0.70. It is hence hypothesized that,
within the constraint that fluidization is not an excellent
process for cleaning, the best cleaning that can be achieved is
by expanding the bed to these porosities.
3. Only qualitative and semiquantitative attempts have been made to
unravel the reasons for the optimum in the turbulence parameters.
These have been based on the assumption that dispersion is
caused by (a) eddies around individual particles and (b) the
movement of porosity fluctuations through the bed.
The theory developed for optimum backwashing in the next section of
this chapter is entirely new and original; however, it draws suste-
nance principally from the conclusions and results reported and sum-
marized in this literature review.
A New Theory of Optimum Backwashinp
by Water Fluidization Only
It has been shown in the previous sections that filter cleaning dur-
ing backwash is due to the turbulence of the fluidized bed and
31
-------
hydrodynamic shear forces. The voluminous fluidization literature
quoted has shown that turbulent diffusion is maximized at the poros-
ity of 0.65 to 0.70. Some mathematical theories have sought to iden-
tify this maximum with the eddies around particles and with the fluc-
tuations in porosity which travel up the fluidized bed.
The following new theory shows that hydrodynamic shear forces in a
fluidized bed reach a maximum at this porosity of 0.65 to 0.70 for
most systems considered. Even if the turbulence parameters cannot be
directly related to the shear forces at the present time, on the
basis of the discussion in the previous section the cleaning in a
backwashed filter bed necessarily reaches an optimum with the maximum
shear. Since shear and turbulence parameters are inseparably related
it should be possible as a future extension of the theory to obtain
an analytical model which will relate the maximum in the hydrodynamic
shear to the maxima in the turbulence parameters. This is also pred-
icated on the fact that the actual fluid velocities in a fluidized
bed are in the transitional regime and that only the interaction with
the solids produces a system having behavior similar to that of a
turbulent field.
Consider a one-dimensional analysis of the motion of an elemental
volume of fluid as in Fig. 3. Let p be pressure intensity, S shear
stress, V velocity, and Ax, Ay, and Az the dimensions of the cube.
The work done by shearing stresses is irreversible and is dissipated
as heat. This loss in energy corresponds to what is frequently called
head loss or friction loss in fluid flow problems.
Let the power dissipated by the torque composed of the shear forces
due to S, by P^,
P.. = torque X angular velocity
(SAxAy)Az X
dV
S ——AxAyAz .
By definition
dV
S = p, -:— where jj, is absolute viscosity,
Therefore,
32
-------
(pAyAz)
(a) FORCES
SAxAy
Az
dV
Ax
V1
(b) VELOCITIES
Fig. 3. Shear forces on an elemental volume of fluid.
33
-------
Let the hydraulic gradient in the z direction or the head loss per
unit length be (dh/dz) .
Hence the power dissipated by the element of height Az and area of
cross section AxAy, moving with a velocity V' , and having mass den-
sity p is
P2 = AZ ' AxAyPg ' V'
Since the power dissipated by the shear forces corresponds to the
head loss, Eqs. (3) and (4) give
= (^AxAyAzpgV,
that is,
1
IhM 2 .
lz>J (5)
Now, [(dh/dz)pgV] is the power dissipated per unit volume, say P/C,
and G is the velocity or shear gradient defined as (dV'/dz); hence
Eq. (5) reduces to
I
G = ye/ (6)
Equation (6) is the familiar form of Eq. (5), originally derived by
Camp and Stein [27] as the general power dissipation function in a
three-dimensional treatment. The derivation of Eq. (6) in condensed
form is also presented in Fair, Geyer, and Okun [46].
For a fluidized bed with a superficial velocity V, where V1 = V/e,
the hydraulic gradient (dh/dz) = i, and (j,/p is kinematic viscosity
, Eq. (5) becomes
(7)
Equation (7) has previously been presented by Camp [25] for calcu-
lating the shear forces during filtration and backwashing.
Equation (5) will now be put in the form most useful for the follow-
ing development of the theory of optimum backwashing:
dV f V /dh\l2
34
-------
that is,
-.1
"'-"l)2 (8)
where
S = shear intensity
V = superficial fluid velocity
(-:— J = head loss gradient.
An important property of fluidized beds arises from the fact that
particles suspended in a fluid require that the frictional drag of
the fluid exactly counterbalance the pull of gravity. In effect,
this leads to the requirement that the head loss across a fluidized
bed must equal the buoyant weight of the particles [Eq. (1)]. Two
of the earliest researchers to report this well known property were
Fair and Hatch [47]. In differential form this result is
dhpg = dz(po - p)g(l - e)
S
that is
/dh\ (PS "
(S)
p (1 - e) (9)
Equations (8) and (9) above and Eq. (10) below, form the principal
equations of the optimum backwashing theory.
Richardson and Zaki's equation, modified by Amirtharajah and Cleasby
[6] for graded and irregular particle systems, is
V = Ken (10)
where
K = f(V ,ty,d/D ) = constant for a particular system
S t
\|l = sphericity
d = particle diameter
D = diameter of tube or bed
V = unhindered subsiding velocity of the particle
35
-------
n = expansion coefficient in Richardson and Zaki's equation
(presented below) .
The coefficient n is a function of the flow regime and the dimensions
of the apparatus but is constant for a particular system. For the
flow regimes of interest under filter backwashing conditions,
n
/4
I
45 + 18 ;
D
for 1 < Re < 200
o
(11)
^ '
where
PV d
Re = = Reynolds number based on unhindered subsiding
^ velocity.
It should be pointed out that Eq. (11) was developed from fluidiza-
tion studies of spherical particles and is not directly applicable to
nonspherical particles as Amirtharajah and Cleasby had presumed [6].
Nevertheless, K and n are constants for a particular filter media.
Substituting Eqs. (9) and (10) in Eq. (8) gives
S =
n
12
-e)J
that is,
where
(p. -
-a (e11-1 - en)
n2
(12)
r i
tt = UsK (PS " P)
= constant for a particular system.
The above equation is the basic equation of the writer's theory. It
is a relation between the shear stress and the porosity in a fluidized
bed.
Let us analyze this function by classical optimization techniques to
determine the stationary points of the function as the porosity e
changes.
36
-------
2
The simplest analysis is to consider the function S :
O O / _ 1 _v
1,
i < 0 when e =
de
Thus the stationary point is a maximum. Alternatively, the following
simpler analysis gives the same result.
37
-------
Consider the sign of dS/de as it passes through the stationary point.
From Eq. (13) ,
dS _ f (n - 1)
3— > 0, for e < s '
de n
dS
de < 0, for e > ~
n
Hence the stationary point is a maximum.
For a typical filter sand with an effective size of 0.45 mm and a
uniformity coefficient of 1.47, n for the top 3 in. of the graded
sand is 3.54 at 25 °C. For a uniform sand of 0.66-mm size, n = 3.2
at the same temperature by Eq. (11).
For the graded sand, maximum shear stress S occurs at
e - (° - 1) = (3.54 - 1) =
*>— ~~ A f. / — I/ • / ^ «
n 3.54
For the uniform sand,
? 7
e.^f-0.69
Thus, a maximum shear stress S occurs in a fluidized bed at the po-
rosity e = (n - l)/n, which corresponds to porosities of 0.69 to 0.72
for real sand systems. This is the main result of the optimum back-
washing theory.
The above result, derived entirely from the theory, indicates that
optimum cleaning of the filter by maximum hydrodynamic shear forces
occurs at the porosity of about 0.70. This theory, in combination
with the literature cited in fluidization, which reviewed several
experimental studies indicating an optimum diffusion at the porosity
0.65 to 0.70, provides an excellent theoretical framework for experi-
mentally studying optimum backwashing. Detailed experimental studies
which provide confirmation of this theory are presented in the next
section of this report.
The above theory is developed from three equations which are valid
for all types of flow regimes in fluidization. Camp and Stein's
equation is valid for viscous as well as turbulent flows since it
only equates the energy, dissipated by shear to the head loss. The
constant head loss equation is valid for all fluidized beds, and
Amirtharajah and Cleasby's equation is a modified form of a power
function of porosity which is valid for all shapes and sizes of par-
ticles. Hence the theory developed is applicable for all particu-
lately fluidized systems in all regimes of flow.
38
-------
As a fitting closure to the above theory, it is necessary to antici-
pate the results that can be derived by applying this theory to back-
washing in practice. The theory predicts an optimum in cleaning at
a porosity of 0.70. Consider a uniform sand bed of depth Jlo with a
fixed bed porosity of 0.43. For the porosity to ,be 0.70 in the ex-
panded state of depth SL,
SL (1 - 0.70) = jeQ (1 - 0.43);
therefore,
Si = 1.9 A .
o
Hence the expansion required is about 90%. This can rarely be a-
chieved in practice. However, for a graded system the particle diam-
eters at the top of the bed are a fraction of the diameters in the
deeper sections. Thus an expansion much smaller than 100% will cause
the porosities to be 0.70 in the top layers. Since these layers are
the ones that remove most of the suspended matter in filtration, it
can be rationally expected that optimum cleaning of the top layers
will produce the best cleaning for the system. Expansions higher
than that producing the porosity of 0.70 in the top layers will tend
to increase the porosities of the layers on top but will simultane-
ously cause the lower layers to reach the optimum porosity of 0.70,
hence we would expect only a negligibly small decrease in the opti-
mum cleaning. This would cause a nearly asymptotic curve of optimum
cleaning to be produced for graded systems.
Experimental Support for the Optimum Theory
Experimental Apparatus
A schematic layout of the experimental apparatus is shown in Fig. 4.
The arrows in Fig. 4 trace the path of water from the tap supply to
the outlet drain for filter F3 during a filtration run. The main
pilot plant consisted of the university tap water supply (hot and
cold) blended in a thermostatically controlled mixing valve A
(Lawler Automatic Controls, Inc., Mt. Vernon, New York). The blended
water passed through a centrifugal pump B, used as an in-line booster.
After being metered in the flowmeter C, the water passed through a
dual outlet; one end of this outlet fed the supply water to the mix-
ing tank D, while the other end provided the backwash water supply.
Each of these outlets was used singly, and the pump B was only used
during high rates of backwash since the normal tap pressure was
sufficient for most uses.
The influent to the filters was mixed in tank D with the chemicals
being added from constant head capillary feeders. The influent water
was pumped from the mixing tank by a centrifugal pump E, which was
driven by a variable speed DC motor. This enabled influent control
39
-------
BACKWASH
TAP WATER SUPPLY DRAIN —
HOT COLD
I -I •!
i i
CHEMICAL
FEEDER
INFLUENT
LINES
BACKWASH
LINE
MAIN
1 INFLUENT
LINE
OVERFLOW
F3
MIXING TANK
F2
*—_
kl
r— SIPHON BACK-
WASH LINE
TYPICAL
PIEZOMETER
CONNECTION
Fl
: T
DRIP
SAMPLER
FLOWMETERS
H,
EFFLUENT
RATE
CONTROLLERS
NEEDLE VALVES
TO DRAIN
Fig. 4. Schematic layout of experimental apparatus.
40
-------
to be achieved. The main influent line trifurcated to the filters
via the filter valve system. The effluent from each of the filters
Fl, F2, and F3 passed through its own flowmeter and then discharged
freely into a float operated effluent rate controller. Eleven pi-
ezometer connections enabled the head losses to be determined at
every 3-in. depth of the filter. The expanded height of the filters
was determined by a scale placed along the side of the filter.
For backwashing the filters, blended tap water was used and metered
in flowmeter C. The backwash line passed via the valve system and
discharged in a drain.
The filters consisted of 6-in. inner diameter, 1/2-in. thick plexi-
glass tubes 4 ft 5 in. deep with a 3-in. high calming section at the
bottom. The water was fed through 59 orifices of 1/16-in. diameter
in a 1-in. thick underdrain plexiglass plate. Sets of orifices were
staggered from one another so as to provide a uniform matrix of ori-
fices on the entire plate. The calming section was filled with
1/2-in. diameter glass marbles. Two pressure gauges, one attached
to the center of the calming section and the other to the cover plate
of the filter, were provided to serve as safety gauges to warn of
dangerous pressures.
The solid particles composing the bed were supported on two stainless
steel meshes (No. 50 over No. 10) placed above the 1-in. plexiglass
plate with the orifices. Sixteen pressure taps were located on two
rows on diametrically opposite sides of the filter to permit obser-
vation.
A mixing tank 3-ft high and 3 ft 6 in. in diameter equipped with a
slow speed paddle was used to provide the necessary detention time
for completion of the iron precipitation reaction. Allowing 6 in.
of freeboard, this 180-gal. tank provided a 30-min theoretical deten-
tion time at a pumping rate of 6 gpm.
A 1.5-in. outlet diameter, self-priming centrifugal pump (Teel Self-
Priming Centrifugal Pump, Model 1P746, Dayton Electric Manufacturing
Co., Chicago, Illinois) was used to pump the influent water from the
mixing tank to the filters. The pump was driven by a 2 by 3.5-in.
pulley drive from a 1/3-HP DC motor equipped with a variable speed
control.(Westinghouse Hi-Torque Speed Control with DC Motor, Westing-
house Electric Corporation, Springfield, Massachusetts). This system
enabled the influent pumping rate to be increased simultaneously on
all three filters by a gradual increase in the speed of the centrif-
ugal pump. Thus influent control was achieved; any changes in flow-
rate affected all three filters to the identical extent.
The total flow to the mixing tank during filtration and the total
backwash rates were metered by a rotameter C which had a range up to
13 gpm.
41
-------
The effluent from each filter passed through a rotameter suitable for
water flow measurement from 0.2 to 2.0 gpm. These flowmeters, Gi,
G£, and 63, were calibrated by determining the time to fill a liter-
measuring cylinder.
The effluent from each filter passed via each flowmeter into a float-
operated, rate-of-flow controller. The controller maintained a con-
stant rate of filtration by holding a constant head on a needle valve
outlet. As the head loss through the filter increased during a run
the float valve gradually opened to maintain a constant filtration
rate. These rate-of-flow controllers functioned remarkably well.
Granular filter sand (Fine sand, from Northern Gravel Co., Muscatine,
Iowa) was used for this study. The original sand had an effective
size of 0.455 mm, a uniformity coefficient of 1.52, a specific grav-
ity of 2.648 and a porosity of 0.412. The graded sand used in the
latter stages of the study was this original sand. The uniform sand
used in the earlier stages of this study was prepared by sieving in a
Gilman set of United States Standard sieves on a mechanical shaker
for 10 min. The uniform sand used was that sand, 100% of which
passed sieve no. 30 and was retained on sieve no. 35. This sand had
an arithmetic mean size of 0.548 mm or a geometric mean size of 0.545
mm on the basis of the adjacent sieve openings. .Special precautions
were taken to insure that the sand was identical in all three filters.
Laboratory Analytical Procedures
The influent suspension was prepared by dripping a stock solution of
ferrous sulphate (FeS04'7H20) from a constant head capillary feeder
to the mixing tank, into which a metered quantity of water (4.5 gpm)
was added continuously. This flow rate was sufficient to apply to
the three filters at 7 gpm/sq ft, and allowed a continuous bleed from
the bottom of the mixing tank as well as continuous overflow. The
stock solution had an iron concentration of approximately 0.2 M in an
acid solution of strength 0.1 N to prevent precipitation in the
bottle.
Table 1 gives a typical analysis of the University tap water used in
the studies. This is a hard well water of high alkalinity and total
dissolved solids and of relatively constant quality. When ferrous
sulphate in acid solution was added to this water a yellowish brown
precipitate formed. A sample of the suspension from the influent to
the filters, after the normal detention in the mixing tank, was fil-
tered through a 0.45-|o,m millipore filter, and the filtrate was ana-
lyzed for any dissolved iron. It was found that within the accuracy
of the equipment used no dissolved iron was detectable. This indi-
cated that the precipitation was complete in the mixing tank. It
should be noted that no aeration or addition of air was required for
the formation of the precipitate. All the runs were made without
the addition of any air except that which occurred at the water sur-
face of the tank which was open to the atmosphere.
42
-------
Table 1. Analysis of University tap water.
Characteristic of water Concentration, mg/1
Total dissolved solids 680
Total hardness as CaC03 365
Calcium hardness as CaC03 254
Magnesium hardness as CaCOj HI
Total alkalinity as CaCOs 270
Calcium as Ca"*"1" 102
Magnesium as Mg** 27
Bicarbonate as HC03~ 330
Chlorides as Cl~_ 17.5
Sulphates as 804"" 160
Fluorides as F~ 0.9
Manganese as Mn 0.0
Iron as Fe"^ 0.03
The actual nature of the precipitate formed by addition of ferrous
sulphate to high alkalinity water has been the subject of considera-
ble controversy during the last decade [30,52,114]. It has been
supposed by various workers that the precipitate could be ferric
hydroxide, ferrous carbonate, or a combined precipitate of both. It
is suggested as a future project that wet chemical analyses be per-
formed on the precipitate to identify its character more definitely.
The main series of experiments consisted of 18 runs, 12 made on uni-
form sand filters and 6 on graded sand filters. The influent and
effluent qualities were evaluated on the basis of iron content.
During these 18 runs nearly 5400 analyses for iron were made; thus
a simple and accurate method of analysis was required.
During the initial series (run 1 to run 6) the standard method in the
water supply field, namely the 1, 10-phenanthroline method was chosen
[117]. The procedure was considerably simplified by using the pat-
ented single-powder formulation (Hach Chemical Company, Ames, Iowa -
FerroVer) which dissolves and reduces the iron without any heating.
Since very few interfering ions were present the results were un-
affected by such interferences. The developed color was observed at
510 nm on a Beckman Model B spectrophotometer and the iron content
read on a calibration curve.
The above procedure for measuring the effluent iron had the following
weaknesses: (1) The procedure for measuring the quantity of reagent
using a scoop had possible errors from analysis to analysis; (2) the
transmittances of the final effluent at the 12-in. depth was too
high; (3) the molar absorptivity of phenanthroline limited the possi-
ble accuracy of small changes in iron concentration.
43
-------
In order to alleviate these weaknesses the following modified proce-
dure was used in all the runs after run 6, and it improved the accu-
racy of the analyses considerably. A new patented reagent (Hach
Chemical Company, Ames, Iowa), disodium salt of 3-(2-pyridyl)-5,6-
bis(4-phenyl sulfonic acid)-l,2,4 triazine, hereafter called Ferro-
Zine, having a molar absorptivity of 27,900 was used. This compound
reacts with divalent iron to form a stable magenta complex species
which is very soluble in water and may be used for the spectrophoto-
metric determination of iron. The absorption spectrum of the complex
has a sharp peak at a wavelength of 560 nm and is uniform in develop-
ment over the pH range from 4 to 10. Further details of interference
studies and statistical data on multiple laboratory studies of Ferro-
Zine can be seen in Stookey [120].
The above reagent as a single-solution formulation, FerroZine Solu-
tion 1 (Hach Chemical Company, Ames, Iowa), can be added directly to
iron hydroxides or carbonates for reduction and dissolution. The
procedure for analysis was to add 0.5 ml of FerroZine Solution 1 to
25 ml of water and read the transmittance on a spectrophotometer
after allowing 5 min for development of color.
In order to obtain increased accuracy of the transmittance readings on
the spectrophotometer, a dual standard procedure, based on the method
of ultimate precision as described in Ewing [44] was used. This
precision calibration method nullified all the stated weaknesses of
the FerroVer method considerably and increased the accuracy of the
effluent quality analyses in the second and third series of runs.
During the runs, the influent suspension was sampled every half hour.
These samples had iron concentrations of 7 mg/1, and hence they had
to be diluted for obtaining readings within the range of the calibra-
tion curves.
All the samples of influent and effluent were collected every half
hour. Also, the initial effluent quality was analyzed at frequencies
of nearly a minute during the first 10 min of a run to study the
initial degradation and improvement of effluent quality. These sam-
ples were also analyzed by the FerroZine test.
While backwashing the filters after a dirtying run, samples of the
backwash water were collected periodically. These samples had con-
siderable amounts of suspended iron floe, some as high as 800 mg/1.
These samples were also analyzed using the phenanthroline procedure,
the samples being diluted to obtain reasonable readings on the spec-
trophotometer.
During the course of the research project it was realized that addi-
tional evidence of the effectiveness of backwash could be obtained by
analyzing the amount of iron left as a coating on the sand after the
wash. Not only would this provide comparative evidence for studying
44
-------
various expansions, but it would also prove beyond any shadow of a
doubt that abrasion in the fluidized bed was negligible. This of
course, was already anticipated from theory and other experimental
results quoted in the earlier chapters.
It was decided to evaluate two methods of washing the sand: (1) a
physical wash and (2) a chemical acid wash. The physical wash pro-
cedure was selected as outlined below.
The filter was fluidized, and a long-handled scoop was used
to remove a sample of sand from the top layers of the fluid-
ized bed. Each sample of sand weighed approximately 25 g.
This sand was washed from the scoop into a 250-ml beaker
using exactly 100 ml of distilled water. The beaker had al-
ready been weighed, empty and dry. It was now weighed con-
taining the sand, the water drawn by the scoop and the dis-
tilled water. A 1.25-in. magnet was placed in the beaker
with the sand and water, and the sand was washed by the
magnetic stirrer for 10 min at a fixed speed. It was found
that the distilled water turned quite dark and cloudy and
that considerable amounts of iron had been removed from the
sand by abrasion between sand particles as well as by abra-
sion with the magnet. The supernatant iron suspension was
stirred by the tip of a pipette and 25 ml were withdrawn and
delivered into a 500-ml volumetric flask. Distilled water
was added to make up to 500 ml and this diluted solution
(1:20) was analyzed for iron concentration using the Ferro-
Zine standard calibration technique.
The beakers containing the sand and the iron suspension
(approximately 75 ml) were placed in an oven. The water
in the beakers was evaporated, and the beakers containing
the dry sand vere cooled to room temperature and weighed
again.
From the above readings the amount of iron removed from the sand in
mg/g can be determined from the following formula:
Iron removed (mg/g) - [Cone, of iron in diluted solution (mg/1)]
[Dilution Factor] x [Weight of water (g)l
x 1000 x [Weight of sand (g)] (15)
Note that the weight of water in grams is assumed to represent the
volume of water in milliliters, and it is the total water in the
beaker including whatever water is drawn with the sand from the flu-
idized bed. This formula gives the iron removed from the sand quite
accurately.
The above was the procedure used in all the analyses run on samples
of the graded sand during the third series, runs 20 to 25. At the
45
-------
end of each sand analysis, and prior to the beginning of the subse-
quent run, the withdrawn samples of sand were returned to their re-
spective filters. Thus, the sand in all the runs was identical and
no sand losses were allowed to develop.
Chronology and General Descriptions of Filter Runs
Twenty-five filter runs, of which 18 were dual runs, were made. The
dual runs consisted of a dirtying run designated A, the backwash of
A, and a filtration run designated B. In the following, a reference
to a run means the complete dual run. Runs A and B are designated
as such. The first 12 runs (series 1: 1A, IB to 6A, 6B; series 2;
7A, 7B to 12A, 12B) were made on filters with 12-in. depth uniform
sands of 0.548-mm mean size. The runs in series 1 and 2 were made
at 7 gpm/sq ft, and each run (A or B) lasted for approximately 5 hr.
Runs A or B had to be terminated due to the fact that the head losses
developed were nearly 8 to 9 ft, and this was the maximum differen-
tial height that could be measured on the piezometer boards.
The reagent used for iron analysis in series 1 was FerroVer, In the
first series of runs, the dirtying runs A were made on the first day,
and the backwash and the filtration runs B were made on the follow-
ing day. Since it was possible that some physical and chemical
changes could have occurred in the solids removed, due to overnight
standing, and also because in actual treatment plants backwashing is
performed soon after a filter is removed from service, the two runs
A and B, and the backwash of runs A in series 2 were made on the same
day. Each dual run including backwash required about 15 hr. and two
experimenters were required to work continuously to take the readings
and make the analyses of iron. In run 1 the samples of water at
depths other than the full depth of 12 in. were collected by a single
experimenter and kept for about two to three days before all the
analyses were completed. It was found that iron in suspension tended
to be deposited on the sides of the plastic sampling bottles and that
the effluent quality at 9-in. depth of the filter measured on a later
date was better than the effluent quality measured at the 12-in.
depth on the day the run was made. These results were invalid and
were not used in the analyses. All iron analyses from run 2 onward
were made within a couple of hours after sampling, and rational
readings were obtained. This could only be accomplished because two
experimenters worked. The reagent used for series 2 and 3 was Ferro-
Zine.
The last six dual runs (run 20A, 20B through run 25A, 25B) were made
on 18-in. depth graded sands with an effective size of 0.45 mm and a
uniformity coefficient of 1.47. This set of runs was called series 3.
These runs were made using identical procedures to those of series 2.
However, in addition to ferrous sulphate, a nonionic polyelectrolyte
(Dow Chemical Company, Midland, Michigan - Separan) was also added to
the influent suspension to obtain a concentration of 0.10 mg/1 of
46
-------
polyelectrolyte in the feed to the filters. The polyelectrolyte was
added in the hope of magnifying the differences of backwashing at
different expansions.
The parameters used to study the effectiveness of backwash in all
three series were (1) initial effluent quality, (2) head loss in-
crease, (3) cumulative effluent quality in the run following back-
wash, (4) the backwash water quality during backwash, and (5) the
backwash water volume.
For series 3 the additional parameter based on the iron which is phys-
ically removable from samples of the backwashed sand was also used.
The variation of each of these parameters with porosity was studied
at six different porosities, 0.55, 0.60, 0.65, 0.70, 0.75, and 0.78,
on each of the filters. The expansions needed to obtain these poros-
ities for the uniform sand were approximately 33, 50, 70, 100, 140,
and 190% respectively. Since each run was made on a bank of three
filters and each series consisted of six runs, the effective number
of points for each series was 18. The experiment was designed such
that each filter was studied at the six different porosities during a
series. Thus by considering all 18 readings of a series of six runs,
the small variations between filters and the small variations from
run to run were averaged out. Table 2 illustrates the format of the
experimental design for series 2, and how the backwash at the differ-
ent expansions was studied for the three filters Fl, F2, and F3. A
similar format was used for series 1.
Table 2. Experimental design for series 2.
Porosity during backwash
0.55 0.60
K.un
number
7
8
9
10
11
12
F2
Fl
F3
Fl
F2
F3
Filter
Fl
F3
F2
number
F2
F3
Fl
F3
Fl
F2
F3
F2
Fl
For the series 3, the format of runs 7 and 8, as shown in Table 2 for
series 2, was repeated three times. This procedure was adopted to
enable a particular run to be studied with the backwash expansions of
30, 50, and 75% or 15, 40, and 60%. This format avoided two adjacent
expansions such as 40 and 50% being studied in a single run, and the
differences for purposes of comparison were enhanced. This was
thought advantageous because of the anticipated smaller differences
in effectiveness of backwash at different porosities for graded
47
-------
systems. The cause for the anticipated smaller differences in back-
wash for graded systems has been discussed previously.
The expansions needed to obtain porosities of 0.58 to 0.82 in the top
layers of the graded sand ranged from 15 to 75%, and these were much
less than the expansions required for the uniform sand. All back-
washing expansions were controlled on the basis of the expanded
heights of the fluidized bed during backwash.
In order to reproduce identical conditions with a clean filter at
the beginning of each dual run, the following standardized procedure
was used for all runs. The filter was expanded to the anticipated
optimum at a porosity of 0.70 and washed for 15 min. During the runs
1 to 10 the filters were washed twice before each run began, once at
the completion of the previous run and again immediately preceding
the start of a run. However, from run 11 onwards, in order to achieve
identical conditions to those in a treatment plant, the solids re-
moved in run B of the preceding run were not washed until the follow-
ing run A. Immediately before the dirtying run A, the filter was ex-
panded to the anticipated optimum at a porosity of 0.70 and washed
for 15 min.
In the case of series 1 and 2 the time of backwash was the same for
all expansions; for the graded sand studies of series 3, however, the
operation was modified slightly. In series 3, the backwash at dif-
ferent expansions was done for different times, so that the same
volume of washwater, approximately 36 gal., was used during the total
sequence of washing (i.e., valve opening, washing, and valve closing).
Also, while the bed was fluidized, one experimenter collected two
samples of the backwashed sand at different times. The samples were
collected after approximately 10 and 21 gal. of washwater had been
used for backwash. A resume of the backwash sequences and sand col-
lection times for the graded sand at different expansions is pre-
sented in Table 3.
Table 3. Backwash procedures for graded sand.
Wash sequence
Expansion,
%
15
30
40
50
60
75
Porosity in
top 3 in.
0.58
0.67
0.70
0.74
0.77
0.82
Valve
opening,
min
1.0
1.0
1.0
1.0
1.0
1.0
Wash
time,
min
10.5
•6.5
5.0
4.0
3.25
2.75
Valve
closure,
min
2.0
2.0
2.0
2.0
2.0
2.0
Sand collection times
Sample 1,
min
4.5
3.0
2.75
2.5
2.0
2.0
Sample 2,
min
9.0
6.0
5.0
4.5
3.75
3.5
48
-------
The observations made during the course of a run can be grouped into
two categories, (1) data for analysis and (2) data for quality con-
trol. For series 1 and 2 the following readings were taken as data
for analysis.
During runs A:
a. The initial effluent quality for each of the three filters at
intervals of 1 min each for the first 10 rain of a run.
b. Piezometer readings at depths of 0, 3, 6, and 9 in. for each of
the three filters at frequencies of one-half hour.
c. The effluent quality at the total depth of 12 in. for each of
the three filters at every half hour. The samples were collected
at the outlets flowing into the effluent rate controller cham-
bers.
d. The effluent quality at intermediate depths of 3, 6, and 9 in.
for each of the three filters at intervals of every hour begin-
ning with the first sampling at 0.5 hr after the run began.
These samples were collected from the continuous drip samplers.
During backwash:
e. The heights of the expanded bed.
f. The backwash flow rate.
g. The piezometer readings at every 3-in. depth of the expanded bed.
h. The temperature of the washwater.
i. The backwash water quality from each filter at times of 0.5,
1.0, 2.0, 3.0, 4.0 and 5.0 min during the washing time of 5 min.
During runs B:
Similar readings to those taken during runs A, and indicated
above by a, b, c, and d were taken.
For purposes of maintaining identical conditions from run to run, the
following data were taken for purposes of quality control, during
runs A and B.
j. Influent iron concentration for one of three filters at fre-
quencies of one-half hour.
k. The flow rate through the three filters was monitored and ad-
justed if necessary every half hour. Adjustments were only re-
quired during the latter halves of runs A or B.
49
-------
1. The room temperature and the water temperature were monitored
every half hour. Any small changes of water temperature were
adjusted.
A similar set of readings were taken during series 3 for the graded
sand, subject to the following modifications.
a. In order to obtain the peak of the initial effluent quality
curve, the samples of water were collected at 0.5, 1.0, 1.5,
2.0, 3.0, 4.0, 5.0, 6.0, 8.0, and 10.0 min, respectively.
b. Since the depth of sand was 18 in., piezometer readings were
taken at 0, 3, 6, 9, 12, and 15 in.
c. The effluent quality was measured at the total depth of 18 in.
d. The effluent quality at intermediate depths was measured at 3,
6, and 12 in.
The sampling for backwash water quality from each filter was variable
depending on the duration of the washing sequences as shown in Table
3. However, the sample collection times were preplanned so that
seven samples were collected from each filter at times corresponding
to usage of equal volumes of washwater.
In order to evaluate the degree of segregation and the distribution
of particle sizes in layers of the fluidized bed, the following ex-
periment was made on the graded sand. The graded sand bed was fluid-
ized to 507o expansion and allowed to stabilize for nearly one half
hour and then approximately 3-in. layers of the fluidized bed were
siphoned off. These sections of sand were then oven dried and cooled.
The total dry volumes were measured, and then representative samples
from each layer for purposes of sieving were obtained by repeatedly
reducing the total volume of each layer in a two-way splitting sampler
to about 500 g. Using the balance sand from each layer after select-
ing the 500 g, bed porosities of each layer were measured by proce-
dures described elsewhere [4]. From the measured volumes and the ex-
panded heights of the same layers, the porosity of each layer when
the total bed expansion was 5070 was also calculated.
For purposes of quality control and to check the general form of the
curves, the following curves were plotted for most of the runs:
(1) head loss against time and (2) ratio of effluent to influent con-
centration against time (i.e., C/CO vs t). Some typical results are
shown in Figs. 5 and 6. From the head loss curves it can be seen
that the behavior is reasonably linear for the uniform sand. The
ratio of effluent to influent curves indicate that even after back-
wash at various expansions the variation from filter to filter is
very small. Thus, for using the effluent quality as a parameter to
indicate the effectiveness of wash needs a cumulative effluent quality
curve as described by Johnson and Cleasby [67].
50
-------
8
2 4
RUN 3B, TOP 6 in
Fl O
F2 D
F3 A
TIME,hrs
Fig. 5. Head loss curves for run 3B, top 6 in. of filter media.
Results
The results in this section indicate the variation of the following
parameters with the porosity provided during the controlled backwash
following the dirtying run: (1) effluent quality at various depths,
(2) head loss increases, (3) backwash water quality, (4) backwash
water volumes, and (5) sand wash analysis.
Cumulative effluent quality and porosity. Figure 7 shows typical
plots of the cumulative differential iron of the effluent with time
(run 8, series 2). The ordinate is the accumulated difference be-
tween the iron in observation run B and dirtying run A for the
effluent from the full depth of the filter. If the ordinate is nega-
tive, the cumulative iron in run A was greater than that in run B, or
mathematically Z Iron (B-A) < 0. In each run all three filters
time
were dirtied under identical conditions, as explained in the previous
chapter. The filters were then backwashed at different expansions,
and the quality was studied in the following run. Study of these
figures towards the end of the run will indicate which filter
51
-------
o
IU
tu
0.06
0.04
RUN 10B
F10
F2A
F3D
S o.oo
o
So 1 2 3
TIME, hrs
Fig. 6. Variation of the ratio of effluent to influent iron with time.
performed the best in a particular run. The filter with the smallest
positive, or the most negative iron produced the filtrate of the best
quality in the filtration run compared to the dirtying run.
An index called the effluent quality index was defined to compara-
tively grade the filters in each run. In each run the filter pro-
ducing the best cumulative differential effluent quality was given an
index value of 3, the next best quality was given an index value 2,
and the worst quality was given an index 1. If two filters performed
almost identically in a run then the two effluent quality index
grades were divided between the two corresponding filters.
The values of the index for all the runs from 7 to 12 and the corre-
sponding porosities in each case are shown in Table 4.
Originally, it was thought that the differential cumulative iron could
be summed from run to run to give the basis for comparison. In prac-
tice it was found that even with the best possible experimental con-
trols it was impossible to reproduce identical dirtying runs in the
series, due to the inevitable fluctuations in flow rate and iron con-
centrations. However, within a single run, such changes occurred on
all three filters simultaneously, and comparisons within filters
52
-------
o) -62.5
*»
Z
Z -125
LU
ID
u_
u_
LU
S-187.5
LU
U.
S
A
1A\
\l\
- \
\\
\v
- \V
ti \X.
\v -
NN 0^
RUN 8 POROSITY
0*5*5 P9 /\ — —
0.75 F3 D
^oJl^^±^-^---^---A,
Urf
^ -250 —
U
-312.5
Fig. 7. Cumulative differential effluent iron vs time, run 8.
during a run were the best possible means of study. Using the above
artificial effluent quality index and repeating the runs so that each
filter was subjected to all the expanded porosities during backwash,
differences due to variations in the sands of the filters were aver-
aged and hence eliminated. Thus the index provides the best means of
comparison of backwashing effectiveness between different expansions.
The variation of the cumulative effluent quality index (i.e., summing
the index for all runs) with porosity for runs 7 to 12 shown in
Table 4 is plotted graphically in Fig. 8. The results show a maximum
in the cumulative effluent quality index at a porosity of 0.65 to
0.70. The maximum indicates that the best effluent quality in the
run following backwash is produced by backwashing the dirty filter at
the expanded porosity of 0.65 to 0.70.
The same type of analysis was done for the effluent qualities mea-
sured at depths of 3, 6, and 9 in. during series 2. The variations
of the cumulative effluent quality index with porosity for all the
depths of 3, 6, 9, and 12 in. are shown in Fig. 9.
53
-------
Table 4. Effluent quality index and porosity for series 2.
Rim
XVU1J
number
7
8
9
10
11
12
Expanded porosity
0.55 0.60 0.65 0.70 0.75 0.78
Backwash ing index
2 3* 1
1 2 3
1 3 2
2 1 3
3 2 l
2 3 i
Cumulative
effluent
quality
index
The best performance in effluent quality was given in index 3, the
next best 2, and the worst 1.
The results for runs 1 to 6 for the 12-in. depth are presented graph-
ically in Fig. 10. The results at other depths were not analyzed due
to the fact that the readings of effluent iron in run 1 were invalid
due to storage of the samples for too long a period before analysis,
as already mentioned under experimental observations. Since a com-
plete set of readings is required for an unbiased analysis at all
porosities, the results for series 1 could not be analyzed for the
intermediate depths.
The results of the variation of the cumulative effluent quality index
with expansion for the graded sand for the full depth of 18 in. and
for all depths of 3, 6, 12, and 18 in. are shown in Figs. 11 and 12.
Also marked on figures are the experimentally determined porosities
of the top 3 in. of the graded sand bed while in the fluidized state.
All the results of series 1, 2, and 3 shown in Figs. 8 to 12 indicate
quite clearly that in every case the best effluent quality in the run
following backwash is obtained by expanding the bed to porosities of
0.65 to 0.70 during backwashing.
Initial effluent quality and porosity. A technique similar to above
was used to study the variation of initial effluent quality with po-
rosity during the preceding backwash. The cumulative initial efflu-
ent index was given a value of 3, 2, or 1, depending on the cumula-
tive differential iron between runs B and A during the first 10 min
of a run. The method was identical to that used to evaluate the
54
-------
8
t
g
o
»-
LU
u
RUNS 7 THROUGH 12
UNIFORM SAND
12-in. DEPTH
I
I
I
I
0.55
0.60 0.65 0.70 0.75
AVERAGE POROSITY DURING BACKWASH
0.80
Fig. 8. Cumulative effluent quality index vs porosity, series 2, 12-in.
depth.
effluent quality as already described. The results indicate no re-
lationship between the initial effluent quality in the run following
backwash and the porosity of the expanded bed during backwashing.
It was felt that the initial effluent quality was not dependent on
the backwash expansion but was a function of the rate of closure of
the backwash valve.
Head loss increases and porosity. A study was also made on the
effect of backwash on the head losses in the run following backwash.
Again comparison was made between filters based on the difference of
head loss in run B over that of the dirtying run A. The results were
not conclusive and are not presented here.
Backwash water quality and porosity. The parameter which provided
data that was the most consistent in all the runs was the backwash
water quality. It enabled comparisons between different backwash
porosities to be made on the basis of usage of equal quantities of
washwater, even though the actual washwater used in the series 1 and
2 was dependent on the constant wash duration of 5 min. For series 3
55
-------
28
26
24
>;
<
O
UJ
Z>
22
20
D
U
SERIES 2
RUNS 7 THROUGH 12
UNIFORM SAND
ALL DEPTHS
SERIES 1
RUNS!THROUGH 6
UNIFORM SAND
12-in. DEPTH
0.55 0.60 0.65 0.70 0.75
AVERAGE POROSITY DURING BACKWASH
0.80
Fig. 9. Cumulative effluent quality index vs porosity, series 2,
all depths, (above)
Fig. 10. Cumulative effluent quality index vs porosity, series 1,
12-in. depth, (below)
56
-------
2
2
8
I
C
UJ
2 4
RUNS 20 THROUGH 25
GRADED SAND
18-in. DEPTH
I
20 30 40 50
PERCENT EXPANSION
J I I I
60
70
80
0.58
0.67 0.70 0.74 0.77
POROSITY OF TOP 3 in.
0.80 0.83
Fig. 11. Cumulative effluent quality index vs expansion, series 3,
18-in. depth.
the total volume of washwater used for the different expansions was
maintained the same for all the runs by varying the durations of
wash.
Figures 13 and 14 illustrate the backwash water quality for series 1
in terms of the iron concentration in mg/1 in samples of washwater as
a function of the total volume of washwater used up to the time of
sample collection. Using the time of collection of samples and the
flow rate during that particular wash the total washwater used was
calculated and plotted as the abscissae. The plotted points are from
different filters and different runs but are grouped together to in-
dicate the variation of backwash water quality with porosity. The
apparent scatter in the points towards the end of the backwash is due
to the graphs being plotted on logarithmic coordinates. The loga-
rithmic coordinates were necessary to show the variations in backwash
water quality which range from 1000 to 0.2 mg/1. However, for pur-
poses of analysis the most relevant sections of these primary curves
shown in Figs. 13 and 14 are the lower curved portions before the
curves reach asymptotic values. Magnified curves of these sections
57
-------
31
27
O
t—
LU
23
r 19
<
U
15
RUNS 20 THRO UGH 25
GRADED SAND
ALL DEPTHS
1
1
20
30 40 50
PERCENT EXPANSION
1
1
60
1
70
80
0.58
0.67 0.70 0.74 0.77
POROSITY OF TOP 3 in.
0.80 0.83
Fig. 12.
Cumulative effluent quality index vs expansion, series 3,
all depths.
for series 1 are shown in Fig. 15 in arithmetic coordinates. The
lines drawn are smoothed curves through the means of the values from
the three filters. The curves represent the mean variations of back-
wash water quality with volume of washwater at the different poros-
ities.
The smoothed curves of Fig. 15 were used to prepare secondary curves
showing the variations of final backwash water quality with porosity
for constant volumes of total washwater (Fig. 16). The points
plotted are the intersections of ordinates at washwater volumes of 20
and 25 gal., respectively, with the smoothed curves drawn in Fig. 15.
58
-------
100
60
o 30
ce.
10
§ 6.0
UJ
3.0
<
0.6
0.3
0.
POROSITY RUN
0.55
0.65
3 Fl ®
2 F2 •
4 F30
2 Fl A
5 F2A
6 F3 A
0.75 4 Fl B
6 F2 •
2 F3 D
1
10 20 30
WASHWATER VOLUME, gal.
40
Fig. 13. Backwash water quality vs washwater volume, series 1.
59
-------
200
TOO
60
z
o
>: io
g 6.0
LLJ
S
X
3.0
0.6
0.3
0.1
POROSITY RUN
0.60
1
4
5
0.70
0.78
1 F2
3 F3
5 Fl
- 3 F2
1 F3
OMIT
F2
F3
6 Fl OMIT
A
S
I
I
10 20 30
WASHWATER VOLUME, gal.
60
Fig. 14. Backwash water quality vs washwater volume, series 1,
60
-------
3
Z 2
9
^
> i
t
_i
<
z>
Q£
LLJ
1 °
i
<^^ O
3
g
0
CD
<
2 2
^^
S
i—
1
0
1 \ I
1 \ U POROSITY RUN
l \ r^
— \ \ i \ 0.60 1
1 i \ 1
\ \ \ »
\ \ \ 0.70 6
% * I
- \ \ \ — 1
\ \ \ 0.78 5
\ \ \ — — — 3
x^ \ \ '
S^ ^N A
B ^^-\
•^ "^m
**• ^NSJ
* ^33^.^^
i i ^""^
i ' \
1 » \
i i \
> 1 \ 0.55 3
— i \
\ » \ 4
^ \ \ 0.65 2
\V\ —2
\ \ 0.75 4
\ \ 6
\\
\ \A
\\
V^
^^i,
^•^-^r
i i i"
10 20 30
F1 OMIT
F2 •
F3 0
Fl OMIT
F2 A
F3 A
Fl B
F2 •
F3 D
f51
' 1
Fl ®
F2 •
F3 0
Fl A
F2 A
F3 A
FT B
F2 •
F3 D
1
40
TOTAL WASHWATER, gal.
Fig. 15. Backwash water quality vs washwater volume, series 1.
61
-------
2.4
Z
S£ 2.0
1.6
O
O£
LU
I
I
<
1.2
0.8
0.4
0.0
I
I
TOTAL WASHWATER USAGE
• 25 gal.
A 20 gal.
I
0.55 0.60 0.65 0.70 0.75
AVERAGE POROSITY DURING BACKWASH
Fig. 16. Terminal backwash water quality vs porosity, series 1,
62
-------
The results show once again that the best terminal backwash water
quality is achieved by backwash at the porosity of 0.70. This has
resulted from analyses which consider backwashing to different ex-
pansions, but using a constant total volume of washwater. The re-
sults indicate that most effective backwash is achieved by expansion
to porosities around 0.70.
An alternative graph, also derived from Fig. 15 by considering the
different volumes of washwater needed at the different expanded po-
rosities to achieve a given terminal backwash water quality, is shown
in Fig. 17. A family of curves for terminal backwash water qualities
of 0.75, 1.00, and 1.50 mg/1 of iron is shown. These graphs show
again that the minimum quantity of washwater necessary to obtain a
given terminal backwash water quality occurs at the porosity of 0.70.
The above analyses should be restricted to the lower sections of the
curves when the quality changes become small, since only in these
sections are the results meaningful. Identically, similar graphs
resulted in all the experiments of series 2 for the uniform sand and
of series 3 for the graded sand. In every graph a minimum in the
total washwater volume usage or the terminal washwater quality oc-
curred around the anticipated porosity range of 0.65 to 0.70. Thus
both of these parameters have provided still further evidence of the
optimum theory developed in a previous chapter.
Physical sandwash and porosity. As already recorded, an extra param-
eter to evaluate the effectiveness of backwash was proposed based on
the amount of iron removable from the sand by a physical wash. The
washing procedure was simple abrasion using a magnetic stirrer under
standard conditions. Cpnsiderable amounts of iron were removable
from the sand by this method, providing final evidence for the fact
that negligible collisions and abrasions between particles occur in a
fluidized bed. If there were considerable abrasion in the fluidized
state, it should not be possible to remove these large amounts of
iron by a physical wash.
The iron removable from the graded sand in mg/g as a function of ex-
pansion is shown in Fig. 18. The points plotted are for the first
two runs on the graded sand - runs 20 and 21. These runs were the
initial runs made on the new graded sand after it had been subjected
to one unnumbered run for purposes of coating the new sand with at
least a small layer of the iron floe. Though similar measurements
were made for all the runs of series 3, it was found that the results
of runs 22 to 25 were subject to considerable error due to the fol-
lowing cause. In series 3, the influent suspension contained 7 mg/1
iron and 0.10 mg/1 of a nonionic polyelectrolyte. As series 3 pro-
gressed from run to run, mudballs started building up due to the
added polyelectrolyte. These were about 0.5 to 2.0 mm in size and
consisted entirely of globules of the precipitated iron without any
sand within them. They floated on top of the sand layer during flu-
idization, and every time the sample of sand was drawn for analysis
63
-------
o
o>
o
85
I
to
I
26
24
22
20
10
16
I
TERMINAL WASHWATER QUALITY
0.75 mg/l IRON •
1.00 mg/l IRON A
1.50 mg/l IRON •
I
0.55 0.60 0.65 0.70 0.75
AVERAGE POROSITY DURING BACKWASH
Fig. 17. Backwash water volume vs porosity, series 1.
64
-------
a
§
oe
o
z
o
c*
0.10
0.09
0.08
0.06
0.05
0.04
0.58
• RUN 21
O RUN 20
10 20 30 40 50 60
PERCENT EXPANSION
I I I I I I
0.67 0.70 0.74 0.77
POROSITY OF TOP 3 in.
70
0.80
Fig. 18. Iron removable by physical abrasion test vs expansion, runs
20 and 21.
65
-------
considerable amounts of these mud balls or globules were drawn with
the sand sample and caused the iron removable readings to be erratic
from run to run. We found that as the series of runs progressed the
mud balls became bigger and bigger, causing larger and larger errors.
Even though a modified larger expansion wash, nearly 100%, was used
for a few minutes during the standard cleaning at the start of each
run, removing substantial amounts of the mud balls was still not
possible.
For the above reason the physical abrasion test results of runs 22
to 25 were invalid and are not presented. However, in runs 20 and 21
the mud balls were still few in number and small in size and did not
substantially affect the readings.
The results shown in Fig. 18 reconfirm the fact indicated in the
other sections that an expansion of about 40 to 50% produces the
cleanest sand in the graded filter. These expansions cause porosi-
ties of about 0.70 to 0.74 in the top 3-in. layer of the expanded bed
of graded sand.
Analysis. The results presented in the preceding pages prove beyond
little doubt that an optimum backwash occurs in a system expanded to
achieve a porosity of approximately 0.65 to 0.70 in the layers con-
taining the most amounts of suspended matter. In the case of a uni-
form sand bed an expansion to a porosity of 0.70 is equivalent to
nearly 90% expansion of its height. For a theoretical uniform sand
consisting of identical particle sizes this expansion is an exact
and fixed condition. However, in filtration practice uniform sand
beds are never used, nor is it feasible to provide for 100% expan-
sions because of the unreasonably large backwash flows required. We
are fortunate that both these limitations are simultaneously removed
by use of a graded sand. A 40% expansion of a typical graded filter
sand can be shown to cause the porosity of the top 3-in. layer to
reach a value of 0.70, and use of three different parameters has
shown that the optimum backwash for the graded system occurs at
expansions of about 40 to 45%. These values of expansion can be ob-
tained in practice.
A few points need to be made regarding this optimum of 40 to 50% for
graded systems and the results reported in the sanitary engineering
literature. The fact that effective backwash requires an expansion
of about 50% for graded systems has been a well known rule of thumb
in filtration design and practice. This rule developed originally
from the work of Hulbert and Herring [62] in 1929.
Several workers have suggested during the last decade that expansions
of 20 to 25% may be sufficient for effective backwash [10,26,67,132].
None of these papers, however, provide fundamental considerations or
experimental results which are valid to draw this conclusion. Baylis
[10] suggested the figure without any experimental work. Camp et al.
[26] reported this expansion as suitable for all filters on the basis
66
-------
that serious problems did not occur in the operation of the Billerica
water treatment plant. But note that the Billerica plant had multi-
media filters, and expansions of 20 to 25% easily give porosities of
about 0.70 in the top coal layer. This can be seen in the results
reported by these workers themselves. This, in fact, is additional
evidence for the hypothesis of this report.
Thus, careful analysis indicates that the results reported in the
literature [26,62] are consistent with the theory and experimental
work of this report. Also, remember that effective backwashing does
not necessarily mean optimum backwashing, and due to the rather flat
nature of the shear stress maximum it is possible to backwash filters
effectively even though the optimum condition is not obtained. In
case one may be tempted to run away with the idea that a lower ex-
pansion may result in a saving of washwater, Fig. 17 needs to be re-
membered. It clearly shows that backwashing at lower expansions than
the optimum necessarily results in the usage of larger amounts of
washwater to achieve a given bed cleanliness.
A concluding summary. The results summarized and the analyses pre-
sented in this section give a complete picture of optimum backwashing
by water fluidization alone. The experimental results are entirely
consistent with the theory of optimum backwash developed in a pre-
vious section and provide excellent confirmation of the theoretical
results. Optimum backwash has been shown to simultaneously provide
these advantages: (1) a better effluent in the following run, (2) a
minimum usage of washwater, and (3) a minimum growth of the coatings
on the sand. This plurality of advantages should considerably improve
the performance of most filtration plants, if optimum backwashing
is put into operation.
67
-------
VII. WASTEWATER FILTRATION AND BACKWASHING - LITERATURE REVIEW
Filtration has been used in the United States as a liquid and solids
separation process for wastewater since 1883, but was not widely im-
plemented over the years because settling and biological treatment
processes were considered adequate for the needs of the times. How-
ever, with stringent federal and state effluent standards presently
reflecting the demands of the citizenry for more complete treatment
of wastes, filtration is becoming increasingly popular. Similar re-
quirements for better waste treatment were necessary in Britain nearly
25 years ago, so the majority of what has been accomplished in waste-
water filtration and backwashing progress since 1949 has been the
result of British research, development, and experience.
Experience with filters used in water treatment has demonstrated the
need for effective media cleaning techniques. A similar need exists
with respect to wastewater filtration, but the problem is greatly
compounded because of the variable characteristics of sewage. Often
sticky and gelatinous, the solids removed by the media are much more
resistant to cleaning procedures than those normally encountered in
water treatment.
The following literature review summarizing wastewater filtration
and backwashing experience is a summary of a more comprehensive re-
view prepared by Rice [99].
A diverse array of designs is available for wastewater filtration,
varying in flow configuration, bed depth, media type and gradation,
and performance. Cleaning methods, however, generally rely on the
application of water or air and water to remove entrapped solids from
the bed (chemicals such as chlorine occasionally are used as cleaning
aids). The purpose of this section of the review is to review in a
general way the wastewater filter designs and cleaning techniques.
Types of Wastewater Filters and Cleaning Techniques
Conventional Rapid Sand Filtration
Because of their use for years in filtering water for potable use, it
is not surprising that rapid sand filters were among the first to be
used in wastewater filtration. Bed depths of 6 to 36 in. have been
reported [13,41,87,121,136] for full-scale applications, but current
practice favors depths of 24 in. or greater. The gradation of sand
used in the installations varies widely with location, but research
and operating experience have demonstrated that the relatively fine
sand, approximately 0.5 mm, used in water treatment is unsuitable for
wastewater filtration because of rapid head loss buildup and conse-
quently shorter filter runs. Considerable pilot scale research has
shown that 1 to 2-mm media size will produce good effluent quality
and allow reasonable filter runs [59,68,129].
68
-------
Design of these facilities closely parallels that employed for years
in waterworks, the filters placed in rectangular concrete boxes
equipped with perforated block, nozzle type, or header and lateral
underdrain systems [41,87,136].
Backwash water is usually drawn from a filtered wastewater storage
tank although unfiltered water is occasionally used. In either case
the washwater is pumped through the bed in an upward direction, carry-
ing the accumulated solids upward to washwater collection troughs.
Washwater is ordinarily applied at a rate sufficient to expand the
sand bed 10 to 20% during fluidization, usually at rates from 15 to
25 gpm/sq ft.
In Britain, air is almost universally used in wastewater filtration
plants as a media scouring aid prior to or in the course of intro-
ducing the washwater. The air agitates the media and helps to break
up agglomerations within the bed, thus allowing the water backwash to
more easily remove the entrapped solids. Although installed less
frequently, rotary surface washers have been used for the similar
function [87,88]. No reports of backwashing studies testing the
effectiveness of either of these two scouring techniques were dis-
covered, and since wastewater filtration plants in the United States
are rare, little direct evidence exists as to which method best
cleans the media. Indications from scattered passages in filter per-
formance reports are that omission of either air scour or rotary sur-
face wash has led to difficulties [59,68,118].
Dual- and Triple-Media Filters
Dual- and triple-media filters are being used to provide filtration
from coarse to fine media size in the direction of flow, and thus to
achieve longer filter cycles without detriment to filtrate quality
[38,59].
The most common type of dual-media filter is an anthracite and sand
design although other combinations such as activated carbon and sand,
resin beds and sand, and resin beds and anthracite have been re-
ported [82]. Studies [59,127,128] appear to conflict somewhat on the
degree to which solids penetration and efficient bed utilization
occur in a dual-media filter. Bed depth, flow rate, media sizing,
and the nature of the filter influent (activated sludge versus trick-
ling filter effluent) appear to be variables influencing both the
penetration and degree of removal in a dual-media filter [38].
As with any granular media filter, efficient backwashing is essential
to prevent deterioration of the filter bed condition, which results
in filter cracks, mud balls, high initial head losses and reduced
filter runs or poorer filtrate quality. Backwashing techniques com-
monly used for dual-media filters are essentially the same as those
described for rapid sand filters. Air scour prior to backwashing is
common practice and is introduced through the nozzle underdrain
69
-------
systems common with this type of filter [58] . The use of a series of
air-scour and rest cycles, referred to as pulsed air scour, has been
recommended as a means of improving the efficiency of media agitation
[58,86]. Initially the air will follow the path of least resistance
and may completely skirt agglomerations in the bed if continued. By
pulsing the air and allowing the media to resettle, areas of the bed
resistant to break up are continually lifted and dropped, resulting
in better separation of the entrapped solids from the filter media.
Triple-media filters are the result of a logical extension of the
principles outlined for dual-media filters. The most frequently used
design incorporates anthracite, sand, and garnet, which are sub-
stances having approximate specific gravities of 1.7, 2.65, and 4.2,
respectively. Gradation from coarse to fine follows, obviously, the
same sequence, and it is claimed that the filter is less susceptible
to shock from rapid fluctuations in suspended solids concentration
[111]. It has also been suggested that triple-media filters are
superior to deep-bed filters using a single coarse media [111], but
reliable pilot-scale studies have not demonstrated superiority in
effluent quality or process reliability [68].
Promoters of the triple-media filter have expressed concern about
the use of air scour in this and other filters [111]. Disadvantages
of air scour are listed as increased downtime, possible media loss or
bed upset, and complication of the backwash cycle. The use of
rotary surface washers is recommended as an alternative. Others have
suggested that air scour is perhaps the only practical method for
cleaning deep bed filters like the dual- and triple-media units
[38,86], and that increased scouring efficiency is possible with air.
Filtered water is, in any case, generally used as washwater and is
applied at a rate of about 15 gpm/sq ft.
The use of triple-media filters for waste treatment is growing in
popularity in the United States, with over 50 installations in opera-
tion to date. Most designs have favored bed depths of 36 to 42 in.
and hydraulic loading rates of 5 to 6 gpm/sq ft. Pressure filters
are normally used for treating secondary effluent from plants with
flows of less than 5 mgd and gravity filters for treatment works with
flows in excess of this figure [43].
Immedium Upflow Filter
Although relatively unknown in the United States, the Immedium filter
was invented in 1961 by a Dutchman named Pieter Smit and was subse-
quently developed by the Boby Corporation in Britain [9,13], It is
a deep-bed, high rate, upflow filter using a patented grid placed
several inches below the top of the bed to prevent the expansion of
the media during filtration. Typical arrangements for both pilot-
and full-scale Immedium filters are presented in Figs. 19 and 20.
The grid is the key to the success of the design and consists of a
series of parallel bars at 4 to 6-in. centers which normally provides
70
-------
ACTIVATED
SLUDGE
WORKS FINAL
EFFLUENT
WET WELL
PUMP
FLOW
METER
MANOM-
ETER
OVERFLOW
EFFLUENT
WEIR
ft COARSE
&SAND (1-2 mm)$
•COARSER GRAVEJ.
UNDERDRAIN-
BACKWASH
OUTLET
DRAIN
MEDIA
SURFACE
SAMP
TA
Lll^
NK
JG
TO
WASTE
AIR BLEED
Fig. 19.
Pilot-scale Immedium filter used at West
Hertfordshire, England [142].
71
-------
CONCRETE
FILTERED
WATER
WASTE
TnTTTTmiLmTTTTTTTmrrniiirimniniiiiTTiin
'SAND RETAINING GRID
NOZZLE DISTRIBUTION SYSTEM
AIR
-XJ—
INLET
-txj-
WASH
OPEN TYPE
SAND RETAINING
STEEL SHELL
\
i n
rfv.
FILTERED
WATER
WASTE
N
INLET
WASH
NOZZLE DISTRIBUTION SYSTEM
Fig. 20. Immedtum filter arrangements for full-scale
installations, open and pressure [l3].
72
-------
openings 100 to 150 times the size of the smallest particle contained
in the bed [13,58]. During the course of filtration the sand forms
compression arches in the vicinity of the grid which resist the tend-
ency to lift or fluidize the bed, even at rates up to 6 gpm/sq ft and
at pressure drops equivalent to 20 ft of water.
Intense interest in the Immedium upflow filter in Great Britain has
resulted in at least five independent studies [68,83,86,131,142],
which are discussed in more detail later. Typical media gradations
and depths, in the direction of flow through the filter, are 4 in. of
10 to 15-mm gravel, 10 in. of 2 to 3-ram gravel and 60 in. of 1 to
2-mm sand, and the filter can be operated in a pressure or open type
housing as shown in Fig. 20 [9,13,58].
Backwashing is usually initiated when the head loss reaches 6 ft and
is preceded by draining the water above the filter to a level just
above the media surface and air scouring the filter at a rate of
5 cfm/sq ft. The air serves also to break up the sand compression
arches at the grid, and the backwash water is then turned on to ex-
pand the bed before the arches can reform. Since backwash water is
applied at 16 gpm/sq ft in the same direction as the influent, the
latter is conveniently used as washwater in most instances. Expan-
sionsof 27o [9] to 20% [13] have been reported as common during back-
washing, but it is the writer's opinion that the former value is more
accurate for 1.0 to 2.0-mm sand fluidized at 16 gpm/sq ft.
The first reported full-scale Immedium filter plant became operation-
al in 1969 at the East Hyde Works in Luton, England following a
series of pilot-scale studies. As a result of the pilot-scale stud-
ies, the backwashing procedure was modified to provide a series of
alternating high and low rate air scours and washwater applications,
both of which have been automated [9,82].
Deep-Bed, Coarse Sand Filters
The use of deep-bed filters with a single media of coarse sand has
been developed in Germany. According to Jung and Savage [70] its use
is widespread in potable water treatment, with over 200 existing in-
stallations. They are also being promoted for wastewater treatment
with media depths of 4 to 6 ft and media sizes of 1 to 2, 2 to 3, and
3 to 6 mm for wastewater filtration [108], Sand is very uniform,
with a uniformity coefficient of 1.25 or less. Backwash is first,
with air and water simultaneously followed by water alone. The 'back-
wash rate is 6 gpm/sq ft during the simultaneous air-water phase and
8 gpm/sq ft during the water only phase. Air-scour rates are 6 scfm/
sq ft for the 2 to 3-mm and 3 to 6-mm sand and 3 scfm/sq ft for the
1 to 2-mm sand. Adequate freeboard to overflow level (20 to 30% of
bed depth) is required to prevent media loss with the finer sand.
Backwash is for extended periods of 15 to 20 min so that total wash-
water used per backwash is not reduced compared to higher rates used
in United States dual-media filters.
73
-------
The backwash described above is well below fluidization velocity for
the media of the sizes indicated. Thus, the mechanism of cleaning
comes into question. Jung and Savage [70] describe the air and water
delivery systems and the presumed action in some detail. Air is dis-
tributed evenly by a pipe network. Water is delivered by a no (very
low) head loss precast block underdrain floor. The rising columns of
air act as airlift pumps to ensure uniform water distribution. The
localization of the air columns causes the velocities of air and
water flow to vary from zero to quadruple the mean and results in
vigorous pulsating washing action.
Quantitative studies demonstrating the effectiveness of this back-
washing system were not found.
Automatic Backwash Filters
In order to backwash any of the wastewater filtration systems pre-
viously discussed, it is required that they first be removed from
service. However, the automatic backwash (ABW) filter, as manufac-
tured by the Environmental Elements Corp. (formerly the Hardinge
Division) of Koppers Company, Inc., allows both filtration and back-
washing to occur in the same bed. A full-scale version of this sys-
tem, similar to that shown in Fig. 21, is now in operation at
Chicago, Illinois, and is filtering chemically treated secondary
effluent in two 12-in. deep sand beds having a total surface area of
1329 sq ft. The filter beds consist of a series of 8-in. wide, con-
tiguous compartments separated by steel plates which run perpendi-
cular to the long axis of the filter. A carriage assembly traveling
on rails mounted on the filter box walls suspends a cleaning hood
above the compartmented bed for the full width of the filter [116].
Backwashing is accomplished by centering the cleaning hood over a
given compartment, thus isolating it and allowing the rest of the bed
to continue filtration. A port on the side of the isolated compart-
ment is opened to the moving backwash supply pump and shoe, and the
washwater is pumped into the compartment underdrain and up through
the sand. Another pump mounted on the cleaning hood withdraws the
spent washwater and discharges it to waste. The time required to
completely wash all the compartments in one of the 53.2 by 12.5 ft
beds at the Chicago installations is 57 min [116].
The principal investigator recently (1975) visited this Chicago in-
stallation at"Hanover Park after it had been in service about seven
years. During that period the filters had been used for direct fil-
tration of a prechlorinated secondary effluent from an activated
sludge plant which was sometimes coagulated prior to filtration. The
plant superintendent, Mr. Robert A. Ziols reported satisfaction with
the filters and no particular problems related to inadequate back-
wash. The automatic backwash occurs about nine times per day. They
have routinely replaced about 1 in. of filter sand each year. In-
spection of the media at such times had not revealed typical dirty
74
-------
CARRIAGE DRIVE
MOTOR tk
WASHWATER
PUMP
BACKWASH
PUMP
WASHWATER
LAUNDER
UQDTD
CONTROL
LEANER
WATER LEVEL
•EFFLUENT CHANNEL INFLUENT CHANNEL
Fig. 21. Environmental Elements Corp. (Koppers) full-
scale automatic backwash filter (from
manufacturer's brochure).
filter problems such as mud balls and filter cracks. An expansion of
the plant will include additional filters of the same type. The
manufacturer now uses sand with an effective size of 0.6 to 0.65 mm
for wastewater filters and reports 25 wastewater installations be-
tween 1966 and 1974.
Moving-Bed. Continuously Cleaned Filters
In the next few years wastewater filtration practice may encompass
the use. of moving-bed filters which are at present in the research
and development phase. One filter of this type is basically a verti-
cal,- upflow unit in which provisions have been made to.continuously
remove a small portion of the bottom-most sand, the media making
first contact with the influent. The sand is then cleaned and lifted
to the top of the'filter where it is redeposited. In this manner
the flow through the filter contacts progressively cleaner media as
it travels through the bed, and the dirtiest portion of the media is
constantly being drawn off and cleaned. Preliminary results indicate
that raw water quality has little effect upon performance for flow
rates comparable to conventional upflow and downflow units [58],
75
-------
A variation of the moving-bed filter known as the Simater filter is
being manufactured by Simonacco Limited of Carlisle, England, and was
tested in a pilot-scale study in Derby, England [68], Figure 22
shows the basic arrangement of the unit, which differs from the pre-
viously described filter in that the unit distributes the influent
radially from the center of the bed (A and B). Vertical perforated
pipe laterals whose orifices are covered with filter cloth are spaced
around the outer circumference of the filter housing to collect the
effluent, which is then discharged through lines D and E. Dirty
media is withdrawn in line F and air lifted to the top of the filter
where it is discharged into the upper chamber, G. During the lifting
processes the solids are removed from the media by the scrubbing
action so that the sand falling to the top of the filter has been
cleaned. The solids are then withdrawn from the upper chamber at
point G [68].
Case Histories
The following case histories will present the backwashing procedures
and their effectiveness as reported in various wastewater filtration
studies from the literature. In many of these papers, substantial
filter performance data are also presented but will not be repeated
in this backwashing report. The performance data have also been sum-
marized by Rice [99].
Early Practice According to Streander
Perhaps one of the unsung pioneers of wastewater filtration was
Philip B. Streander, a New York City consulting engineer whose in-
terest in sewage filtration began in the late 1920's. In a solitary
article published in 1935 [121] and a subsequent series of three ar-
ticles published in 1940, Streander [122,123,124] outlined the state-
of-the-art of sewage filtration in this country and abroad.
Streander proposed a mechanically cleaned downflow filter [121] of
silica sand or crushed anthracite graded to 0.59 to 0.84-mm particle
size. Filtration rates of 2 gpm/sq ft were recommended, and cleaning
was accomplished by a moving, full-width cleaning head. The head was
equipped with two rows of hollow rake teeth, positioned to break up
the surface and subsurface portions of a 6 to 18-in. deep filter bed.
The teeth were equipped with orifices through which high pressure
washwater was pumped, causing an expansion of the media directly
under the cleaning heads equivalent to a rise rate of 16 in./min.
Solids entrapped by the media were released and carried above the ex-
panded bed to an upper zone where they were withdrawn in the spent
washwater by a pump mounted on top of the hood.
Although it was thought that the combined mechanical and hydraulic
action would be sufficient to thoroughly clean the filter, perfor-
mance in actual installations proved otherwise. It was discovered
that the cleaning method was highly efficient in the upper part of
76
-------
SEE TEXT FOR EXPLANATION
BAFFLE
H
SURFACE AT MEDIA -
HOUSING.
INFLUENT
A
*^*~*~i*~i~*~*j(
WATER LEVEL
CENTRAL INFLUENT
DISTRIBUTOR
MEDIA
Fig. 22.
AIR LIFT
Pilot-scale "Simater" radial-flow, moving-bed sand
filter by Simonacco Ltd. of Carlisle, England [68].
77
-------
the bed but could not effectively clean the media directly above the
supporting screen. As a result bacterial slimes developed which ad-
versely affected the capacity and efficiency of the filter [122],
Much of Streander's thinking was directed along the lines of shallow-
bed silica sand filters employing mechanical cleaning devices. He
felt these designs had definite merit as a secondary treatment system
for the removal of solids escaping the primary settling process.
Because of the automatic backwash feature, installations of this
equipment for direct filtration of primary effluent had consistently
removed 50% of the incoming suspended solids regardless of solids
concentration or filter hydraulic loading (apparently 2 gpm/sq ft was
maximum) [124]. Streander recommended bed depths of 6 to 12 in. and
hydraulic loadings less than or equal to 2 gpm/sq ft.
Although a proponent of shallow-bed, mechanically cleaned filters,
Streander was also aware of the potential of backwash type, deep
granular filters in sewage treatment. He reported on the full-scale
installation of such a plant in Wuppertal, Germany, a city of about
410,000 people at that time [122], As a result of extensive pilot-
scale studies of filtration rate, sand size, and bed depth, the de-
signers of the Wuppertal plant chose a 1.0 to 2.0 mm sand placed to
a depth of 28 in. over a 4-in. layer of coarse and fine gravel. Ten
filter beds were provided, each 26 ft 3 in. by 123 ft long, giving a
combined filtering area of 32,280 sq ft for a flow rate of 24 mgd.
At the relatively low hydraulic loading rate of 0.5 gpm/sq ft and
filtering primary effluent, suspended solids removals averaged only
40£. Streander felt that the low efficiency was the direct result
of a poor backwashing procedure which did not thoroughly clean the
filter. Basically the backwash sequence consisted of air scour of
unspecified rate and duration followed by a low rate backwash of 6 to
8 in./min (vertical superficial velocity). Streander speculated that
the air agitation loosened the heavier trapped solids, but the low
wash rate could not effectively carry these away nor remove the
finer, more tenaciously bound solids.
Regardless of the depth of bed employed, Streander realized the need
for proper sand size selection and the dependence of such upon the
nature of the influent solids, the concentration of the solids, and
the availability of washwater [122]. He stressed that media selec-
tion should be carefully considered to prevent premature clogging and
to promote distribution of the suspended solids removal throughout
the filter bed. Furthermore, he recognized the necessity of adequate
media cleaning techniques with respect to proper filter operation.
Streander predicted the increased use of filtration as an effluent
polishing application for treating activated sludge and trickling
filter plant effluents. Placed after final settling tanks, filters
could, he believed, be used to remove the lighter, less settleable
solids, thus permitting a smaller clarifier or overloading of an
78
-------
existing unit. The following quote [124] summarizes his views on the
potential of wastewater filtration:
It is the writer's belief that this matter of dollars and
cents value of effluent strainers has only begun to be
realized, and that the future will find these worthy ad-
juncts to treatment plants coming into widespread use as
an economy feature as well as being producers of more
even and dependable quality of plant effluent.
British Studies
Luton. British wastewater treatment literature abounds with refer-
ences to the Luton Corporation Works filtration plant and to experi-
mental filtration studies conducted there. In 1949, prior to the in-
stallation of filtration facilities, initial studies were begun by
Pettet et al. [92,93] using pilot-scale equipment. Treatment at that
time consisted of primary settling, activated sludge and trickling
filters in series, and final settling. The investigations were con-
ducted using two pilot-scale pressure filters operating at a hydrau-
lic loading of 2 gpm/sq ft, which filtered plant effluent over an 11-
month period. Filter housings consisted of vertical, cast iron cyl-
inders 3 ft high and 2 ft in diameter capped with a flanged dome and
seated with a flanged cone. The overall bed depth in each filter was
24 in.; however, one was equipped with sand graded 0.85 to 1.7 mm and
the other with 10 in. of 2.1 to 6.0-mm anthracite under 14 in. of
0.85 to 2.1-mm anthracite.
Backwashing was scheduled on a 24-hr interval and was applied at a
rate of 14 gpm/sq ft using filter effluent. Spent washwater, com-
prising approximately 3% (300 gal.) of the throughput, was discharged
through a valve in each housing and returned to the treatment plant
settling tank. To assist in the breakup of the clogged bed, each
filter housing was equipped with a handwheel-operated rake whose
teeth extended 1 ft below the media surface. As the study progressed,
daily head loss increased from 4.5 to 30 ft because of inefficient
cleaning. An examination of the filter interior revealed that the
filter walls and sand were covered with gelatinous coatings of 1/2
in. and 1/4 in., respectively. Sodium hypochlorite solutions were
applied in varying doses and for several contact times, the most
successful combination being 0.05% sodium hypochlorite as chlorine
with a 2-hr contact time. After this cleanup procedure, no addition-
al chlorination was required.
Immediately following the conclusion of the first Luton study, Pettet
et al. [92,93] began a second pilot-scale study at a nearby trickling
filter plant, the Finham Works in Coventry. Again, separate sand and
anthracite filters were investigated except that they were now grav-
ity operated. One 12-in. diameter housing contained 24 in. of sand
graded 1.0 to 2.0 mm and the other contained 3-in. of 2.6 to 6.0-mm
anthracite supporting 24 in. of coal graded 1.0 to 2.0 mm.
79
-------
Backwashing requirements were approximately 27=, of the filter through-
put, and each bed was air scoured prior to introducing the washwater.
During the latter four months of the Finham study the terminal head
loss of the sand filter approached 8 ft on occasion where it normally
did not exceed 2 ft. This problem was reduced by using approximately
three times the normal washwater volume during the backwashing fol-
lowing every third run. At no time was chlorination required to con-
trol gelatinous accumulations within the filters. Based upon the
data available from both studies Pettet concluded that air scour was
worthy of additional study as a backwashing aid.
Upon completion of the Finham study, Pettet et al. [93,94] returned
to the Luton Corporation Works to conduct additional experiments with
pressure and gravity filtration. Equipment for the pressure filter
study were the same as previously described for Luton. Media type,
depth, and gradation were changed several times during the study.
Initially one filter contained 24 in. of 0.35 to 1.7-mm sand and the
other, 24 in. of 1 to 2-mm anthracite. These two filters were oper-
ated at rates varying from 1.2 to 4.0 gpm/sq ft over a seven-month
period. Backwashing of each filter was conducted on a 24-hour basis
and was preceded by three minutes of air scour at 18.3 scfm (5.8
scfm/sq ft). Washwater was drawn from the filter effluent tank and
pumped at rates of 14 gpm/sq ft and 8 to 10 gpm/sq ft for the sand
and coal filters, respectively. Total washwater volume was approxi-
mately 2 to 3% of throughput. Throughout the comparison of sand and
coal filters, no appreciable differences in effluent quality were ob-
served, but effluents from each deteriorated markedly after six
months of operation. An inspection of the media in both filters re-
vealed that the individual particles were heavily coated with a bio-
logical slime which was believed to be the source of the problem.
Pettet et al. [93,94] considered this the result of inadequate air
scour, so the media was chlorinated and, subsequently, the air-scour
time was increased to 10 to 15 min. The effluent quality of both
filters was restored to previous levels and the increased air scour
appeared to maintain the cleanliness of the media so that no addi-
tional chlorination was required.
The next phase of the study was directed at determining the effect of
depth on effluent quality. The old media in each filter was replaced
with a single size range (0.85 to 2.06 mm) of hand-graded sand placed
in one filter to a depth of 24 in. and in the second to a depth of
42 in. It was thought that the deeper bed would prevent solids
breakthrough at higher hydraulic loadings, but it was discovered that
the shallow bed reached terminal head loss before exhibiting break-
through. The filters were operated for two months at hydraulic load-
ings up to 5.7 gpm/sq ft, but no appreciable differences in effluent
quality were observed below loadings of 4.6 gpm/sq ft and only slight
improvement by the deep filter at rates above this figure. Backwash
requirements for the shallow sand bed were the same as those for the
sand bed of the preceding phase: 14 gpm/sq ft and 2 to 3% of the
80
-------
filtered volume. For proper cleaning of the deeper bed, however, the
wash rate had to be increased to 20 gpm/sq ft, and the total volume
of washwater used also increased [93,94]. Each was air scoured at an
unspecified rate and duration prior to the water backwash.
Based upon the data collected by Pettet et al. [92,93,94] in the ex-
tensive experimental studies previously discussed, six rectangular
rapid sand filters were installed at Luton in 1951 [41,42]. Each
filter was operated at a hydraulic loading of 4 gpm/sq ft, which was
maintained by a rate controller on the filter effluent line. The
media in each filter consisted of 36 in. of sand graded so that 9070
of the grains were within the 0.85 to 1.67-mm size range. Water
backwashing was preceded by air scour at a rate of 1 scfm/sq ft for
an unspecified duration. Washwater was then pumped at a rate of
14 gpm/sq ft from a filtered water storage tank, each filter normally
requiring a washwater volume equal to 2-1/2% of its throughput.
Spent washwater (approximately 15,000 gal. per filter backwash) was
collected in a second storage tank and fed back gradually to the head
of the plant. The filters ordinarily exhibited an initial head loss
of 1 to 1.5 ft following backwashing and were allowed to attain a
head of 8 to 9 ft before cleaning was again initiated. Under average
conditions the filters required backwashing twice daily [41,42,86,94].
In 1954 the -filter plant capacity was increased by 50% with the addi-
tion of three more units, which corresponded to a simultaneous 507<,
expansion of the activated sludge treatment units. The new filters
were essentially the same as the original units although the sand was
apparently graded somewhat finer.
Data gathered over the years since the installation of the units has
shown that the rapid sand filters generally performed well although
they were susceptible to effluent deterioration from poor backwashing
or shock loads of sewage containing high suspended solids concentra-
tion [86]. It had also been observed that media was being lost as a
result of slime coatings which decreased the effective particle den-
sity and, thus the grains were more easily carried away with the
spent washwater. Backwashing procedures were changed to those as
presented in Table 5. Because of these difficulties and of more
stringent effluent standards recently imposed upon the Luton plant, a
new study was initiated.
The objectives of the investigation were to directly compare a pilot-
scale Immedium upflow filter with one of the existing rapid sand fil-
ters and to determine the best means of backwashing each. The re-
sults of the study demonstrated the superiority of the upflow unit
for the following reasons [86] :
1. Suspended solids removals were consistently better even at flow
rates 50% higher.
81
-------
2. Greater utilization of the bed for solids removal resulted in
longer runs and greater throughput.
3. High influent suspended solids concentrations did not cause
severe upsets and were readily handled at slightly reduced load-
ings (3.4 gpm/sq ft).
4. Backwash can be conducted readily with filter influent.
5. Reduced capital costs resulting from elimination of high filter
walls (no static head requirements).
6. Less down time for backwashing.
Backwashing procedures are summarized in Tables 5A and 6A, which are
copied from the original references. The pulsating air scour is re-
portedly very effective in breaking up the agglomerated media of the
upflow filter following a run. Apparently, the agitation is quite
severe, and the disturbance travels upward through the bed in a wave-
like motion.
A visit to the Luton works in November, 1975, revealed the following
developments. Sand in the rapid sand filter after backwashing had
heavy organic slimes surrounding the sand grains, but no mud balls
were visible. The slimes in the upper sand layers after backwashing
caused about 12% dry weight loss when a sample was ignited in a stan-
dard loss in ignition solids test. The backwash procedure had been
changed to that shown in Table 5B incorporating a pulsed air scour.
Sand loss was still a problem, and periodically sand is added to
maintain desired sand depth. The dirty sand did not appear to affect
the filtrate quality. Run lengths were only about 7 hr due to high
influent suspended solids (about 35 mg/1) coming from overloaded
final settling tanks.
The upflow filters have also experienced sand loss, which at times
has completely exposed the hold down grid. During high flows, this
has resulted in uplifting of the bed and breakthrough of solids at
head losses of greater than 6 ft. The backwash procedure had been
modified in 1975 as shown in Table 6B to attempt to improve backwash
effectiveness. The change included an increase in the high rate wash
from 15 to 19 gpm/sq ft. The condition of the deep media was not
observed due to difficulty of sampling.
Immedium filter studies. The favorable results obtained at Luton
sparked interest in the Immedium upflow filter as a wastewater filter
and spawned three additional studies over the next four to five years.
The first of these studies was conducted by Woods et al. [142] using
a pilot-scale unit at the West Hertfordshire Works and activated
sludge plant effluent. Details of media sizing and bed depth were
not presented; however, the pilot Immedium filter was described as a
82
-------
Table 5A. Manual backwashing procedure used on full-scale rapid
sand filters at Luton [86],a
Minutes after
stoppage of Description of operations
inflow
0 Inflow stopped. Drain down commenced and, if
possible, continued to completely drain down the
water through the filter sand into the effluent
channel.
15 Backwash outlet valve opened to remove any water
remaining above the level of the backwash weir.
20 High-rate air scour and slow-rate backwash com-
menced together.
30 Air scour stopped. High-rate backwash commenced.
40 High-rate backwash stopped. Normal inflow
commenced.
filtered water used for backwashing, water and air rates were not
specified.
Table 5B. Backwashing procedure in full-scale rapid sand filters
at Luton in November 1975.
Minutes after
inflow stoppage Operation
0-28 Drain down to bottom of sand level.
28-32 Low-rate air and water backwash simultane-
ously to a water level 9 cm above the fixed
bed surface. Air at 1.23 cfm/sq ft and
water at 2.1 gpm/sq ft
32-41 Intermittant air scour. Air on at 1.23
cfm/sq ft for 45 sec and off for 45 sec, and
cycle repeated for 9 min
41-49 High-rate backwash at 13.7 gpm/sq ft
83
-------
Table 6A. Automatic backwashing sequence used on pilot-scale
Immedium upflow filter at Luton [86].a
Minutes after
initiation of
backwash cycle
Description of operations
0
19
26
Backwash cycle initiated by pressure switch.
Inflow stopped. Drain down commenced.
Drain down completed. High rate air scour (4.8
cfm/sq ft) and slow-rate backwash (2.4 gpm/sq ft)
commenced together.
Slow rate backwash stopped by electrode 9 in.
above sand level. Pulsating air-scour commenced
(3.3 cfm/sq ft for 45 sec alternating with 0.5
cfm/sq ft for 45 sec).
Pulsating air scour stopped. High-rate backwash
(15 gpm/sq ft) for 30 sec, then dropped to slow
rate backwash for 10 sec, then increased to high-
rate backwash. (This pulsing of the backwash
rates ensures that any entrapped air bubbles
rise up out of the sand bed.)
High rate backwash stopped. Normal inflow
c ommenc ed .
water used for backwashing.
Table 6B. Backwash procedure for full-scale upflow filters at
Luton in November 1975.
Minutes after
inflow stoppage
Operation
0-27
27-31
31-37
37-47
Drain down to bottom of sand.
Air and water wash together to a level 9 cm
above sand surface. Air at 2.85 cfm/sq ft and
water at 2.5 gpm/sq ft
Intermittent air scour at 2.14 cfm/sq ft, 45
sec on,.45 sec off for 6 min
High rate backwash at 19 gpm/sq ft
84
-------
typical unit, which would mean that it consisted of a 1-ft layer of
gravel topped by 5 ft of 1 to 2-mm sand.
Operational advantages of the unit were: excellent ability to handle
hydraulic shock loads, high suspended solids and biochemical oxygen
demand (BOD) efficiency at higher flow rates per unit area, and ease
of backwashing. The latter item was accomplished by draining down
the filter to the media surface and air scouring at unspecified rate,
then applying filter influent at a rate sufficient to break the com-
pression arches in the bed and fluidize the media. This backwashing
procedure was a departure from that recommended by Boby and Alpe [13]
because the backwash water was introduced after the air scour. No
leaves, paper, or fat was observed lodged in the base of the filter,
which is the point of initial contact with the influent and backwash
water [142]. The backwash water removed 1.33 Ib of entrapped solids
per square foot of filter area and required an average 1056 gal. for
each wash.
At approximately the same time, Truesdale and Birkbeck [131] were
conducting a similar Immedium filter study at another activated
sludge plant in Letchworth. Their pilot-scale apparatus consisted of
a 2.5-ft diameter housing containing a 1-ft layer of graded support
gravel and 5 ft of 1 to 2-mm sand. The media-restraining grid con-
sisted of a series of parallel bars at 4-in. centers located 2 in.
below the sand surface. When the head loss reached a predetermined
value, cleaning was initiated with an air scour at unspecified rate
followed by backwash with filter influent at 14 gpm/sq ft for 15 min.
The total volume of washwater used averaged 5.7% of the throughput.
Michaelson [83] conducted a third study on upflow filtration using
trickling filter effluent from the plant at Ashton-Under-Lyne. Few
details about the backwash were presented except that backwash was on
a daily basis and used 1.25% of the throughput.
Derby. An extensive, pilot-scale investigation of wastewater filtra-
tion was begun in 1966 by Joslin and Greene [68] using trickling fil-
ter effluent from the treatment works at Derby. During the next two
years, seven separate filters using three different flow configura-
tions were studied:
1. A 24-in. deep downflow sand filter with media graded 1.2 to
2.4 mm.
2. A 36-in. deep downflow sand filter with media graded 1.2 to
2.4 mm.
3. A 24-in. deep downflow sand filter with media graded 1.2 to
1.7 mm.
85
-------
4. A 24-in. deep triple-media downflow filter consisting of anthra-
cite, sand, and garnet graded 1.4 to 2.4 mm, 1.2 to 1.4 mm, and
0.71 to 0.85 mm, respectively.
5. A 24-in. deep upflow sand filter graded 0.71 to 2.40 mm.
6. An Immedium upflow sand filter with 26 in. of graded gravel and
48 in. of 1 to 2-mm sand.
7. A radial flow, moving bed sand filter containing 0.5 to 1.0-mm
sand.
Joslin and Greene demonstrated that with respect to filtration of
Derby effluent, flow direction was insignificant in comparison to
proper media depth and gradation. This was in direct contrast to
several other studies where the Immedium upflow filter had demon-
strated superiority to conventional downflow sand filters [83,131,
142]. Although flow configuration appeared to play no role in im-
proved suspended solids removal, increased bed depth did improve re-
moval, as evidenced by the superior efficiency of both the Immedium
upflow (4-ft depth) and deep-bed rapid sand filters (3-ft depth).
Head loss development was less rapid in the filters containing 1 to
2-mm graded sand than those, such as the triple-media filter, con-
taining finer particles. Because of the subsequently longer filter
runs and the ability to meet desired effluent standards at high
loading rates, the overall conclusion of the study was that deep-bed
filters of coarse 1 to 2-mm sand were most suitable for a full-scale
application.
Backwashing experiences with the filters at the Derby installation
were interesting and were reported in some detail. An inadequate air
supply precluded the use of air scour prior to applying washwater to
the media, so an attempt was made to compensate for this deficiency
by backwashing at higher rates. Difficulties were encountered in all
the units, however, with mud ball and agglomerate formations which
resisted breakup by the water (only) backwash. This problem was par-
tially relieved by draining the water level to within 1 in. of the
surface and then directing a jet of water onto the surface of the
filter prior to backwashing. Mud balls which escaped breakup by the
jet tended to settle deeper in the filter. Difficulties were also
encountered at the start of each backwash because the filter media
tended to rise as a plug upon initial application of the washwater
and caused fluctuation on the backwash rate indicator.
Minworth. Activated sludge effluent from the Minworth treatment
plant was used as filter influent in an investigation conducted by
Tebbutt [129]. Three pilot-scale filters were assembled. Total bed
depth in each filter was 24 in., with the first filter containing 0.5
to 1.0-mm sand, the second filter containing 1.0 to 2.5-mm anthracite,
and the third containing half 0.5 to 1.0-mm sand and half 1.0 to 2.5
mm anthracite. Flow rates to each filter were varied from 1.7 to
86
-------
10.3 gpm/sq ft, and no substantial drop in efficiency with increasing
rate was observed.
A second series of in-plant tests was also conducted using 24-in. bed
depths, the three filters having one of the following specifications:
1.0 to 2.5-mm anthracite, 1.2 to 2.4-mm sand, 2.4 to 4.7-mm sand.
Flow rates applied to the filters were the same as for the first in-
plant series.
Backwashing details provided in the text of the article by Tebbutt
are scanty, but no air scour was used, and difficulties were encoun-
tered with the media rising as a "plug" at the start of the wash, a
problem also reported by Joslin and Greene [68]. Filter beds con-
taining anthracite were most pronounced in exhibiting this problem,
while the sand beds were generally easier to maintain. Coarse sand
(1.2 to 2.4 mm) was not expanded but did clean readily. Tebbutt felt
that the rising plug problem could be eliminated in a full-scale
plant with the use of air scour and that this would place the anthra-
cite media in an operating advantage since the Minworth studies
showed less washwater was required for anthracite filters in spite of
the plugging.
South African Studies
Ancor. In 1947 Vosloo [136] reported the results of an investigation
in which a pilot-scale pressure filter was used to treat the effluent
from the Ancor waste disposal plant. Although the nature of the
treatment of the Ancor plant was not described by Vosloo, Huang [59]
described it as having secondary treatment. A reasonable assumption,
considering the climate and other installations in the area [87], is
that the plant consisted of primary settling, trickling filters, and
final clarifiers.
The pilot filter was contained in a 2-ft diameter steel casing having
a height of 4.5 ft. The filter media consisted of a 24-in. layer of
graded 0.5-mm sand and 5 in. of 0.8 to 1.7-mm sand, both layers of
which were supported by 7 in. of graded gravel. Flow configuration
was downward, and effluent from the filter was collected by nozzle
mounted in a steel plate false bottom. In a manner similar to the
arrangement on the early Luton pressure filters [92], the housing for
the Ancor filter was equipped with a handwheel-operated rake to break
up the media surface during backwashing. Unfortunately, no other de-
tails about backwashing effectiveness or backwashing problems were
presented.
Pretoria. Since 1954 full scale gravity filters have been used at
the Pretoria Purification Works to remove suspended solids from a 3
mgd flow at the trickling filter plant [87]. The five filters are
each 13.5 ft in diameter and can develop a maximum head of 9 ft. Air-
scour capability for backwashing has been provided on the four filter
beds equipped with a steel plate false bottom containing 1-in.
87
-------
diameter nozzles at 6-in. centers [72,88]. Compressed air is intro-
duced below the false bottom of these filters by a 1-1/2-in. diameter,
T-shaped head and is distributed through the nozzles to the media.
The remaining filter is equipped with a United States-manufactured
carborundum block false bottom and rotary surface washers. Each fil-
ter contains sand media graded to 0.55 to 0.85 mm for 21 in. and 0.85
to 3.2 mm for 6 in., and those equipped with the steel plate (nozzle)
underdrain have been provided with a 12-in. layer of support gravel.
Designed for a maximum rate of 3 gpm/sq ft, the filters have been
operated at an average rate of 2.6 gpm/sq ft to a terminal head loss
of 6.5 ft [72], During an average 8 to 10-hr run, the filters con-
sistently reduced the influent suspended solids level of 22 mg/1 from
the treatment works to less than 5 mg/1 [102]. Backwash pumps were
sized to provide a maximum rate of 45.6 gpm/sq ft, which is suffi-
cient to expand the sand 50%; however, a rate of 25.7 gpm/sq ft has
been found to be adequate [ 111]. Washwater volume normally comprised
192,000 gal. or approximately 9.5% of the treated flow. Interesting-
ly, Nicolle [87] has reported that the filter equipped with rotary
surface washers had slightly longer runs and appeared quite clean
compared to the filters equipped with air scour. Several comments by
Nicolle [87,88] with respect to design and operation of the filters
are important:
1. Filter influent at Pretoria is chlorinated at a dosage 4.7 mg/1
to control biological growths and deposits and is considered
highly beneficial.
2. The filters have demonstrated a distinct susceptibility to hy-
draulic shock caused by stopping, starting, or altering the
application rate and have required backwashing soon thereafter.
3. The use of a circular shape for wastewater filter housings should
be avoided because of increased space requirements and diffi-
culty in altering washwater trough placement.
88
-------
VIII. EXPERIMENTAL COMPARISON OF BACKWASH METHODS
IN WASTEWATER FILTRATION
Pilot-Plant Equipment
The Ames Plant
Experimental work comparing different backwashing methods was con-
ducted at the Ames Water Pollution Control Plant using a pilot-scale
filter plant designed and built for this study. The Ames treatment
plant facility employs biological treatment via three standard rate
trickling filters. The plant was completed in 1951 and at the time of
this study (1973-75) was somewhat overloaded. Raw sewage enters the
plant through a comminutor pit and then is lifted to an aerated grit
removal chamber. Effluent from the grit chamber is then passed
through four rectangular primary sedimentation tanks which provide
about 2 hr detention time at a flow rate of 4 mgd. The primary tank
effluent is then applied to the trickling filters by rotary distri-
butors and thence to three circular final settling tanks. Although
chlorination facilities have been provided, they have never been used
because of hydraulic difficulties in discharging from the tank to the
Skunk River. Therefore after final clarification, the flow is col-
lected and discharged to the Skunk River.
Primary sludge and scum are collected and directed to an anaerobic
digester. Final sludge is drawn off from each of the tanks and mixed
with raw sewage at the head of the plant.
Serving both the University and the city, the plant treats sewage
which is primarily domestic in nature and not subject to unusual load
or strength variations from large industrial water users. Variations
are observed in sewage strength, however, during the regular school
year and the summer vacation, when the university population fluc-
tuates widely. Furthermore, the sewage collection system is quite
susceptible to infiltration and inflow, and high flows of dilute
sewage are not unusual during wet weather.
General Arrangement
The pilot plant was arranged as shown in Fig. 23 for the influent,
flow-splitting method of rate control. In this arrangement the in-
fluent to each of the three filters was provided by throttling the
supply pump discharge and then splitting the pump discharge equally
into thirds. Outlet lines from the filter housings were connected
to a single effluent line which was placed at an elevation above the
surface of the media in the filters and defined the low static
water level in the filter boxes. As head loss developed in the fil-
ters during the course of a run, the water level above the media sur-
face continued to rise until it reached an elevation equal to the
89
-------
FLOW SPLITTER
.ROTAMETER-^
SPLITTER BOX
BYPASS LINE
BACK-
WASH
WASTE
NORTH
FILTER
ROTAMETER
-ex 1 | SUPPLY
I PUMP
BACKWASH
WASTE
SOUTH
FILTER
SURFACE
WASHER
SUPPLY
r
BACKWASH
WASTE
WEST
FILTER
EFFLUENT
PUMP
_J* ?
* 1 i I
FLOWMETER
PRESSURE
GAUGE
ROTAMETER
BACKWASH PUMP
COMPRESSED AIR
BACKWASH
STORAGE
TANK
PILOT PLANT EFFLUENT
Fig. 23. Schematic representation of pilot-scale filter plant
used in experimental investigation.
90
-------
height of the bypass line in the influent flow-splitting device.
This latter elevation defined the high static water level, or termi-
nal head loss, which was available in the pilot plant. When filter
head loss reached the terminal stage, part of the influent was by-
passed automatically, and the filter ceased to operate at a constant
and equal rate and thereafter declined in rate until it was back-
washed.
Chronology of Experiments
The experimental work with the pilot plant was conducted in several
phases in which different filter media and or backwashing techniques
were used. The phases are summarized in Table 7. In addition to the
differences shown in the table, there was also a difference in source
of backwash water. During Phases I and II the filtered effluent was
used as the backwash supply by passing it through the backwash stor-
age tank of Fig. 23. In Phases III through VI, the unfiltered sec-
ondary effluent was used as a backwash source.
Equipment Details
The west filter was completed during the middle of Phase I. Since
a complete set of data were not collected for that filter on Phase I,
the data for that filter in Phase I are not presented herein.
Flow Splitting
Two different flow splitters were used. During the first half of
Phase I, a standard 1-in. pipe tee was used. Globe valves on either
branch were used to adjust the split. This arrangement was easy to
construct and functioned reasonably well. During the last part of
Phase I, it was desired to split the flow three ways so that the West
filter could be used. This was accomplished by use of an influent
splitter box which was used for all subsequent Phases. The splitter
box consisted of two fabricated steel boxes, or halves, that fit on
top of each other. The influent flow was split equally to the three
filters by identical orifices located in the bottom of the upper box.
A float valve on the influent pipe was used to maintain a constant
water level on the orifices in the upper box. Discharge from each of
the three orifices dropped through an air gap into one of the inlet
lines to the filters which were connected to the bottom of the lower
box. When one or more of the filters reached terminal head loss,
water would back up in the lower box and overflow to waste. The fil-
ter inlets and overflow line were arranged so that the other filters
continued to function without interruption or change in flow rate.
This arrangement allowed completely independent operation between the
three filters. Several sets of identical orifices were available
which could be inserted in the flow-splitter box to achieve various
desired filtration rates. The actual filtration rate to each filter
was measured with a rotameter on the overflow line. The flow to any
91
-------
Table 7. Summary of experimental phases for wastewater filtration backwashing study.
Phase
I
II
III, IV
and V
VI
Period
5/17/73-
8/21/73
8/28/73-
10/20/73
5/16/74-
ll/ 2 Ilk
6/4/75-
8/4/75
Filter
Influent
Alum
treated
sec. effl.
Secondary
effl.
Secondary
effl.
Secondary
effl.
Media
Na sa wa
Identical dual
media
Same
Dual Coarse Mixed
media sand (tri)
media media
Identical coarse
sand
N
Water
only
n
Air
scour
type 2b
Backwash
S
Air
scour
type lb
n
Air and
water
together
type 3b
Identical air and
type 3b
W
Fixed
surface
wash
Rotary
surface
wash
Surface
and
subsurface
wash
water
&N, S, W refer to the north, south, and west filters, respectively.
Type 1 air scour means air first followed by water with full-bed fluidization and expansion.
Type 2 air scour means air and water together briefly during the rising water level followed by
water only with full-bed fluidization and expansion.
Type 3 air scour means air and water together during most of the backwash while overflow was
occurring. Used only for coarse sand media which was not fluidized or expanded during the backwash.
-------
filter was purposely bypassed to the rotameter to measure the rate
whenever desired.
Alum Treatment Equipment
During Phase I, the settled secondary effluent was pumped to an upflow
solids contact unit where it was treated with alum for phosphate
precipitation and settled before filtration. Figure 24 shows the
upflow solids contact unit. This unit was mounted in a truck and
formerly served as part of a mobile water purification plant for the
United States Army. The conical shaped settling and reaction tank is
referred to as an erdlator. It had a volume of approximately 530
gal.
Raw water was pumped through a pair of nozzles into two troughs.
This aerated the water just before it overflowed into a vertical,
cylindrical, center column in the erdlator. Alum solution was added
in one of these troughs. The center well contained an agitator that
mixed the alum solution with the raw water and provided some floccula-
tion. Rapid mixing was not provided. After flowing downward through
the center column, the water moved upward to a collection trough.
The flocculant precipitate comprised of A1P04 and A1(OH>3 separated
in a floe blanket where the upward water velocity became equal to the
hindered settling velocity of the floe. The blanket of floe provided
an environment that aided the reactions with the alum and encouraged
flocculation or the agglomeration of smaller floe particles into
larger ones. Either sedimentation or enmeshment in the floe blanket
removed much of the material suspended in the erdlator influent. The
surface of the floe blanket was easily observed through the clear
water above. The blanket itself appeared to be fairly thick, but its
lower surface could not be observed.
The clear water above the floe blanket was collected in a circular
trough. Orifices on both sides near the top of the trough provided
for even collection of water. From the collection trough the water
flowed to a clear well.
As the floe in the erdlator formed and as sewage particles were en-
meshed in it, a considerable volume of sludge was created. A slot in
the side of the erdlator drew the sludge from the top of the blanket
into a sludge concentrator tank. The clear supernatent from the con-
centrator was drawn off into the clear well and the concentrated
sludge was drawn to waste by a positive displacement pump.
A dual head, diaphram type, positive displacement pump was used to
feed the alum solution. The pump drew from two 150-liter containers
where the alum was prepared.
Water was drawn from the clear well by a self-priming centrifugal
pump. A rotameter and a globe valve in the pump discharge line
93
-------
COLLECTION
TROUGH
TO POP FLOC
BLANKET
15 in.
UPFLOW
SETTLING
ZONE
39 In.
CENTER
COLUMN
MIXING ZONE
CENTER
COLUMN
COLLECTION
TROUGH
Fig. 24. Pilot plant solids contact unit.
94
-------
enabled quick flow measurement and adjustment. The pump then dis-
charged to the flow splitter.
Filters
Housing details. Each of the three filter housings used in this ex-
perimental investigation was fabricated from 3/16-in. steel plate and
was equipped with a 1/2-in. thick plexiglass front for viewing the
interior during the backwashing cycle. Vent pipes were installed on
each housing to permit the filters to operate by gravity for this
study, although each unit can be completely sealed for pressure fil-
ter operation. The interior horizontal cross sectional area of each
filter was 2.25 sq ft, and other pertinent interior dimensions are
shown in Fig. 25. Access to the interior of the housing was provided
by a top mounted, 6-in. diameter, hand hole. At the beginning of
Phase VI, all three filter housings were extended in height to 10 ft
so that deeper media could be used.
Located near the top of each housing was a 1-1/2 by 4-in. inverted
bell mouth which distributed the influent during the course of a fil-
ter run and collected the washwater during the backwash cycle.
Each filter housing was equipped with an underdrain system consisting
of a steel plate false bottom and five General Filter Company media-
retaining strainers (or nozzles). During a normal run the nozzles
collected the filtrate and directed it to the plenum below the false
bottom where the effluent lines carried it to the backwash storage
tank. When cleaning was required, water and air (if used) entered
the plenum and were distributed uniformly to the media through the
nozzles. The media-retaining nozzles had openings of 0.4 mm during
Phases 1 and II. There was evidence of partial clogging of these fine
openings by the end of Phase II. Therefore, nozzles with 1-mm slots
were installed at the beginning of Phase III and graded gravel was
needed for the north and west filters to prevent loss of filter media.
A nozzle clogging incident occurred during Phase III which required
termination of that phase. During rebuilding of the two filters, new
nozzles with 4.5-mm slots were installed. The incident will be de-
scribed later in the results section. Nozzles with 4.5-mm slots
were used in all three filters in Phase VI.
Head loss development in the filters was monitored by seven piezom-
eters attached to taps located vertically along one side of the fil-
ter housing. The taps were located at 4-in. centers starting just
above the underdrain plate. Each tap consisted of a 3-in. length of
1/4-in. copper tubing soldered to 1/4-in. brass fittings which could
be threaded directly into the side of the housing. The interior end
of the tap extended slightly into the media and was covered by a fine
stainless steel mesh. The exterior end of each tap was connected to
a length of 1/4-in. clear plastic tubing which was then mounted to a
piezometer board. Thus, using the false bottom as a datum, the
pressure at several points within the filter media was determined by
95
-------
—
^s
\
-«-
—
•J
•^ —
-^ —
O—-0
1 ff 6 \nT
-*-PIEZOMETER BOARD
UNDERDRAIN PLATE
|10 1/2 in
o o
o
o
STRAINERS
PIEZOMETER TUBES
•PIEZOMETER TAPS
VENT PIPE
4:
»o
in.
•AIR INLET
•FILTRATE OUTLET AND BACKWASH INLET
•DRAIN AND SAMPLING POINT
— 1 ft 6 in.-j
Fig. 25. Details of filter boxes,
96
-------
observing the water level in each of the clear plastic tubes. Dif-
ferent piezometer tap positions were used in the various phases de-
pending upon the depth of media in service.
Surface washer. The west filter was chosen for installation of a
rotating surface washer consisting of an inverted lawn sprinkler
which threaded into a coupling on the back plate of the filter
housing. Water was supplied to the surface washer from a yard hy-
drant supplied by the treatment plant non-potable well water system.
Hydrant line pressure was usually 60 to 70 psig and surface washer
operating pressure generally was from 40 to 45 psig.
After about 30 filter runs in Phase II using the improvised rotary
surface washer, a two-armed model incorporating two nozzles was in-
stalled. The new rotating surface washer more closely resembled
systems in actual use and produced strong, highly directional jets.
The two arms of the surface washer consisted of 1/2-in. diameter
brass nipples, were each 6-in. long, and were positioned 180° apart
in a horizontal plane. One 1/8-in. diameter nozzle (Leopold Corpora-
tion) was placed on a 45° elbow at the end of each arm and positioned
toward the media surface at an angle of 15° to the horizontal.
This system was used essentially as described for the remainder of
Phase II except for a slight shortening of the rotating arms to
better direct the end nozzles to the media. At the beginning of
Phase III, a subsurface washer was added to the west filter, and a
booster pump was added to the surface wash supply line, which in-
creased the operating pressure to about 80 psig. Both nozzles were
positioned down at a 15° angle to the horizontal for the surface
washer while the subsurface washer had one nozzle pointed up and one
down at the same angle. Two 1/2-in. pipes in the filter supported
the washers. At the start of Series IV, both surface washers were
modified. Both washer arms were changed to 1/4-in. nipples approxi-
mately 5 in. in length. A 1/4-in., 90° elbow was attached to the
end of each arm, and a Leopold nozzle was fastened on the end of each
elbow. Both nozzles of the surface washer were pointed down at a
15° angle. The subsurface washer had rubber capped nozzles, one up
and one down, both at a 15° angle. The vertical position of the
washers was changed to coincide more closely with the coal-sand in-
terface for the subsurface washer, and the surface of the coal for
the surface washer.
At the beginning of Phase VI, the surface and subsurface washer were
removed from the west filter.
Filter media. The filter media was obtained from several sources and
in several sizes for the various phases as summarized in Table 8. In
Phase I and II, it was desired that the dual media be as identical
as possible in each of the filters. To achieve this objective the
following precautions were used at the beginning of each phase when
placing new media in the filters. The total volume of a particular
97
-------
media layer required for the desired finished depth plus a 3-in.
skimming depth was determined, and the required number of bags were
poured onto a concrete slab. After the media had been thoroughly
mixed, it was split and placed in each housing. Next the media was
backwashed at a rate of about 15 to 20 gpm/sq ft for 15 rain to allow
the fine particles to work their way to the surface. Finally, with
the backwash rate reduced to the point that the coal was just flu-
idized, the top 3-in. layer was removed from the housing with a si-
phon. The sand layer in Phase I was not skimmed in this fashion.
After the layer had been placed and skimmed, a core sample of the
layer was taken from each filter and subjected to a sieve analysis,
the results of which are presented in Table 8.
A similar procedure was used in Phase III and IV to ensure that the
coal and sand of the dual- and triple-media filters were identical.
The only difference was that the skimming depth of each layer was
1/2 in., in accordance with the suppliers recommendation. The coarse
sand of the south filter was not skimmed in this fashion since it was
too coarse to fluidize with backwash rates that were available, and
also because fluidization was not to be practiced in the routine
backwash operation of that filter. The splitting precautions without
skimming were also used in Phase VI.
Gravel support. No gravel support layers were used in the filters
during Phases I and II, when media retaining underdrain strainers
were used. However, the clogging problems encountered in the strain-
ers led to the use of larger openings in the strainers in Phases III
through VI and necessitated the use of gravel support in some cases.
It was desired to test the stability of a double reverse graded
gravel during air scour to determine if gravel movement could be
avoided, and to compare that behavior with a conventional graded
support gravel. Therefore, at the beginning of Phase III, a double
reverse graded gravel support system was installed in the dual-media
filter and a regular graded gravel system in the mixed-media filter.
Originally, the double reverse graded system was accidentally in-
stalled upside down, and the gravel started to mound immediately.
The gravel was then installed properly in the following manner:
Size (in.) Thickness (in.)
Top 1/2 X 3/4 3
1/4 x 1/2 1.5
1/4 x 1/8 1.5
1/8 x #10 2
1/4 x 1/8 3
Later, when the filter was rebuilt at the beginning of Phase IV, the
underdrain gravel was modified to accommodate strainers with larger
98
-------
Table 8. Filter media details for wastewater filtration pilot studies.
VO
Layer Details Size Information3
Media
Phase Filter type Type
I N and S Dual Coal
Sand
II N, S, Dual Coal
and W
Sand
III N Dual Coal
Sand
W Triple Coal
Sand
Garnet
S Coarse Sand
sand
IV &V NandW Same as Phase III
VI N Coarse Sand
sand
W
S "
Effective Uniformity 90% Specific
Thickness size, coeff. finer, gravity
11. 8a 0.94 1.31 1.4 1.67
12 0.38 1.50 0.73 2.65
12a 0.92 1.52 1.5 1.67
12a 0.38 1.50 0.73 2.65
15a 1.03 1.57 2.03 1.7
9" 0.49 1.41 0.82 2.65
(Same as N filter)
(Same as N filter)
3a 0.27 1.55 0.54 4.2
46 2.0 1.52 3.6
except new media installed and coal depth changed to 17 in.
24)
47) 2.5 1.28 3.7
60)
Supplier
Carbonite Filter Corp.
Del ana, Fa.
Northern Gravel Ca.
Muscatine, la.
Carbonite Filter Corp.
Northern Gravel Co.
Neptune Microfloc Corp.
Corvalis, Or.
Neptune Microfloc Corp.
Neptune Microfloc Corp.
CX Products Corp.
Brady, Tx.
CX Products Corp.
••
^epth and size after placement, stratification by backwashing and skimming fines from the surface of the layer with a siphon.
Coarse sand was not skimmed.
-------
(4.5-mm) slots. The support system used for the rest of the study
was:
Size (in.) Thickness (in.)
Top 3/4 X 1/2 3
1/4 x 1/2 1.5
1/4 x #10 5
1/2 x 1/4 1.5
3/4 x 1/2 1.5
The mixed-media filter employed a normally graded support system.
The original system and the rebuilt one are the following:
Original Rebuilt
Top 3 in. of Coarse Garnet 3 in. of Coarse Garnet
(-12+16 mesh)
2 in. of 1/8 X #10 3 in. of 1/4 x #10
2 in. of 1/4 x 1/8 2 in. of 1/2 x 1/4
2 in. of 1/4 x 1/2 2 in. of 3/4 x 1/2
Bottom 3 in. of 3/4 x 1/2
As will be reported later, the double reverse graded gravel was not
stable in Phases III through V, so a similar but revised support was
used in Phase VI in the north filter. The gradations in that support
were:
Size (in.) Thickness (in.)
Top 1/2 x 3/4 4
1/4 x 1/2 2
1/4 X #10 2
1/4 x 1/2 2
Bottom 1/2 x 3/4 2
Backwash Supply
A centrifugal pump mounted on the backwash tank pumped washwater to
the filters in Phases I and II. The flow rate of the backwash water
was measured by a type of Venturi meter. This meter was fabricated
by replacing a portion of the 1-1/2-in. diameter backwash line with a
section of 1-in. diameter pipe using two reducing fittings. Pressure
taps on the 1-1/2-in. and 1-in. pipes were connected to opposite legs
of a mercury filled manometer. The meter was calibrated volumetri-
cally by measuring the rate at which water rose in the filters for
100
-------
different manometer readings. In Phases III through VI, the filter
influent supply pump was used also for backwashing with unfiltered
secondary effluent. An additional pump discharge connection was made
from the pump to the same Venturi meter described above.
Air Supply
A Speedair belt driven, dual piston compressor supplied compressed
air for the air scour sequence. A pressure switch automatically
started the compressor when the pressure in the storage tank fell
below 40 psig. A Fisher Model 95L pressure reducer, reduced the
storage pressure to an operating line pressure of about 8 psig. The
amount of air flow was regulated by adjusting a globe valve and
reading the percent of full flow through a Brooks rotameter. The
manufacturer's calibration of the rotameter at standard conditions
was accepted, but was corrected to the operating temperature and
pressure of the air supply at the rotameter. Near the end of the
research, an approximate calibration was achieved by noting the time
for a given pressure drop on the air storage tank while the air was
being used at a constant rate and the compressor was shut off. Ap-
propriate calculations assuming the ideal gas law was applicable
verified the manufacturer's calibration.
Samplers
In Phase I, composite samplers were used to collect samples of efflu-
ent from both filters and the erdlator (Surveyor Sampler, N-Con
Systems Co., Inc., New Rochelle, N. Y.). During Phase II, the same
three samplers were used on the three filter effluents. A fourth
sampler was obtained late in Phase II which was then always used on
the filter influent (the Ames plant secondary effluent). Up until
that time, the secondary effluent samples were either grab samples or
composite samples collected manually by the Ames plant operators.
Each sampler consisted of an electric motor driven positive displace-
ment pump actuated by a timing mechanism. The sampler contained two
arms on a timing face, one of which set the pump running time and one
of which could be set to activate the pump from 3 to 20 times per
hour. To ensure that a fresh sample was obtained each time the pump
was running, the waste line on the discharge side of the pump was
1/2-in. diameter while the sample line was 1/4-in. diameter. This
meant that there was a delay before the filtrate was pumped to the
sample container, thus allowing the system to first flush itself
through the waste line. Vents were provided in both the waste and
sample lines to prevent siphoning into or out of the polyethylene
collection containers, which were kept refrigerated in an ordinary
domestic refrigerator to preserve the sample during the compositing
period. The compositing period for each phase was somewhat different
and will be described later in the presentation of the results.
101
-------
To evaluate the effectiveness of the various backwashing techniques,
core samples of the media were taken periodically following the nor-
mal backwash cycle for a particular filter. The core sampler was
constructed of a 4-ft length of 1-1/2-in. diameter copper tubing
fitted with a shear gate at one end. With the gate fully open, the
sampler was lowered to the desired depth in the media, at which time
the gate was closed and the sample removed. Because it is well known
that the majority of solids removal takes place in the upper layers of
a deep bed granular filter, the penetration of the core sampler was
restricted to approximately the top 12 in. of the filter media.
Sometimes, the backwash water was turned on slightly, not enough to
fluidize or expand the bed, but to ease the entrance of the sampler.
After removal of the core, the backwash water was turned on at the
normal rate for about 1.5 min to even out the surface of the media.
Analysis of Samples
The analysis of samples was the same during all phases of the study
except for the media abrasion test. The procedures are summarized
in the following paragraphs.
Filter Media
Sieve analysis. A set of United States standard sieves was used to
determine the media size. Samples were dried and then split with a
riffle type sample splitter until approximately 300 g of media were
obtained. The media samples were hand shaken through the sieve.s
until 1 min of additional hand shaking changed the weight of media
retained on any sieve by less than 1% of the weight on that sieve.
The weights of sieves empty and with media were measured and recorded
to the nearest 0.05 g.
Media density and specific gravity. The density of the media was
determined by a water displacement technique with a 100-ml pycnometer.
The procedure is presented in more detail with illustrative calcula-
tions in a later chapter.
Media abrasion. The media abrasion procedures changed several times
during the research. However, in all cases, the purpose of the test
was the same: to determine the relative amount of dirt or suspended
solids remaining on the media after the normal backwash procedure
for the filter.
The procedure in Phase I was as follows. Samples were reduced in
size by dividing them in half with a riffle type sample splitter.
Then each half was divided twice again so that two samples, each one-
eighth of the original sample, were obtained. The abrasion test was
then performed in duplicate. When samples were taken and split, a
portion of the dirt attached to the media was undoubtedly loosened.
Therefore, as the samples of sand and coal were split, the liquid
102
-------
portions were also split and the dirt present included in the calcu-
lation of dirt removed from the media.
After the sample was split, it was placed in an 800-ml beaker and
stirred at 200 rpm for 10 min with a 2-in. diameter, three-bladed
propeller (Cole Farmer Constant Speed and Torque Control Unit) im-
mersed in the media. After stirring, the dirt was rinsed into a
graduated cylinder. Care was taken so that no sand or coal was
transferred with the dirt. The-media was swirled and rinsed repeat-
edly with distilled water and the dirty rinse water decanted into
the graduate cylinder, until the rinse water was clear. The total
amount of water used to wash the media was recorded. Suspended
solids concentration of the wash water was determined on triplicate
samples according to Standard Methods [117] by filtering an aliquot
through Whatman GF/C glass fiber filter paper. The media samples
were then dried overnight at 103 °C and weighed. The dirt removed
was calculated by the following formula and results expressed as mg
of suspended solids removed per gram of filter media.
(SS of washwater, mg/1)(Volume of washwater. 1)
(Weight of filter media, g)
= mg SS removed
g media
A different abrasion test procedure was used in Phases II and III
through V. The main differences were the deletion of the splitting of
the sample on the riffle type splitter (since that was not considered
very suitable for a sample containing water) and the extension of the
period of abrasion to 30 min (because the media was dirtier in the
direct filtration of secondary effluent). A motor driven support
stand was added to rotate the beaker in the direction opposite that
of the propeller. The other details of the procedure are as follows;
The entire core sample and any water collected with it was put in a
2000-ml beaker and placed on the revolving stand (Driven by a Gerald H.
Keller Co. variable speed direct current motor). The sample was
stirred for 30 min with a 2-in. diameter, three-blade propeller ro-
tating at 200 rpm. The propeller and the stand, revolving in opposite
directions, were positioned so the propeller would traverse through
the media in a circle, abrading all the sample. After mixing, the
supernatant was poured into a 6000-ml flask. The media was rinsed
repeatedly with distilled water and the dirty rinse water decanted
into the 6000-ml flask until the rinse was clear. The total amount
of washwater in the 6000-ml flask was then recorded. The remainder
of the procedure was unchanged from that in Phase I.
The final modification of the abrasion test procedure was adopted at
the beginning of Phase IV due to concern that the 30-min abrasion was
causing excessive abrasion of the coal itself, and not merely of the
103
-------
solids adhering to the coal. The abrasion period was reduced to 5
min. The details were as follows:
As before, the core sample was placed in a 2000-ml beaker, but this
time distilled water was added to bring the level up to an easily
read mark. The sample was then mixed in the same way as previously
mentioned but only for 5 min. The coal or sand was allowed to settle
for 5 sec before a 100-ml sample of the supernatant was pipetted off
and transferred to a 1000-ml volumetric flask. The flask was then
filled to the mark (10:1 dilution), and the suspended solids test
previously mentioned was run on this suspension. If the media was
very dirty, a higher dilution factor was used to speed up the fil-
tering step in the suspended solids test. The media was dried and
weighed as before. The abrasion test result was then figured as
follows:
Abrasion Test _ (Avg SS)(Dilution factor)(Supernatant Vol)
Result (mg/g) ~ (Weight media)
Even though the abrasion test procedure for Phases IV and V was
shorter and less complex, it still provided the data needed to evalu-
ate the relative effectiveness of the backwashing procedures.
Naturally, the absolute values obtained were lower due to the shorter
abrasion period.
Abrasion tests were not conducted during Phase VI.
Water Quality Analyses
During the various experimental phases, extensive testing of filter
influent and effluent samples were performed to see if the backwash-
ing methods being used had any affect on filter performance.
Observation filter runs in which detailed observations were recorded
and composite samples were collected were made at regular intervals.
During these "observation runs," grab samples of filter influent and
effluent were collected for immediate turbidity measurement in the
field. Composite samples were collected over the observation run and
transported to the laboratory for further analysis. Most of the
analyses were conducted there by-personnel of the Analytical Services
Laboratory of the Engineering Research Institute. Suspended solids
analyses were done by project personnel in Phases I, II, and VI, and
by the laboratory personnel during Phases III through V.
A typical composite sample was subjected to several analyses: five-
day biochemical oxygen demand (BOD), soluble BOD, suspended solids
(SS), total organic carbon (TOC), soluble organic carbon (SOC), and
ammonia (NH3). Periodic composite samples were also analyzed for
the following: nitrite (N02), nitrate (N03), total Kjeldahl nitrogen,
total phosphate/ and orthophosphate. Soluble BOD and SOC were not
104
-------
analyzed in Phases I and II. The analytical procedures were as
follows.
Suspended solids. Suspended solids tests were conducted on filter
influent samples, filter effluent samples, and abrasion test wash-
water aliquots in accordance with Standard Methods [117]. Disposable
aluminum weighing tins and Whatman GF/C glass fiber filtering discs
were placed in a 103 °C oven for 15 min to drive off moisture and
then cooled and weighed to 0.1-mg precision. The filter discs were
then placed on a Millipore filter holder which was connected to a
Millipore vacuum pump. Under a steady vacuum of 15 in. of mercury,
the following sample amounts were filtered through the discs;
1. Filter influent, 100 ml or 250 ml depending upon the level of SS
2. Filter effluent, 500 ml
3. Abrasion test washwater, 25 ml
After filtration had been completed, the discs and weighing tins were
returned to the oven and dried overnight at 103 °C. Finally, the
discs and tins were cooled and reweighed, and the subsequent gain in
weight was divided by the volume filtered to yield the suspended
solids concentration in mg/1.
Turbidity. Turbidity measurements were made using a Hach Model 2100
turbidimeter during Phase I and a Hach Model 2100 A during the re-
mainder of the studies. Both turbidimeters have four ranges gradu-
ated in Formazin turbidity units (FTU). Formazin standards provided
by the manufacturer were used to properly calibrate the instrument in
each range. For the purpose of this study, only the 0 to 1, 0 to 10,
and 0 to 100 FTU ranges were required. Glass cells were rinsed with
distilled water before each use, and care was taken to dry the outside
of each cell before inserting it into the instrument.
TOG and SOC. Total organic carbon (TOG) and soluble organic carbon
(SOC) determinations were conducted on a Beckman Model 1R315 carbon
analyzer using the tentative procedure outlined in Standard Methods
[117]. Fifty-milliliter portions from the filter influent and efflu-
ent samples were transferred to small polyethylene bottles, treated
with several drops of concentrated 112804, and then refrigerated until
analysis. The depressed pH, as well as the refrigeration, retarded
bacterial decomposition of the organic carbon present.
At the time of analysis, the containers were purged with nitrogen to
eliminate carbonate and bicarbonate interferences (in conjunction
with the acidification step). Samples of 25 p.1 size were analyzed in
triplicate by comparison with acetic acid standards. Soluble organic
carbon samples were filtered through Whatman GF/C glass fiber paper
prior to analysis. The carbon analyzer was down for repairs for ex-
tended periods during the course of these measurements and the
105
-------
acidified samples were held under refrigeration for long periods
while repairs were made. This is a questionable procedure and thus
the TOG and SOC results reported herein are not considered to be of
uniform reliability.
BOD. Determinations of BOD and soluble BOD were conducted using a
dilution technique as outlined in Standard Methods [117]. A con-
tainer of primary effluent from the Ames Water Pollution Control
Plant was supplied with each new set of samples for use in seeding
the dilutions. Because of a misunderstanding between the laboratory
and the research personnel, nitrification was not suppressed in the
BOD analyses in Phases I and II. Nitrification was suppressed there-
after using "N-serve" from Hach Chemical Company. Soluble BOD sam-
ples were filtered through Whatman GF/C glass filter paper prior to
analysis.
Other analyses. Procedures for the remaining three analyses were
automated using a Technicon Autoanalyzer. Total Kjeldahl nitrogen,
which included both organic and ammonia nitrogen, was conducted as
described in Standard Methods [117]. The test differs from the clas-
sic Kjeldahl determination of organic nitrogen only in that ammonia
is not removed from the sample before digestion and distillation.
For use on the autoanalyzer a colorimetric technique using the tenta-
tive phenate method was employed for the final nitrogen measuring
step of the procedure.
Both phosphate determinations were also conducted in accordance with
procedures described in Standard Methods [117], Orthophosphates were
measured using the tentative ascorbic acid procedure, which involved
a colorimetric determination preceded by filtration of the sample.
The digestion step of the total phosphate analysis was modified by
using both perchloric and sulfuric acids in the presence of a sele-
nious acid catalyst to convert the various phosphate forms to ortho-
phosphate. A colorimetric determination using the vanadomolybdo-
phosphoric acid method was employed for final phosphate measurement
in the autoanalyzer.
Operation and Results - Phase I
Dual-Media Filtration of Alum-Treated
Secondary Effluent
Phase I was a comparison of two backwashing methods during the fil-
tration of alum-treated secondary effluent. The alum treatment was
to reduce the phosphorous level of the secondary effluent to about
1 mg/1 phosphorous. Both filters were equipped with dual media.
One was washed by water fluidization only. The other was washed with
air scour followed by water fluidization, hereinafter referred to as
air-scour auxiliary.
106
-------
Operation - Phase I
A filter run started with a backwashed filter and ended when the fil-
ter had operated for a cycle and had then been backwashed and was
ready for service again. Filter runs were numbered consecutively
starting with number 1. Since both filters were backwashed at the
same time, and placed in service simultaneously, a particular run
number refers to both filters.
The first filter run of Phase I started on May 17, 1973, and the last
one using alum-treated water ended on August 21. The filters were
operated continuously during this period except during equipment
failures and when major maintenance work was needed. The filters did
not operate for a total of about 1-1/2 weeks.
Solids contact unit. Commercial grade, granular alum (approximately
Al2(S04>3 • 14 H20, approximate molecular weight 600) was mixed at the
rate of 37.8 g/1 of water. This solution was fed to the erdlator to
achieve an alum dosage of 200 mg/1 when the erdlator was operating
at 10 gpm. For runs 1 through 45, the first 7-1/2 weeks of operations
the erdlator operated at approximately 10 gpm, but for runs 46 through
57, the next 3-1/2 weeks, it was operated at 12 gpm and therefore the
alum dosage was only 167 mg/1. For the last three weeks, after run
57, the erdlator was operated at approximately 15 gpm, and the alum
feed was increased so that the dosage was 200 mg/1. The detention
time in the upflow portion of the erdlator was 53 min at 10 gpm and
35 min at 15 gpm. The detention time in the center mixing column was
11 min at 10 gpm and 7 min at 15 gpm. The surface loading was 0.37
gpm/sq ft at 10 gpm and 0.56 gpm/sq ft at 15 gpm based on the area
outside of the center column. Sludge was wasted at a rate of about
1 gpm.
The use of alum was discontinued after run 73, 13-1/2 weeks of opera-
tion. For another week, or eight filter runs, the trickling filter
effluent flowed through the erdlator without any chemical treatment.
Filters. The filtration rates averaged 2 gpm/sq ft for runs 1 through
45 and for runs 59 and on. The filters were operated at 1.6 gpm/sq ft
for runs 46 through 57.
After about 12 weeks, run number 65 for the south filter and run
number 66 for the north filter, a layer of fine sand was skimmed from
the top of the coal. This layer was removed to see what effect it
was having on the development of head loss in the filters.
Backwash. A combination of time and head loss was used to determine
when the filters were backwashed. If one of the filters approached
the maximum available head loss of 7 ft, both filters were back-
washed. Frequently the filters were backwashed after 24 hr of ser-
vice even though neither filter was near the maximum head loss.
107
-------
For runs 1 through 26, air scour was not used on either the north or
the south filter. The backwash for run 27 consisted first of the
normal water fluidization backwash. Samples of filter media were
taken for the abrasion tests. Then the south filter was air scoured.
Following the air scour, a second normal water backwash was conducted
and waste washwater samples were collected at various intervals of
time during the backwash. Another sample of media from the south
filter was then taken. From run 27 to the end of Phase I (run
78), the north filter received water backwash only, while the south
filter received air scour followed by water backwash.
After run 78, the very end of Phase I, a series of special back-
washes was used in an effort to restore the filter media to a clean
condition. After the normal backwash, the following procedures were
used on both filters in the order given.
1. A normal air scour followed by a normal water backwash.
2. Simultaneous air and water backwash (as the water rose from the
media to near the overflow) followed by a normal water backwash.
3. Step 1 repeated.
4. Step 2 repeated.
5. Step 2 repeated.
During each step samples of waste backwash water were composited for
later analysis.
The normal water backwash procedure at any time was the same for both
filters. For runs 1 through 28 a constant rate of 20 gpm/sq ft for
5 min was used. For runs 29 through 63, both filters were backwashed
for 5 min at rates that expanded the beds 38 to 40%. The backwash
rates to achieve these expansions changed from one backwash to another
and varied between the two filters. Rates from 18 to 23.8 gpm/sq ft
were required. After run 63, a constant backwash rate of 18.6 gpm/
sq ft was adopted, again for a 5 min duration. Filter bed expansion
then varied from 25 to 38%. When air scour was used, it consisted of
9 scfm [scfm (standard cubic foot per minute at 14.7 psi and 70 °F)]
or 4 scfm/sq ft of air for 5 min.
i
The backwash procedure included the following details. The flow to
both filters was stopped. Then effluent valves were closed, the
valves to the composite samplers were closed, and the backwash waste
drain valve was opened. If air scour was to be used, the effluent
valve on that filter was left open until the water level had drained
at least 1 ft below the backwash water outlet. This was to prevent
loss of media due to the violent agitation caused by the air scour.
After the air scour was terminated, the water wash was started at a
very low rate until all of the trapped air had escaped. This usually
108
-------
took 1 to 2 min. The filter was then backwashed with water. The
backwash water inlet valve was gradually opened over a 20-sec period.
The water backwash continued for 5 min, and then the inlet was closed
over a 10-sec period. It was closed in as nearly as possible the
same manner each time.
After both filters were backwashed, flow was started through them
simultaneously. Piezometer readings and turbidity measurements were
taken about 15 min later. This allowed time for head loss and flow
to reach an equilibrium with each other. Also any backwash water
left in the filter was flushed out by then.
Because the primary purpose of this study was to evaluate backwashing
techniques, careful notes were recorded during the backwashing of
each filter after every run. Bed expansion, backwash rate, air-scour
rate (where applicable), surface washer line pressure, water tempera-
ture, and media heights before and after backwashing were noted for
each filter. Comments concerning the condition of the bed before,
during, and after cleaning were also recorded. The comments were
particularly important because they contained information with respect
to surface cake formation, cracks in the media, mud ball or agglom-
erate formations, and other unusual qualitative observations.
Maintenance. Periodic maintenance was required in various parts of
the pilot plant to ensure proper operation and collection of meaning-
ful data. Because the sewage being filtered was a rich source of
nutrients and the pilot plant was constantly exposed to the sun,
algal growths flourished if not properly controlled. To prevent
algal growths from occurring in piezometer tubes and composite sam-
pler lines, a 10 to 20-ml dose of 57, NaOCl (household bleach) was
introduced to each line on a once weekly basis. After allowing a
15 to 20 min contact time, the lines were drained and thoroughly
flushed.
Because the filter housings had plexiglass fronts, some means of con-
trolling algal growths within the filter housing was necessary. To
retard algal accumulations, therefore, removable 1/4-in. plywood
covers were constructed to fit over the glass front of each filter
housing and prevent the passage of sunlight.
Although the covers were effective in preventing algal growths within
the filters, they could not prevent the accumulation of biological
solids on the inside face of the plexiglass fronts. If left unat-
tended, such growths became thick enough to prevent one from viewing
the media. It was necessary to scrape the growths from the glass at
least once a week using a rubber squeegee lowered into the 6-in.
handhole in the top of each filter housing.
A particularly troublesome and persistent problem encountered was the
clogging of the piezometer taps. This was probably caused by an ac-
cumulation of biological solids on the fine steel mesh soldered over
109
-------
the internal end of each 1/4-in. copper tube. It was found that one
could generally free the tap by fitting a filled water bottle, with
stem removed, over the end of the tap and applying hand pressure. If
this was unsuccessful, a length of 1/4-in. tubing was fitted to the
tap and connected to the well hydrant.
Biological solids coated the surface of the influent rotameter tube
and float. Solids accumulating on the float increased its drag,
causing it to rise higher in the tube and give erroneously high read-
ings. Daily cleaning was required to enable the meter to function
properly, and this proved particularly inconvenient because of the
construction of the rotameter. Therefore, at the end of Phase II the
rotameter was removed from the flow scheme, and flows were monitored
only after the flow splitter, as shown in Fig. 23, for all remaining
phases of the study.
Results_^ Phase I
Various parameters were used to compare the backwash effectiveness
between the two filters. The most direct measurement was the abra-
sion test, which revealed the relative amount of solids left on the
media after a backwash. Visual observations of the filter media
through the transparent filter wall also directly indicated the ef-
fectiveness of the backwash. Indirect measures included the initial
head loss of the filter media after backwash, the head loss develop-
ment patterns and the filtrate quality. Excessively dirty filter
media or the presence of agglomerates and mud balls should result in
higher initial head losses, more rapid head loss development, shorter
filter runs and potentially poorer filtrate quality. In addition,
dirty filter media will alter the backwash rate required to achieve
a particular degree of expansion. Coatings should reduce the effec-
tive density of the grains and reduce the required wash rate. Ag-
glomerates will cause erratic behavior because portions of the bed
may not fluidize above the agglomerates, and because part of the
media is held in the agglomerate and is not free to be fluidized.
The observations related to these measures of backwash effectiveness
are presented in the following paragraphs.
Visual observations. Small mud balls that appeared to be mixtures of
sand, coal, and alum floe began forming during the first few filter
runs. The mud balls started out on the surface of the coal. By the
end of the sixth filter run and backwash, the mud balls had grown
from 1/16 in. in diameter until 1/4 to 3/4-in. balls covered over 907»
of the filter bed surface. Before run 10 some of the mud balls were
up to 1-1/2 in. in diameter and had sunk into the coal layer. By
the end of run 14 almost all of the mud balls had sunk into the coal
and very few remained on the surface. A layer of fine sand had
worked its way to the surface of the coal, presumably due to reduc-
tion of hydraulic subsiding velocity of the sand by the coatings
which had developed on the sand grains.
110
-------
Just before the air scour was used (run 27), both filters had large
numbers of mud balls up to 3 in. in diameter at the sand-coal inter-
face. The first use of air scour dramatically reduced the number of
mud balls in the south filter, and the largest one left was 3/4 in.
in diameter instead of 3 in. After the backwash for run 29, three
backwashes after the first air scour, only one mud ball was observed
in the south filter, while numerous mud balls up to 2 in. in diameter
remained in the north filter. The layer of fine sand had also dis-
appeared from the surface of the south filter.
Starting with the backwash for run 51, a layer of fine sand was once
again noticed over the coal in the south filter. Also, a portion of
the media near the plexiglass front did not easily fluidize in the
water wash. However, only a few small mud balls were seen in the
coal from then until the end of Phase I.
The north filter developed a heavy layer of fine sand over the coal.
As the head loss increased during filter runs, large cracks up to
3/4-in. wide and several inches deep formed at the media surface.
While most of the cracks extended along the filter walls, some smaller
ones appeared toward the center of the media surface. The south fil-
ter had developed some cracks at the media surface, but they were
much smaller and occurred less frequently than those in the north
filter. After the fine sand was skimmed from above the coal, the
cracks at the media surface stopped forming. However, another layer
of fine sand soon worked its way to the top of the coal and the cracks
at the surface of the north filter reappeared.
The sand used in the filters was somewhat small and undoubtedly con-
tributed to the difficulties encountered in keeping the filter media
clean and to the migration of fine sand to the surface. The use of
a larger sand would have been better. Also, it should be recalled
that for this experimental Phase I, the sand was not skimmed to re-
move fines, which may have contributed to the sand migration problem.
Abrasion tests. Figure 26 shows the amount of suspended solids re-
leased from the core samples of filter media as indicated by the
abrasion test results. Two values are shown for the south filter at
run 27 because the normal water fluidization backwash was followed
by an air-scour assisted backwash as previously described. A core
sample was taken after each backwash. The abrasion test results
started out at about 4 mg/g and increased to 22 mg/g. The use of
air scour greatly reduced the abrasion test results in the south
filter where values dropped from over 11 mg/g at run 27 to less than
2 mg/g at run 34. After run 45, the abrasion test results for the
south filter started to increase. The lines shown in Fig. 26 are
least square fits using all of the values for the north filter and
the values from runs 45 through 72 for the south filter.
The benefit of the air-scour auxiliary in cleaning the filter is
clearly evident from the abrasion test data. However, the gradual
111
-------
20
|
O
CO
O
_j
O
co
10
Fig. 26.
o NORTH FILTER - WATER FLUIDIZATION ONLY
~
SOUTH FILTER -AIR-SCOUR AUXILIARY o
oo o
10 20 30 40 50
FILTER RUNS
60
70
I
1
i rn r r i n i i i r
11132022261 10 18 25 1 7 15 20
JUNE JULY AUGUST
MONTH AND DATE, 1973 (NONLINEAR SCALE)
i
26
MAY
80
Abrasion test when filtering secondary effluent treated
with alum for phosphorus reduction in Phase I.
deterioration evident toward the end of Phase I indicates that even
the air-scour auxiliary as used in this phase was not totally effec-
tive in keeping the media clean.
Initial head loss. As mud balls form, the volume of void space in
the filter is reduced. This results in greater water velocities be-
tween the grains of filter media and therefore greater initial head
losses. Therefore, an increase in initial head loss would indicate
the accumulation of mud balls in a filter.
At the beginning of Phase I when both filters were clean, the initial
head loss was 0.44 ft for both filters. Table 9 summarizes the means
and standard deviations of the initial head losses for various peri-
ods of filter operation in Phase I. During the 3-1/2 weeks at the
beginning of the project, the values of initial head loss progressive-
ly increased. In the week immediately preceding the start of air
scour in the south filter (filter runs 22 through 27), the average
initial head loss was 0.82 ft for the north filter and 0.80 ft for
the south filter. The difference between the filters was negligible.
112
-------
Table 9. Initial filter head losses during various portions of the
study (ft of water).
N filter, water S filter, air-scour
Runs fluidization only auxiliary
22-27
28-45
46-57a
58-65
68-73
Mean
Std. dev.
Mean
Std. dev.
Mean
Std. dev.
Mean
Std. dev.
Mean
Std. dev.
0.82
0.096
0.64
0.141
0.55
0.057
0.72
0.144
0.54
0.062
0.80
0.105
0.44
0.036
0.37
0.061
0.52
0.114
0.41
0.029
n
During this period the flow rate was only 1.6 gpm/sq ft instead of
2 gpm/sq ft as for the other periods.
After the south filter was backwashed using air as auxiliary agita-
tion, the initial head loss on the following filter run dropped
nearly to that observed at the very beginning of the project. In
the next 3-1/2 weeks, the initial head loss for runs 28 through 45
averaged 0.64 ft for the north filter and 0.44 ft for the south fil-
ter. From the eleventh week of operation through the twelfth week
of operation the initial head losses for both filters increased.
For these eight runs, 58 through 65, the north filter averaged 0.72
ft and the south filter averaged 0.52 ft. By this time a quantity
of fine sand had worked its way to the top of the coal layer. It was
felt that this fine sand was adversely affecting the head loss char-
acteristics of the filters, therefore, it was skimmed out of the fil-
ters. Not surprisingly, the initial head losses for both filters
decreased. For the next six runs, 68 through 73, the north filter
averaged 0.54 ft and the south filter averaged 0.41 ft of initial
head loss. The initial head losses in the filters after skimming out
the fine sand should not necessarily,be the same as the initial head
losses at the beginning of the experimental work because the media
characteristics were changed somewhat by the skimming operation.
The initial head loss data above supports the visual and abrasion
test observations. The benefit of the air-scour auxiliary compared
to water fluidization only is clearly demonstrated. However, a
slight deterioration is evident in the south filter, even with the
air-scour auxiliary.
Head loss development. Before the use of air scour was started, the
head loss in each filter was similar with respect to time and to
113
-------
depth in the filter. Figure 27 shows typical head loss versus time
curves for various intervals of media depth, measured from zero depth
at the top surface of the media.
After air scour was begun on the south filter, it developed head loss
at a slower rate than the north filter. Figure 28 is a typical exam-
ple. During the two weeks following the beginning of the air scour
(runs 28 through 44), the filters were generally backwashed before
they reached 4 ft of head loss. Later they were operated to higher
head losses. Starting with run 45, terminal head losses for both
filters were about equal, with the south filter frequently having a
slightly greater head loss. This was particularly true when higher
head losses were reached. Figures 29 and 30 are typical examples.
After run 65, 8-1/2 weeks after air scour was started, the layer of
fine sand that had worked its Way into the coal was removed from the
filters. Figure 31 shows a typical curve of head loss versus time
after the fine sand was removed.
The effect of the air-scour auxiliary on head loss patterns is
clouded by the variable extent of surface cracks in the filters. If
surface cracks were absent and mud balls were present, the dirtiest
filter should have the highest rate of head loss development. A
higher rate of head loss development for the dirtier north filter was
not consistently observed in this Phase. This anomaly is explained
as follows.
During the first four weeks after air scour was started, the filter
runs were generally terminated when the head loss reached only 3 or
4 ft. It was then observed that, at greater head losses, the head
loss in the south filter would approach and even surpass that in the
north filter. This was attributed to the fact that the north filter
experienced more extensive surface cracking during the filter runs.
These cracks allowed the surface to be bypassed and solids removal to
take place in deeper layers of the filter. The head losses shown in
Figures 29 and 30 clearly indicate that most of the removal in the
south filter took place in the top 4 in. of the media, while in the
north filter the removal took place over the upper 8 in. This indi-
cates that the greater head loss in the south filter on some occa-
sions was not due to accumulated dirt, but was rather due to the fact
that the dirtier media in the north filter resulted in more extensive
surface crack formation.
Backwash rates required. For runs 31 through 63, both filters were
backwashed to a constant expansion of 38 to 40% and the backwash rate
needed to achieve this was recorded. For runs 31 through 49 the
north filter required an average of 19.4 gpm/sq ft, and the south
filter required an average of 22.2 gpm/sq ft. For runs 51 through 61
the required backwash rates were 19.7 gpm/sq ft for the north filter
and 20.0 gpm/sq ft for the south filter. Table 10 lists the means
and standard deviations of the backwash rates of the filters for the
114
-------
NORTH FILTER
(WATER FLLUDIZATION ONLY) 0 to 24 in.
GO
«/»
B
a
&
X
4 to 24 in
8 to 24 in.
12 to 24 in.
16 to 24 in.
I I
SOUTH FILTER
(WATER FLUIDIZATION ONLYY
0 to 24 in.
I I
4 to 24 in.
05 10 15 20
TIME, hrs
25 0 5 10 15 20
TIME, hrs
25
Fig. 27. Head loss vs time at various media depths, run 27.
-------
NORTH FILTER
WATER FLUIDIZATION ONLY)
3_
4>
oo
to
Q
0 to 24 in.
I
I
4 to 24 in.
8 to 24 in.
12 to 24 in.
I
SOUTH FILTER
(AIR-SCOUR AUXILIARY)
0 to 24 in.
4 to 24 in.
8 to 24 in.
012 to 24 in.
I
10 15 20
TIME, hrs
25
10 15 20
TIME, hrs
25
Fig. 28. Head loss vs time at various media depths, run 42.
-------
g 4
in o
«/> o
O
NORTH FILTER (WATER FLUIDIZATION ONLY)
0 to 24 in.
4 to 24 in.
8 to 24 in.
I I I I I
SOUTH FILTER (AIR-SCOUR AUXILIARY)
0 to 24 in.
4 to 24 i
in.
8 to 24 in.
I
05 10 15 20 25 30 35 40
TIME, hrs
0 5 10
15 20 25 30
TIME, hrs
35 40
Fig. 29. Head loss vs time at various media depths, run 59.
-------
oo
NORTH FILTER (WATER FLUIDIZATION ONLY) SOUTH FILTER (AIR-SCOUR AUXILIARY)
.0 to 24 in.
,0 to 24 in.
4 to 24 in.
8 to 24 in.
12 to 24 in.
10 15 20
TIME, hrs
25
10 15
TIME, hrs.
20
25
Fig. 30. Head loss vs time at various media depths, run 63.
-------
4)
1
t/J
uo
2
Q
NORTH FILTER (WATER FLUIDIZATION ONLY) , SOUTH FILTER
(AIR-SCOUR AUXILIARY)
0 to 24 in.
5-
4 to 24 in.
8 to 24 in.
12 to 24 in.
to 24 in.
4 to 24 in.
8 to 24 in.
12 to 24 in.
0
10 15 20 25 30 35 40 0
TIME, hrs
10 15 20 25 30
TIME, hrs
35 40
Fig. 31. Head loss vs time at various media depths, run 71.
-------
Table 10. Backwash rates required to achieve 38 to 40% bed
expansion (gpm/sq ft).
N filter, water S filter, air-scour
Runs fluidization only auxiliary
31-49
50-61
Mean
Std. dev.
Mean
Std. dev.
19.4
0.88
19.7
1.85
22.2
0.80
20.0
1.07
two periods. The backwash temperature averaged 22.9 °C for runs 31
through 49 and 23.1 °C for runs 50 through 61.
The data presented support the other observations concerning the
cleanliness of the filter media. Shortly after initiation of air
scour, the media of the south filter was cleaner and thus had a
higher average density than the north filter. The higher density re-
quired a higher backwash rate to achieve a given expansion. Later,
as coatings developed on the air-scoured media, the average density
decreased and the required backwash rate decreased. Thus, the re-
sults of Table 10 are in harmony with the abrasion test results and
initial head loss results. All three tests demonstrate the superi-
ority of air-scour auxiliary over water fluidization alone, but all
three also indicate some deterioration of the air-scoured filter
towards the end of Phase I.
Water quality. The means and standard deviations of the various
water quality parameters are given in Tables 11 and 12. Results for
each parameter were divided into two periods. The first period
(Table 11) covers the initial 3-1/2 weeks of filter operation when
both filters were backwashed by water fluidization only. The second
period (Table 12) covers the remaining 8-1/2 weeks of filter opera-
tion when the south filter received air-scour auxiliary. Except for
turbidity, the values for the solids contact unit influent may not
be used to calculate actual removals through the treatment process.
Samples of influent to the solids contact unit were grab samples,
while the other samples were composited over approximately one day.
All of the turbidity measurements were made on grab samples taken at
approximately the same times for each point in the treatment process.
From the data on filter effluent quality, no apparent differences
are evident between the two filters. One would expect the dirtier
filter to yield poorer filtrate due to surface cracks permitting
deeper penetration of solids, and due to higher interstitial veloci-
ties caused by mud balls and agglomerates. The absence of detriment
to filtrate quality in this work must be attributed to the fine grain
size of the media, the low filtration rates and the low terminal head
120
-------
Table 11. Results of analyses during alum treatment (Phase I) for samples from May 17 to July 11,
1973, when both filters were washed by water fluidization only. (All results from com-
posite samples except solids contact influent.)
Solids contact
influent
Avg suspended solids (mg/1)
CTa (N - 16)
Avg turbidity (FTU)b
aa(N = 55)
Avg BOD5C (mg/1)
<7a(N = 13)
Avg TOG (mg/1)
aa(N = 15)
Avg total P04 (mg/1)
aa(N = 8)
Avg ortho P04 (mg/1)
'
-------
r-o
Table 12. Results of analyses during alum treatment series (Phase I) for samples from July 11 to
August 20, 1973, when air scour was being used on the south filter. (All results from com-
posite samples except solids contact influent.)
Solids contact
influent
Avg suspended solids (mg/1)
aa (N = 4!)
Avg turbidity (FTU)b
a (N = 207)
Avg BOD5C (mg/1)
CT (N = 14)
Avg TOG (mg/1)
-------
losses. Had one or more of these filtration variables been increased,
the detriment may have occurred.
Statistical comparisons to determine differences in the effect of
backwashing on filter performance efficiency betveen the filters are
not reported since a proper statistical base did not exist in this
experimental study. Since the filter media was not returned to
exactly the same condition following each backwash, the individual
runs on a filter cannot be considered independent, which is a funda-
mental assumption in statistical theory. Because it was desired to
determine the cumulative effects of the various backwashing methods
over a period of continuous operation of the filters, it was not
practical to thoroughly restore the media after each filtet run, and
it was equally impractical to replace it. The minimum effort alter-
native for the application of statistical comparisons would be to
run at least two separate eight-week studies using identical media,
switching the backwashing techniques on each housing, and replacing
or completely restoring the media at the end of each eight-week
series. Time and expense considerations precluded the additional
study, so no base exists for making statistical inferences.
Clean up operations at the end of phase I. The results of the sus-
pended solids concentrations versus quantity of backwash water for
the first backwash following the use of air scour (run 27) are shown
in Fig. 32. Since a normal backwash using only water fluidization
immediately preceded this backwash, the area under the curve repre-
sents the additional suspended solids released by the first air scour.
The suspended solids removed by the air scour was 109.6 g per sq ft
of filter area. If the suspended solids are assumed to come from a
12-in. layer of coal and a 1-in. thick layer of sand, 4.13 mg sus-
pended solids were removed per gram of filter media.
The amounts of suspended solids released from the filter media in
the final cleanup operations after run 78 are shown in Table 13.
Based on the same amount of filter media as was assumed before, 13.3
and 3.2 mg of suspended solids were released per gram of media for
the north and south filters respectively. All mud balls were broken
up during the cleanup operation. The series of cleanup steps was
continued until the media appeared to be in new condition and the
last step appeared to release very few additional solids. The sub-
stantial difference in total suspended solids released clearly shows
that the south filter which had routinely been washed with air-scour
auxiliary was in much cleaner condition than the north filter. It
should be borne in mind that this final cleanup operation came after
a week of filtering uncoagulated trickling filter effluent at the end
of the alum treatment series.
Summary and Conclusions - Phase I
The objectives of this phase were to evaluate the effectiveness of
filter cleaning by water fluidization backwash alone as compared to
123
-------
CO
loooh
800h
60°
10
co
9
_i
O
CO
Q
z
UJ
a.
CO
:D
CO
400K
200h
0 10
Fig. 32.
20 30 40 50 60 70 80
VOLUME OF BACKWASH WATER USED, gal./sq ft
90
100
Suspended solids concentration of backwash water vs quantity
of backwash water used, run 27, second backwash of the south
filter immediately following the first application of air
scour.
-------
Table 13. Suspended solids released from filters in special back-
washes after run 78.
g of SS removed/sq ft
of filter area
Backwash used
North filter
South filter
Air followed by water
Air and water combination
Air followed by water
Air and water combination
Air and water combination
Total
87
144
57.5
47
28. 2b
393.7
37.3
28.2
a
14.4
8.5
88.4
Step omitted.
Backwash conducted but no suspended solids value obtained therefore,
the value indicated is based on the comparison of turbidity with
the other samples.
air scour followed by water fluidization backwash, and to compare the
performance and operation of filters backwashed by these methods when
filtering wastewater. For this study two dual-media filters were
used following alum treatment of domestic wastewater for phosphate
removal. One filter used air scour followed by water backwash, while
the other used water fluidization backwash alone.
The media in the filter backwashed with water only became very dirty,
while the media in the filter with air scour during backwash was much
cleaner. However, toward the end of the project, the media in the
air-scoured filter also showed some deterioration. Few mud balls
were observed in the air-scoured filter while numerous, large mud
balls were observed in the other filter. When operated at only 3 to
4 ft of head loss the filter backwashed with water only had greater
head losses than the filter with air scour. When operated to greater
head losses this was not true due to the formation of surface cracks
in the filter receiving water backwash alone. Little difference in
effluent quality was observed between the two filters.
The study of the filtration of domestic wastewater which had been
subjected to secondary treatment and subsequent alum treatment for
phosphate reduction leads to the following conclusions concerning
backwashing:
1. Filter media were kept cleaner by air scour than by water fluid-
ization backwash alone as evidenced by the results of visual
125
-------
observations, media abrasion tests, initial head losses, back-
wash flow requirements, and the final cleanup operation.
2. Some minor deterioration of the condition of the media of the
air-scoured filter was evident toward the end of this experi-
mental phase as evidenced by the results of the media abrasion
test, initial head losses and backwash flow requirements.
3. Air scour prevented the formation of mud balls, while water
backwash alone did not.
4. The two methods of cleaning resulted in little difference in
effluent quality. The use of a coarser filter media, a higher
filtration rate, or higher terminal head loss may have demon-
strated that a cleaner media would give better filtrate quality.
5. The sand used in the experimental filters was smaller than de-
sirable (0.38-mm effective size) and was not skimmed of fines
on placement. This contributed to backwashing difficulty and
sand migration to the surface of the coal of the dual-media
filters.
Operation and Results - Phase II
Dual-Media Filtration of Secondary Effluent
Phase II was a comparison of three methods of backwashing during the
direct filtration of secondary effluent at the Ames, Iowa, trickling
filter plant. All three filters were equipped with dual media. One
was washed by water fluidization only. The second was washed by air
scour followed by water fluidization, hereinafter referred to as air-
scour auxiliary. The third was washed with a rotary surface wash
auxiliary which operated before and during the water fluidization
backwash.
Prior to commencing Phase II, the west filter with surface wash
auxiliary was installed and new filter media was installed in all
three filters. This was done to ensure that Phase II would not be
influenced by any carry-over effects from Phase I.
Operation - Phase II
The secondary effluent of the Ames trickling filter plant was pumped
directly to the influent splitter box and thus became the filter in-
fluent as illustrated in Fig. 23.
Filters. Because it was desired to observe the effects of continuous
operation on filter efficiency, the pilot plant was operated without
major interruptions from August 28 to October 30, 1973. The entire
series of runs was designated Phase II, with individual runs being
numbered consecutively from 1 to 64. Normal procedure dictated one
126
-------
observation run per week at which time the head loss development,
influent and effluent turbidities, and flow temperature were moni-
tored, and composite samples were collected. During the remaining
runs of a given week, only initial readings immediately following
backwashing were recorded. The backwashing observations outlined
in Phase I were recorded for each backwash.
When the series was begun it was hoped to be able to operate the
filters at a loading rate of 2 gpm/sq ft. Piping losses and solids
accumulations in the lines gradually reduced this figure as operating
time increased to 1.6 gpm/sq ft near the end of Phase II. Even
though the rate declined gradually, the flow split to the three fil-
ters was equal, so that valid comparisons between the filters are
possible.
Backwashing. Ordinarily one would backwash filters when the head
loss had reached some maximum permissible value or on a time cycle
prior to attainment of the maximum permissible head loss. Neither of
these procedures was deemed practical during Phase II because the
runs were quite short and too much operational time would be re-
quired. Typically, the filters required approximately 10 to 12 hr to
attain maximum head loss (point of incipient bypass in the splitter
box). Furthermore, divergence in run lengths was expected to become
more pronounced as operation time increased. Therefore, a schedule
was adopted whereby the filters were backwashed and placed in service
each morning as a group. This meant that by late evening the filters
had begun to bypass through the lower half of the splitter box and
that by the next morning nearly all of the influent was being by-
passed.
Although each of the filters was backwashed using a different tech-
nique, all three filters required similar initial preparations as
described previously in the Phase I operational discussion.
A nominal water fluidization backwash rate of 20 gpm/sq ft was se-
lected for use on all three filters, this value being an upper limit
for normal backwashing of filters used in water treatment. The rate
was easily obtained during the first 20 runs, but after that the
attainable rate decreased, presumably due to solids accumulations in
the backwash lines and partially clogged underdrain strainers. From
run 21 on the rates varied from 18.5 to 20 gpm/sq ft, with a pre-
ponderance of values in the 19.0 to 19.5 gpm/sq ft range. Attempts
to clean the distribution nozzles were not made because to do so
would have required all media to be removed from the filter housing.
Backwashing of the north filter consisted of a water fluidization
(only) backwash at the 20 gpm/sq ft rate for a duration of 5 min and
was not preceded by air scour or surface wash. Bed expansion
during fluidization varied from a low of 8.3% to a high of 37.5%,
with an average value-of 25.6%. The slightly reduced backwash rates
had no noticeable influence on bed expansion, which appeared to be
127
-------
more a function of bed condition, i.e., number of mud balls and
agglomerates.
The south filter was subjected to air scouring at a rate of 3.72
scfm/sq ft for 5 min prior to the water (only) backwash. Before the
air line valve to the filter was opened, the filter was drained down
through the effluent line until the water level was 6 in. below the
backwash water collection trough. This prevented the loss of media
out the backwash waste line. By run 36 this procedure had proved
unsatisfactory because a surface coating formed on the anthracite
which made draining the water above the filter nearly impossible.
More importantly, the filter itself was draining below the surface
cake, causing a negative head condition in the interior of the media.
This caused air to be drawn in through the piezometer tubes, which
air bound the filter and impaired the air-scour agitation. To combat
these difficulties in subsequent backwashes, the procedure was al-
tered by first "bumping" the bed with a short, low volume application
of air. This practice broke the surface layer of the media suffi-
ciently to allow proper draining without negative head development,
after which the air scour could proceed in the normal fashion.
Upon completion of the air scour, the air line valve was closed and
the backwash line valve opened to a rate of 20 gpm/sq ft for a total
duration of 5 min. Expansion for the south filter ranged from 25.5
to 42.6%, with an average value of 34.0%.
The west filter backwashing procedure was initiated with a 2-min solo
operation of the rotating surface washer. Next the 20 gpm/sq ft
water backwash rate was applied to the filter and was operated in
combination with the surface washer for a total of 3 min. Finally,
the surface washer was shut off and the water (only) backwash was
continued at the initial rate for an additional 2 min so that both
the surface wash and water backwash were operated for 5 min each.
Expansion of the west filter during the water fluidization (only)
phase ranged from 25% to 50%, with an average value of 37.4%.
Sampling procedures and data collection. The frequency of observa-
tion runs in Phase II was less than in Phase I to reduce budget ex-
penditures. Observation runs in Phase II were conducted once weekly
and consisted of a careful monitoring of performance parameters
throughout the duration of the run. Flow rates, influent and efflu-
ent turbidities, flow temperatures, and head loss development were
monitored at 1 to 2-hr intervals. Prior to the beginning of each ob-
servation run the normal maintenance, described in a later section,
was performed to ensure the collection of meaningful data. The auto-
matic composite samplers were also turned on and checked before
placing the discharge lines into refrigerated sample receivers. At
the conclusion of each observation run, core samples were collected
for the abrasion tests and influent and effluent composite samples
collected for laboratory analyses.
128
-------
In all cases the filter effluent samples were composited for the du-
ration of the filter observation run. However, for the first six ob-
servational runs (2, 14, 22, 29, 36, 43) the composite samplers were
left on for the entire 24-hr period even though the filters began to
bypass after about 12 hr. Because it was thought that this procedure
may have generated misleading data, the routine was changed for the
last two observation runs so that each composite sampler was shut
off when the filter first started to bypass.
Filter influent samples were not composited for the majority of the
runs because filter performance was not considered a prime objective
of the research, and because adapting the available sampler to the
pilot plant posed some difficulties which were not solved until later
in Phase II. A grab sample was taken at the conclusion of each ob-
servation run for numbers 14, 22, 29, 36, 43. Admittedly, this was a
weak procedure, even for a parameter of secondary importance, but the
influent data were, fortunately, later obtained from the log sheets of
the Ames Water Pollution Control Plant (composite samples). For the
final two observational runs the filter influent was automatically
composited, and the sampler was shut off after the last filter began
to bypass. The influent was also composited during the first obser-
vation run (run 2), which occurred while the influent was still being
pumped through the erdlator of the mobile water purification unit used,
in Phase I. This practice was discontinued after run 3 of Phase II.
Although the influent was composited over the entire 24-hr period
during observation run 2, no unusual differences were noted between
the sample analyses of this run and the last two runs.
Other details of sampling, data collection, and maintenance were
identical with Phase I.
Results - Phase II
Visual observations. Although the condition of the media varied sub-
stantially among the three filters, certain characteristics were
common to all three. One common trait, the accumulation of mud balls
and agglomerates, was a problem encountered in varying degrees
throughout the study. Because of the downflow configuration of the
filters and the relatively small particle size of the anthracite, the
majority of the suspended solids removal took place in the uppermost
4 to 6 in. on the anthracite layer. As each run progressed, the
solids in the influent filled the interstices of this upper layer and
simultaneously compressed the layer into a tightly packed mat as the
head loss across the layer increased.
Because of the biological nature of the influent solids and the ad-
sorptive properties of the anthracite, the particles of media in the
matted layer were bound tenaciously together, forming a plug at the
surface which required violent agitation for its complete disintegra-
tion. Since this was not always accomplished by a given backwash
technique or within a given backwash cycle, the matted layer was more
129
-------
often split into fragments which descended to the sand-coal interface
during fluidization and were then designated as mud balls as they
floated at the interface of the fluidized bed. Those fragments which
affixed themselves to the sides of the filter housing were designated
as agglomerates. Nomenclature aside, all fragments observed during
fluidization were spawned from the matted layer as a direct result of
the overall ineffectiveness of the backwash technique being used.
When present in sufficient concentrations, the mud balls and agglom-
erates caused similar types of problems, in varying degrees of
severity, in all three filters. If not broken up by the preceding
backwash cycle, these accumulations shortened filter runs by increas-
ing the initial head loss and reducing interstitial volume of the bed
available for subsequent solids removal. During the backwash cycles,
the mud balls and agglomerates prevented uniform distribution of water
or air and forced them to form high velocity jets or streams. Be-
cause of this channeling effect, areas of poor fluidization developed
above the fragments which reduced the effectiveness of the wash and
created conditions favorable for the formation of additional mud
balls.
Not once in the course of this study were mud balls observed to have
originated in the sand layer of any of the filters. All were com-
posed originally of anthracite and solids trapped during the course
of the filter run, and were formed as previously described. Sand was
observed in some of the mud balls and agglomerates, but this was
caused by the jets of water which tended to lift the sand high into
the coal layer where it became trapped.
Surface layers of solids on the very top of the media were observed
frequently on all the filters just prior to backwashing.
Surface cracking was observed in all three filters but was most pro-
nounced and sustained in the north filter. Cracks in the north fil-
ter were generally 1/4-in. wide, 3 to 7-in. long, and usually appeared
to result from the media pulling away from the filter walls as the
bed compressed. More severe cracking in the north filter was observed
on three occasions when the cracks were from 1/4 to 1/2-in. wide and
extended the full 18-in. width of the filter. Surface cracks in the
south filter were less severe and less frequent than those in the
north filter. Slight cracking, usually 1/16-in. wide and 1 to 2-in.
long, was noticed, but only about one-third as often as in the north
filter. The west filter was essentially free of surface cracks
throughout the study.
Dead spaces were observed at the bottom center of the sand layer in
each filter. The sand in this location, between two outside adjacent
underdrain distribution strainers, did not fluidize during the back-
wash. No difficulties were encountered as a result of this.
130
-------
Visual observations - north filter. Because of the relative inef-
fectiveness of the water fluidization (only) backwash, the accumula-
tion of mud balls and agglomerates in the north filter was rapid and
sustained. During the backwashing cycle of the first five filter
runs, the mud balls observed ranged in size from about 2 to 6 in. in
diameter and appeared in number of about 10 to 20. By the end of the
fifteenth run, however, the size and concentration of the mud balls
and agglomerates had increased to the extent that fluidization and
stratification during backwashing did little more than shuffle and
shift the mud balls slightly. Actual cleaning of the media was
virtually halted. From run 16 until the completion of the study,
mud balls and agglomerates consistently composed 50 to 7070 of the
volume of the anthracite layer visible at the window of the filter.
Typical comments as recorded in the data book are presented below for
runs 17 and 50, respectively.
Heavy surface coating. No cracking. 4-in.
penetration into bed. Large mud ball at
start 18 in. x 8 in. (almost entire coal
layer). Smaller mudballing also present.
Bed poorly stratified.
1/4-in. cracks, length of glass, heavy sur-
face coating. Intermixing and agglomerates
70-80% of bed. Fluidization poor, huge
agglomerates - 24 in. x 6 in. x 8 in., and
6-12 in. diameter - causing jets and inter-
mixing. Large agglomerate is only 3 in.
from bottom of filter. Bed essentially
intermixed - agglomerates did not break.
Occasionally the concentration of mud balls and agglomerates was even
higher although this higher level (80 to 9070) was not sustained for
any length of time. The complete clogging of the media was probably
prevented in part by anaerobic decomposition within the large masses
in the filter. The distinct odor of hydrogen sulfide was detected
frequently while backwashing the filter and lent some qualitative
support to this possibility.
On 10 occasions (runs 14S 17, 21, 39, 49, 51, 53, 56, 60, and 62), the
weakness of water (only) backwashing technique as applied to sewage
filters was dramatically demonstrated. When the backwash water was
applied to the media, the entire matted layer, a block 18-in. square
and 4 to 6-in. deep, rose as a plug in the filter housing. Three
times (run 14, 49, 62) the water backwash failed to break the layer
(even slightly), and it settled as one large mass to the sand-coal
interface. In the remaining instances the matted layer was broken
into large chunks which formed mud balls or agglomerates.
Although the high concentration of mud balls and agglomerates in the
north filter had no apparent effect upon the quality of the filtrate,
131
-------
it did cause the most severe instances of high initial head loss and
shortened runs. Initial head loss readings averaged nearly 1 ft
greater than those observed for the other two filters, and filter
runs were noticeably shorter as will be shown later. The large,
solid masses often covered one-half to two-thirds of the bed width
and caused the backwash water to channel and to lift sand high into
the anthracite layer where it became trapped. Dead areas occurred
above the mud ball layers and the unequal distribution of backwash
water left the media mounded and piled after the backwash. Fluidiza-
tion of free media was poor, and stratification of the coal and sand
layers became virtually non-attainable. Frequently, large gaps and
holes were left around the mud balls and agglomerates after the wash
which failed to fill in with filter media.
Visual observations - south filter. The media of this filter under-
went a series of significant changes throughout the course of this
investigation. Through run 5 the build up of mud balls and agglom-
erates in the bed had been minimal, roughly a half dozen mud balls
were observed during each backwash ranging in size from 1 to 2 in. in
diameter. During this initial period the 5-min air scour appeared to
be thoroughly breaking up the matted layer and intermixing the sand
and coal layers. Air-scour agitation was most pronounced during the
first minute, during which time layers intermixed. The bed quickly
subsided to a steady, pulsing action, primarily at the surface, for
the remainder of the scour.
Starting with run 6, however, much larger mud balls and agglomerates
began to appear during the backwash. While the bed was being fluid-
ized these large chunks settled to the sand-coal interface and chan-
neled the backwash water to the sides of the housing. As it had in
the north filter, the channeling action caused sand to be carried and
trapped in the upper anthracite layer. Additionally, the effective-
ness of the air scour had decreased markedly, and the air appeared
to agitate only the top 1 to 2 in. of the anthracite. The air-scour
agitation was carefully observed during the backwash following run
10. The air was channeled readily through the media instead of
being uniformly dispersed, and the bed as a whole exhibited a gelati-
nous or cohesive character. By the end of the air scour the agita-
tion had usually begun to be most effective, eroding a small portion
of the bed and piling it on the surface. Even more perplexing, the
media fluidized and stratified fairly well following the seemingly
ineffective scour.
Compared to results obtained from Phase I on alum-treated, secondary
effluents, the air-scour agitation results were quite poor. Initially
this was attributed to the much higher solids loading on the filter,
but a slight procedural change in the backwash cycle of run 36 proved
otherwise. Until this particular run, the procedure had been to
lower the water level 6 in. below the washwater trough to prevent
media loss. Because of the highly clogged nature of the upper anthra-
cite layer, this was a relatively slow process, so the procedure was
132
-------
altered for run 37. During this backwash cycle the water level was
not lowered below the overflow weir, and the air was applied at the
standard rate. The results were surprising because the air scour
suddenly regained its past effectiveness. The matted surface layer
was effectively broken up as was a very large 4 by 18-in. agglomerate
at the interface. Furthermore, the sand and coal thoroughly inter-
mixed during the air scour and then stratified excellently during
fluidization of the bed.
A subsequent investigation revealed the reason for the poor air
agitation during the previous 26 backwashes of the south filter.
Both the surface coal and the matted anthracite layer severely re-
stricted the ability of the bed to drain the water above, as evi-
denced by the long period required to do so. This caused the water
within the media below the matted layer to drain faster than the
water above could pass through the mat. Therefore, a negative head
condition occurred in the filter which, in turn, drew air into the
bed through the piezometer taps. The entrained air tended to bind
the filter media, causing it to appear gelatinous and to resist
break up by the air agitation.
A simple procedural alteration was incorporated in the south filter
backwashing sequence for subsequent runs. Before draining the media,
the bed was subjected to a short, low volume application of air to
break up the compressed surface layer. This allowed the filter to
drain freely without inducing negative head and air binding of the
media.
Immediately following the procedural change, the condition of the bed
improved with respect to mud ball and agglomerate concentration.
However, smaller 2 to 3-in. diameter mud balls began to accumulate
and cause backwash water distribution problems during runs 39 and 40,
although these were quickly dispersed and reduced in number by run
41. The next 20 runs were characterized by relatively few mud balls,
but were hampered by an inability of the air agitation to completely
disintegrate the upper clogged layer in the anthracite. Typically,
the air scour broke the left and right one-third portions of the mat,
but left the middle fragment which settled to the sand-coal inter-
face. This may be the result of poor air distribution by the filter
underdrain systems with strainers on about 7-in. centers. These
fragments varied in length from 6 to 12 in. and in depth from 2 to
3 in., and interfered with proper fluidization and stratification of
the media. Usually the agglomerate was broken up during the next
backwash cycle, but was immediately replaced by another. The back-
washing entry for run 54 typified this period of operation.
Minor surface cracking present. Air scour
does not break up mat. At media interface
mud balls sinking into sand. Approx. 1/3
of bed is not washing. Bed frees during
wash and media is well stratified.
133
-------
For the remaining four runs, the above trend disappeared, and the air
scour appeared very effective in breaking the mat. Several small mud
balls were observed but caused no serious problems with backwash dis-
tribution, intermixing, or fluidization and stratification.
Visual observations - west filter. Except for the first two runs of
the study, the west filter was equipped with a rotating surface
washer as described previously in detail. For the first two runs the
bed was subjected to surface wash from a fixed nozzle washer. For
the next 29 consecutive runs a rotating jet washer was used which
consisted of an inverted, three-armed lawn sprinkler. Finally, a
two-armed, two-nozzled, rotary washer was installed using nozzle
jets similar to those in actual surface washers.
Backwashes following runs 1 and 2 were characterized by one or two
large agglomerates (6 by 5 in. and 12 by 3 in.) accompanied by a host
of smaller 2 to 3-in. diameter mud balls. The fixed nozzle washer
was breaking the mat into chunks which remained fairly well intact
for the rest of the wash. When first used following run 3, the make-
shift surface washer appeared to do an excellent job in breaking up
the matted layer leaving only three, 2 to 3-in. diameter mud balls
visible in the bed. The very next backwash revealed a highly plugged
surface layer which effectively resisted break up by the surface
washer. When the bed was fluidized, the matted anthracite layer rose
as a plug with cross section equal to that of the housing and a depth
of about 8 in. Below this plug approximately five 2-in. diameter mud
balls floated at the interface of the sand and coal layers. Even-
tually the plug broke up into large agglomerates which sank to the
interface.
The condition of the bed stabilized during the next 10 backwashes to
one of several small mud balls, 3 in. in diameter or less, plus one
or two large agglomerates at the interface, sized approximately 8 by
2 in. to 6 by 2 in. After a brief period of almost no mud balls or
agglomerates, this pattern was consistently observed from run 17 to
run 32. Although these accumulations caused some channeling of back-
wash water and intermixing of media, the media continued, generally,
to fluidize and stratify quite well, as evidenced by this entry for
run 30:
7-in. penetration into bed. Heavy surface
coating. Mud balls in bed larger than in
south filter. 4-in. dia., 2-in. dia., three
3-in. dia., 8 in. x 2 in. Bed is generally
well stratified.
Preceding the backwash for run 32, the third surface washer was in-
stalled. Although the action of the newly installed washer did not
appear particularly violent, the condition of the bed continued to
improve for the next three runs. Then the familiar pattern of several
large agglomerates accompanied by approximately six small mud balls
134
-------
returned with the backwash following run 36. Except for a brief
period from run 46 to 49, during which the bed was relatively free of
mud balls and agglomerates, this trend continued for the balance of
the phase and is exemplified by the entry for run 43:
No visible cracks, cannot see surface coating
(blocked by support). No appreciable inter-
mixing. Bed is well stratified. One 5 in.
x 1 in. agglomerate visible at sand-coal in-
terface. Surface washer working, violent
swirling in area of washer when bed fluid-
ized. Several large mud balls observed:
3 in. x 12 in. x 12 in., 6 in. x 6 in. x 3 in.,
(2) 2-in. dia., (2) 1-in. dia.
Regardless of the washer used, therefore, the condition of the bed
remained relatively constant throughout the study - not good but not
exceptionally poor as in the north filter. Several periods of signi-
ficantly improved condition of the media were observed, but were also
short-lived. Fluidization and stratification of the media were not
severely hindered except when the accumulations became heavily con-
centrated, which was also infrequent. Surprisingly, the action of
the surface washer was most effective after the bed had been fluidized
and the rotating arm was turning through the fragments of the mat,
and not when it was attempting to break up the surface of the static
layer.
Abrasion tests. The abrasion tests were considered of prime impor-
tance in evaluating the effectiveness of the various backwashing
techniques because they directly indicate the condition of the filter
media. As previously described, the abrasion test contained a step
in its procedure in which the core sample was subjected to a vigorous
30-min mechanical mixing. Obviously, the mixing action tended to
break up some of the anthracite sample, so a background level was
determined for the test by subjecting an unused sample of anthracite
(out-of-bag) to the procedure. The test was run twice on the same
sample of clean anthracite and yielded values of 0.61 mg/g and 0.22
mg/g, respectively. The latter valve was accepted as the background
level because the former was influenced by coal dust which initially
adhered to the media.
The results of the abrasion tests are displayed graphically in Fig.
33. Unquestionably, the most striking trend is that exhibited by the
north filter. Except for a slight dip after run 20, the abrasion
test values for this filter climbed steadily throughout the filter
run series. The rapid and sustained build up of solids on the north
filter media, as evidenced by these tests, is a direct result of the
ineffectiveness of the water fluidization (only) backwashing tech-
nique which was used to clean the filter. The test results are also
quite in line with visual observations of the north filter, which
135
-------
OJ
.O)
o
LU
I
LU
"
o
_i
O
CO
36
32
28
24
20
O NORTH FILTER - WATER FLUIDIZATION ONLY
a SOUTH FILTER - AIR-SCOUR AUXILIARY
A WEST FILTER - SURFACE WASH AUXILIARY
10 15 20 25
30 35 40
RUN NUMBER
45
50 55 60 65 70
Fig. 33. Standardized abrasion test results (Phase II) during direct
filtration of secondary effluent.
-------
qualitatively labeled it as the bed in worst physical condition as
described previously.
The lower of the two values plotted for each filter at run 64 was the
abrasion test result at the conclusion of a clean up operation which
will be described in more detail later. Interestingly, the north
filter was still in worse condition (in terms of solids remaining on
the media) at the end of its extensive clean up procedure than the
other two filters were before the start of their clean up operations.
The abrasion test data is certainly less definitive with respect to
differences in condition of the media between the south and west fil-
ters. The graphs for these two filters cross each other four times,
although each displays an average upward trend. The west filter ex-
hibited a sharp peak at run 28, but immediately dropped with the next
test. This point is viewed with suspicion and is thought to be in
error, although nothing in the test record indicated an error.
Unfortunately, conclusions about differences in the cleanliness of
the media in the south and west filters are not warranted based upon
the available abrasion test data. It is apparent that the filters
should have been operated for a longer period of time to establish
definite trends; however, this was impossible because of the onset of
winter.
Initial head loss. Initial head loss readings were recorded at the
start of each run for each filter. The procedure was standardized by
allowing 15 min after opening the filter influent valve before taking
readings on the piezometer tubes. The initial head loss, as plotted
for each filter in Fig. 34, represents the total loss in head through
the filter media after 15 min of operation.
By rearranging the data for each filter without regard to time and
plotting the percent of observations exceeding a given head loss
value on normal probability paper, one obtains three curves as shown
in Fig. 35.
Clearly, the north filter exhibits substantially higher initial head
losses than either the south or west filters. One would expect such
results since throughout the filter run series, the accumulation of
mud balls and agglomerates was most rapid and sustained in the north
filter. Because of the ineffectiveness of the water fluidization
(only) backwashes, many of these masses were not broken up during the
backwash and remained in the bed at the start of the next run. Since
much of the bed remained plugged by these solid masses, the filter
influent was forced to find paths around them, causing increased flow
velocity through the cleaner portions of the bed and proprotionately
increased initial head loss through the filter.
Substantial differences in initial head losses between the west and
south filters were not apparent. Although the west filter data
137
-------
oo
2
§2
o NORTH FILTER - WATER FLUIDIZATION ONLY
a SOUTH FILTER - AIR-SCOUR AUXILIARY
A WEST FILTER - SURFACE WASH AUXILIARY
10
Fig. 34.
20
40
50
30
RUN NUMBER
Initial head loss data for north, south, and west filters for
entire Phase II study during direct filtration of secondary
effluent.
60
-------
J: 2
o
to
V)
2
o
10
o NORTH FILTER - WATER FLUIDIZATKDN ONLY
D SOUTH FILTER-AIR-SCOUR AUXILIARY
£ WEST FILTER - SURFACE WASH AUXILIAR>
I
I
I
I
I I I
0.01 0.1
1 5 10 20 30 40 50 60 70 80
PERCENT OF OBSERVATIONS LESS THAN INDICATED VALUE
90 95
99
Fig. 35. Frequency plot of Initial head loss data, Phase II.
-------
plotted slightly higher in Fig. 35 and had nearly twice as many ob-
servations above 1 ft, the differences were not appreciable. For the
majority of the runs the initial head loss values were quite close
and followed similar trends.
A typical pattern developed for all three filters wherein the initial
head loss exhibited a cyclic trend. Although,there was no quantita-
tive means of correlating initial head loss data with the visual ob-
servations of the backwashing procedures, the latter demonstrated a
similar trend. Mud balls and agglomerate accumulations appeared to
attain a maximum concentration, decrease somewhat, and then resume
their build up. This trend was most noticeable in the north and west
filters, but was also observed in the south filter to a lesser ex-
tent. It seems reasonable, although admittedly speculative, that the
cyclic trend of initial head loss was paralleling the cyclic solids
build ups in the filters.
Head loss development. Studies by Cleasby and Baumann [31] have re-
vealed that surface layers which form on sand filters used in water
treatment cause a characteristic exponential shape to the head loss
versus time curve. The accelerated head loss development with time
was attributed to the surface layer or "cake" because the layers act
as filters themselves. Once established, the layers remove more
influent suspended solids and increasingly compress during the course
of a run, so that by the time terminal head loss is attained, the
layers form dense mats which are resistant to cleaning.
As described earlier in the section on media appearance, layers of
organic matter on the filter surface were frequently observed prior
to backwashing the filters in Phase II. One might reasonably expect
an exponential shape for the total head loss curve, although the data
as plotted in Figs. 36 through 43 exhibited no such tendency. The
nearly straightline development of head loss by the filters could
have been indicative of solids removal occurring much deeper in the
bed, particularly since penetration of solids was consistently visi-
ble 4 to 6 in. below the surface in all filters. However, this does
not seem plausible since surface layers were observed; therefore, it
seems as though some other phenomenon was acting to alter the shape
of the curve.
An explanation of the linear head loss development may be hypothe-
sized by observing the head loss development in the 4-in. layer of
media directly below the surface and by noting that surface cracking
was observed primarily in the north and south filters throughout the
study, as described earlier. During the course of a run, the bed
depth was consistently compressed from 1.0 to 1.5 in. in each filter,
which was likely the result of the effect of compressible coating on
the media due to inefficient backwashing. The compression can cause
the media surface to crack -and to pull away from the sides of the
filter housing walls or to break, as in a beam failure, over the pro-
truding piezometer tap. When surface cracks open up as the filter
140
-------
CO
2
NORTH FILTER
(WATER FLUIDIZATION ONLY)
6.62 ft at 12.75hr
0 to
SOUTH FILTER
(AIR-SCOUR AUXILIARY)
T
WEST FILTER
(SURFACE WASH AUXILIARY)
3.26 ft at 12.75 hr
16 to 24 in.
12 to 24 in.
4 to 24 in.
10 0 2 4 6 8 10
TIME FROM BEGINNING OF RUN, hrs
8 10
Fig. 36. Chronological head loss development at various media
depths, run 2.
-------
10
I/)
3
NORTH FILTER
(WATER FLUID IZATION,
ONLY)
0 to 24 in. & 4 to 24 in.
8 to 24 in.
0 2
12 to 24 in.
16 to 24 in.
SOUTH. FILTER
(AIR-SCOUR AUXILIARY)
0 to 24 in.
WEST FILTER
(SURFACE WASH AUXILIARY)
0 to 24 in.
6 8 10 024 68 10 024*
TIME FROM BEGINNING OF RUN, hrs
Fig. 37. Chronological head loss development at various media
depths, run 14.
8 10
-------
NORTH FILTER
"(WATER FLUIDIZATION ONLY)
0 to 24 in.
to
CO
O
*
X
to 24 in.
to 24 in.
16 to 24 in.
SOUTH FILTER
(AIR-SCOUR AUXILIARY)
0 to 24 in.
24 in.
>
to 24 in.
12 to 24 in.
•016 to 24 in.
WEST FILTER
(SURFACE WASH AUXILIARY)
0 to 24 in.
4 to 24in.
8 to 24 in.
«^ol2 to 24 in.
Q—pQ|6 jo 24 in.
10 0 2 4 6 8 10 0
TIME FROM BEGINNING OF RUN, hrs
4
8 10
Fig. 38. Chronological head loss development at various media depths, run 22.
-------
NORTH FILTER
(WATER FLUIDIZATION ONLY)
0 to 24 in
jj 5
1
to
CO
o
4 to 24 in.
8 to 24 in.
12 to 24 in.
SOUTH FILTER
(AIR-SCOUR AUXILIARY)
0 to 24 in.
4 to 24 in.
24 in.
12 to 24 in.
24 j.
WEST FILTER
(SURFACE WASH AUXILIARY)
>0 to 24 in.
4 to 24 in.
8 to 24 in.
to 24 in.
916 to 24 in.
8 10 0 2 4 6 8 10
TIME FROM BEGINNING OF RUN. hts
8 10
Fig.
39. Chronological head loss development at various media depths, run 29.
-------
O
oo o
^) O
2
Q
NORTH FILTER
(WATER FLUIDIZATION ONLY)
0 to
to 24 in.
SOUTH FILTER
(AIR-SCOUR AUXILIARY)
0 to
24 in.
24 in.
-o] 2 to 24 in
016 to 24 in
WEST FILTER
(SURFACE WASH AUXILIARY)
0 to 24 in.
2 to 24 in.
6 to 24 In.
68 10 02468 10 024
TIME FROM BEGINNING OF RUN, hrs
Fig. 40. Chronological head loss development at various media
depths, run 36.
8 10
-------
*. 4
o
NORTH FILTER
(WATER FLUIDIZATKDN ONLY)
0 to 24 in.
12 to 24 in.
16 to 24 in.
SOUTH FILTER
(AIR-SCOUR AUXILIARY)
0 to 24 in.
I
to 24 in.
WEST FILTER
(SURFACE WASH AUXILIARY)
0 to 24 in.
24 in.
to 24 ii
Pl6 to 24 i
in.
to 24 in.
2 to 24 in.
1°16 to 24 in.
8 10
8 10
02468 10 0246
TIME BEGINNING OF RUN, HR
Fig. 41. Chronological head loss development at various media depths, run 43.
-------
_
o
it
t/T
i/-»
2
a
_ NORTH FILTER
(WATER FLUIDIZATION
" ONLY)
0 to 24 in, "
24m.
24 in.
o-o-
•o-
12 to 24 in.
•O—O—O—-o
16 to 24 in.
I
SOUTH FILTER
(AIR-SCOUR AUXILIARY)
Oto24 in.
in.
24 in.
to 24 in.
WEST FILTER
(SURFACE WASH AUXILIARY)
i
0 to 24 in.
4 to 24 in.
8 to 24 in.
12 to 24 in.
16 to 24 In.
802468 10 024
TIME FROM BEGINNING OF RUN, hrs
8
10
Fig. 42. Chronological head loss development at various media depths, run 54.
-------
•P-
oo
NORTH FILTER
'(WATER FLUIDIZATION ONLY)
0 to 24 in.
I 5
V
SOUTH FILTER
(AIR-SCOUR AUXILIARY)
0 to 24 in.
24 in.
24 in.
12 to 24 in.
16 tp 24 ?p.
WEST FILTER
(SURFACE WASH AUXILIARY)
0 to 24 in.
24 in.
>
to 24 in.
12 to 24 in.
to 24 in.
8 10 0246 8100
TIME FROM BEGINNING OF RUN, hrs
8 10
Fig. 43. Chronological head loss development at various media depths, run 64.
-------
run progresses (as they did in this study) they may preclude the for-
mation of an exponential total head loss pattern since the influent
is being introduced deeper into the bed through the cracks. The net
effect would be to increase the head loss in the media layer directly
below the surface layer, which was frequently observed in the north
and south filters.
In some cases, a crack must have formed by beam failure over the
first protruding piezometer tap at 4-in. depth, which caused the head
loss curve at that depth (4 to 24 in.) to converge on the curve for
the piezometer above the media (0 to 24 in.). This is evident on
several of the curves for the north filter, which had the most ex-
tensive surface cracking.
Table 14 illustrates the initial total head loss and the time re-
quired to reach 3 ft and 4 ft of total head loss for each filter. It
is apparent that the initial head loss was higher for the north fil-
ter, which reduced the head available for solids accumulation during
the filter run. Consequently, the average run length for the north
filter was reduced compared to the other filters. Differences be-
tween the south and west filters are not great or consistent, so
conclusions are not warranted.
Water quality. The means and standard deviations of the various
water quality parameters are given in Table 15. No statistical com-
parison between the filters was attempted for the reasons explained
in Phase I. There is no apparent difference between the filters,
even though the condition of the north filter was substantially
poorer than that of the other two filters.
As the result of sampling difficulties, composited samples of the in-
fluent were not made for the majority of the observation runs.
Therefore, reporting removal efficiencies on the basis of influent
grab samples was not considered meaningful or proper. Fortunately,
the Ames Water Pollution Control Plant personnel had conducted sus-
pended solids and BOD determinations on composited samples collected
over a period corresponding closely to the observation runs of this
study; these are included in Table 15.
At roughly 2-hr intervals during the observation runs, grab samples
were collected from the effluent of each of the filters and from the
influent, and turbidity determinations were made. These values were
treated as a hand-sampled composite, averaged for each run, and re-
ported in Table 15.
Filter clean up operations. At the conclusion of the regular filter
run series in Phase II, special cleanup operations were conducted on
each of the three filters. The purpose of these operations was to
determine if the media in each filter could be restored to its origi-
nal state by a series of consecutive backwashes using varied tech-
niques. Core samples were taken at the beginning and at the end of
149
-------
Table 14. Summary of head loss development during observation runs of Phase II, during direct fil-
tration of secondary effluent.
Run
2
14
22
29
36
43
57
64
Avg
Initial
N
1.00
2.53
2.58
2.77
1.55
2.19
4.76
1.95
2.42
head lossa
S
0.33
0.62
0.54
1.31
0.81
0.89
0.95
0.84
0.79
, ft
W
0.38
0.96
0.79
1.05
1.09
0.63
0.95
0.88
0.84
Time to 3
N
5.8
1.8
1.5
0.6
4.2
1.6
2.4
2.0
2.5
ft head
S
11.8
5.2
6.3
2.6
6.0
4.8
3.5
3.0
5.4
lossa, hr
W
9.6
5.4
7.2
4.0
5.6
5.8
2.9
3.2
5.5
Time to 4
N
8.4
3.8
3.6
1.6
6.4
3.0
3.6
3.6
4.2
ft head
S
15. 6b
7.4
8.8
4.0
8.3
6.8
5.2
4.0
7.5
lossa, hr
W
12.8
7.8
8.8
6.0
8.2
7.8
3.6
4.6
7.5
Total head loss across the filter media in each case.
Extrapolated value.
-------
Table 15. Results of analyses during direct filtration of secondary effluent (Phase II) from August
30 to October 30, 1973. (All composite samples except as noted.)
Suspended solids (mg/1)
cr (N = 8)
Turbidity (FTU)
a (N = 9)
BOD5 (mg/l)b
a (N = 8)
TOG (mg/1)
CT (N = 7)
Total P04 (mg/1)
CT (N = 8)
Ortho PC>4 (mg/1)
cr (N = 8)
Total Kjehldahl nitrogen (mg/1 as N)
(T (N = 8)
Filter
influent
30. 5C
7.6
17.4
2.2
40. 1C
15.4
17. 4d
4.1
20.6d
3.9
19.3d
3.3
12. 7d
4.4
N filter, water
f luidization
only
4.05
2.80
4.84
1.33
12.1
3.1
12.7
2.8
23.9
4.5
22.7
3.7
15.4
2.6
Filter effluent3
S filter,
air -scour
auxiliary
3.63
1.58
4.69
1.42
14.2
4.9
12.0
2.9
24.6
3.3
22.8
2.4
15.5
2.6
W filter,
surface wash
auxiliary
4.10
1.6
4.61
1.36
12.7
4.6
14.2
5.3
24.4
3.7
22.5
2.5
15.1
2.6
Average filtration rate 1.8 gpm/sq ft
Nitrification not suppressed in BOD test except in the samples run by Ames WPG Laboratory.
'Two of the observations are from grab samples, and two from composites run by Ames WPC Laboratory.
Four of the observations are from composite samples, the remainder from grab samples.
-------
the clean up procedure to evaluate the overall effectiveness of the
operations. Additionally, samples of the backwash wastewater were
taken at 30-sec intervals and composited throughout each step of the
three clean up procedures. The results of these procedures are sum-
marized in Table 16 and discussed in detail below.
Initial head loss readings, abrasion test results and visual observa-
tions all indicated the media of the north filter to be in the worst
condition of the three filters. A visual check just prior to the
start of the clean up operation, further confirmed the poor condition
of the north filter. The media had a heavy surface layer, and 1/4 to
1/2-in. wide cracks of varying lengths were visible at the surface.
The sand and coal layers had become highly intermixed as the result
of poor fluidization and poor stratification in previous backwashes.
Finally, two 6-in. diameter agglomerates were clearly visible.
To initiate the clean up procedure for the north filter, the media
was subjected to a normal water fluidization (only) backwash. At
the beginning of the fluidization of the media, the top 6-in. layer
of anthracite rose as a plug and then settled to the sand-coal inter-
face. The backwash was successful only in breaking up this layer
into large chunks and could not effect further breakdown. A core
sample was taken at this stage, and the abrasion test results indi-
cated a high value of 33.7 mg/g. This initial step was necessary to
place the media in the state in which it would normally be at the
conclusion of its standard backwashing procedure and, therefore, to
establish a starting point for the evaluation of subsequent and
varied backwashes.
The second step in the procedure consisted of a subfluidization back-
wash with simultaneous air scouring, the only introduction of air to
this media since the beginning of the Phase II. Prior to the back-
wash, the water level in the filter was drained to the media surface.
Air was then applied at 3.72 scfm/sq ft and water at 11 gpm/sq ft,
until the water level reached an elevation 6 in. below the washwater
trough. At this point the air was shut off and the water rate in-
creased to 20 gpm/sq ft, which fluidized the media by itself. The
combination air and water wash seemed to produce substantial agita-
tion in the media and effectively disintegrated many of the mud balls
in the media. However, many mud balls were still present during the
5-min water (only) backwash which immediately followed, and one
large, 5-in. mud ball was observed deep in the sand layer near the
conclusion of the wash.
The third clean up step was essentially a repeat of the previous
step, and it continued to improve the condition of the media. During
the combined air and water wash, the water rate was reduced slightly
to 10 gpm/sq ft, but the agitation of the media was quite good. The
combined action was particularly effective because the media exhibit-
ed no tendency to "pack" after 30 sec to 1 min of operation, as it
did when using air scour alone as on the south filter during the
152
-------
Table 16. Data summary of clean up operation at end of Phase II.
Steps
Brief description of cleanup procedure - see text
Solids Media
released abrasion
(g/sq ft) (mg/g)
1.
2.
3.
4.
5.
6.
1.
2.
3.
4.
1.
2.
3.
4.
5.
6.
North Filter
Normal water (only) backwash.
Air and water combination (water at subsidization rate)
as water rose from 1 in. above media to 6 in. below trough.
Follow with water (only) backwash.
Repeat Step 2 with a reduced wash rate during air and
water combination.
Repeat Step 3 except continue air scour after combination
air and water. Follow with water (only) backwash.
Air and water combination, water above fluldlzation.
Follow with water (only) backwash.
Repeat Step 5, but leave air on 3 min. after air water
combination. Follow with water backwash.
Total
South Filter
Normal air (only) followed by water (only) sequence.
Combination air and water with water just at fluidiza-
tlon rate. Follow with air (only) (2 min.) and then
water (only).
Repeat Step 2.
Repeat Step 2.
Total
West Filter
Water (only) backwash.
Subsurface wash and water wash simultaneously.
Repeat Step 1 without drawdown of water above media.
Repeat Step 2.
Air and water combination, follow by 2 min. air (only),
and 'finally 3.5 min. water (only) backwash.
Repeat Step S, except for 5 min. water (only) backwash.
Total
33.7
314
163
205
84
67 15.3
833 - (35.1mg/g)a
9.0
100
42
_17 1.9
161 - (6.8 mg/g)«
11.3
52
18
12
120
_26 2.9
228 - (9.6 mg/g)a
*Based on 12 In. coal media Involved in filtration and weight of coal media of 23,742 g/sq ft.
153
-------
series. Layer stratification was more pronounced during the water
fluidization backwash which followed, and the number of mud balls was
further reduced. However, some mud balls persisted at the sand-coal
interface, and one large, 5-in. diameter agglomerate was observed
near the end of the water (only) backwash.
The fourth cleaning procedure consisted of the previously described
air and water combination followed by air (only) for 2 min at the
same rate and concluded with water backwash at 20 gpm/sq ft. Careful
observation during the combined action revealed that violent agita-
tion took place mainly in the upper 18 in. of the bed while the lower
12 in. became fairly packed with little movement. As expected, the
agitation during the air (only) scour was fairly good for the first
minute until the bed packed, but overall agitation appeared much less
effective than the combined action. One large, 6-in. diameter mud
ball and several smaller ones were observed at the conclusion of the
water backwash.
The final two steps in the north filter clean up operation were es-
sentially the same and differed from the fourth step only in that the
water rate used during the combined air-water scour was increased to
14 gpm/sq ft. However, this change in rate caused the media to be
fluidized and improved the media agitation by extending the action
throughout the bed. By the end of the water backwash of the fifth
step, the remaining agglomerates had disappeared, although two
"clusters" of four to six 1-in. diameter mud balls persisted and were
not broken up even at the conclusion of the final step. In each of
the last two steps, the air (only) scour produced the now typical
result: good action during the first minute but little movement
after 1 min because of intermixing and packing of sand and coal
layers.
Throughout the regular filter run series, the south filter was
cleaned using a 5-min air-scour auxiliary at 3.72 scfm/sq ft followed
by a 5-min water fluidization backwash at 20 gpm/sq ft. This proce-
dure was used to initiate the cleanup operations on the south filter.
The media was not in nearly as poor condition as the north filter,
but during the last 30 sec of the water wash, six 3 to 4-in. diameter
mud balls were observed. At the conclusion of the initial step, a
core sample was taken and subjected to an abrasion test. The test
results indicated the media to be approximately four times cleaner
then the north filter at the same stage. The clean up procedure
consisted of an air and water combination wash with the media slight-
ly fluidized (14 gpm/sq ft), an air (only) scour for 2 min at 3.72
scfm/sq ft, and a water (only) backwash at 18 gpm/sq ft. Expanded
bed depth remained constant at 32-1/2 in. for the remaining three
steps, and fluidization and stratification were consistently good.
As the clean up continued, the number of mud balls observed during
the water backwashes diminished. At the conclusion of the fourth and
final step, only one small mud ball was observed, and the physical
appearance of the bed was excellent. A final core sample taken at
154
-------
this time was analyzed in the laboratory and was found to verify the
visual observations: the abrasion test value was determined as 1.88
mg/g which indicated the bed was in very good condition.
As one will recall from an earlier discussion, the west filter had
been equipped with a rotating surface washer which operated both in-
dependent of and in conjunction with the water backwash during the
backwash sequence of the filter series. The routine wash was not
used to begin the west filter clean up operation; instead, a water
(only) backwash rate of 19.5 gpm/sq ft for 5 min was used. The sur-
face washer had been modified prior to the start of the wash, and
while the media was fluidized, the rotating washer was pushed to the
fluidized interface of the sand and coal layers for use in subsequent
steps. At the conclusion of the water wash, a core sample was taken,
and a subsequent abrasion test resulted in a value of 11.26 mg/g,
very close to that of the south filter.
Steps 2, 3, and 4 were combination rotating subsurface wash, water
backwash, with the former operating for 3, 4, and 3 minutes, respec-
tively. The rotating washer had been lowered into the bed to see if
it could effectively break up the mud balls which accumulated at the
sand-coal interface. By the end of the second step no mud balls were
seen even though several large (4 to 8-in. diameter) mud balls had
been observed at the start of step 2. Only three small mud balls,
approximately 1 in. in diameter, were seen during the third step, and
none at all were observed during the fourth step — the last combina-
tion subsurface wash, water backwash used in the cleanup. The rotat-
ing subsurface washer, therefore, appeared to do an excellent job of
breaking up the large agglomerates and mud balls in the media.
At the conclusion of the fourth step, however, there were signs that
the media was not thoroughly clean even though the mud balls had been
eliminated. Dirt remained floating on the surface of the coal after
step 4, and the anthracite itself had a grayish cast, as though it
were still coated. This seemed an excellent opportunity to change
the backwash procedure for step 5 to include the use of air scour and
see if it could further clean the coal layer.
The procedure for step 5 was, therefore, changed to provide combina-
tion air and water wash with the water wash rate set at 12.5 gpm/sq
ft to fluidize the media as the water rose from the media to near the
overflow. This was to be followed by 2 min of air (only) scour at
3.72 scfm/sq ft and, lastly, by 5 min of water (only) backwash at
20 gpm/sq ft. Prior to step 5 the composited backwash waste water
had shown a decline in suspended solids concentration equivalent to a
dirt released value of 12 g/sq ft after step 4. The composited sam-
ple for step 5, however, indicated a nearly ten-fold increase in the
dirt released value to 120 g/sq ft. These results seemed, at first
glance, even more significant because a dwindling supply of backwash
water limited the actual water (only) backwash to 3-1/2 min.
155
-------
Step 6 in the west filter clean up was identical to step 5, although
the water (only) wash was extended 5 min after building up the back-
wash supply. It was noted that, again, the composited backwash waste
sample had a suspended solids concentration twice that of step 4.
During this final step of the west filter cleanup, no mud balls were
observed, all floating material had disappeared from the surface, and
the anthracite regained its rich, black lustre. An abrasion test of
a core sample taken after this step was completed yielded a value of
2.19 mg/g, which further confirmed the clean condition of the bed.
The results from steps 5 and 6 initially appeared to demonstrate the
superiority of the air and water combination wash over the rotary
surface wash backwash auxiliary. However, as noted in step 4, much
dirt was visibly released from the media which was not carried out
during the water (only) backwash, and large solids were also noted
adhering to the sides of the filter housing. At the conclusion of
step 6 the floating material had disappeared entirely, and some of
the wall solids had also been removed. Therefore, one can state con-
fidently that the air and water combination backwashes (steps 5 and 6)
were definitely more effective in loosening solids from the filter
housing, and in disintegrating solids so they could be transported
out of the filter during the water (only) backwash, but one cannot
state categorically that it was more efficient in releasing dirt from
the media.
The data presented in Table 16 clearly demonstrate the benefit of
both auxiliaries in maintaining the filter media in cleaner condi-
tion. However, it is not possible to choose which auxiliary is bet-
ter from the data. The fact that the air and water used simultane-
ously were able to release substantial additional solids from the
south and west filters, which had been routinely washed with air-
scour and surface wash auxiliaries, respectively, implies the superi-
ority of air and water together as a backwash method. One should be
careful about jumping to such a conclusion, however, since the re-
search of Phase II was not designed to prove that point. To prove
the superiority of air and water together over the other two backwash
auxiliaries, it would be necessary to conduct an entire research
phase in which the three methods of backwash auxiliary were compared
in parallel.
Summary and Conclusions - Phase II
The objectives of the experimental investigations in Phase II were to
determine the effectiveness of three different backwashing techniques
on dual-media filters and to compare the performance of the three fil-
ters while filtering secondary effluent. The backwashing techniques
used were as follows:
1. North filter, backwash by water fluidization alone.
156
-------
2. South filter, two-phase sequence consisting of air (only) scour
followed by water fluidization backwash (i.e., air-scour auxil-
iary) .
3. West filter, three-phase sequence consisting of surface wash
(only), surface wash and water backwash, and water fluidization
backwash.
The experimental data were collected over a nine-week interval at the
Ames, Iowa, trickling filter plant using pilot-scale equipment. The
following are the conclusions of the phase:
1. None of the three methods of backwashing was able to keep the
filter bed completely free of mud balls and agglomerates.
2. Based upon higher initial head losses, steadily increasing solid
accumulation observed in abrasion test results, visual observa-
tions of the media condition, and results of the clean up opera-
tion, the north filter was clearly in the worst condition of the
three beds. Therefore, water fluidization (only) backwashing is
ineffective in maintaining the filter media in good condition,
and some means of auxiliary cleaning is required.
3. On the basis of the items mentioned in conclusion 2, no conclu-
sive differences were observed between the south and west filter
in the effectiveness of their cleaning techniques, air-scour
auxiliary and surface wash auxiliary, respectively.
4. No apparent differences were observed in the effluent qualities
among the three filters, particularly in the primary removal
efficiency parameters BOD, suspended solids, and turbidity.
This indicates bed condition played little part in removal ef-
ficiency in this study; however, this may have been due to the
choice of a fine filter media, low filtration rate or low termi-
nal head loss in this research.
5. The use of some form of air-scour auxiliary or some form of sur-
face wash auxiliary is essential to the satisfactory functioning
of wastewater filters. The methods used in this research did
not completely eliminate all dirty filter problems, but both
auxiliaries reduced the problems to acceptable levels so that
filter function did not seem to be impaired.
Operation and Results - phases III, IV and V
Single-, Dual-,and Triple-Media Filtration
of Secondary Effluent
Phases III through V were prompted by the deficiencies in water flu-
idization backwashing demonstrated in Phase I and II, even when as-
sisted by air-scour or surface wash auxiliary. These phases were
157
-------
also prompted by the implied superiority of simultaneous air and
water backwash revealed in the cleanup operations at the end of
Phase I and II. There were other questions raised by Phase II which
needed evaluation so Phases III through V were designed to try to
answer those questions.
During Phases III through V, the north filter was equipped with dual
media. The backwash included three steps, including air scour alone,
air and water simultaneously for a very brief period without over-
flow, and finally, water fluidization backwash alone. Coarser under-
drain strainer openings were used which required the use of gravel
below the media. A double reverse graded gravel was used to resist
movement by the simultaneous air and water backwash action.
The south filter was equipped with a deeper bed of coarse sand (6 to
10 mesh, 2 to 3.36-mm sieve range) and was washed at subfluidization
rates with air and water simultaneously during overflow for a rather
extended period, followed by a brief period of subfluidization water
backwash.
The west filter was equipped with triple media (i.e., commercially
obtained "mixed-media" from the Neptune Microfloc Corporation). The
media was underlain by a graded gravel, and the water fluidization
backwash was assisted by a surface and subsurface washer. The sub-
surface washer was added in this phase because of the problem of mud
balls sinking to the coal-sand interface and floating there out of
reach of the normal surface washer.
The details of media sizes and depths, gravel gradation, and .ider-
drain strainers have been presented previously. Details of the back-
wash are presented in the following pages.
Operation - Phases III through V
The first filtration run of the testing period was on May 16, 1974,
and operation continued daily, except when the equipment malfunc-
tioned, to the cleanup operations on November 2, 1974. The five-
month period was divided into three phases coinciding with changes in
media and sampling technique. The first runs were designated as
Phase III, a continuation of the notation started the previous year,
and were continued to July 18, 1974, when Phase IV began. Due to an
incident of clogging of the uriderdrain strainers (to be described in
detail later) and the rebuilding of two of the filters, Phase IV was
started here to signify these changes. It continued to August 27,
1974, or to the start of Phase V. New sampling containers were pur-
chased and a more careful sample bottle washing procedure was used
starting on this date, so a phase change was felt to be appropriate.
Each particular day of filtration was designated as a run. Phase III
ran from run 1 to run 50; a new run sequence was started at Phase IV
but not at Phase V, so from July 18 to November 2, 1974, a single
sequence of runs was used (run 1 to run 99). Rather than report the
158
-------
results by phase and run number, which is misleading for Phase V
since a new run number sequence was not started, the actual number of
the day of the year the run was performed will be used to designate
the occurrence. For example, Phase III began on May 16 or calendar
day no. 136 and the cleanup operations were November 2 or day no.
306. This notation will be used to discuss the results of Phases III
through V. Another reason for this choice of notation was the method
used to figure averages and to plot data points over the entire test-
ing period. It was simpler to have the abscissa as one continuing
sequence rather than as various run and phase numbers.
The flow rate for Phase III was 2.1 gpm/sq ft. The flow rate varied
slightly between runs in Phase III but was evenly split between fil-
ters in any particular run. At the start of Phase IV and continuing
through Phase V, the flow rate was changed to 3.2 gpm/sq ft (7.8 m/hr),
a 50% increase. Here again the flow rate between runs was slightly
variable. The flow rate for the runs discussed will be presented
individually in a later section.
Backwashing. One objective of this study was to compare the effec-
tiveness of three different backwashing procedures. One backwashing
procedure was assigned to each filter for the length of the study,
and it was not changed although minor changes in duration of the vari-
out backwash operations were adjusted as they seemed necessary. Al-
though each filter was backwashed by a different technique, they all
were prepared for backwashing by the same series of steps as was used
in prior phases.
The dual-media filter was designated for an air-scour cycle and a
fluidized waterwash. After some initial experimenting in the first
few runs, the following backwash procedure was decided upon for the
dual-media filter.
Step 1. Drain water from above the media to within 1 in. of the sur-
face.
2. Add air at 3 scfm/sq ft (standard cubic ft/min of air at
70 °F and 1 atmosphere pressure) and water at 13 gpm/sq ft
in combination until 6 in. below overflow trough.
3. Shut air off and continue water alone wash at 23 gpm/sq ft
for 5 min.
There were a few days where the air wash rate was 4 scfm/sq ft, but
the above procedure was run as indicated for approximately 2-1/2
weeks. At that point, the water alone wash rate was cut to 21.5 gpm/
sq ft, and that change was continued to run 38 (day 183) of Phase III.
The following backwashing procedure was the one used from run 38 to
the end of the Phase V.
159
-------
Step 1. Drain down as before.
2. Apply air scour alone at 3 scfm/sq ft for 2 min.
3. Add water at 13 gpm/sq ft in addition to the
air until the water is 6 in. below the overflow.
4. Shut off air and continue water at 21.5 gpm/sq ft for 5 min.
The only minor change to the above procedure came at run 32, Phase IV
(day no. 232), when the time of the air scour was increased to 5 min
and continued to the end of the study.
The mixed-media filter was equipped with a surface and subsurface
auxiliary washer. The backwash technique for this filter began by
starting the surface washer and continuing it for 2 min. The surface
washer was left on, and then the backwash water was started at 15
gpm/sq ft, fluidizing the bed, and followed by the starting of the
subsurface washer. The subsurface washer failed to spin at all until
the bed became fluidized. After 3 min, the surface and subsurface
washers were stopped, thus providing 5 min and 3 min total operating
time, respectively. The water alone backwash continued for 4 min to
end the cycle. Throughout the first 16 runs of Phase III, the water
only backwash lasted for only 2 min, and the water rate used during
surface and subsurface washer cycle varied from 13 to 15 gpm/sq ft.
However, by run 17, the backwash cycle first mentioned was established
and used until run 30 of Phase III. There had been some trouble get-
ting the surface washer to break up the surface mat. The washer was
just a little too high to fully agitate the surface of the media.
Beginning with run 30 (day 173), a very small amount of backwash
water was added during the first 2 min to provide a slight expansion
of the bed, enabling the surface washer to do a more effective job.
The coarse sand filter was backwashed with air scour and water simul-
taneously at washwater rates far below fluidization for the media.
The backwashing sequence involved only two steps. The first step ap-
plied an air-water combination wash at rates of 7 scfm/sq ft and 9.7
gpm/sq ft, respectively, for 15 min. The second step was a waterwash
only cycle at the same rate for 5 min to expel as much air as possi-
ble from the media.
Sampling procedures and data collection. Sampling procedures and data
collection were similar to those described for Phases I and II, with
the following exceptions. An influent composite sampler was in oper-
ation during Phases III through V in addition to the effluent samplers
which had been in service during prior phases.
Observation runs were conducted two times per week during Phases III
through V. As before, an observation run consisted of carefully mon-
itoring and recording of the performance parameters throughout the
duration of the run. In most cases, all three filters were observed
160
-------
up to the point when the head loss reached the splitter box elevation
and the filter began bypassing. The run lengths for the dual- and
mixed-media filters were much shorter than for the coarse sand media.
Therefore, during observation runs, filters were backwashed whenever
they reached terminal head loss and put back in service so they would
operate at constant rate for the same total amount of operating time
as the coarse sand filter. Detailed observations were not continued
for these two filters after the first backwashing, but the composite
samplers were continued in operation (except during backwashing) to
obtain a sample from each filter over the same time period.
Results - Phases III through V
Visual observations - dual-media filter. The dual-media filter
(north filter) showed good backwashing in the first stages of the
study. The filter had been in operation five days, day no. 144, be-
fore the first mud ball appeared, 6 by 1 in. The small mud balls
grew over the next week until the air-water combination wash failed
to disperse the surface mat. Since the air-water wash could be run
just until the backwash water was near the overflow trough, the bed
received approximately 1 min of the combination wash. When the sur-
face mat was thick and highly compacted, this was not enough time to
completely break up the mat. However, the air-water wash did provide
excellent agitation throughout the entire bed when being applied.
The coal and sand were completely mixed after only the short time the
combination wash was applied. Complete fluidization was then required
to restratify the filter media. With the clean filter media in the
first part of this study, the fluidization and stratification was
easily accomplished. However, as the size and population of the mud
balls increased, jetting action occurred along with dead areas in the
bed. Sometimes sand remained mixed with the coal in various areas
above excessive mud ball deposits. A typical comment recorded in the
data book over the next month of operation was 5 to 6 mud balls, 2 to
3 in. in diameter, excellent air-water wash agitation followed by
complete fluidization and restratification. After over a month of
operation, two successive days of observations indicated no mud balls
present in the dual-media filter. However, the backwash for day no.
177 failed to break up an 8 by 3-in. agglomerate. The agglomeration
sunk to the interface, channeling the backwash water around it and
producing a "dead" space above it. The following backwashes had
similar results with the formation of an increasing number of mud
balls and less effective backwashing. Water channeling became com-
mon, and cracking was evident 1-1/2 in. into the bed. During the
next few days, the top 2 to 3-1/2 in. of media became heavily packed
with solids, which made the air-water combination almost totally in-
effective. The bed was in extremely poor condition on day no. 181,
with numerous dead areas and layers of previous surface mats resting
in the coal. On day no. 183 the backwashing procedure was modified
in an attempt to improve the condition of the bed.
161
-------
The new procedure added a step which applied air only. The complete
backwashing technique was now 2 min of air scour, air-water wash to
just under overflow, and, finally, 5 min of waterwash only. This
modification immediately improved the condition of the bed. Since
the coal was not so agglomerated now, the air-water wash efficiency
was improved as well. It was noted that the action of the air scour
alone was most effective during the first 30 sec of application;
after that the scouring action was confined to only the top 2 to 3
in. of media. The condition of the bed improved, as indicated by the
decreasing number and size of mud balls and agglomerates until on
day no. 188, the comment recorded was, "...a few mud balls present
but not to any great extent." It would seem the addition of the air-
scour cycle improved the bed condition of the dual-media filter. It
was at this time, trouble developed with the strainers, and the fil-
ter was rebuilt, signaling the commencement of Phase IV.
After the filter was rebuilt, the same backwashing procedure was
still used. The first week showed no problems or mud ball formation,
rather complete mixing during the combination wash, good fluidization,
and very distinct restratification. Small mud balls (2 to 3 in. in
diameter) were observed the following week until during the third
week of operation large pieces of the surface mat (2 by 6 in.) were
observed falling into the bed. Day no. 221 and 222 note large mud
balls (3 by 4 in. and 4 by 5 in.) falling to the interface. Typi-
cally, along with these large agglomerations, several smaller (1/2 to
1 in.) were noted as well.
On day no. 232, a special backwashing sequence was performed to try
and rid the dual-media filter of its high surface layer head loss.
The first step of the special backwash was simply the routine back-
wash procedure combined with turbidity measurements of the dirty
washwater collected at 30-sec intervals. After completion of step 1,
it was noted there was a considerable amount of gray organic matter
on the surface of the filter along with some 2 by 3-in. mud balls 8
in. below the surface. Step 2 of the special backwash was: (a) 5
min of air alone, (b) air-water wash for approximately 30 sec as the
water rose to the overflow trough, and (c) 5 min of water alone, all
at the same rates as previously used. Six small (1 by 2 in.) mud
balls were observed following Step 2. Following this day, the rou-
tine backwashing sequence was changed to that of Step 2.
On day no. 235 it was noted that the top 8 in. of fluidized coal
were individual grains covered with a hairy, stringy slime. When
f luidization was stopped, this 8 in. of media was compacted into the
top 2 to 3 in. of media, causing excessive initial head loss. Spe-
cial observations were made on day no. 241 which described the string-
ers made of organic matter, 1/16 to 1/8-in. long, and fuzz-like in
appearance. A special backwash was conducted on day no. 242 to at-
tempt to correct this problem.
162
-------
The special backwash consisted mainly of simultaneous air-water wash
at low rates for an extended period of time. The media was subjected
to 1.5 scfm/sq ft of air and 7.5 gpm/sq ft of water for 21 min. The
media that was carried out of the filter was collected in a garbage
pail to be replaced later. Samples of the waste backwash water were
collected every minute. Figure 44 presents the suspended solids and
turbidity for the samples versus the time they were taken. The
stringy growths were reduced but not eliminated. The initial head
loss for the two days prior to this experiment was 1.19 ft and 1.28
ft, while after this day it was 0.86 ft.
The hairy growths were still a problem five days later, so another
special backwash was conducted. The lost media was again collected
and returned, and the bed was then restratified. No samples were
taken, but these.observations were made. The stringly attachments
did not appear as long or thick as before; however, the initial head
loss following the experiment was 1.38 ft. No positive changes were
seen in the condition of the dual-media filter, so chlorination was
tried. The filter was drained down to within 1 in. of the surface
and 200 ml of household bleach (5.257. sodium hypochlorite by weight)
was added, stirred in by air scour, and let set for 10 hr before
backwashing. No immediate results were seen. It was at about this
time, day no. 255, that a similar stringy coating was noted on the
mixed-media filter but in a much milder concentration. This condi-
tion was recorded for only a few days for that filter, and day no.
269 was the last day the stringy attachments were mentioned on
either of the two filters.
Following the disappearance of the stringy attachments, the operation
of the dual-media filter was rather routine to the end of the testing
period. The air scour alone provided good mixing and bed agitation
for the first minute, then the air passed upward through the bed in a
channelized manner, causing little action, except as the air bubbles
came out of the bed surface, which violently mixed the top inch of
media. The air-water wash accomplished excellent mixing, but the
average duration was only 30 sec. Typical mud ball observations were
4 to 5 in number and 1 to 3 in. in size. Every few days the surface
mat would be extremely thick, causing pieces of it to tumble down to
the sand-coal interface where they would disintegrate or reduce in
size.
Visual observations - mixed-media filter. As previously stated, the
mixed-media filter was equipped with a surface and subsurface washer.
The main observation which relates to mud ball formation and bed con-
dition was to note if the washers were working properly. Following
one day of operation, the surface washer turned freely with a 30 psig
line pressure. The subsurface washer did not turn, but the jet ac-
tion of the washer nozzles lifted coal layers within the bed. The
next week of operation caused excessive mud ball formation, still
without the turning of the subsurface washer. On day no. 146, a
booster pump was installed on the water supply to the surface and
163
-------
300-
RESULTS OP SPECIAL BACKWASH
DAY NO. 242
o SUSPENDED SOLIDS
TURBIDITY
80
70
60
50
40
30
20
A /BACKWASH"]10
0 I SUPPLY
I
I
I
0
6 8 10 12 14 16
DURATION OF BACKWASH, min
18
20 22
Fig. 44. Results of special backwash, day no. 242.
-------
subsurface washer line to increase the line pressure and possibly
free the "hung-up" washers; the line pressure was increased to 60 to
80 psig. This did not solve the problem, and the following day a
large 3 by 4-in. agglomerate and 10 to 12 small,1-in. mud balls were
observed. These mud balls fell to the interface and were still not
broken up because the subsurface washer was below the interface.
Finally, the rubber caps on the subsurface washer were removed and
the washer rotated by hand, freeing the washer so that it worked
properly for the next two weeks of operation.
The action of the surface washer was described in detail for several
observations. The washer violently mixed the top 3 to 4 in. of the
coal. Ineffective action was noted in the corners of the filter
housing, as could be expected with a rotary washer in a square
housing.
Sometimes the surface cake was broken up by this mixing action, while
at other times it simply broke into smaller pieces which, upon fluid-
ization, fell into the bed. Also during this same period, considera-
ble amounts of fine silica and garnet sand were observed working
their way into the coal layer. The condition of the bed gradually
deteriorated until channeling of the backwash water was common and
numerous mud balls, 4 to 6 of 2 to 3 in. in diameter, were observed.
On day no. 173, a backwashing procedure change was made to increase
the effectiveness of the surface washer. During the surface washer
only cycle, a small amount of backwash water was now added to expand
the bed so the surface washer was submerged in the top 1 to 2 in. of
media. At the low rate of application, the backwash had trouble
raising the thick, heavy surface mat to the surface washer, so the
ineffective wash was still present at times. The surface washer
worked best when the bed was completely fluidized, but since several
large mud balls (5 by 2 in., 4 by 2 in., 10 by 2 in.) were observed
on day no. 180, complete cleaning of the bed was still not accom-
plished. Shallow cracks were observed at the top of filter media.
Solids penetration into the bed in general was 2 to 4 in. The lower
12 in. of the coal contained silica and garnet sand and was highly
agglomerated on day no. 191, just before the filter was rebuilt. Mud
balls and agglomerates were so dense that complete fluidization and
stratification was not attainable.
The media depths were changed during the rebuilding so the washers
were relocated to better coincide with the interface and surface.
Also, modifications of the washers themselves were made, with changes
in the washer arms and nozzles.
The line pressure was commonly 60 to 80 psig after the modifications,
which helped break up mud balls and surface mats in the first few
runs. Typically, both the surface and subsurface washers worked well
at times if all the water was channeled through each one separately.
However, both washers would not turn at the same time unless almost
165
-------
all of the water went through the subsurface washer; this condition
required a very delicate balance and was not reproducible from day
to day. The bed remained in good condition for the first two weeks;
no mud balls were observed on day no. 210 and 211. Both the washers
worked fine in the following days, and on. day no. 216, pieces of the
surface mat that were not destroyed by the surface washer, fell and
were broken up by the subsurface washer. One or two mud balls of 1
to 2 in. in size were typically observed during this period of opera-
tion.
For the next several weeks of operation, the equipment worked fine,
and a very few small mud balls were observed. Smaller mud balls (less
than 1 in. in diameter) were seen floating between the washers, too
small to fall down and be broken by the subsurface washer. No cracks
were observed, and penetration of solids was estimated at between 4
and 5 in. The media was reported to be very clean for day no. 235,
244, and 247. Both the fluidization and stratification were easily
accomplished with the clean, almost mud ball-free media. It was at
this time, day no. 256, that the stringy attachments similar to those
observed on the dual-media filter were noticed.
A special observation was performed on day no. 241 to document in
detail the operational characteristics of the washers. When the
surface washer was on by itself, the line pressure was about 70 psig
with the washer rotating at approximately 80 rpm. A steady-state
revolution condition was hard to maintain when both washers were
operating; 27 rpm was typical for both. The subsurface washer alone
operated at 75 psig and 27 rpm. The surface washer provided violent
agitation to the top 6 in. of media when it was immersed in the flu-
idized bed, but the water fluidization (only) cycle provided very
little action. On day no. 263 trouble with the subsurface washer was
noted; only 20 rpm was observed. The surface washer revolved at 100
rpm but was throttled down to 60 rpm for the backwash.
The first night in October, day no. 274, was extremely cold, causing
the surface washer booster pump to freeze and split. A used pump was
installed, but the washers turned at a slower rate thereafter, and it
was feared dirt particles in the replacement pump plugged a portion
of the washer nozzles. Because the surface washer turned much slower
now and sometimes not at all, mud balls began to develop in the bed.
Four to five, 2-in. and 10 to 12 less than 1-in. mud balls were com-
mon. In the remaining month of operation several days of no washers
rotating or only one of the two washers rotating were noted. The
bed condition quickly showed signs of the loss of the auxiliary
washers. Finally, on day no. 293, a length of pipe was used to manu-
ally reach into the filter and turn the washers. The surface washer
was freed quickly, but the subsurface washer failed to turn at all.
In the process of trying to free the subsurface washer, the bed was
fluidized and stirred for approximately 30 min, breaking up all the
166
-------
mud balls present. The subsurface washer failed to turn again for
the remainder of the study.
Visual observations - coarse sand filter. As previously stated, the
coarse sand filter was backwashed for a longer period at wash rates
below minimum fluidization velocity. The backwashes were all very
similar and rather "dull" as compared to the other two filters. No
mud balls were formed, fluidization did not occur, and extensive
descriptions of the backwash sequence were not warranted. Very sim-
ple, short data notes completely described all the action taking
place. The first backwashes provided only 5 min of air-water wash
followed by 3 min of water only. The dirt could be seen moving up-
ward and out of the bed. The typical backwash pulsed the top 6 to
8 in. of the bed. Pronounced air channeling causing jetting which
mounded the top couple of inches of media and produced an uneven
surface for the rest of the study was noted. The mounds moved about
during successive backwashes. Early in the study, the air-water wash
was changed to 15 min in duration and the waterwash to 5 min. Nine
days into Phase III, the bed began to show signs of dirt accumula-
tion.
Dark grey areas were starting throughout the bed except in the top 4
in., which retained their original appearance. The pulsing action was
noted in the top 12 to 18 in. as streaks of dirt were seen on the
plexiglass window being carried away. The bed cleaned up nicely
after each wash, particularly the top 18 in. On day no. 157, darker
areas appearing in the bed were reported. A 2 to 3-in. vertical
strip along both edges of the window in the corners of the filter
box was covered with this matter. The dark patches at the window
also started to develop at the two-thirds depth and below. On day
no. 165, the dark areas shifted to the lower parts of the filter,
covering the bottom 15 in. of media. Penetration of solids into the
filter was recorded between 12 and 15 in. on day no. 185. The filter
condition remained very constant for the remaining four months of the
testing period. The bed cleaned very well with the pulsating action
except for the bottom 15 in. of the filter, scattered areas of anaero-
bic deposits which shifted position as the filter runs progressed,
and areas along the corners of the filter box.
A special investigation was conducted to determine if the bottom 15
in. of media showed any progressive increase in initial head loss
due to its dirty condition. Figure 45 shows the head loss in the
bottom 16 in. of the coarse sand filter. As can readily be seen, no
distinct pattern was indicated. The head loss increased when the 50%
rate increase was made at the start of Series IV. No definite con-
clusions can be drawn about this dark 15 in. in the bottom of the
filter. A possible explanation could be simply a wall effect in con-
junction with the strainers which were located at about 4 in. from
the window and 10.5 in. apart. They created a dead space near the
walls and between the strainers which received ineffective washing.
167
-------
00
o
£
«i
O
o
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.0
- o
INITIAL HEAD LOSS IN BOTTOM
16 in. OF COARSE MEDIA FILTER
PHASE III
2.1 gpm/sq ft
«— PHASE IV >k PHASE V
<— 3.2 gpm/sq ft
O O O
~ O
8
o
o o
°0
-------
The visual observation and operation of the coarse sand filter was
very routine. However, very late in the study, the left and lower
half of the bed became dirtier. By the growth of the dark areas it
was obvious the left strainer near the window was partially or fully
plugged. The study was terminated due to cold weather without any
further changes.
Abrasion tests. The abrasion test was also used in Phases III
through V as a direct measure of backwashing effectiveness. The re-
sults of the abrasion tests for the entire test period cannot be com-
pared because two different tests were performed. As previously
described, Phase III employed a complex, rather long abrasion test
procedure while Phases IV and V used a short modified version of the
same basic test. For that reason, the abrasion test results for
Phase III are presented in Fig. 46 and the results for Phases IV and
V in Fig. 47.
An abrasion test was performed on the clean filter media to obtain a
control or background level to see if the media itself was being
abraded. An unused sample of both the anthracite coal and coarse
sand was subjected to the longer abrasion test used for Phase III.
First, however, the coal was mixed for 45 min with the abrasion pro-
peller and then flushed with water until the rinse water remained
clear. The sand sample was subjected to only the rinsing water be-
fore testing. The mixing and rinsing of the filter media prior to
the abrasion test was to remove any dust remaining on the media from
shipment. The coal and sand samples were then subjected to the abra-
sion test procedure. The results for the new media using the Phase
III procedure were 1.38 mg/g for coal and 0.94 mg/g for sand. Sever-
al days later, the same media sampler were again subjected to the
abrasion test procedure, yielding 3.17 mg/g and 0.39 mg/g for the
coal and sand, respectively. It was this high value, 3.17 mg/g, for
the coal that prompted the change of the abrasion test procedure in
Phase IV. Too much abrasion of the coal itself was taking place
during the 30-min mixing period of the Phase III test procedure.
Since the abrasion testing procedure was modified at the start of
Phase IV, a new standardization was necessary. Here again, an unused
(out-of-bag) sample of both the coal and sand was mixed and flushed
as before. The abrasion test results using the new procedure were
0.04 mg/g for coal and 0.018 mg/g for sand.
Despite the fact that the testing procedure was modified during the
testing period, the same pattern can be seen for all three series.
The abrasion test results for Phase III, Fig. 46, shows the coarse
sand filter always had cleaner media than the other two filters.
Also, except for the very first and last values, the mixed-media fil-
ter was cleaner than the dual-media filter. Furthermore, the graph
shows an erratic buildup of solids for the dual-media filter. The
coarse sand and mixed-media filters show a rather level result, ex-
cept that there is some erratic behavior in the mixed-media filter in
the latter days of the series. Based on the abrasion test results of
169
-------
16.00
14.00
12.00
•s
s 10.00
o
LLJ
2 8.00
6.00
CK
CO
O
4.00
2.00
0.00
ABRASION TEST RESULTS
PHASE III
O DUAL MEDIA - AIR-SCOUR WASH
A MIXED-MEDIA - SURFACE WASH
D COARSE MEDIA - SUBFLUIDIZATION WASH,
150
160 170
CALENDAR DAY, 1974
180
190
Fig. 46. Standard abrasion test results for Phase III.
-------
ABRASION TEST RESULTS
PHASES IV & V
O DUAL MEDIA - AIR-SCOUR WASH
MIXED-MEDIA-SURFACE WASH
COARSE MEDIA - SUBFLUIDIZATION WASH
245 255 265
CALENDAR DAY, 1974
Fig. 47. Standard abrasion test results for Phases IV and V.
305
-------
Phase III, the simultaneous air and water wash of the coarse sand
below fluidization velocities maintains the filter media in cleaner
condition than the air scour and water wash for.the dual-media filter
and the auxiliary surface wash of the mixed-media filter. Also, it
would appear the auxiliary surface wash used on the mixed-media fil-
ter resulted in better backwashing than the air scour and water wash
as used on the dual-media filter up to day no. 180 of the series, and
thereafter they were comparable.
The results for Phases IV and V are shown in Fig. 47. The same basic
patterns similar to Phase III exist here as well. The simultaneous-
air and water wash of the coarse sand below fluidization velocities
was superior during the entire period of Phases IV and V. During
the first half of operating period, up to about day no. 252, the
auxiliary surface wash filter maintained a cleaner filter bed than
the air scour, fluidized water wash method of backwashing. After
that day, the test results vary a great deal. No pattern exists be-
tween the dual- and mixed-media filters as before. The only constant
aspect of behavior during this period was in the coarse sand filter
where very little, if any, buildup of solids is shown by the test
results. However, the other two filters and corresponding method of
backwashing show a gradual buildup of solids on the media. Some
operational difficulties were experienced throughout this period with
the surface and subsurface washers, especially after day no. 274.
They either turned very slowly or failed to turn at all at times.
This might explain the erratic results for the filter using surface
and subsurface washers. No explanation can be given to qualify the
results for the air-scour, water wash except to say it was caused by
the simple failure of the backwashing method to maintain the media
in clean condition. The effectiveness of the combined air-water
scour on the dual-media filter was no doubt hampered by brief time of
application, approximately 30 sec due to the short rise distance from
the media to the overflow trough. One can conclude from the abrasion
test results that the simultaneous air and water wash below fluidiza-
tion velocities is the superior backwashing technique. Also, when
working properly, the surface and subsurface washer backwashing aux-
iliary and the air-scour auxiliary as used in the dual-media filter
are roughly equivalent in effectiveness, with a possible edge in
favor of the former.
Initial head loss. The initial head loss readings throughout the
testing period were also recorded in an attempt to evaluate the ef-
fectiveness of the three backwashing techniques. The initial head
losses were monitored for each day of filter operation. All the
readings were standardized by allowing 15 min of filtration before
reading the piezometer tubes; later in the testing period they were
read after the head loss readings had stabilized, which ordinarily
took less than 15 min. Figure 48 graphically displays the initial
head loss readings of all the filter observation days and several
additional data points for non-observation runs around periods of in-
terest.
172
-------
U)
k.
C
u-
o
*
CO
o
_J
§•
UJ
X
_l
<
1—
z
2.0
1 1.5
UJ
J 1.0
UJ
X 0.5
5
0.0
1 Q.
. . O \SP
-* . 2.1 gpm/sq ft— *U-~ 3. 2 gpm/sq ft ^ Oo
r» « O
^> ^,0 ° 0 0 <>
~ °b o o ^ o ^ o^ ° ° "
Q cP oo OQ cP 0 0
3 °
^ I 1 1 1 1 1 1 I.I
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
O ' 150 170 190 210 230 250 270 290 310
u
CALENDAR DAY, 1974
Fig. 48. Initial head loss data for each filter for entire study.
-------
Before one can attempt to evaluate Fig. 48, the reader must be aware
of several points. There were several changes between Phase III and
the rest of the study. First of all, the media depths and piezometer
locations were changed for the dual- and mixed-media filters. There-
fore, the points representing the initial head loss in Phase III are
not under the exact same conditions as Phases IV and V. For the dual-
media filter, the initial head loss represents the head loss across
the entire bed, 27 in. that is, from the top of the filter, including
15 in. of coal, 9 in. of sand, and 3 in. of supporting gravel. For
the mixed-media filter, the values represent the head loss across the
filter bed from 3 in. into the bed to 27 in. into the bed. Inadver-
tently, there was no piezometer tap above the media bed, so the head
loss in the top 3 in. (approximately) of coal was excluded from the
measurement. Therefore, the total head loss presented corresponds
to the head loss across the 12 in. of coal, 9 in. of sand, and 3 in.
of garnet sand below and excludes the head loss in die first 3 in. of
coal at the top of the bed. Therefore, valid conclusions or trends
can not be made for the mixed-media filter in Phase III. The initial
head loss for the coarse sand filter is the head loss across the
entire 46 in. of media; there was no change of media or piezometer
tap location for this filter throughout the length of the testing
period. At the start of Phase IV, new bed depths and piezometer taps
were established in the dual- and mixed-media filters. The initial
head loss for the dual- and mixed-media filter now corresponds to the
head loss across the total bed depth. The media used in Phase III
was discarded and new media of the same gradation was installed.
Also, the filtration rate was changed from approximately 2.1 gpm/sq
ft for Phase III to approximately 3.2 gpm/sq ft for Phases IV and V.
Because of these changes, and because of the scatter of the measured
values, it is difficult to draw firm conclusions from the initial
head loss data.
Several observations can be made of Fig. 48. As stated earlier, an
increase in initial head loss is indicative of mud balls and agglom-
erates due to poor backwashing. During Phases IV and V, when valid
trends can be identified because there were no changes in experimen-
tal routine, the mixed-media filter with auxiliary surface and sub-
surface wash exhibits a more pronounced upward trend, especially at
the end of the testing period, than the other two filters. The dual-
media filter with air scour and water wash and the coarse sand filter
washed below fluidization velocities show little if any buildup in
initial head loss. The surface and subsurface wash equipment of the
mixed-media filter experienced operational difficulties in the latter
stages of Phase V as already discussed. Since the water fluidization
backwash alone could not maintain the bed in clean condition, the
initial head loss increased. On day no. 293, the filter bed was
fluidized and the surface and subsurface washers were turned manually
by reaching down into the bed with a length of pipe. The bed was
fluidized approximately 30 min, and the washers were turned around
50 times with the washer supply turned on, in an effort to free the
washer. In the process, the stirring action broke up all the mud
174
-------
balls and agglomerates within the bed. This cleaning of the bed
corresponds to the drop in initial head loss for the next few days.
Since the washers were not freed, the buildup of mud balls returned
shortly and continued to the end of the testing period, and the higher
values of initial head loss returned. Thus, the rising initial head
loss in the mixed-media filter at the end of the study must be at-
tributed to the non- functioning of the surface and subsurface wash
auxiliary.
Head loss development. Some typical head loss curves during Phases
III through V are presented in Figs. 49 to 55. Only a limited number
are presented because they are all very similar in general appearance.
The curves show the head loss across various depths of media within
the filter bed (with zero depth at the top surface of the media) .
The initial bed depth of each filter was measured before each run and
was used to determine the actual depths of media covering the piezom-
eter taps. Since the initial bed depths varied throughout the study,
the depths of media reported on the head loss curves varied as well.
All the depths were measured to the nearest 1/4 in. and then rounded
to the closest 0.1 in. for presentation. Attention is again called
to the inadvertent omission of a piezometer tap above the media in
Phase III for the mixed-media filter as is evident from the top head
loss curves for that filter during Phase III.
A casual review of the curves shows the marked advantage of the coarse
sand "filter in terms of run length. Since all three filters were
operated at identical filtration rates, production would be directly
proportional to the run length. A comparison between the filters
with regard to head loss performance can be made in a number of
way?, e.g., run length to a given head loss, volume of production per
unit increase head loss, or solids capture per unit increase in head
loss. The latter method was selected because it can be used to com-
pare filtration studies conducted at different times and on various
types of influents and is sometimes used to predict head loss develop-
ment in filter design [133]. The influent and effluent suspended
solids concentrations, run length, flow rate, and the initial and
terminal head loss are the operational data needed to calculate a
solids capture by the following equation:
Solids capture value (SS . f i ~ ss f f j) (Run length) (Flow rate/unit area)
-- (Head loss increase) -
Table 17 presents average solids capture values for all the observa-
tion runs reaching a total head loss of at least 4 ft. Phase IV is
omitted from the table because of uncertainty about the suspended
solids analysis results and will be discussed later. As would be
expected from the head loss curves which have been presented, the
coarse sand filter has a higher solids capture value for both series.
The average solids capture values for the dual- and mixed-media fil-
ters are a factor of two or three times smaller than the coarse sand
175
-------
DUAL MEDIA - AIR SCOUR WASH
DAY NO. 155
6 — FLOW RATE 2.15 gpm/sq ft
5 —
BED DEPTH
INITIAL 35.5 in.
FINAL 34.5 in.
0-27.5 in.
MIXED-MEDIA - SURFACE WASH
DAY NO. 155
- FLOW RATE 2.15 gpm/sq ft
BED DEPTH
INITIAL 38.8 in.
FINAL 37.8 in.
024 6 8 10 120 2 4 6 8 10 12
TIME FROM BEGINNING OF RUN, hours TIME FROM BEGINNING OF RUN, hours
J4
o
4:
COARSE MEDIA-SUBFLUIDIZATION WASH
DAY NO. 155
FLOW RATE 2.15 gpm/sq ft
BED DEPTH
INITIAL 45.0 in.
FINAL 45.0 in.
10 12 14 16
TIME FROM BEGINNING OF RUN. hn
18
20
0 to 45 in.
22
24
Fig. 49. Chronological head loss development at various media
depths, day no. 155, Phase III.
176
-------
6 -
5 -
£
031-
DUAL MEDIA - AIR-SCOUR WASH
DAY NO. 183
FLOW RATE 2.05 gpm/iq ft
BED DEPTH
INITIAL 35. Jin.
FINAL 34.3 In.
0 to 27.5 in.
3.5to 27.5 In.
MIXED MEDIA - SURFACE WASH
DAY NO. 183 F
_ FLOW RATE 2.05 gpV«l ft
BED DEPTH
INITIAL 39,0 in.
FINAL
to27in
7 to 27 in.
1 to 27 In
27 In
O23to27ln
2 4 6 8 10
TIME FROM BEGINNING OF RUN, hn
12
2 4 6 8 10
TIME FROM BEGINNING OF RUN, hn
12
COARSE MEDIA - SUBFLUIDIZATION WASH
DAY NO. 183
- FLOW RATE 2.07 gpn\Aq ft
BED DEPTH
INITIAL 44.0 in.
_ FINAL 43.3 in.
§3
0 to 44 in.
048
16 20 24 28 32
TIME FROM BEGINNING OF RUN, hn
40
Fig. 50. Chronological head loss development at various media
depths, day no. 183, Phase III.
177
-------
i
*/> -
03
Q
DUAL MEDIA- AIR-SCOUR WASH
DAY NO. 239
FLOW RATE 2.92 gpm/sq ft
BED DEPTH
INITIAL 38.3 in.
FINAL 36.5 in.
0 to 30.3 in.
1 —
6.3 to 30.3 in.
14.3 to 30.3 in.
.3 to,30.3 in,|
MIXED MEDIA - SURFACE WASH
DAY NO. 239
FLOW RATE 2.96 gpn/sq ft
BED DEPTH '
INITIAL 39.0 in.
FINAL 37.8 in.
Ote27 in.
3 to 27 in.
7 to 27 in.
O—O-o 19 to 27 In.
23 to 27 ip.
2 4 6 8 10
TIME FROM BEGINNING OF RUN, hn
12 0
2 4 6 8 10
TIME FROM BEGINNING OF RUN, hn
12
.. 5 •
93
COARSE MEDIA -SUBFLUIDIZATION WASH
DAY NO. 239
FLOW RATE 2.96 gpm/sq ft
BED DEPTH
INITIAL 43.0 in.
FINAL 43.0 in.
8 10 12 14 16
TIME FROM BEGINNING OF RUN, hn
20
22
24
Fig. 51. Chronological head loss development at various media
depths, day no. 239, Phase IV.
178
-------
23
o
DUAL MEDIA - AIR-SCOUR WASH
DAY NO. 240
FLOW RATE 3. 16 gpm/sq ft
BED DEPTH
INITIAL 38. 5 In.
FINAL 36. 5 in.
0 to 30.5 in.
.5 to 30.5 in.
30.5 in.
-cyo22.5to30.5in.
MIXED MEDIA - SURFACE WASH
DAY NO. 240
FLOW RATE 3.20 gpm/«q ft
BED DEPTH
INITIAL 39.5 in.
FINAL 37.8 in.
J) to 27.5 in.
3.5 to 27.5 in.
7.5 to 27.5 in.
,19.5 to 27.5 in.
23.5 to 27.5 In.
2 4 6 8 10
TIME FROM BEGINNING OF RUN, hn
12 0 2 4 6 8 10
TIME FROM BEGINNING OF RUN, hn
12
COARSE MEDIA - SUBFLUIDIZATION WASH
DAY NO. 240
FLOW RATE 3.24 gpm/sq ft
BED DEPTH
INITIAL 43 in.
FINAL 42 in.
6 8 10 12 14
TIME FROM BEGINNING OF RUN, hn
22
24
Fig. 52. Chronological head loss development at various media
depths, day no. 240, Phase V.
179
-------
8,
_i 3
jOto 29 in.
DUAL MEDIA - AIR-SCOUR WASH
DAY NO. 267
FLOW RATE 3.16 gpm/sq ft
BED DEPTH
INITIAL 37.0 in.
FINAL 35.5 in.
0 to 27 in.
to 27 in.
MIXED MEDIA - SURFACE WASH
DAY NO. 267
FLOW RATE 3.20 gpm/»q ft
BED DEPTH
INITIAL 39.0 in.
FINAL 37.0 in.
7 to 27 in.
19 to 27 in.
23 to 27 in.
24 6 8 10
TIME FROM BEGINNING OF RUN, hn
12
2 4 6 8 10
TIME FROM BEGINNING OF RUN, hn
12
$5-
I
COARSE MEDIA - SUBFLUIDIZATION WASH
DAY NO. 267
FLOW RATE 3.24 gpm/sq ft
BED DEPTH
INITIAL 42.0 in.
FINAL
6 8 10 12 14 16
TIME FROM BEGINNING OF RUN, hn
20
22
24
Fig. 53. Chronological head loss development at various media
depths, day no. 267, Phase V.
180
-------
0 to 28.8 in.
DUAL MEDIA - AIR-SCOUR WASH
DAY NO. 269
FLOW RATE 3.04 gpnvAq ft
BED DEPTH
INITIAL 36.8 in.
FINAL 35.5 in.
4.8 to 28.8 in.
to 28.8 in.
20.8 to 28.8 in.
0 to 26.3 in.
2.3 to 26.3 in.
MIXED MEDIA-SURFACE WASH
DAY NO. 269
FLOW RATE 3.08 gpm/sq ft
BED DEPTH
INITIAL 38.3 in.
FINAL 37.3 in.
6.3 to 26.3 in.
14.3 to 26.3 in.
22.3 to 26.3 in.
2 4 6 8 10
TIME FROM BEGINNING OF RUN, hn
12
24 6 8 10
TIME FROM BEGINNING OF RUN, hn
12
I
S.
COARSE MEDIA - SUBFLUIDIZATION WASH
DAY NO. 269
— FLOW RATE 3.12 gpmAq ft
BED DEPTH
INITIAL
FINAL
42.0 in.
42.0 in.
0 to 42 in.
2 to 42 in.
10 to 42 in.
to 42 in.
8 10 12 14 16
TIME FROM BEGINNING OF RUN, .hn
18
20
22
24
Fig. 54. Chronological head loss development at various media
depths, day no. 269, Phase V.
181
-------
0 to 28.5 In.
DUAL MEDIA - AIR-SCOUR WASH
DAY NO. 281
FLOW RATE 3.08 gpm/tq ft
KD DEPTH
INITIAL 36.5 in.
FINAL 35.0 in.
4.5 to 28.5 In.
8.5 to 28.5 in.
16.5 to 28.5 in.
0 to 26.5 in.
2.5 to 26.5 in.
MIXED MEDIA - SURFACE WASH
DAY NO. 281
FLOW RATE 3.08 gprn/m ft
BED DEPTH
INITIAL 38.5 In.
FINAL 36.5 in.
6.5 to 26.5 in.
18.5 to 26.5 in.
22.5 to 26.5 in.
24 6 8 10
TIME FROM BEGINNING OF RUN. hn
12
2 4 6 8 10
TIME FROM BEGINNING OF RUN, hn
12
COARSE MEDIA - SUBFLUIDIZATION WASH
DAY NO. 281
FLOW RATE 3.20 gpm/x, ft
BED DEPTH
INITIAL 42.0 in.
FINAL 41.5 in.
0 to 42 in.
6 8 10 12 14 16
TIME FROM BEGINNING OF RUN, hn
18
20
22
24
Fig. 55. Chronological head loss development at various media
depths, day no. 281, Phase V.
182
-------
Table 17. Solids capture per unit head loss results for direct fil-
tration of trickling filter effluent, 1974.
Solids capture value8
Phase III
Phase V
Dual
media
0.081
0.042
Mixed
media
_b
0.038
Coarse
sand
0.156
0.140
a
Ib SS captured/sq ft/ft head loss increase.
Omitted because no piezometer in operation above the media surface
during Series III.
filter. This higher value for the coarse sand filter indicates a
slower rate of head loss development and longer run lengths than for
the dual- and mixed-media filters. The dual- and mixed-media filters
were approximately the same in Series V, indicating similar head loss
development patterns. Therefore, no conclusion about backwashing
effectiveness between these two filters is warranted from the head
loss development patterns.
The shape of the head loss curves also is an interesting finding.
Most of the curves for the dual? and mixed-media filters show a steep,
straight increase (i.e., day no. 269) or an exponential curve (i.e.,
day no. 240). The exponential curve is indicative of partial surface
filtration of compressible solids. The straightline head loss curve
found in this study is believed to be caused partly by surface fil-
tration as well, but bed compression and surface cracking clouded the
results as discussed in the Phase I results. However, the head loss
curves for the coarse sand filter develop at a much slower rate.
An interesting result is shown in Fig. 52 on the head loss curve for
the coarse sand filter on day no. 240. The initial head losses were
recorded, but as the filtration run progressed the head loss or re-
sistance decreased gradually then increased gradually. This pattern
of decrease followed by increase is present in every single head loss
curve for the coarse sand filter. One possible explanation for this
strange behavior is the following. After the air and w.ater wash is
finished, a water only wash is applied to flush out the dirt parti-
cles within and above the media and to force out the air remaining
in the bed. It was observed that even after this water wash there
were a large number of air bubbles still remaining in the bed. The
bubbles take up volume, which reduces the available space for water
passage, thereby increasing the actual flow rate and initial head
loss. It is hypothesized that throughout the filtration run the
air is dissolving into the wastewater (which is generally below the
183
-------
saturation level in oxygen) while the head loss is increasing due to
the accumulation of solids. At the beginning of the run the air is
dissolving at a faster rate than head loss buildup which would pro-
vide an increased filtration area and a lower head loss value. This
occurs up to a point where the oxygen is completely dissolved or the
rate of dirt accumulation is greater so the head loss development
curve would then have an increasing pattern..
Water quality. The means and standard deviations of the various water
quality parameters are given in Tables 18 through 20 for Phases III
through V. The analytical tests were performed on composite samples
of the influent and effluents taken throughout each observation run.
Turbidity measurements were taken approximately every two hours for
the entire length of the filtration run. All of the values for a
particular run were averaged to arrive at one number representing a
turbidity value for each observation run. The first six tests re-
ported in the tables, that is, BOD, soluble BOD, suspended solids,
TOC, SOC, and turbidity, represent parameters commonly used to mea-
sure the removal efficiency of the filters. For these parameters,
the average lab test result value, which appears in Tables 18, 19
and 20, was calculated omitting partial data. If the analytical re-
sults for any particular observation run were missing a parameter,
i.e., BOD, TOC, etc., for one or more filters, then all the other
data for that parameter, and run, partial data, were excluded in
averaging the lab test results. The remaining parameter averages,
that is, NH4, N03, N02, ORG N, OP04, and TPO^ were calculated in-
cluding all the data obtained from the lab tests.
The same sampling containers were used for Phases III and IV. During
Phase IV, the lab results seemed erratic, so new sample containers
were purchased and more rigorous washing of sample containers was
commenced at the beginning with the start of Phase V. It was be-
lieved that the old sample containers were not being adequately
washed during Phase IV, so the conclusions drawn about filter perfor-
mance will be based on Phases III and V.
Excluding Phase IV, the coarse sand filter effluent proved to be of
slightly, but consistently, poorer quality than the dual- or mixed-
media filter effluent. When comparing the dual- and mixed-media
filters, no such pattern existed. Comparing just these two filters,
virtually half of all the effluent quality tests were low for each
filter. Also, because of the small sample number and the large stan-
dard deviation, one would be hesitant to declare that even the coarse
sand filter produced the poorest effluent quality. The effluent
quality data between the filters are close, yet so variable, that no
firm conclusions can be drawn concerning performance differences.
Filter bed compression. An additional observation made throughout
the study was the amount of compression experienced throughout the
filter run. Excessive compression is indicative of dirty filter
media caused by inefficient backwashing procedures. When dirt
184
-------
Table 18. Summary of analytical test results for Phase III.
Filter effluent
BOD5 (mg/1)
N = 13a
Soluble BOD5 (mg/1)
N = 10
Suspended solids (mg/1)
N = 14
TOC (mg/1 as C)
N = 12
SOC (mg/1 as C)
N= 12
Turbidity (FTU)
W- 16
Ammonia (mg/1 as N)
Nitrate (mg/1 as N)
Nitrite (mg/1 as N)
Organic nitrogen
(mg/1 as N)
Orthophosphate
(mg/1 as P04)
Total phosphate
(mg/1 as PO^)
« = number of observation
O = standard deviation.
Influent
14.61
0=6. 00b
3.88
0=1.79
37.49
0=12.03
12.51
0=4.53
7.54
0=1.67
16.38
0=4.31
4.62
0=1.77
N«17
5.53
0=1.21
N=14
0.43
0=0.11
N=12
0.70
0=0.42
N=2
10.99
0=2.79
N=13
12.17
0=3.53
N=6
runs averaged,
Dual
media
3.73
0=1.72
1.97
0=0.96
6.84
0=3.23
8.06
0=2.24
6.97
0=1.43
2.38
0=0.97
4.48
0=1.74
N=17
5.92
0=1.50
N=14
0.68
0=0.33
N=12
2.30
0=1.98
N=2
10.18
0=3.06
^=13
9.27
0=2.92
N=5
Mixed
media
4.11
0=2.03
2.20
0=0.99
6.31
0=3.87
8.19
0=1.87
7.21
0=1.40
2.20
0=0.56
4.35
0=1.59
N=14
5.70
0=1.55
N=13
0.74
0=0.39
N=12
2.90
0=2.26
N=2
10.43
0=3.34
N=13
10.60
0=3.82
N=6
Coarse
sand
4.73
0=2.56
2.34
0=1.11
7.92
0=5.80
7.07
0=1.88
7.44
0=2.01
2.89
0=1.10
4.26
0=1.99
N=17
5.17
0=1.41
N-1A
0.65
0=0.34
N=12
4.65
0=5.16
N=2
10.03
0=2.88
N=13
10.49
0=3.03
N=6
185
-------
Table 19. Summary of analytical test results for Phase IV.
Filter effluent
BOD5 (mg/1)
N = 8a
Soluble BOD (mg/1)
N = 8
Suspended solids (mg/1)
N = 8
TOG (mg/1)
N = 8
SOC (mg/1)
N = 8
Turbidity (FTU)
N = 8
Ammonia (mg/1 as N)
Nitrate (mg/1 as N)
Nitrite (mg/1 as N)
Organic nitrogen
(mg/1 as N)
Orthophosphate
(mg/1 as PO.)
^T
Total phosphate
(mg/1 as PO,)
*T
Influent
17.04
i
C7=4. 21b
4.71
0-1.92
36.41
(7=9.22
11.84
-------
Table 20. Summary of analytical test results for Phase V.
Filter equipment
BOD5 (mg/1)
N = 15a
Soluble BOD5 (mg/1)
N = 15
Suspended solids (mg/1)
N=14
TOC (mg/1)
N = 10
SOC (mg/1)
N = 10
Turbidity (FTU)
N = 15
Ammonia (mg/1 as N)
Nitrate (mg/1 as N)
Nitrite (mg/1 as N)
Organic nitrogen
(mg/1 as N)
Orthophosphate
(mg/1 as PO.)
M*
Total phosphate
(mg/1 as PO,)
M-
Influent
30.38,
a=!4.52b
9.67
0=3.76
34.08
0=16.87
19.86
cr=8.03
13.41
0=3.22
17.60
CT=6.18
21.08
cr=5.30
N=15
3.09
o=i.77
N=6
0.48
0=0.24
N=6
2.13
o=l.28
N=2
24.61
(7=4.17
N=5
23.41
a=0.69
N-=4
Dual
media
12.68
or=6.88
7.21
0=3.72
7.05
0=4.27
12.02
0=4.16
12.00
a=3.98
4.80
CT=2.28
20.00
a=6.24
N=15
2.84
a=2.05
N=6
0.52
a=0.24
N=6
2.17
(7=1.95
N=2
24.67
0=6.09
N-5
20.66
a=3.10
N=4
Mixed
media
12.99
(7=6.82
7.27
(7=3.61
6.82
a=3.io
12.77
(7=3.20
11.83
(7=2.60
6.78
(7=3.01
19.79
(7=5.76
N=15
2.25
(7=1.78
N=6
0.46
(7=0.28
N=6
2.49
0=1.51
N=2
25.04
a=5.05
N^5
21.31
(7=2.69
N=4
Coarse
sand
14.46
a=6.56
7.78
0=3.57
9.46
(7=4.53
12.99
(7=3.96
12.98
(7=3.47
4.66
0-2.12
20.69
(7=5.47
N=15
2.65
0=1.27
N=6
0.44
0=0.15
N=6
1.81
0=0.65
N=2
25.62
0=5.14
N=5
23.39
0=0.75
N=4
N = number of observation runs averaged.
O = standard deviation.
187
-------
adheres to the filter media, the actual grains do not contact each
other but rest upon the dirt layers that surround the grains. As
head loss develops during filtration and thus applies a load to the
filter bed, these dirt layers compress, which results in a noticeable
compression of the entire bed. Clean media show no compression
during filtration since the grains touch one another and no compres-
sible layers exist. The head loss developments curves (Figs. 49
through 55) indicate the initial and final depth of the entire filter
bed. As can readily be seen, the dual and mixed filters have approxr
imately 1-in. compression in the early stages of the study progressing
up to 2-in. compression in Phase V. The coarse sand filter shows no
compression at all early in the study, with minor compression later.
The recorded compression of the coarse sand filter is less reliable
because an uneven surface results from the combined air and water
wash below fluidization velocity. Thus, the depths recorded are
average depths by visual estimation and subject to greater uncer-
tainty. These observations of compression reinforce the conclusion
previously established that the coarse sand filter was cleaner than
the other two filters. The dual- and mixed-media filters exhibited
roughly comparable degrees of compression, indicating both were in
somewhat dirty condition,
Filter cleanup operations. After 170 days or over five months of
testing, the three filters were removed from normal service and sub-
jected to clean up operations. All three filters were first subject-
ed to a normal backwashing sequence. This was followed by a new
backwashing technique to see if more dirt could be removed from the
media. Samples of the dirty backwash water were collected and com-
posited during the new backwash routine. The total solids of the
three composite samples were then determined.
The cleanup operation consisted of a prolonged air-water combination
wash below fluidization velocities. The water wash rate was 8 gpm/
sq ft and the air wash rate was 4 scfm/sq ft. The dual-media filter
was the first filter subjected to this new sequence. The filter
overflow was collected in a 30-gal. garbage pail to catch the media
being washed into the overflow weir. The dirty washwater samples
were collected at the weir overflow in the filter. For the first 2
min of backwashing, when samples were taken every 30 sec, approxi-
mately 20 ml of sample was composited. After the first 2 min,
samples were taken every minute, approximately 40 ml of sample was
composited for each collection. This was continued for 13 min until
the washwater was quite clear in appearance leaving the filter. Tur-
bidity measurements were attempted on all the samples, but the water
was so dirty it was doubtful the readings were within the capabili-
ties of the instrument, and they are therefore not reported. The
total solids of three aliquot samples of the washwater composite was
measured. The average result of the three aliquot samples for the
dual-media filter was 1019 mg/1. The underdrain plate was also re-
moved to see if any sand was evident from leakage through the dis-
turbed gravel. None was evident.
188
-------
The mixed-media filter was treated to the same cleanup procedure as
the dual-media filter. The dirt released in the first few minutes
was phenomenal. The water was literally black with solids and de-
scribed as much dirtier than the dual-media filter. The media quick-
ly worked its way about 6 in. into the gravel. The bed was complete-
ly mixed in a few seconds, and as the wash progressed the garnet
became more evident throughout the bed. Samples were taken and com-
posited in the same manner as in the dual-media filter but continued
for 20 min because it took longer for the washwater to clear up. The
30-sec sample was diluted 10 to 1 and still recorded a 60 FTU turbid-
ity reading, indicating a 600 FTU condition. However, since the tur-
bidity readings were so doubtful, the remaining samples were not
measured. The total solids of the washwater composite for the mixed-
media filter was 1415 mg/1.
The same backwashing procedure and sampling technique was performed
on the coarse media filter except that the air rate was increased to
the usual level for that filter (7 scfm/sq ft). The backwash was
very similar to those seen throughout the study for that filter and
was continued for only 10 min. The discharge water was very clear
throughout the wash compared to the dual- and mixed-media filters.
The total solids for the coarse media filter was only 616 mg/1.
Finally, all three filters were drained and air blown through them in
hopes the media could be left through the winter without breaking
the plexiglass window. No damage was noted the following spring.
Table 21 presents a summary of the data obtained during the filter
cleanup operation. As indicated by the visual observations, the
Table 21. Data summary for cleanup operation at the end of the
operating period in 1974.
Dual media
Mixed media
Coarse sand
Duration of
special backwash
procedure,
min
15
20
10
Total solids of
composite samples
of backwash water,
mg/1
1,019
1,415
616
Total dirt
released,
g/sq ft
1,041
1,928
419
mixed-media filter released the greatest total amount of dirt during
the cleanup operation, indicating it was in the dirtiest condition
prior to the cleanup. The difference between the filters would have
been more dramatically evident if suspended solids analyses had been
measured rather than the total solids which were measured inadver-
tently. The total solids values include soluble inorganic solids es-
timated at about 300 to 400 mg/1 which have nothing to do with the
dirt released in the cleanup operation. Nevertheless, the cleanup
operation results do show a marked advantage for the simultaneous air
189
-------
and water backwash of the coarse sand compared to the other two fil-
ters. The results also show an advantage of the air-scour auxiliary
for the dual-media filter compared to the surface and subsurface
auxiliary of the mixed-media filter. This is partly due to release
of solids from the gravel layers, and from the inside walls of the
mixed-media filter upon the first application of air-water wash, as
was also noted in Phase II.
Other backwash design investigations. A number of other observations
were made in Phases III through V to try to answer some important
design questions related to backwashing. Due to strainer clogging
problems in Phases I and II, strainers with coarser slots were used
in Phases III through V. These strainers necessitated the use of
gravel in the underdrains with the potential problem of gravel move-
ment, especially when air and water backwash were used simultaneous-
ly. Unfiltered secondary effluent was used for backwashing. The
advantage of this arrangement is that recycled backwash water does
not increase the hydraulic load to the filters. The disadvantages
are the increased hazard of strainer clogging and the potential ef-
fects of the dirtier backwash water on the underdrain gravel. The
details of these investigations follow.
Filter influent "feedwater" was used as a backwash supply for all
three filters in Phases III through V. That supply was the normal
secondary effluent in this case. Two problems arose as a result of
this backwash supply.
The first problem arose after 51 days of operation when there was a
slight increase in the underdrain pressure for the dual-media filter.
Two days later, it exceeded 15 psig (the gage limit) when normally
it was approximately 3.5 psig. Only 75% of the backwash rate pre-
viously used could be attained even when the control valve was wide
open. The next day there was a popping noise, and water streamed
upward along one side of the filter during the backwash cycle.
Originally, it was thought a nozzle had failed, but closer examina-
tion showed a gasket around the underdrain plate had failed.
The media was removed and the nozzles examined. A 1/4-in. thick ring
of coal was found around the inside of the nozzle slits. The back-
plate for the underdrain plenum was removed, and more coal and sand
was found. Although no failure had occurred in the mixed-media fil-
ter, a similar increase in underdrain pressure led to the same work
on that filter. Coal and sand were found there as well. The same
modifications and cleanup were performed on that filter also. The
coarse sand filter had no evidence of strainer plugging at the time
of this incident. The low wash rates were believed to be the reason
for this. Late in the Phase V, however, some plugging of one strainer
was evident at the window, as reported before.
It was believed that some media had been washed out of the filter and
pumped back into the underdrain. Since the influent pump and backwash
190
-------
waste discharge meet in the same collection box, this was a believ-
able explanation. An apparatus was installed which carried the back-
wash waste discharge out of the collection box, and no similar prob-
lems occurred.
The second problem observed was the accumulation of substantial black
solids in the supporting gravel layers of the dual- and multi-media
filters and in the bottom 15 in. of the coarse sand filter. It is
not possible to attribute the solids solely to the backwash water or
to the filtered water reaching those depths. Nevertheless, the gravel
layers were very dirty and gravel movement did occur as described in
more detail in the following paragraphs.
A double reverse graded gravel underdrain (coarse to fine to coarse)
was used in the dual-media filter as previously described to avoid
movement from the combined air and water backwash. As previously
stated, the double reverse graded underdrain was installed upside
down, and the backwashing with air and water simultaneously caused
almost immediate mounding of the fine gravel. Excessive mounding
required attention, so the media was removed and the gravel was
placed correctly in the filter two weeks later. No shifting or mound-
ing of the gravel was noted immediately following or for the next six
weeks following the change until a strainer clogging incident com-
pletely disrupted the bed. Even larger strainer openings were used
in the filter rebuilding (4.5 ram), so a revised gravel support was
used to accommodate the larger slots as previously described. No
gravel mounding or shifting was observed during the next three weeks.
However, some temporary cracks (horizontal, 1/2 by 6 in.) were noticed
in the finest gravel layer during backwashing throughout these weeks.
At about that time, some of the fine middle gravel escaped to the top
of the coarser gravel above. The details of the escape were not ob-
served or recorded, but it may have happened in a single wash, and
further progressive escape did not seem to occur. The escaped fine
gravel first appeared on the sides, but later it was observed to have
moved to the center as well. During backwashing the fine gravel was
carried up into the silica sand by jets, and some gravel remained
there after fluidization had ceased.
Since the gravel escape and mounding came so suddenly, one would
suspect that the desired water wash rate was exceeded during the com-
bination air-water wash cycle. The control valve was a 1/4-turn plug
valve, and the flow rate was difficult to control precisely. The
data book had no evidence of this, however, and no quick jumps or ex-
plosions in ths gravel were noted.
Nevertheless, the potential hazard of gravel movement was demon-
strated, and the opening of horizontal cracks in the fine gravel dur-
ing backwashing was recorded. Such cracks in a large filter could
result in an upset to the gravel layers. The accumulation of solids
in the gravel no doubt contributed to the pressure drop across the
191
-------
fine gravel during backwashing, and thus to the formation of the
cracks.
A conventional graded gravel was used in the mixed-media filter. It
proved to be unstable, even though washed with water only at normal
backwash rates reported as follows.
Within 20 days, the top fine gravel of the regular graded support
gravel was intermixing with the coarse support garnet layer. The
migration of the gravel upward and the coarse garnet downward was
very pronounced two days later. The coarse garnet progressively
worked through 6 in. of gravel while the fine gravel broke through
the coarse garnet layer and began mixing with the silica sand. The
fine garnet layer between the silica sand and the coarse garnet layer
had already mixed up into the silica sand in the first filter wash
and effectively disappeared from view. Finally, on day no. 168,
after 32 days of operation, the coarse garnet and the top fine gravel
were "totally mixed" and mounding. The mounding and shifting caused
channeling, leading to more mounding up to the point when the filter
was rebuilt due to the strainer clogging incident.
In the very first backwash after rebuilding, coarse garnet migrated
into the gravel base until just eight days later mounding of the
fine gravel and extensive mixing of the coarse garnet and gravel was
noted. Once again, the gravel mounded on the sides as before causing
jetting action. On day no. 234 the course garnet had worked to with-
in 3 in. of the false bottom. The same reports of coarse garnet and
fine gravel migration were recorded to the end of the study. Unlike
the double reverse graded gravel, the underdrain problems encountered
with the regular gravel were well documented and occurred gradually.
One can draw several wastewater filter design conclusions from the
foregoing observations. First, there are inherent dangers in the use
of feedwater for backwashing which must be recognized. For example,
if gravel is used in the underdrain, and if the backwash water should
accidentally contain an unusually high concentration of suspended
solids, these solids may be partially removed in the fine, gravel.
This could result in sufficient pressure drop across the gravel layer
to lift the gravel and cause an upset of the gravel.
Second, the double reverse graded gravel design used in the research
did not prove adequately stable to resist movement when backwashed
with air and feedwater simultaneously. The solids in the backwash
water may have contributed to the instability, and the research
should be repeated using filtered water for a backwash supply.
Third, the conventional graded gravel support was unstable when
washed with feedwater alone. This observation reiterates the weak-
ness of the conventional design reported by prior workers [10,11].
192
-------
Therefore, the use of underdrain strainers without supporting gravel
is a desirable design arrangement, but the use of fine slots of less
than 1 mm should be avoided, and feedwater should not be used as a
backwash. The advantages of using feedwater do not justify the risks
which result therefrom.
Summary and Conclusions - Phases III through V
The objectives of the experimental investigations of'these phases
were to compare the effectiveness of three backwashing techniques on
three different types of filters while filtering secondary effluent,
and to look at the problems of underdrain strainers and supporting
gravel when backwashed with unfiltered secondary effluent. The fil-
ters and backwashing techniques were as follows:
1. A dual-media filter backwashed with air scour followed by water
fluidization backwash. The filter media was supported on a
double reverse graded gravel.
2. A mixed (triple) media backwashed with a surface and subsurface
wash auxiliary before and during water fluidization backwash.
The filter media was supported on a conventionally graded
gravel.
3. A coarse sand media with a deeper bed backwashed with air scour
and water simultaneously at subfiLuidization velocity. This fil-
ter was supported directly on the underdrain strainers without
the use of gravel.
The experimental data were collected over a five-month period of con-
tinuous operation of the filters at the Ames, Iowa, trickling filter
plant, using pilot-scale equipment. All three filters were back-
washed with unfiltered secondary effluent (i.e., feedwater) through-
out the course of the study. The following conclusions resulted from
the study.
1. Of the three backwash methods, simultaneous air scour and sub-
fluidization water backwash of the coarse sand media proved to
maintain the cleanest filter media based on the abrasion test
results, bed compression data, visual observations, and a termi-
nal cleanup operation.
2. The use of air scour and subfluidization water backwash simul-
taneously on coarse sand media was able to keep the filter bed
completely free of mud balls, but there were dirty regions in
the bottom 15 in. of the filter and along the vertical corners
of the bed. These dirty regions did not impair the functioning
of the filter, and those in the corners would be inconsequential
in full-scale filters.
193
-------
3. The surface and subsurface backwash auxiliary in the mixed-media
filter and the air-scour auxiliary in the dual-media filter pro-
vided equivalent backwashing effectiveness. Both filters ex-
perienced mud ball problems and other indications of dirty fil-
ter media. Comparing these two methods alone, the air-scoured
filter was cleaner based on terminal cleanup observations,
whereas the filter with surface and subsurface auxiliary was
cleaner based on visual observations of the media and slightly
lower abrasion test results.
4. Both the double reverse graded gravel support and the conven-
tional graded gravel proved unstable as used in this research.
The work should be repeated using filtered water as a backwash
supply before deciding on the suitability of the use of gravel
in wastewater filters.
5. Feedwater is not recommended as a backwash water source because
of the danger of clogging underdrain strainers and/or gravel.
The advantages of using feedwater do not justify the risks which
result therefrom.
6. The underdrain orifice or strainer system should have suffi-
ciently large openings so that solids in the backwash water do
not cause progressive clogging problems. Media-retaining
strainers with slots less than 1 mm are not recommended. This
recommendation dictates the use of a sufficiently coarse filter
media or supporting gravel to prevent loss of media to the
underdrains.
7. Comparing the three filters of this study, the coarse sand fil-
ter produced a filtrate slightly poorer in quality than that
produced by the dual- and mixed-media filters but provided sub-
stantially more filtrate to a common terminal head loss. These
differences can not be attributed to the backwashing methods
used, but rather to the differences in the filter media.
Operation and Results - Phase VI
Coarse Sand Filtration of Secondary Effluent
This final research phase was devoted to further studies of coarse
sand filters backwashed with air and water simultaneously at subflu-
idization velocity. The favorable results with one filter of this
type in Phases III through V prompted this final phase to answer some
additional questions about such filters. It was concluded in Phases
III through V that this type of filter and backwashing routine re-
sulted in a cleaner filter bed than the dual- and multi-media beds
which were studied. However, there was a layer of dirty sand ob-
served in the bottom of the filter about 15 in. deep. It was also
observed that the filtrate was slightly poorer from the coarse sand
filter than from the dual- and multi-media filters.
194
-------
Therefore, Phase VI was conducted to determine (1) the effect of
depth of media on the extent of the dirty zone at the bottom of the
filter and (2) the effect of depth of media on the filtrate quality.
To study these questions, the three pilot filters were equipped with
new filter media of slightly coarser size, with depths of 24, 47, and
60 in. The shallowest filter media was supported on a double reverse
graded gravel of revised design to gain further experience on the
stability of the gravel when backwashed with air and water simultane-
ously. The three filters were operated for two months in 1975 fil-
tering secondary effluent at the Ames, Iowa, trickling filter plant.
The secondary effluent was used for backwashing throughout Phase VI.
Operation - Phase VI
Operating routine and sampling details were identical with Phases III
through V. The filters were backwashed every 24 hr. Observation
runs were conducted twice each week, with detailed observations re-
corded and composite samples collected for chemical analysis. De-
tails of media size, gravel gradations, and underdrain strainers have
been described previously in the equipment section.
Backwashing. The backwashing routine for the first 35 runs of Phase
VI included the simultaneous use of air at 7 scfm/sq ft and water at
8 gpm/sq ft for 15 min followed by water alone at 8 gpm/sq ft for 3 min,
This was a slightly lower water flow rate than used in Phases III
through V and was selected because this is the normal recommendation
of one filter manufacturer who promotes this type of filter in the
United States (Dravo Corporation). It was visually evident from the
start that the three filters were not as clean after backwashing in
Phase VI as the single coarse sand filter had been in Phases III
through V. Therefore, beginning with run 36 of Phase VI, the back-
wash flow rate was increased to 15 gpm/sq ft during both the combina-
tion air-water wash and the water wash (alone) which followed. The
period of combination wash was reduced to 10 min to maintain the
total water usage for each backwash approximately unchanged. This
routine continued to the end of Phase VI at run 51.
Results - Phase VI
The results of Phase VI will not be presented in as much detail as
were prior phases because the objectives were more limited.
Visual observations. The condition of the filter media was observed
during and after each backwash. Prior to run 35 (lower rate of water
backwash) the condition of the three filters was roughly comparable.
The descriptions of the media as reviewed through the plastic window
after the backwashes had been completed are characterized by the
following comments. The top 6 to 18 in. of the media appeared clean
except for dirty strips along the vertical corners of the filters.
These strips varied in width from 2 to 5 in. Below the clean area of
195
-------
the media, the sand was progressively dirtier toward the bottom of
the filter. At times, the clean region was not continuous with, for
example, a 6-in. clean area at the surface and another similar area a
bit deeper, separated by a horizontal dirty region in the bed. Small
black anaerobic strips and spots were generally evident but changed
position with time in the dirtier regions of the bed. No mud balls
were observed on the surface or anywhere within the bed.
The air-water backwash action was described as a feeble pulsing of
the top 10 to 15 in. of the bed, which was most noticeable in the
first couple of minutes of the air-water backwash.
Due to the relatively dirty condition of the bed reported above, the
water backwash rate was increased beginning with run 35, as previous-
ly described. The condition of the media improved immediately and
continued in a steady good condition for the remainder of the study.
The descriptions of the media for all three filters at the completion
of each backwash are similar. The media was consistently clean to
within 12 to 15 in. of the filter bottom except for dirty strips
along the vertical corners of the filter. These strips varied from
0 to 2 in. in width. The bottom 12 in. was quite dirty except for
two clean penetrations immediately above the two underdrain strainers
closest to the window. One or both of these penetrations would reach
nearly to the strainer level at the bottom.
The north filter with supporting gravel was slightly different. The
gravel remained dirty for its full 12-in. depth. The clean penetra-
tions above the strainers would reach the gravel surface, but a dirty
region between them would reach 6 in. above the gravel.
Small black anaerobic regions were occasionally reported in the dirty
strips in the vertical corners of the filter. They were seldom ob-
served in the bottom 12-in. dirty region of any of the three filters.
The air-water backwash action was described as a good pulsing action
throughout the bed except for the bottom 12 in. and in the vertical
corners. These regions remained dirty as a result.
In view of the marked difference observed before and after run 35,
when the rate change was made, some detailed observations of the
media movement were made at the end of Phase VI. The water rate was
increased stepwise from 8 to 20 gpm/sq ft. The media action was
carefully recorded.
The general behavior was an increase in the vigor and extent of
pulsing as the water rate was increased. In addition, and more im-
portantly, it was noted that a circulation of the sand occurred and
that the rate and the circulation increased with the water flow rate.
Apparently, the rising air and water moved sand up in the center of
the bed, because the sand was observed to move down slowly at the
196
-------
window. The observed details at different rates are summarized in
Table 22.
Table 22. Action of simultaneous air and water backwash on coarse
sand at subfluidization velocities,a (Air rate =
8 scfm/sq ft for all observations.)
Water rate,
gpm/sq ft Action of media
8 Slight pulsing on right side in top 3 ft. No
downward circulation patterns.
10 A bit more vigorous pulsing in a little wider zone
on right side in the top 4 ft of the bed. No
downward circulation patterns.
12 Pulsing full width of bed in top 1 ft, in left
side of top 3 ft and in right side of top 4.5 ft.
Very slow downward media circulation in the center
at about 0.5 in/min.
15 Pulsing about same as at 12 gpm/sq ft. Downward
circulation of media over entire width (except for
2-in. wide strips at corners) at a rate of 2 to
3 in./min. Movement downward continues to 18 in.
from the bottom.
20 Similar but faster action with downward circula-
tion of media at 4 to 10 in./min., the lower rate
observed where the rising pulsations are more pro-
nounced.
aAs observed in a 1.5 by 1.5 ft filter with 5 ft of sand and 2.5 to
3.7-ram size, water temp = 21 °C. No fluidization of the bed was
observed at any of the flow rates in the table.
It is believed that this circulation of media is essential for good
cleaning because it moves the media periodically through regions of
intense upward pulsing and movement in which the better cleaning ac-
tion occurs.
At the completion of the observations reported in Table 22, the sand
was removed from the filters, with special attention to the condi-
tions in the bottom dirty region of the bed. It was noted that cir-
cular zones of clean media were found directly above the five strain-
ers in the dirty region of the bed. In the north filter with gravel
support, the above pattern was noted from about 3 to 10 in. above the
gravel. In the west filter, without gravel, it was noted about 12 in.
197
-------
from the bottom of the filter. No details were reported for the
south filter. It was difficult to tell the upper extent of the zone
because part of the sand fell out of the filter spontaneously when
the back plate of the filter box was removed to facilitate removal of
the sand.
It was also noted that some filter sand had escaped into the under-
drain plenum in the west and south filters, which were not equipped
with gravel. About two pounds was reported in the west filter. This
observation is not surprising since these filters had strainers with
4.5-mm slots, and the media was 2.5 to 3.7 mm in size range. In
fact, it is surprising that more media did not move downward through
the strainers. The use of the 4.5-nm strainer for this sand was an
oversight; it should not have been used.
Underdrain gravel stability. The double reverse graded gravel in the
north filter again proved to be not completely stable. On about 407«
of the backwashes, a horizontal crack was observed to open in the
finest gravel layer at the beginning of the air-water backwash. The
thickness of the crack varied from 1/8 to 3/4 in. and the length from
a few inches to the full width of the filter. It generally contract-
ed as the wash proceeded.
The gravel did not upset in Phase VI as it had in Phase IV, but the
potential for upset remains evident in these observations. There-
fore, the prior conclusions drawn from Phases III through V about the
use of gravel in filters washed with air and feedwater simultaneously
remain unchanged.
Water quality. Fewer water quality parameters were measured in Phase
VI than in previous phases due to budget limitations. The means and
standard deviations for the analyses of the composite samples are
reported in Table 23. There was no apparent difference between the
performance of the three filters of different depth. There appears
to be a very slight advantage in the 60-in. depth as measured by sus-
pended solids and turbidity, but it would be impossible to prove it
statistically.
By comparing the values of each run using a ranked analysis, some re-
inforcement of that conclusion is obtained. In this analysis, the
rank of 1 is assigned to the lowest value, the rank of 2 to the mid-
dle value, and the rank of 3 to the highest value for a particular
filter run. Comparing the turbidity values in this fashion, the 60-
in. deep filter had seven lowest, three middle, and one highest
value for a total score of 16. The 24-in. filter had a total score
of 22, and the 47-in. filter a total score of 28. The 24-in. filter
was supported on 12 in. of gravel in a rather dirty condition, which
may have contributed to the filtration and resulted in its exceeding
the performance of the 47-in. filter. Using a ranked analysis in the
manner above also placed the 60-in. filter in first position in sus-
pended solids removed and BOD5 removed.
198
-------
Table 23. Results of analyses during direct filtration of secondary
effluent (Phase VI) from June 24 through August 2, 1975,
using course sand filters of different depths.
Filter effluent
Suspended solids (mg/1)
a (N = il)c
Turbidity (FTU)
a (N = ll)
BOD5 (mg/1)
a (N - ll)
Soluble BOD5 (mg/1)
a (N = ll)
Filter
influent
31.3
9.7
12.6
3.14
15.6
4.7
5.3
1.7
N filter,
24-in.
depthb
5.9
2.1
3.30
1.21
6.5
2.8
3.9
1.3
W filter,
47-in.
depth
6.4
2.3
3.38
1.14
7.1
2.5
4.0
1.3
S filter,
60-in.
depth
5.7
1.8
3.14
1.14
6.6
2.5
3.8
1.3
Filtration rate 3.0 gpm/sq ft.
Filtrate from 24 in. of sand and 12 in. of supporting gravel.
^
CT = standard deviation, N = number of observations.
Looking at the mean values in Table 23, one must conclude, however,
that there was little gained by using the deeper media. In the fil-
tration of secondary effluent, one would need to use alternate ap-
proaches to reach a higher quality filtrate, either chemical pre-
treatment or a finer filter media. Deeper media should prolong the
period of acceptable filtrate if higher filtration rates and/or ter-
minal head losses had been used. In this work, however, deteriora-
tion of the effluent at the end of the filter run was not observed
on any of the filters.
Head loss patterns and initial head loss. The shape of the head loss
development curves in Phase VI were very similar to those observed
for the coarse sand filter in Phases III through V. Therefore, no
additional curves are presented for Phase VI. Furthermore, there was
no apparent difference between the curves for the three filters of
different depth in Phase VI, except for a different initial head
loss. As expected, the initial head loss was higher for the deepest
filter since they were all operated at the same filtration rate.
An analysis of the head loss developed to a common time for each ob-
servation run is summarized in Table 24. A total of 11 observation
runs are included, the same 11 runs for which water quality data were
199
-------
Table 24. Mean total head loss during filtration of secondary
effluent on coarse sand filters during Phase VI.
N filter,
24 in.
sanda
Mean total head loss , ft 4.49
CT 0.48
W filter,
47 in.
sand
4.35
0.55
S filter,
60 in.
sand
4.54
0.49
&Head loss includes 12 in. supporting gravel.
Head loss for an average run length of 16.18 hours (o = 1.89 hours).
presented in Table 23. It is apparent that very little difference in
head loss was observed between the three filters. This would be ex-
pected since they have the same media size and the suspended solids
removed was nearly the same for three filters, as shown previously.
The initial head loss for the three filters was different due to the
different depths of media. One observation of interest is the change
in initial head loss that occurred when the backwash flow rate was
increased after run 35. The average initial head loss values for a
few filter runs before and after the change in backwash rate are pre-
sented in Table 25.
Table 25. Average initial head loss for three coarse sand filters
before and after run 35 in Phase VI, when increase of
backwash rate was adopted.
Period
Runs 26-34
Runs 36-42
Runs 43-47
(inclusive)
(inclusive)
(inclusive)
N filter,
24 in.
sand*
0.84b
0=0.07
0.45
0=0.03
0.64
0=0.04
W-filter,
47 in.
sand
0.89
0=0.04
0.55
O=0.00
0.64
0=0.07
S filter,
60 in.
sand
1.15
O=0. 09
0.69
O=0. 03
0.82
0-0.05
loss includes 12 in. of supporting gravel.
All values in feet.
200
-------
It is clearly evident from Table 25 that a marked reduction in ini-
tial head loss occurred at the higher backwash rate. This indicates
that the bed was somewhat dirty before the change and that the new
backwash routine achieved a cleaner bed.
There was again some deterioration indicated by the initial head loss
in the later runs of Phase VI, but the series was not continued long
enough to see if that trend continued.
It would be desirable to compare these initial head loss data with
values for the clean media at the beginning of Phase VI. Unfortu-
nately, the operating routine used prior to Run 26 was not such as to
obtain reliable initial head loss data. This weakness did not in-
fluence the head loss data for the bulk of each run, only the initial
value recorded.
The mean removal of suspended solids, the mean run length and in-
crease in head loss (total head loss - initial) and the filtration
rate were used to calculate the solids capture value as explained in
the Phase III through V discussion. The average value for all these
filters in Phase VI is 0.16 Ib SS captured/sq ft/ft of head loss in-
crease. This value compares favorably with the value obtained in
Phases III and V in Table 17.
Media loss in air-water backwashing. The primary advantages of
coarse sand filters washed with air and water are (1) more effective
backwash and (2) lower headloss development, and thus longer filter
runs. The primary danger of this type of filter and backwash is the
potential loss of media which can occur during overflow due to the
violence of the air-water action.
Some preliminary observations on this question were made in the lab-
oratory using a 5.5-in. diameter filter column of transparent plastic
construction. Sands of two sizes were observed in the column while
being washed with air and water simultaneously at various rates. The
sand depth was 24 in. The height to which the sand grains were
thrown by the backwash was observed visually. This was admittedly
difficult to judge because of the presence of the air bubbles and the
violence of the motion.
The results of the observation are summarized in Table 26 for the two
sand sizes. The data are admittedly very limited and should be re-
peated covering a broader range of flow rates and media sizes. They
are presented here for the record, merely to call attention to the
danger of media loss, and the need for adequate distance from the
media to the overflow in the filter design.
No loss of sand was observed during the two months of observation of
the three filters in Phase VI. No loss was expected because of the
large distance which existed from bed surface to overflow, a minimum
of 3 ft in the filter with 5 ft of media.
201
-------
Table 26. Height that sand is thrown by simultaneous air and water
backwash.
Height sand thrown
water rate,
gpm/sq ft
2
2
4
4
6
6
Air rate,
scfm/sq ft
4
8
4
8
4
8
1 to 2- mm sand, in.
11-12
12-13
11-12
13-14
15-16
16-17
2 to 3 . 6-mm sand , in .
nil
nil
5-6
5-6
5-6
5-6
Some small loss of media appeared to have occurred in prior Phases
III through V on the single course sand filter then in use. During
five months of operation, the recorded depths of filter media de-
creased 2 to 3 in. The uneven surface of the sand made the surface
depth measurement difficult, and thus the loss reported above is only
approximate. The distance from bed surface to overflow in Phases III
through V was only about 12 in. In view of this reported loss, a
distance of more than 12 in. should be provided with 2 to 3.6-mm sand
to prevent loss.
Summary and Conclusions - Phase VI
The objectives of the experimental investigation of Phase VI were to
observe the effect of depth of media on the backwashing effectiveness
of deep coarse sand filters washed with air and water simultaneously
and to observe the effect of depth on filtrate quality. Three pilot
filters were operated in parallel while filtering secondary effluent
at the Ames, Iowa, trickling filter plant. The filters were equipped
with 24 in., 47 in., and 60 in. of coarse sand in a size range of 2.5
to 3.7 mm. Double reverse graded gravel was used to support the sand
in the filter with 24 in. of sand. All three filters were backwashed
with secondary effluent (i.e., feedwater) throughout the two-month
period of operation.
The following conclusions have resulted from the investigations in
Phase VI:
1. A backwash routine including simultaneous use of air at 7 scfm/
sq ft and water at 8 gpm/sq ft for 15 min was not effective in
keeping the filter sand in clean condition.
2. A backwash routine including the simultaneous use of air at
7 scfm/sq ft and water at 15 gpm/sq ft for 10 min was effective
in keeping the filter sand in clean condition except for about
202
-------
a 12-in. layer at the bottom of the filters. The extent of
this dirty layer at the bottom was independent of the depth of
media and probably is a function of the underdrain strainer
type and spacing.
3. In backwashing of coarse sand filters with air and water simul-
taneously, the rates of flow of air and water should be selected
to ensure that a modest circulation of the sand occurs in the
filter, upward above the underdrain strainers and downward be-
tween the strainers.
4. The filtrate quality produced by the three filters of different
depth was nearly the same when filtering secondary effluent.
Thus, one cannot achieve much improvement in filtrate quality
with such filters and feedwater merely by increasing the depth
of media.
5. The double reverse graded gravel backwashed with feedwater
proved somewhat unstable again in Phase VI, and the prior con-
clusions about the suitability of feedwater and gravel are thus
unchanged as a result of the Phase VI experiments.
6. The principal advantages of coarse sand filters washed with air
and water simultaneously were again demonstrated in this phase,
namely, lower head loss development and thus greater solids
capture per unit head loss development, and better backwash ef-
fectiveness. The principal hazard in this backwash arrangement
is the potential loss of filter media during backwash overflow
due to the violence of the air-water action. Adequate freeboard
must be provided and the rates of air and water flow must be
selected to ensure no loss of media.
203
-------
IX. EXPANSION AND INTERMIXING OF MULTI-MEDIA FILTERS
Introduction
Granular filters are widely used for water treatment and are gaining
importance for tertiary treatment of wastewater because of the recent
higher effluent quality standards. In a conventional single-media
rapid sand filter, the sand particles are hydraulically graded during
backwashing, resulting in the finest particles being in the upper
layer . This stratification remains after backwashing. In a strati-
fied filter bed, the pore openings between the particles vary direct-
ly with the particle size. Because of this, most of the material
removed by the filter during filtration is at or near the surface of
the filter. An ideal filter would have this stratification reversed
so that the pore size would be the largest in the top layer and
steadily decrease in size to the bottom layer of the filter. The
recent development of multi-media filters with media of anthracite
coal, silica sand, and garnet sand approach this ideal pore size
arrangement. The densities of these three media vary with the an-
thracite coal having the lowest density, the garnet sand the highest
density, and the silica sand between the two extremes. The average
particle sizes of the anthracite coal, silica sand, and garnet sands
are selected to decrease in that respective order. With proper se-
lection of particle sizes, the filter will stratify with the anthra-
cite coal on the top, the garnet sand on the bottom, and the silica
sand in the middle. Although the pore size may increase with depth
within each individual media, the overall effect will be a decrease
in pore size of the filter with increasing depth. This decrease in
overall pore size with increased depth will greatly increase the
penetration of filterable solids into the filter bed, thereby in-
creasing utilization of the filter bed and extending the filter run
length, hopefully without detriment to the filtrate quality.
With the addition of anthracite coal and garnet sand in dual and
multi-media filters, new problems have arisen. What degree of expan-
sion of the individual media will provide optimum cleaning during
backwashing? Is there an optimum degree of intermixing at the inter-
face of the individual media? Can the degree of intermixing of the
media be predicted and controlled through selection of backwashing
rate and media size?
The specific aims of this research are fourfold:
1. to evaluate the effectiveness of existing expansion models for
predicting the expansion of garnet sand, silica sand, and coal,
and to suggest new or modified models if necessary,
2. to test the validity and sensitivity of two available models for
the prediction of intermixing between silica and garnet sands,
and between coal and silica sands,
204
-------
3. to observe the fixed-bed hydraulic profiles of various coal and
sand filters selected to produce different amounts of intermix-
ing in order to study the effect of the intermixing on the per-
meability of the intermixed zone of the filter, and
4. to observe the effect of interfacial intermixing on the perfor-
mance of dual-media filters.
Dual-Media and Multi-Media Filtration Literature
Media Design Characteristics
Baylis et al. [3] gave these typical dual-media design characteris-
tics. Sand medium usually has an effective size (ES) of 0.40 to 0.55
mm and a uniformity coefficient (UC) of 1.3 to 1.7. The sand medium
should be clean and well graded and have a specific gravity greater
than 2.65. The depth of the sand bed is normally 6 to 12 in. The
top 1/2 in. or more of sand is usually removed after hydraulic grad-
ing to prevent the head loss at the surface from being excessive.
Similarly, the anthracite coal layer usually has an effective size
of 0.8 to 1.2 mm, a uniformity coefficient of 1.3 to 1.7, a specific
gravity greater than 1.5, and a hardness factor on the MOH scale be-
tween 2.0 and 3.5. The coal medium should be clean and free of all
thin or scaly pieces, often prevalent in smaller sized coal parti-
cles. Depth of the coal medium is partially dependent on the uni-
formity coefficient.
According to Camp [24], the top layer of anthracite coal has inter-
stitial spaces approximately 20% greater in volume than those of the
top sand layer. The larger void capacity of coal absorbed more sol-
ids per volume of filter medium. As a consequence, Camp [24] suc-
cessfully used filtration rates up to 6 gpm/sq ft and achieved longer
filtration runs. Walker [137] suggested a dual-media filter is best
designed by first choosing a favorable coal size for filtering the
influent water. The sand size should be chosen to complement the
coal size.
Baylis et al. [3] considered that the coal layer should contain the
bulk of the filtering capacity. The sand layer was then used to fur-
ther polish the water after it had passed the coal layer. Coal par-
ticles should not be present at the bottom of the sand layer since
larger pore spaces occur around the coal particles. These large pore
spaces lessen the effectiveness of the sand layer as a good filter.
Observed Intermixing in Dual-Media Filters
Conley [32] found that the intermixing of coal and sand avoided the
rapid buildup of head loss at the interface, while acceptable water
quality was still achieved. Camp [23] disagreed with Conley [32] by
recommending a non-intermixed filter and suggesting that the good
water quality results obtained by Conley were due to excellent
205
-------
chemical control. According to Camp [23], removal of suspended sol-
ids was more efficient if the fine sand was not intermixed with the
coarse anthracite coal.
Later, Robeck and Kreissl[103] studied the effect of intermixing in
coal and sand filters at Erie, Pennsylvania. They considered inter-
mixing beneficial for increasing the run length. They demonstrated
that run lengths increased with increasing size of the surface coal.
This was achieved by removing the coal finer than 1.0 mm from the
source medium.
The amount of intermixing at the interface was studied by using hy-
draulic profiles of three different graded-media filters. The dual-
media filter contained 18 in. of 1.14-mm ES coal and 6 in. of 0.43-mm
ES sand, while the two, single-media filters contained 24 in. of
0.75-mm ES coal and 24-in. of 0.43-mm ES sand, respectively. The
coal and sand media were unskimmed, and both created a large head
loss at the surface. Limited data concerning the hydraulic profiles
through the three filter beds were presented for downflow with clean
water. Cumulative head losses were 1.83 ft for the unskimmed coal,
4.75 ft for the unskimmed sand, and 2.50 ft for the dual-media filter
at 14.5 gpm/sq ft. The intermixing zone depth could be determined
since the hydraulic profiles gave evidence that the head loss per
unit depth gradually increased from the coal to sand layer through
a 6-in. bed depth. Robeck and Kreissl [103] also showed that perme-
ability in the upper sand layer can be controlled by varying either
media to produce different amounts of intermixing or by varying the
backwashing shutdown procedure. An instantaneous shutdown shifted
the level of maximum head loss upwards 2 in. from the level of maxi-
mum head loss for the slow shutdown. In addition, the instantaneous
shutdown gave a greater maximum head loss. Thus, the instantaneous
shutdown gave less available capacity.
Brosman and Malina [19] studied intermixing of dual-media filters
made from 18 in. of 0.43-mm ES silica sand having a UC of 1.32 and a
specific gravity of 2.64; and 6 in. of 1.00-mm ES anthracite coal
having a UC of 1.40 and a specific gravity of 1.77. Four of the many
conclusions were: (1) intermixing gave more uniform bed porosity
with depth; (2) increased intermixing decreased head loss at the in-
terface; (3) increased initial downflow head loss accompanied in-
creased backwash rate and shorter backwash valve closure time for
intermixed filters; and (4) greater size ratios of anthracite to sand
gave greater intermixing. The intermixing zone was defined as the
filter length which has, within each infinitesimal section, quanti-
ties of media equalling at least 20% of the dry weight of each medi-
um. Intermixing was not considered to have occurred if the inter-
mixing zone was less than 5% of the total bed height since flow pat-
terns and bed instabilities probably caused any observed intermixing.
Brosman and Malina [19], in studying filter performance, concluded
that intermixing in dual-media filters resulted in longer filter
206
-------
runs, more uniform head loss with depth, and better filtrate quality.
However, in the opinion of the authors of the present report, this
conclusion may not be valid because the initial cumulative head
losses in the various dual-media filters were unequal; thus valid
comparisons of filter performance were impossible.
Backwashing of Granular Filters
Since the advent of the rapid sand filters around the turn of the
century, the study of hydraulics of the filtration process has pro-
gressed. The prediction of the expansion of granular filters during
backwashihg usually was approached from an extension of filtration
hydraulics.
This extension of the fixed-bed hydraulics to a fluidized bed is
subject to challenge from a theoretical viewpoint. In filtration or
fixed-bed conditions, the particles are not free to move about; while
in the fluidized state, they are suspended in the fluid and free to
move about with little if any contact with other particles for two-
phase, liquid-solid fluidization. However, expansion of the filter
media is seldom greater than 50%. Because of this relatively low
degree of expansion, the extension of the fixed-bed hydraulics to the
fluidized bed has provided models that provide results which are
agreeable with the experimental results [24,29,46].
Recently, Amirtharajah [5] has shown that the optimum porosity for
effective cleaning of silica sand by water backwash is approximately
70%. Expansion required to achieve this porosity would be dependent
upon the initial porosity ratio of the bed and the expansion-flow
rate characteristics of the graded media. For a graded bed of silica
sand, the required expansion would be approximately 45% to expand the
top layer of the bed, where most of the filtered solids are retained,
to a porosity of about 707o.
Flow through a Fixed Bed
Many of the sanitary engineering models for prediction of bed expan-
sion and some of the expressions for determining the minimum fluidi-
zation velocity of granular beds are based in part on equations de-
scribing flow through a fixed granular bed. Fair, Geyer, and Okun
[46] present the classical Kozeny equation for head loss through a
granular bed for laminar flow as,
- u (1 - e)2
P8 3
207
-------
where :
h = loss of head
JL = depth of bed
k = Kozeny's constant
g = acceleration due to gravity
jj, = viscosity of the fluid
p =s mass density of the fluid
V = superficial velocity of the fluid above the bed
e = porosity
d = particle diameter = diameter of equivalent volume spheres
\|l = sphericity - defined as the ratio of the surface of an
equivalent volume sphere to the actual surface area of the
particle.
From Coulson and Richardson [33 (p. 7)] , Carmen modified the Kozeny
equation to apply to transitional and turbulent flow. This equation
is commonly called the Kozeny -Carmen equation,
2 •* A a
p(V) e3 d 8
where:
and
R- = drag force per unit area of particle surface in the
direction of flow motion
V' * — = velocity of the fluid in the pore openings
£
i-j = 5 Re"1 + 0.4Re~0tl
Re. = modified Reynold's number = [Vp/(l - e)p,l [d/6] which
uses specific surface (d/6) for the diameter term.
208
-------
Another equation that describes head loss through fixed granular beds
in the laminar, transitional, and turbulent ranges was developed by
Ergun [23,40] (Clark, Viessman, and Hammer [29, p. 368] incorrectly
called Eq. (19) the Carmen-Kozeny equation).
2
3
ۥ
where:
f- = dimensionless friction factor = 150 ^ " &' +1.75
-L Re
and
Re = Reynold's number = ^— .
P.
Sanitary Engineering Bed Expansion Models
In sanitary engineering practice, three different approaches used for
predicting expansion are of particular interest.
Two of the approaches are given by Fair, Geyer, and Okun [46, Sect.
27, p. 19].
The Kozeny equation [Eq. (17)] and the constant head loss equation
[Eq. (1)] can be equated and solved for the porosity terms resulting
in,
<
where:
k = Kozeny's constant which assumes a value at about 4 for a
fluidized bed of low expansion.
Subscript 'i1 denotes the 'i'th layer of the bed.
The ratio of the expanded height, ^ej> to tne unexpended height, j£c
is from Eq. (2),
* A " ^
C •
l - e
209
-------
and the total expanded bed height Le is
A log-log plot of Rec vs B produces a family of curves in which each
curve represents a different porosity.
The porosity of an expanded layer of the bed at a selected backwash-
ing rate is determined by trial and error as follows. A trial poros
ity is selected, and Rec and B are calculated. The intersection of
the Rec abscissa and the B ordinate on the family of curves is ob-
served to see if it falls on the curve representing that selected
porosity. If not, a new porosity is selected and the process repeat
ed. The expanded bed height is then calculated from Eq. (21).
210
-------
Tesarik's discussion [130] of Camp's paper was critical of the group-
ing of terms for Camp's modified Reynold's number. Tesarik redefined
the Reynold's number in a more conventional way,
VPd
Re a
V>
where:
Re = Reynold's number based on the superficial velocity (V) of
the fluid above the bed.
Tesarik continued his discussion by showing that the expansion of a
granular filter is adequately described by Richardson and Zaki's
[100] expression of V/V^ = en (as presented below) and presented re-
sults of his own work.
Bed Expansion Correlations from Fluidization Literature
Fluidization, although a relatively young field, gained most of its
importance in the early 1940's with the use of fluidized beds in the
catalytic cracking of petroleum. Since that period, extensive stud-
ies of the fluidization process have been published. The work of
Amirtharajah [4-6] is a collection of many of the different aspects
of fluidization that pertain to backwashing of granular filters in
sanitary engineering.
Richardson and Zaki's Correlations
Richardson and Zaki [100] determined that the ratio of superficial
velocity above the bed (V) to the settling velocity of a discrete
particle (Vs) is a function of the Reynold's lumber (ReQ) based on
the settling velocity of a discrete particle, porosity (e), and the
ratio of particle diameter to the tube diameter,
«. 51 (24)
where:
— - ratio of the particle diameter to column diameter.
Under laminar and turbulent conditions, the ratio V/VS is indepen-
dent of
211
-------
They presented their data graphically by plotting log V vs log e
(Fig. 56). Above minimum fluidization, the data plotted as a straight
line with n representing the slope of the line and V. the intercept
of the line at a porosity of 1.0, the mathematical expression for the
line is
log V = log V. + n log e
(25)
or
They also observed empirically that Vs, for spherical particles, and
V"i could be related by the expression
log V
0
^- 10
d/D
(26)
V. = VELOCITY INTERCEPT
n SLOPE
MINIA^JM FLUIDIZATION
VELOCITY (V .)
flflr
LOG
(SUPERFICIAL
VELOCITY, V)
LOG (POROSITY RATIO,
Fig. 56. Relationship between superficial velocity - porosity,
212
-------
combining Eqs. (25) and (26),
V en
77- = —TTcr for spherical particles only. (27)
s 10Q/U
Replacing Vt for Vg in Eq. (24) and from Eq. (25), en is shown to be
n V
e = — for all particle shapes. (28)
vi
Therefore, n slope is independent of e and dependent upon the Rey-
nold's number of a free settling particle (Re0) and the ratio of par-
ticle diameter to column diameter. They developed the following em-
pirical equations for n slope for spherical particles,
for
0.2 < Re < 1
o
n = (4.35 + 17.5 |)Re~0-03 (29)
for
1 < Re < 200
o
n = (4.45+18 j^Re"0'1 (30)
for
200 < Re < 500
o
n = 4.45 Re"0*1 (31)
The above equations are for the transitional range of flow. Where
the inertia forces are negligible (laminar regime), the results were
correlated by
n = 4.65 + 19.5 | (32)
and where the viscous forces are negligible (turbulent regime)
n - 2.39. (33)
The change in n slope for the three different flow regimes is shown
graphically in Fig. 57. In Richardson and Zaki's work, they studied
213
-------
4.65
n
SLOPE
2.39
LAMINAR
REGIME
TRANSITIONAL
REGIME
TURBULENT
REGIME
0.2
500
LOG
REYNOLD'S NUMBER, Re (BASED ON TERMINAL SETTLING
VELOCITY1, V,)
Fig. 57. Relationship between n slope and
Reynold's number, Re .
liquid-solid fluidization and sedimentation of spherical particles of
uniform size greater than 100 microns in diameter with a density
range of 1.06 to 10.6, in all three flow regimes.
They had excellent correlations between calculated Vs and observed Vj
as d/D -» 0, supporting the validity of Eq. (26), for spherical parti-
cles.
Wen and Yu's Correlations
Another expansion correlation was proposed by Wen and Yu [138]. From
force considerations, they showed that porosity is a function of the
following equation,
f(e)
F
ks
(34)
where:
F
g
gravitational force
,3
rrd pg
buoyancy force =
TTd3pj
F, = drag force on a discrete particle
KS
g = Newton's law conversion factor.
214
-------
For the evaluation of F^g, they used Schiller and Naumann's equation,
.2._2
nd pv
(3 Re"1 + 0.45 Re"0'313) (35)
^>
which is valid for Reynold's number from 0.001 to 1000.
Using data from their own work as well as from the literature they
found, f(e) = e~4'7. Combining Eqs. (34) and (35), the following
form of the equation was developed,
e4'7 Ga » 18 Re + 2.70 Re1'687 (36)
where:
^3 , N
d p(ps - p)g
Ga = - = Galileo number (dimensionless).
Effect of Particle Shape on Expansion Correlations
The effect of particle shape on the expansion correlations has not
been studied extensively.
Richardson and Zaki's [100] work with nonspherical particles of regu-
lar shape (cylinder, cubes, plates) was in the turbulent range only.
They tried two different shape factors, sphericity (ty) and volumetric
shape factor (K).
The K factor correlated best for the evaluation of the n slope. The
equation for n slope in the turbulent range was
n = 2.7 K°'16 (37)
where:
d3
K = ? —r = volumetric shape factor
d
P
d - diameter of sphere of equivalent volume of particle
s
d = diameter of a circle of same area as the projected profile
p of a particle when lying in its most stable position.
Lewis and Bowerman [79] reported that the expansion of nonspherical
particles could be correlated by using a modified porosity ratio for
215
-------
the nonspherical particles. The modification is obtained by multi-
plying the porosity ratio (e) by the sphericity (ty) of the particles
OKe).
Whitmore [139,140], working with nonspherical particles, found that
the n slope of the rough particles was higher than Richardson and
Zaki's n slope for spherical particles in the laminar range. Whitmore
stated that this higher n slope increased as particle size decreased.
He presumed that this increase was because a rough particle settling
in a fluid has a layer of attached fluid that smoothes off the irregu-
lar outline of the particle. This trapped liquid gives the particle
a larger effective diameter and a lower density.
Jottrand [69] fluidized uniform crushed sands with water in the lami-
nar range. His plot of log V vs log e resulted in a series of paral-
lel straight lines with an n slope of 5.60. From the experimental in-
vestigation, he found that the fluidization velocity of given parti-
cles at a given expanded porosity closely approximated the hindered
settling velocity of the fluidized bed after the liquid flow was cut
off. In subsequent experiments, he attempted to replace the fluidi-
zation experiments by a more simple measure of the rate of hindered
settling of particles after good agitation. The results of the
hindered settling at a given concentration of particles after good
agitation was found to be about 45% greater than velocities observed
by the first technique for the same particle concentration.
Wilhelm and Kwauk [141] did extensive research on fluidization with
water and air. The particles ranged in size from 5 to 0.3 mm and in
density from 1.125 to 10.792. Their raw data of flow rate, pressure
drop, and porosity is completely reported. Of particular interest is
their work with sea sand, which was described as prismatic with
rounded edges. Jottrand [69] extended the analysis of Wilhelm and
Kwauk's sea sand data with a log V vs log e plot. The results of the
analysis of the sea sands are given in Table 27.
The authors of the present report calculated the n slopes for the sea
sands using Richardson and Zaki's equation [Eq. (30)], and these
values are included in Table 27. The n slopes calculated from Eq. (30)
are somewhat lower than the n slopes from the log V vs log e plots
for the nonspherical particles as reported by Jottrand.
Carvalho [28] , working with crushed anthracite coal of uniform sizes,
plotted log V vs log e and determined the Vj[ intercept and the n
slope. He also experimentally determined the discrete settling ve-
locity, Vs, for the various uniform sized coal particles (Table 28).
The Vg determined experimentally was approximately 25% lower than the
V^ intercept (Table 28). He attributed this discrepancy to the shape
of the crushed particles. Carvalho did not compare the n slopes from
the log V vs log e with the n slopes calculated from Richardson and
Zaki's equation [Eq. (30)].
216
-------
Table 27. Comparison of n slopes of sea sands (using Jottrand's
analysis [69] and Richardson and Zaki's equations).
Sand
1
2
3
Density,
g/cc
2.639
2.639
2.639
Diameter ,
mm Re,
0.373 21
0.556 45.5
1.000 122
Jottrand n
slope from
log V vs
log e plot
3.45
3.15
2.95
n slope
from
Eq. (30)b
3.28
3.04
2.76
Reynold's number based on the velocity intercept at porosity ratio
of one from the log plot of V vs e.
Assuming that Re - ReQ and neglecting the d/D term.
Values of Rei from Carvalho's reported Vj[ and subsequent determina-
tions of n from Eq. (30) were calculated by the writers. The results
are also presented in Table 28.
From the preceding literature, two significant points can be made:
1. The n slopes for fluidized beds of nonspherical particles are
greater than the n slopes of fluidized beds of equivalent sized
spherical particles.
2. The experimentally determined discrete settling velocity of a
nonspherical particle of a bed does not equal the velocity in-
tercept at porosity equal to unity of the log V vs log e plot.
Effect of Particle Size Distribution on Expansion Correlations
The media used for filtration does not consist of one-size particles
but of a range of particle sizes. The narrowest size range of media
that can be prepared conveniently and results in the largest to
smallest particle diameter ratio of about 1.2 is the media that is
retained between two adjacent sieves. Although there is no limit to
the widest range of particle size, for multi-media filters the widest
diameter ratio suggested for any of the individual media comprising
the bed is about 2.5 [34].
Various investigators have used different methods of determining the
representative particle diameter for expansion correlations:
217
-------
Table 28. Comparison of n slope of crushed coal.
vi
fps
n
Average experimentally From log V
size, Temp, observed vs log e
Run3
I
II
III
ivd
V
VI
VII
vine
Runs
1.65
b., ,
mm
1.3
1.1
0.9
1.1
1.3
1.1
0.9
1.1
I to IV
g/cc.
°C
16
16
12
16
16
16
16
16
coal
_ j j
.0
.0
.5
.0
.0
.0
.0
.0
0
0
0
0
0
0
0
0
density
l L.
1_ _ __
fps
.194
.171
.139
.172
.208
.182
.165
.185
= 1.35
plot
0.
0.
0.
0.
0.
0.
0.
0.
g/cc,
265
255
176
203
278
232
222
220
Runs
4.20
4.70
4.70
4.15
3.70
4.00
4.35
3.80
V to
U
based on
V.
95
77
40
61
100
70
55
67
VIII
.2
.7
.3
.9
.0
.6
.5
.0
coal
n
from Re^ 'c
and
Eq. (30)
2.82
2.88
3.07
2.95
2.80
2.89
2.98
2.92
density =
°Assuming Re. = Re and neglecting the d/D term.
Run IV the media for this run was comprised of equal volumes of media
from Runs I to III.
^Run VIII the media for this run was comprised of equal volumes of
media from Runs V to VII.
1. In sanitary engineering practice [24,29,46], the particle diam-
eter is described as the arithmetic mean of adjacent sieve
sizes. Expansion correlations are made by summing the expansion
of individual layers of filter media comprising a graded media
bed.
2. Amirtharajah [4,6], using data from the sieve analyses of graded
sands, plotted the percent passing by weight on a probability
scale against sieve size and used the diameter corresponding to
60% passing by weight in his expansion correlations. One sand
investigated was not normally distributed about the mean size
218
-------
and did not plot as a straight line when plotted in this manner.
This sand was analyzed as two separate components.
Wen and Yu [138] determined the effective diameter of two-component
mixtures by an inverse relationship,
j
inverse n W.
where:
W. « weight fraction of 'i'th layer
d. - mean diameter of weight fraction of 'i'th layer.
Van Heerden et al. [134] , with gas fluidization of fine particles in
the laminar range, also used the inverse definition of particle diam-
eter. He compared the d inverse, d arithmetic mean, and d geometric
mean with minimum fluidization velocity studies and found that the d
inverse gave the best correlation. The arithmetic average gave
higher diameter values and the geometric average gave lower values
than the d inverse diameter.
Leva et al. [77,78] used the arithmetic mean diameter (d_) ,
d. = ZW.d (39)
m i i
for two-, three-, and four-component mixtures in the development of
an equation for minimum fluidization velocity.
Another method of particle size determination used by some authors
[24,28,46] is the equivalent diameter of a spherical particle,
* = = (40)
eq
where:
Y - particle specific weight
s
W = total weight of N particles
N - number of particles.
Fair, Geyer, and Okun [46] point out that when determining particle
size by sieve analysis, the mean or 50% size is determined by weight,
219
-------
but the average diameter as determined by number is more closely rep-
resented by the 10% finer size on a weight basis.
Particle Segregation or Stratification
Particle stratification resulting from backwashing is very apparent
in rapid sand filters. For equal density particles with varying di-
ameters, different workers have reported rather widely different
ratios above which stratification would occur. These ratios have
ranged from 1:1.3 [138] up to 1:4 [96].
Pruden [96,97] did significant work to explain particle segregation
in particulate fluidization on a rational basis. He presumed that
the driving force towards segregation of two different groups of par-
ticles is the difference in bulk density between the groups. His
definition of bulk density (pb) is,
Pb = (1 - e)pg + pe - (1 - e)(ps - p) + p . (41)
The bulk density difference between the large particles x and the
small particles y is then,
(42)
Pruden used Richardson and Zaki's correlations for velocity and poros-
ity [Eq. (27)] and a theoretical equation for the settling velocity of,
a single particle which is applicable in all ranges of flow,
r> / \ 1/m 1/m ,(3-m)/n
Cp(ps - c\ - * ' d^
where;
C_. = drag coefficient (function of Re )
m = index of fluid regime (m = 1 in Stoke's regime, m = 2 in
Newton's regime, m = 1.4 in transitional regime for a
straight line approximation)
n = slope of the log V vs log e plot.
Combining Eqs. (42) and (43), Pruden developed the following equation
for bulk density difference,
* Pb
x y
Y
b
1 - r
nD
(3-m)/mn 10
(mn-l)/mn
Yb
(44)
220
-------
where:
(% -p)
d
yr
r = T~ ratio of particle diameters.
y
The assumptions of the above equation are that m^ - my = m, n^ = ny =
n, and drag coefficients are approximately equal. He tested the
validity of the above equation experimentally by evaluating m, n, and
e from fluidization of the separate components.
He hypothesized that when the bulk density difference is equal to
zero, the components should be completely mixed. A positive or neg-
ative value would mean that stratification would occur. He experi-
mented with equal density and low diameter ratio components and con-
cluded that Eq. (44) does give the correct trend of component strati-
fication, but that a limiting bulk density difference for stratifica-
tion would depend upon properties of the particles. He proposed the
following modification to the bulk density difference: a reduced bulk
density, (3,
(Pb
V
(45)
s
x
where:
0 £ p £ 0.01 mixing was observed
0.01 < (3 £ 0.04 partial segregation occurred
P > 0.04 segregation with an interface occurred.
Amirtharajah [4], in discussing Eq. (44), stated that a separate deter-
mination of ptx and pfcy, by evaluating the e for each component from
Eq. (25) and solving for the bulk density difference in Eq. (42),
would eliminate the uncertainty of some of the assumptions.
Le Glair's thesis [74] of two-component fluidization is a comple-
mentary study to Pruden's particle size segregation. Le Glair ob-
served that when some two-component mixtures x and y, such that
Do > DO and d < d_, were fluidized at low flow rates, the x
rSx roy X J
221
-------
component was below the y component. At an increased flow rate, both
components were intermixed. At still a higher flow rate, a reversal
of the components occurred, resulting in the small size, dense mate-
rial above the larger size, less dense material.
He explained that this phenomena could be attributed to the change
of the individual bulk densities of the two components as flow rate
changes. At low fluidization rates, the x component had a greater
bulk density than the y component, and at increased flow rates, the
bulk densities were equal. With still higher flow rates, the x-
component bulk density was less than the y-component bulk density.
Various densities and sizes of media can be selected which should
display this behavior. The selection would depend upon the expansion-
flow rate characteristics of the individual components.
The porosity at which the particles are completely mixed is called
the inversion porosity (ei) and is approximated by,
1 - Yb
(46)
I " 1/mn (3-m)/mn
Yb - ^ " - Yb
which is derived from Eq. (44) at zero bulk density difference.
Le Clair pointed out the following intermixing problems which are
of particular interest.
1. Equation (46) will give only an approximate value for ej and
corresponding velocity because of all of the assumptions made
for the development of Eq. (44) .
2. Data collected for the individual components will not predict
the e-r precisely because even for a narrow size range, the in-
dividual component data will give the average porosity of the
component, not point porosities within the component.
3. The completely mixed state of two components is reached at a
lower velocity than predicted by the individual component data.
This he attributed to particle size distribution. The small
particles, sized by sieving, have a greater variation in size
and subsequently greater range in bulk density than the larger
particles, also sized by sieving. This difference in change of
bulk densities would cause the completely mixed state to be
achieved at a lower flow rate than predicted from the individual
component data.
4. Inversion of the components would be observed for fluidization
in the laminar range but not in the transitional or turbulent
222
-------
ranges because of turbulence, particle circulation, and macro-
scopic mixing which would destroy the bulk density gradients.
5. The expanded height of a two-component mixture is closely ap-
proximated by the sum of the expanded heights of the two in-
dividual components whether the mixtures are segregated or
^/•mvnl Ai-01 IT ml -vo*1
A
completely mixed.
Another somewhat similar approach to intermixing is given by Camp
et al. [26]. They proposed that the driving force for intermixing
is the relationship between the particle density of the lighter ma-
terial and the bulk density of the more dense particles and water.
The equation used to calculate intermixing tendency was developed
by considering the forces acting upon the grains. The buoyant force
(Fjj) on a grain in a f luidized bed is equal to the weight of the mix-
ture displaced,
Fb = vpgpb (47)
where:
v = the volume of mixture displaced
p, = bulk mass density of the mixture.
The impelling force (Fj_) acting on a floating particle in the mix-
ture is its weight downward and the upward drag force of the wash-
water past the particle. The drag term presented assumes spherical
shaped particles.
weight
term
where :
CD = drag coefficient of the particle, a function of Reynold's
number and particle shape.
Equating F, and F. leads to the following:
223
-------
As applied to a dual-media filter, if p^ of the lower heavier layer
is greater than the right-hand terms for the upper lighter layer,
normal separation of the two media will result. If the reverse is
true, intermixing can be expected. Camp et al. presented very little
experimental evidence in support of the model presented above.
However, using this approach, they concluded that at common back-
washing flow rates of multi-media filters, the silica sand particles
(1.00 to 0.595 mm) will fall into and mix with the garnet sand (0.500
to 0.354 mm), but that the coal (1.41 to 1.00 mm) will be stratified
above the silica sand. The writers observe, however, that the coarse
end of the coal is finer than that commonly used in practice.
Brosman and Malina, in discussion of the Camp paper [26], stated that
intermixing for multi-media filters was more correctly described by
the bulk density approach described by Le Clair [74]. However, in
their closing discussion of the paper, Camp et al. vigorously defended
their model.
Prediction of Settling Velocities
The solution of the Richardson and Zaki expansion correlation, V/Vi e
en, requires the velocity intercept of the log V - log e plot and the
Reynold's number, corresponding to the settling velocity (Vs) for the
determination of the n slope. Two different approaches can be used
to find Vj.: (1) the direct determination from experimental observa-
tion or from tables or graphs found in the literature, or (2) an in-
direct method of using the minimum fluidization velocity and a ratio
of settling to minimum fluidization velocities.
Settling Velocities
Most textbooks in which there is a discussion of settling or sedi-
mentation give a method of calculating the settling velocity of a
discrete spherical particle. The settling velocity is solved by a
direct solution in the laminar or turbulent range. In the transi-
tional range, the calculation consists of a trial and error solu-
tion which simultaneously satisfies the settling velocity equation,
the drag equation, and empirical correlations of the drag coeffi-
cient vs Reynold's number. In the transitional range, which is of
most interest, there is considerable variation from the equations
or graphs presented for the evaluation of drag coefficient of non-
spherical particles. Graphical methods of solving for settling
velocity of nonspherical particles suffer from the same weakness.
One particular method of solving for the settling velocity of a
discrete particle involves the Galileo number (Ga), a term which
occurs frequently in the fluidization literature [17,33,53,69,76,
95,100,141], The development for the Galileo number is as follows.
The resistance force per unit projected area of the particle when
equilibrium is established for a settling particle can be expressed as,
224
-------
Rl J J ^ X V
^ rrd = - rrd (ps - p)g
or,
R1 = dg(ps - p)
where:
R' = resistance force per unit projected area of the particle.
Dividing both sides by pV2 and multiplying both by Re2
yields, s °
R' 2 2 d3
~2 Reo = 3 ~2 P(PS " P)g
PVs M.
= | Ga.
The Galileo number is a dimensionless term that is independent of
velocity and the product of Reynold's number squared and drag force.
Interesting to note is that Ga is equivalent to Camp's dimensionless
backwashing number B, presented previously.
Coulson and Richardson [33] present a table and a graph relating Re
to Ga. They also present a table for a correction factor to be
applied to the Re for nonspherical particles.
Minimum Fluidization Velocities and Ratios of Settling and Minimum
Fluidization Velocities
Most of the formulas for predicting Vmf are developments from the
Kozeny equation [Eq. (17)] and the constant head loss equation
[Eq. (1)]. Leva [76] and Leva et al. [77,78] developed the following
by equating a modified Kozeny equation and the constant head loss
equation. This equation incorporates a Reynold's number relationship
for emf and f,
688 d1'82[Y(Ys - Y)]°-94
Gmf Of (50)
M-
where:
G f « superficial fluid mass velocity at minimum fluidiza-
tion in Ib (mass)/hr sq ft
d = diameter of particle in inches
Y,Y - fluid and particle specific weights in Ib/cu ft
(j, = viscosity in centipoise
225
-------
valid for Remf < 10. For Remf > 10, Leva presents a graphical cor-
rection to be applied to G ~.
Upon expressing Gmf as a superficial velocity (Vmf) , as done by
Amirthara j ah [6] , the equation becomes,
,1.82
V ' °-°0381
mf - CK88
M"
where:
V ,. = minimum fluidization in gpm/sq ft
d = particle diameter in mm
p, = viscosity in centipoise,
Again, valid for Re f < 10.
Zabrodsky [144] gives a multiplication correction factor for the ve-
locity for Re f > 10,
k - - 1.775 Re "°*272 (52)
mf mf ^ '
for 10 < Re , < 300.
mf
Wen and Yu [138], starting with Ergun's equation [Eq. (19)], devel-
oped the following,
Re
mf
^(33.7)2 + 0.0408 Ga - 33.7. (53)
They used their own work plus the work of many others to develop
this equation which is valid for gas and liuqid fluidization with a
Re ,. range of 0.001 to 4000.
m£
Frantz [48], using gas-solid fluidization, made over 400 experiments
with eight different media and extensive analysis of the data deter-
mined that for the solution of the critical fluid mass velocity, (G f),
the theoretical coefficients and exponents give better results than
Leva's empirical correlations [Eq. (50)] and that further refinement
of the coefficients and exponents from the theoretical values was
not recommended for extrapolation outside the range of his experiments.
The theoretical equation for fluid mass velocity is,
5 d Y(YS - Y)
G , = 4.45 x 10 S (54)
mt j,
226
-------
where:
Gmf = fluid mass velocity in Ib (mass)/hr sq ft
d = particle diameter in ft
Y»YS = fluid and solid specific weights, respectively,
in Ib/cu ft
jj, = viscosity in Ib/hr ft
The relationship of free settling velocity and minimum fluidization
velocity has been presented by relating the ratio of Re /Rem£ to the
log Ga by several investigators [17,53,95], Galileo number is a
constant for given particle and fluid properties. Remf and Reo were
calculated from various empirical equations. The results of the
various correlations of the ratio of Re^Re^ vs Ga are presented
graphically in the respective papers.
The above correlations could be used in the solution of the terminal
settling velocity by calculating Ga from the physical properties of
the fluid and the particle, and the Reo/Re^ ratio can be read from
the above correlations. Remf can be determined from Eq. (50), (53),
or (54), or the appropriate equations in the following discussion.
Re0 can then be calculated from Remf and the Re0/Remf ratio.
A brief discussion of the individual papers follows. Pinchbeck and
Popper [95] worked with gas-solid fluidization and plotted R
vs log 1/Ga. Re0 was evaluated from,
ReQ - - 6 + ^36 + | Ga. (55)
Van Heerden's equation for Re,^ [Eq. (56)] was chosen over Leva's
equation [Eq. (50)] because the former equation fit their data better,
Re
-mf = 0.00123 (1 - emf) Ga. (56)
They assumed that emf was a constant of 0.406, as proposed by Van
Heerden et al.[134] for a bed of spheres of homogeneous diameter.
Hence, the constant (1 - emf) would be 0.594 in the above equation.
Pinchbeck and Popper's [951 correlations of Re_/Remf and Ga, using
limited experimental data, could be termed as fair.
Bourgeois and Grenier [17] used Ergun's equation [Eq. (19)] and the
constant head loss equation [Eq. (1)] to develop the following equation
for Remf
227
-------
Re
mf
(
1 + 3.11 x 10"4 Ga
O.5
*
- e -)'
mf'
(57)
They also assumed em£ as a constant 0.406, but pointed out the ef-
fects of a change in porosity from this assumed value and simplified
Eq. (57) further. The terminal settling velocity was evaluated from
an empirical plot of Re vs Ga.
They also found a substantial difference of experimental results be-
tween air and water fluidization and analyzed them separately. They
developed the following analytical expressions for the Re0/Remf vs Ga
correlation for water fluidization,
1. 50 < Ga < 2 x 10
Re
c
Re
= 132.8 - 47.1 log Ga + 4.6 (log Ga)-
(58)
mf
Re
mf
<60
2. 2 X 104 < Ga < 106
Re
Re"
mf
= 26.0 - 2.7 log Ga
(59)
Re
:
mf
3. 10" < Ga
Re
Re
9.0
(60)
mf
228
-------
Godard and Richardson [53] related Re f with Ga. Starting with the
Kozeny-Carmen equation [Eq. (18)] andthe constant head loss equation
[Eq. (1)], they developed,
(1 - em£) 2 (1 - e-)0'1
Ga - 180 Re _ • ^-^- + 2.88 Re . • 7^ (61)
mt J m£ J
Smf emf
They also used Ergun's equation [Eq. (19)] and the constant head loss
equation to develop,
(1 - e f) 7 ,
Ga = 150 Re • , m +. 1.75 Re _ • -~ (62)
mr J mr J
emf emf
They related Ga to the Reynold's number based on the free falling
velocity by the following equations:
Ga = 18 Re Ga < 3.6 (63)
Ga = 18 Re + 2.7 Re1'687 3.6 < Ga < 105 (64)
o o
Ga = \ Re2 Ga > 105 (65)
J o
Equations (63), (64), and (65) were obtained for the laminar, tran-
sitional, and turbulent range, respectively. Equation (64) is ob-
tained from Schiller and Naumann's equation, previously discussed.
They also illustrated the effects of emf on the Reo/Remf ratio. The
higher the emf, the lower the Reo/Re^ ratio for a given Ga. The
curves of Re0/Remf vs log Ga are quite sensitive to changes of emf
at low values of Ga corresponding to the laminar range. However,
this sensitivity diminishes as Ga increases and is almost negligible
at high values of Ga > 10 (turbulent range).
The Re /Re ratio reported in the preceding papers varied from 40
to 110 in Hie laminar range of flow, where the £„,£ effects on the
ratio is most pronounced, down to 7 to 12 in the turbulent range of
flow.
Godard and Richardson [53] extended the use of the ratio of R
to the determination of the n slope.
Equation (27) can be rearranged to the following form of
229
-------
n = -. = 2_ (66)
log e log e f v
(if d/D is negligible). Using this relationship, it was possible to
replace Re Q/R^f with n slope in their correlations with Ga and to
present a series of curves of n slope vs Ga for different values of
emf. The values of the n slope were somewhat higher than corresponding
experimental values from the literature. They attributed the high
values of n slope to the phenomena that at minimum fluidization the
particles become free to orientate in a manner to offer least resis-
tance to flow, but the Ergun and the Kozeny-Carmen equations do not
reflect this change and, therefore, give higher values of Re^, which
then give higher n slopes from Eq. (66).
Existing Models for Predicting the Expansion of Fluidized Beds
Amirtharajah* s Model
Amirtharajah's [4,5] method for predicting bed expansion was a modifi-
cation of the method proposed by Leva [76], Amirtharajah's procedure
was,
1. Experimentally determine e f and p .
m£ s
2. From a probability plot of the sieve analysis, determine the
60% finer size (d.... -. ).
v 60% finer7
3. Calculate V £ from Leva's equation [Eq. (51)]; if Re . > 10 apply
Zabrodsky's correction factor, kmf, [Eq. (52)] to Vfflf.
4. Use the relationship,
to determine V . The above expression is based on the empirical
relationship of drag force on an isolated spherical particle to
drag force on the same spherical particle in a fixed bed,
£
particle isolated _ 71 3 /67\
particle in bed
This ratio was proposed by Rowe [105] and Rowe and Kenwood [106]
and validated by Davies and Richardson [36]. Amirtharajah also
assumed that drag forces are proportional to the square of the
velocities; thus he used the square root of the drag force ratio
in the above expression.
230
-------
5. The n slope can then be determined by the use of Richardson
and Zaki's equations [Eqs. (29) through (33)] where Reynold's
number is based on V .
s
6. Using a modified form of Eq. (27),
V = ke11 (68)
and inserting V^ for V, efflf for e, and n into the equation,
the value k can oe determined. The k value has the same units
as a velocity term.
7. The e of the expanded bed can then be determined for any super-
ficial velocity, V, by reapplying Eq. (68) with the values of k
and n slope previously calculated.
8. The expanded bed height is then found from Eq. (2).
Step 4 of this model is incorrect. The ratio of drag forces used by
Amirtharajah (71.3) is valid for rhombohedral packing of spheres and
can thus be reasonably applied to the bed at the onset of fluidization.
However, from fundamental hydrodynamics, the drag force is proportional
to V in laminar range, V2 in turbulent range, and between vl and V2
over the transitional range. Thus, Amirtharajah's use of drag force
being proportional to V2 is the source of his error and is a misinter-
pretation of Rowe's paper [105]. This error was somewhat self-corrected
in Amirtharajah's subsequent steps. For Amirtharajah1s fine sand A,
his reported value of Ga was equal to 5810 [4, p. 109]. The ratio of
ReQ/Re f or VS/Vf for this Ga should be approximately 20 from the
literature [19,53,95] rather than V 71.3 as used by Amirtharajah. This
would mean that the Vs and the Reo Amirtharajah used for determination
of the n slope from Eq. (30), neglecting d/D, would be lower by a fac-
tor of 20/Y 71.3 <=- 2.5. The correct n slope would, therefore, be larger
by a factor of approximately,
nAmirtharaiah _ 4.45(ReQ) ^ ^
- n i ~ ^ • ' -i.uy.
"correct 4.45(ReQ x 2.5) "*
In his step 6, e f, V -, and n slope were used to calculate a k value.
Theoretically, for spnerical particles, this k value should approxi-
mate^. In his thesis [4, p. 113], the Vg calculated from Vs/Vmf =
Y71.3 and the k value from subsequent calculations were 75.1 gpm/sq ft
and 140.7 gpm/sq ft, respectively. The ratio of these values is
140.7/75.1 =- 2, which is approximately the factor which the Vg/V^ or
Re /Rem£ ratio was assumed to be off. Thus, Amirtharajah used an n
slope which was slightly high, which led to an erroneous k value. Thus,
his expansion model was somewhat self-correcting and gave good results
when compared with his experimental data.
231
-------
Leva's Model
Leva's method of predicting bed expansion is as follows:
1. Experimentally determine e f and pg.
2. From the sieve analysis, define the diameter by the inverse
definition [Eq. (38)].
3. Determine the minimum fluidization mass flow rate, G ,., from
Eq. (50). raf
4. Determine the fluid mass flow rate, Gf, at the superficial
fluidization velocity of interest,
5. From this G^, determine the Reynold's number using Vf correspond-
ing to Gf, and from Richardson and Zaki's equations (Eqs. (29)
through (33)] determine n slope.
6. Calculate V. from the previously determined Vm£, e_f, n slope,
and Richardson and Zaki's Eq. (28), V, = V J(e £)n.
l mi mr
7. Repeat the above calculation using V., n slope, and V to solve
for the expanded e at Gf.
8. The expanded height can be determined by Eq. (2).
Amirtharajah [4,5] pointed out a significant error in Leva's model.
Leva uses the Reynold's number based on the superficial flow rate of
the expanded bed to determine the n slope. This is incorrect.
Richardson and Zaki's n slopes should be determined from the Reynold's
number based on Vi at e of 1.0. For spherical particles, this would
be the point where V = V..
Wen and Yu's Method
Wen and Yu's [138] method of predicting the expansion is very straight-
forward. The expanded porosity can be determined by the fluid and
particle properties and Eq. (36),
e4'7Ga = 18 Re + 2.70 Re1'687 .
The expanded height of the bed can then be calculated from Eq. (2),
* ~ (1 - e)
S, and e having been previously determined.
232
-------
X. EXPANSION AND INTERMIXING EXPERIMENTAL INVESTIGATION
Experimental Apparatus
General Layout of 6-in. Fluidization Column
A schematic layout of the 6-in. fluidization apparatus is shown in
Fig. 58. This apparatus was used in the expansion vs flow rate ob-
servations of garnet sand, silica sand, and anthracite coal for
purposes of developing expansion prediction models. It was also
used in observations of intermixing of silica sand and coal and the
hydraulic gradients which exist in these intermixed beds.
The source of water (hot and cold) was the university tap water
supply. Two sets of water taps were used, A and B. The high flow
rates were metered by flowmeter F]_. The lower flow rates were passed
through a thermostatically controlled mixing value, C, then through
flowmeters Fo and F3 and to the fluidization column, D. The water
temperature was measured by a thermometer placed within the expansion
column.
The fluidizing column consisted of 6-in. inner diameter, 1/2-in. thick
plexiglass tube 4 ft 5 in. deep with a 3-in. high calming section at
the bottom. The water was fed through 59 orifices of 1/16-in. diameter
in a 1-in. thick underdrain plexiglass plate. Sets of orifices were
staggered from one another so as to provide a uniform matrix of ori-
fices on the entire plate. The calming section was filled with 1/2-in.
diameter glass marbles.
The solid particles composing the bed were supported on two stainless
steel meshes (No. 50 over No. 10) placed above the 1-in. plexiglass
plate with the orifices. Pressure taps were located on the column
to permit observation of pressure drop along the depth of the bed.
The first pressure tap was placed in the bottom flange of the column
and projected to within 1/4 in. above the stainless steel screens.
The second pressure tap was placed in the column wall 3 in. above the
screens and directly above the first pressure tap. The rest of the
pressure taps were at 3-in. intervals up the column in the garnet
sand experiments. An identical column with pressure taps at 1.5-in.
intervals was used on the silica sand and coal experiments. The
pressure taps were constructed of 1/4-in. copper tubes. The inner
opening of the pressure taps was covered by a 50-mesh or 100-mesh
stainless steel screen soldered in place. The first pressure tap
protruded 1-1/2 in. into the column. The remainder of the pressure
taps protruded 1/2 in. into the column.
The above pressure taps were connected by plastic tubing to glass
piezometers mounted on boards.
233
-------
CO
UNIVERSITY WATER SUPPLY
A B
HOT COLD HOT COLD
FLUIDIZATION
COLUMN D
PIEZOMETER BOARDS
FLOW
METERS
THERMOS! AT 1C
CONTROL VALVE
C
PRESSURE
TAPS
-* M-
E'
Fig. 58. Schematic layout of 6-in. fluidization column.
-------
General Layout of 2-in. Fluidization Column
A schematic layout of the intermixing column is shown in Fig. 59.
This column was used in expansion vs flow rate observations for
uniform sizes of silica sand, garnet sand, and coal. It was also
used in observation of intermixing tendency between dual media com-
prised of uniform sizes of either garnet and silica sand or silica
sand and coal.
The source of water (hot and cold) was the water taps, B, discussed
in connection with Fig. 58. The mixing valve and flowmeter F, were
also used with the 6-in. column experiments. Flowmeter F^ was
connected in series to flowmeter Fo. Between the two flowmeters was
a 1/4-in. needle valve and a 1/4-in. quick shutoff valve.
The column consisted of the calming section, the fluidizing chamber,
and an overflow structure. The fluidizing chamber was made of a 2-in.
inside diameter plexiglass column 68-3/4 in. in height with 3/4-in.
plexiglass flanges on each end. The inlet to the calming section was
by a 3/8-in. opening. The height of the calming section was 2-1/4 in.,
with an inside diameter of 2 in. The lower 1-1/2 in. were filled with
glass beads 6 mm in diameter. A stainless steel screen of 50 mesh was
placed in the bottom of this section to prevent loss of the glass beads,
The remaining volume of the calming section was filled with 2-mm lead
beads. Between the calming section and the fluidizing chamber, a
100-mesh stainless steel screen was placed.
The overflow structure was an 11-1/4-in. extension of the fluidizing
chamber. The top of this structure was perpendicular to the column
axes, and the water flowed radially out of the column. This water
was collected by a circular trough and drained to the floor drain.
Flowmeters
Four flowmeters were used during the collection of data. They were
the rotameter type and, for the purpose of this research, were desig-
nated as F,, F-, F», and F, . The range of flows and scale of the
flowmeters are as follows:
Flowmeter
Fl
F2
F3
F,
235
-------
UNIVERSITY WATER SUPPLY
B
HOT COLD
u
I i
OVERFLOW
STRUCTURE
FLUIDIZATION
L CO
(yjTHERMOSTATIC
^-^ONTROL VALVE
FLOW METERS
RUBBER HOSE!
5N
rtN
g
N
1
S<
' SCALE
LEAD BEADS
GLASS BEADS
Fig. 59. Schematic layout of 2-in. fludization column.
236
-------
Flowmeters F]_, F2, and F3 were calibrated at the two temperatures
used by a weighting technique. A least squares fit was used in
determining the calibration equation for the flowmeters. These
equations are as follows:
Flowmeter Temperature. C Actual flow, gpm
FX 16.5 • 0.038 + 1.007 (meter reading) X 24
25.0 =0.113+1.003 (meter reading) X 24
F2 17.0 =0.169+1.032 (meter reading)
25.0 =0.197+1.027 (meter reading)
F3 17.0 = - 0.067 + 0.994 (meter reading)
25.0 - - 0.069 + 1.026 (meter reading)
The flowmeter reading of meter F, was checked using a volumetric
technique. The results of this calibration were judged acceptably
close so that the flowmeter reading was accepted to be the actual
flowrate.
Sieves
United States Standard sieves were used in determining the gradation
of the sands and also to separate uniform sizes from wider size ranges
received from the suppliers of the garnet sand, silica sand, and coal.
Filter Media
Two different types of sands, garnet and silica, and crushed anthracite
coal were used in this research. The garnet sands are obtained from
alluvial deposits which are crushed to requirements and sieved to de-
sired specifications. The individual particles are very angular and
vary widely in shape. The recommended sizes of garnet sand for waste-
water filtration range from passing United States Standard Sieve No.
20 to that retained on No. 80 134],
Five different garnet media (Idaho Garnet Abrasive Company, Kellogg,
Idaho) were studied in the garnet expansion experiments. They were:
(1) uniform sized (-14+16) (uniform sizes are defined as passing (-)
and retained United States Standard sieves) separated from manufactur-
ers rating size M-16, (2) uniform sized -25+30 separated from manu-
facturers rating size M-36, (3) uniform sized -50+60 separated from
manufacturers rating size M-60-80, (4) graded sized, as received,
manufacturers size M-60-80 and, (5) graded size, as received, manu-
facturers rating size M-36.
237
-------
For the garnet-silica sand intermixing experiments, a single uniform
garnet sand was fluidized in successive experiments with various
uniform silica sands. The garnet sand was the uniform -50+60 used
in the expansion experiments. The silica sand used in this series
of experiments were uniform sizes of -20+25, -30+35, -35+40, and
-40+45 separated from manufacturers (Northern Gravel Co., Muscatine,
Iowa) rating fine sand. The graded silica sands and anthracite coals
used in the 6-in. column experiments are presented in Table 29.
Table 29. Size and source of raw graded silica sands and coals
studied.
Type
Graded sand
Graded sand
Graded coal
Graded coal
Graded coal
Graded coal
Graded coal
Graded coal
Designation
A
C
A
B
C
D
E
F
Effective
size,
mm
0.54
0.72
0.89
0.98
1.08
1.30
1.46
1.92
Uniformity
coefficients
1.41
1.42
1.53
1.43
1.53
1.47
1.38
1.72
Source
1
1
2
2
3
4
4
3
1. Silica Sand, Northern Gravel Co., Muscatine, Iowa.
2. Philterkol No. 1 Special Anthracite, Reading Anthracite Coal
Company, Locust Summit, Pennsylvania.
3. Carbonite #B, Carbonite Filter Corp., Delano, Pennsylvania.
4. Shamokin Coal Company, Shamokin, Pennsylvania.
Uniform silica sands and coals were used in the 2-in. column to ob-
serve expansion, bulk density, and intermixing behavior as a function
of flow rate. These uniform media were prepared by sieving the above
graded media and included the following sizes: -10+12 and -12+14 mesh
from Sand C; -14+16, -16+18, -18+20, -20+25, -30+35, -35+40, and
-40+45 mesh from Sand A; -4+7, -7+8, -8+10 and -10+12 mesh from Coal
F; and -10+12, -12+14, -14+16, '16+18, -18+20, -20+25, and -25+30
mesh from Coal A.
238
-------
Experimental Procedures
Separation Sieving for Uniform Media
The two sands and the coal that were used were sieved into uniform
sizes by using a stack of appropriate sieves and a Combs Gyratory
machine (Great Western Mfg. Co., Combs Gyratory Sifting Machine,
Leavenworth, Kansas). The sieving time was 5 min for sand and 8
min for coal, and the total load to the sieves was 300 to 350 g
for sand and 200 to 250 g for coal. The sands and coals of each
uniform size were sieved a second time using the same procedure as
in the first separation sieving to improve the uniformity.
Average Particle Size Determination
Sieve analysis. The ASTM Standard Method for Sieve or Screen
Analysis of Fine and Coarse Aggregates [2, p. 93] states the fol-
lowing criteria for sieve analysis:
The sample of aggregate to be tested for sieve analysis
shall be thoroughly mixed and reduced by use of a sample
splitter or by quartering — to an amount suitable for
testing.
... not more than 1% by weight of the residue on any
individual sieve will pass that sieve performed as follows:
Hold the individual sieve, provided with a snug-fitting
pan and cover, in a slightly inclined position in one hand.
Strike the side of the sieve sharply and with an upward
motion against the heel of the other hand at a rate of about
150 times/min, turn the sieve about 1/6 of a revolution at
intervals of about 25 strokes.
The criteria of not more than 1% passing in an additional minute
of hand sieving was not met by using the sieving machine. Because
of this, machine sieving was abandoned for the sieve analysis pro-
cedure, and a modified method of hand sieving was adopted. This
method followed the above ASTM Standard Methods describing hand
sieving except that the sieving time was extended to 3 min. It is
recognized that the hand sieving technique would also have been
better for the separation sieving previously described. However,
hand sieving was not practical for that step due to the large
quantities of media that were separated.
Equivalent diameter of a sphere. The particle size of the three
uniform garnets was also determined by counting and then weighing
particles and calculating the equivalent diameter of a spherical
particle of the same weight and density. The equivalent diameter
of a spherical particle is given by Eq. (40),
239
-------
d
eq.
Density
The densities of the silica and garnet sands were determined by the
water displacement technique using a 50-ml pycnometer bottle. The
sand samples were dried at 110 °C for 2-1/2 hr as a preliminary pro-
cedure prior to the test and were submerged for 1 hr before final
weighings.
The same water displacement method was used for the coal because the
filter coal media during actual operating conditions would be
similarly submerged in water. The samples were dried at 110 °C for
2-1/2 hr as a preliminary procedure prior to the test and were sub-
merged for 24 hr before final weighings to allow the water to pene-
trate the pore spaces.
Porosity
Two methods of porosity determination were used. The first was a
water displacement and simulated fluidization technique, hereafter
referred to as the graduate cylinder technique. Two 1000-ml graduate
cylinders were used. In one cylinder, sand with a dry volume between
200 and 400 ml was measured. In the other cylinder, exactly 500 ml
of water was placed. The known volume of sand was poured slowly into
the cylinder that contained the 500 ml of water. The total volume of
the sand and water and the apparent volume of the sand were measured.
The next step was to simulate fluidization of the particles. But
before this was done, enough water was added to completely fill the
cylinder. There were two reasons for adding the additional water:
(1) to prevent the trapping of air in the settled bed and (2) to pre-
vent some of the particles from sticking to the sides of the cylinder
and rubber stopper. With a rubber stopper placed tightly in the open
end of the cylinder, the cylinder was rapidly inverted a number of times
then quickly set down and the particles allowed to settle. The apparent
volume of sand was then measured. In this method, it was assumed that
the particles would settle in their least dense volume and represent
the same porosity as that at minimum fluidization velocity. Three
sets of measurements were made for each garnet-sand media.
The second method used for porosity determination of a media in a
fixed-bed condition in the column will be hereafter referred to as
the column technique. The volume of media was found from the weight
of the media placed in the fluidizing column and the specific weight
of the media. The total volume of the media and entrapped fluid was
determined from the column diameter and the bed height of the media
after the bed was fluidized, expanded, and slowly contracted. The
column technique was used as a check of the graduate cylinder technique
for the garnet sand and was the only method used for porosity deter-
minations of the silica sand and coal.
240
-------
Settling Velocities
The settling velocities of discrete particles were experimentally
determined for representative samples of the three uniform size
garnet sands. A 5-in. diameter plexiglass column, 56 in. in height,
was used for this experiment. The column was filled to the top with
water. The particles that were placed in the water, rolled gently
between two fingers to completely remove any air attached to the
particle, and then released. They then fell through 16 in. of water
to come to dynamic equilibrium before a stopwatch was started. The
particles then fell through a 30-in. timed interval. The average
settling velocity of each uniform garnet-sand media was determined
at two different water temperatures of 16 to 17 °C and 25 °C. The
water temperature was adjusted when it deviated more than 0.5 °C from
the temperature desired.
Expansion — Flow Rate Experiments
Expansion experiments were made in both the 6-in. column and the
2-in. column previously described. The total bed was fluidized to
about 50% expansion and allowed to contract slowly. Starting with
a fixed bed, the flow rate was incrementally increased to a maximum
expansion. Readings were taken of flow rate, temperature, bed height,
and (for the 6-in. column) the piezometer tubes.
Readings were not made until the water temperature was constant, the
influent temperature matched the effluent temperature, the water
pressure was steady, and the bed height was stabilized. Visual
observations of bed behavior, such as portions of the bed which were
fluidized, and any circulation patterns were also noted.
Upon fluidization of the first garnet sample, run 1 (Series A-13), it
was noticed that the top 1/2 to 1 in. of the bed consisted of a lighter
colored material than the rest of the bed, which was the purple
color characteristic of the garnet sand. The bed was completely
fluidized and then contracted slowly. The light-colored greyish
material was then siphoned off the top of the bed. Removal of this
light-colored, less-dense material was done before any expansion data
were taken for garnet gradations.
Bulk Density and Intermixing of Uniform Sized Media
A full range of uniform-sized media of each type (coal, silica sand,
and garnet sand) were fluidized in the 2-in. column for purposes of
determining expansion vs flow rate. From these data, bulk density
could be calculated at all flow rates. Some of the uniform media
were then observed in dual-media beds comprised of silica sand and
coal or garnet sand to determine their intermixing behavior. The
procedure for the garnet sand-silica sand experiments is presented
here as an example. Essentially the same procedures were used for
the coal and silica sand observations.
241
-------
One uniform size garnet sand -50+60 and uniform size ranges of silica
sand -20+25, -30+35, -35+40, and -45+50 were used. The garnet was
split down to approximately 1000 ± 40 g samples by the use of a sample
splitter. The samples were then adjusted by the removal or addition
of garnet to 1000.0 g ± 0.1.
The silica sand was separated into uniform sizes as described pre-
viously, washed, and dried. Then the sieve analyses were made. The
uniform silica sand samples were then weighed with a precision of
± 0.1 g.
One of the 1000-g garnet sand samples and all of the uniform silica
sand samples were individually fluidized in the intermixing column.
The samples were completely fluidized and slowly contracted to a
fixed bed at zero upward flow. Then the flow rate was incrementally
increased. Readings of temperature, bed height, and flow rate were
recorded during the expansion. Visual observations of the portions
of the bed fluidized and the circulation patterns were also recorded.
After each of the individual silica sand media was fluidized and
observed as described in the above paragraph, a 1000.0-g garnet sand
sample was poured into the intermixing column on top of the silica
sand. The dual media was then expanded 60 to 70% and very slowly
allowed to contract to a fixed bed with no flow. Bed height, relative
location of each media, and qualitative concentration of each media
were noted. Expansion of the bed was started and incrementally in-
creased to a height of about 60 in. (200%). The previously mentioned
bed height, qualitative location and concentration of the individual
media, temperature, and visual observations of circulation patterns
were recorded at several flow rates.
Illustrative Calculations
The illustrative calculations presented are for the uniform garnet
sand -14+16 which had been separated in accordance with the separation
sieving technique previously described.
Sieve Analysis
The results of three sieve analyses are shown in Table 30. The results
of using the mechanical sieving machine and hand sieving are shown.
Sample 1 was sieved for 5 min on the Gyratory sieving machine. Samples
2 and 3 were hand sieved for 3 min. Samples la and 3a are the per-
centage of the material retained on an individual sieve which passes
that sieve in one additional minute of hand sieving. The percentage
that passed the No. 14 sieve in the additional minute of hand sieving
is highly distorted because the amount that was retained in the initial
sieving was very small. The percentage that passed the No. 16 sieve
in the additional minute of hand sieving is slightly higher than the
ASTM recommended value of 1% passing an individual sieve in one addi-
tional minute of hand sieving, but comparison of the mechanical and
242
-------
Table 30. Sieve analysis of garnet sand media (-14+16).
Sample
no.
Sieved
load, g
Sieving
procedure
Sieving
time, min
Sieve
no.
14
16
18
Pan
1
148.99
Machine
5
Sieve 7.
opening, retained
mm
1.41 16.72
1.19 81.32
1.00 1.85
0.11
la
Hand
1
7»
passing3
23.20
11.54
0.08
—
2 3 3a
145.92 154.15
Hand Hand Hand
331
7o 7. 7o
retained retained passing3
0.17 0.31 16.66
88.10 87.35 2.38
11.69 12.30 0.00
0.04 0.04
7, retained
mean of
2 and 3
0.24
87.72
12.00
0.04
% passing
mean of
2 and 3b
99.76
12.04
0.04
0.00
Percent passing in one additional minute of hand sieving expressed as percentage of original amount
retained on that individual sieve.
30ne hundred minus mean percent retained.
-------
hand sieving methods shows the improvement of the hand sieving method
over the mechanical sieving method. The results of sample 1 are rep-
resentative of the results obtained by mechanical sieve analysis of
other samples for sieving times of up to ten minutes. The increase
in mechanical sieving time was found to improve the sieve analysis
only slightly. Because the results of hand sieving method conform
quite closely to the recommended standards of ASTM, as previously
stated, the hand sieving method was adopted, and all sieve analyses
reported herein are by the 3-min hand sieving technique.
Average Particle Size Determination
From sieve analysis. The average sizes are determined by the inverse
definition Eq. (38) and by the arithmetic mean definition Eq. (39)
as follows:
Mean opening Weight
Sieve Sieve opening, between sieves fraction
no.
12
14
16
18
Pan
mm
1.68
1.41
1.19
1.00
0.841
(di), mm
1.55
1.30
1.09
0.92
(wt)
0.24
87.72
12.00
0.04
W.
~ JL
di
Wi/di
0.16
67.48
11.01
0.04
78.69
Widi
0.37
114.04
13.08
0.04
I Wtd - 127.53
Therefore, changing of the weight from percent to a fraction, the
diameter as defined by the inverse definition Eq. (38) is,
Inverse = "iT = 0^869 = U271
and as defined by the arithmetic mean average from Eq. (39) is,
d_ « SW.d = 1.275 mm
m
244
-------
Note: the difference in the two definitions of diameter is small for
a uniform media but is greater as the variation in size of media in-
creases as will be shown in the results section of this report.
Equivalent diameter of a sphere.
Sample I 2
Number of particles 110 110
Weight of particles 0.5798 0.5968
Density of solid, g/cc 4.140 4.140
Equivalent diameter of a spherical particle of the same volume is given
by Eq. (40),
d
eq
[6(0.5798") "I1/3 [6(0.5968)
|_110 TT 4.15J [110 TT 4.15J
de (mm) = 1.343 1.356
avg d (mm) - 1.349
Density
1. Weight of dry pycnometer = 25.9310 g
2. Weight of pycnometer full of water = 125.4432 g
3. Weight of pycnometer with inside wet « 26.1743 g
4. Weight of pycnometer inside wet + garnet sand « 45.7310 g
5. Weight of pycnometer + garnet sand + water to fill = 140.2815 g
6. Temperature of water = 24 C
7. Therefore, weight of sand =4-3 = 19.5567 g
8. Weight of water to fill pycnometer - 2 - 1 = 99.5122 g
9. Weight of extra water to fill pycnometer over the = 94.7938 g
sand =5-1-7
10. Weight of equivalent volume of water - 8 - 9 - 4.7184 g
245
-------
11. Therefore, specific gravity of garnet sand at 24 °C = 7/10
= 4.1448
12. Density of water at 24 °C = 0.99707 g/ml
13. Density of garnet sand = (11) (12) «, 4.133 g/ml
Porosity
Graduate cylinder technique. The following three samples are from the
garnet sand -14+16,
i ii iii
1. Dry volume of sand (ml) 340 390 170
2. Volume of water (ml) 500 500 500
3. Total volume of sand + water (ml) 695 740 610
4. Volume of sand after simulating 360 440 200
fluidization (ml)
5. Porosity [4 - (3-2)]/4 0.458 0.455 0.450
Average of row 5 = 0.454
Column technique.
Dry weight of -14+16 garnet sand removed from column, Ib * 31.5
Fixed bed height, ft = 1.138
Cross-sectional area of column, sq ft » 0.196
Total volume of water and garnet sand, cu ft = 0.2235
Specific weight of garnet sand, Ib/cu ft = 257.9
Volume of garnet sand removed from column, cu ft = 0.1221
Fixed bed porosity - (0.2235 - 0.1221)/O.2235 - 0.453
Settling Velocities
Number of particles dropped * 100
Distance of timed fall, ft =2.5
Temperature of water « 16.5 C
246
-------
Observed settling time of single particles, sec
3.3
4.9
3.4
2.8
3.8
4.4
3.4
3.9
3.6
3.4
Mean
3.
2.
3.
3.
2.
4.
3.
3.
4.
3.
time
Standard
Mean
5
9
6
0
7
3
7
1
0
9
of
3.5
3.2
4.4
2.9
2.6
4.0
3.1
3.1
3.2
3.7
fall
4.1
3.2
3.0
3.0
3.1
3.5
4.4
3.4
3.0
3.6
3.1
3.0
2.9
3.0
4.2
4.3
4.3
3.7
3.5
3.8
2.9
3.0
4.1
3.4
3.5
3.7
3.6
3.0
4.0
4.4
3.9
3.2
3.3
3.0
3.1
3.1
3.5
3.2
3.8
4.3
=
deviation =
velocity =
0.703
fps =
3
.4
3.1
5
3
4
2
3
3
4
4
3.56
.4
.9
.0
.9
.2
.7
.2
.0
sec
3.4
4.2
3.1
3.1
3.2
3.4
4.1
3.8
5.1
3.9
4.2
3.6
3.2
3.2
4.8
3.1
2.8
3.7
3.5
3.2
0.5534 sec
315
gpm/sq
ft
Expansion — Flow Rate and Intermixing
Illustrative calculations for the expansion-flow rate and intermixing
data will be presented along with the presentation and analysis of
the data.
Results and Analysis — Summary
The main series of experimental runs for the study have been summarized
in Tables 31-32. Table 31 lists the downflow and/or upflow runs in the
large column for various single media. Table 32 describes the upflow
and downflow runs in the large column for various dual-media filters.
Table 33 gives the upflow runs in the small column for uniform single-
and dual-media filters.
Results — Media Characteristics
Sieve Analyses
The results of the sieve analyses for the various media studied are
presented in Figs. 60, 61 and 62. Effective size and uniformity
247
-------
Table 31. Upflow and/or downflow experimental runs with the various single
media in the 6-in. column.
Series Media3
A-l
A-2
A-3
A-4
A- 5
A-6
A-7
A-8
A- 9
A-10
A-ll
A-12
A-13
A-14
A-l 5
A-16
A-17
Coal A
Coal UCA
Coal B
Coal C
Coal C2
Coal D
Coal E
Coal F
Sand A
Sand A£
Sand C
Sand C2
Garnet
-14+16
Garnet
-25+30
Garnet
-50+60
Garnet
M-60-80
Garnet
M-36
Description
ES UC
tnm
0.89
1
0
1
1
1
1
1
0
0
0
0
1
0
0
0
0
.38
.98
.08
.16
.30
.46
.92
.54
.60
.72
.75
.20
.60
.25
.17
.47
1.53
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
.12
.43
.53
.49
.47
.38
.72
.41
.28
.42
.43
.08
.07
.10
.71
.43
Initial
depth,
in.
12.75
6.
12.
12.
11.
11.
8.
6.
12.
9.
12.
11.
12.
12.
12.
14.
15.
,38
,75
,25
,12
20
00
63
40
75
38
38
65
35
50
80
45
Downflow Upflow
Rate, Temp, Temp,
gpm/sq ft °C °C
7,8,9 18,22,26,30 18,22,26,30
4,8 18,26 22
4,8 18,26 22
4,8 18 22
4,8 18 22
4,8 18 22
4,8 18 22
4,8 18 22
4,7,8,9 18,22,26,30 18,22,26,30
4,8 18 22
4,8 18 22
4,8 18 22
16,25
16,25
17,25
17,25
17,25
Subscript 2 signifies media remaining after hydraulic grading and skim-
ming 10% of the fines from the raw media of same letter designation.
248
-------
Table 32. Upflow and downflow experimental runs with dual-media filters
in the 6-in. column.
Downflow
Series
B-l
B-2
B-3
B-4
B-5
B-6
B-7
B-8
B-9
B-10
B-ll
B-12
B-13
B-14
Dual
mediaa
AA
AB
AC
AD
A2E
AF
A2A2
AUCA
A2C2
A2F2
A2F2
A2F2
CF
C2F2
Depth, in.
Sand
A =
A =
A «
A =
A2 =
A =
A2 =
A =
A2 =
A2 =
A2 =
A2 =
C =
C2 =
11.25
6.26
10.75
13.25
12.00
6.25
12.00
6.25
9.75
6.63
6.63
6.63
12.25
11.25
Coal
A =
B
C
D
E
F
A2 '
UCA =
C2 -
F2 '
F2 =
F2 -
F
F2 =
12.75
12.63
12.25
12.25
9.00
6.50
11.00
6.25
11.12
7.88
16.00
23.75
12.00
10.87
Rate, Temp,
Total gpm/sq ft °C
23.50
18.50
21.75
24.75
20.50
12.00
22.75
12.00
20.00
12.90
20.00
27.75
22.50
20.38
4,8
4,8
4,8
4,8
4,8
4,8
4,8
4,8
4,8
4,8
4,8
4,8
4,8
4,8
18,26
18,26
18
18
18
18
18
18,26
18
18
18
18
18
18
Upflow
Temp,
°C
22
22
22
22
22
22
22
22
22
22
22
22
22
22
Described by the letter designation for the component sand and coal,
respectively, from Table 31.
249
-------
Table 33. Upflow experimental runs with various uniform single media and
uniform dual media in 2-in. column, 25 °c.
Series
C-l
C-2
C-3
C-4
D-l
Uniform media description
Sand A (-10+12)
Sand A (-12+14)
Sand A (-14+16)
Sand A (-16+18)
Sand A (-18+20)
Sand A (-20+25)
Sand A (-25+30)
Sand A (-30+35)
Sand A (-35+40)
Sand A (-40+45)
Coal F (-4+7)
Coal F (-7+8)
Coal F (-8+10)
Coal F (-10+12)
Coal A (-10+12)
Coal A (-12+14)
Coal A (-14+16)
Coal A (-16+18)
Coal A (-18+20)
Coal A (-20+25)
Coal A (-25+30)
Coal F (-4+7), 5.9 in. and
Sand A (-40+45), 6.6 in.
Coal F (-4+7), 6.7 in. and
Sand A (-35+40), 5.5 in.
Coal F (-4+7), 5.8 in. and
Sand A (-30+35), 6.0 in.
Coal F (-4+7), 5.7 in. and
Sand A (-25+30), 6.3 in.
Coal F (-4+7), 5.5 in. and
Sand A (-20+25), 5.5 in.
Coal F (-4+7), 5.4 in. and
Sand A (-18+20), 5.8 in.
Garnet (-50+60)
Sand A (-20+25)
Sand A (-30+35)
Sand A (-35+40)
Sand A (-40+45)
Sand A (-20+25), 12.75 and
Garnet (-50+60), 11.40
Sand A (-30+35), a and
Garnet (-50+60), 11.40
Sand A (-35+40), 13.85 and
Garnet (-50+60), 11.40
Sand A (-40+45), 9.80 and
Garnet (-50+60), 11.40
Depth, in.
12.20
12.65
11.90
11.25
11.90
12.00
12.85
11.75
11.65
10.20
12.60
11.95
4.85
13.15
12.05
11.60
11.50
11.20
10.90
10.65
12.50
10.0
10.5
10.0
10.3
9.5
10.3
11.40
12.75
13.85
9.80
23.9
22.25
25 10
fmj • iW
20.80
aNot recorded.
250
-------
lOr
E
E
•*.
UJ
LIU
3.0-
2.0
1.5-
1.0
0.6-
0.5
0.4-
0.3
O.ll
0.01 0.1
5
SAND A V
SAND A2 A
SAND C O
I
1 2 5^TO 20 30 50 80 90 95
PERCENT PASSING BY WEIGHT
99 99.9 99.99
5.0
4.0
3.0
^ 2.0(-
1.0
.9
Fig. 60. Sieve analysis of graded sand media.
8
8:
0.7
0.4
0.5
0.4
COAL A
COALB
COALC
COAL D o
COALE
COALF
A
O
COAL C2 D
' 'I I 1 I I I I I I I I III
0.1 0.5 1 2 5 10 20 30 40 50 40 70
90 95 98 99.5
PERCENT PASSING §Y WEIGHT
Fig. 61. Sieve analysis of graded coal media.
251
-------
GARNET - 14 4 16 O
GARNET - 25 + 30 O
GARNET - 50 4 60 D
GARNET M-60-80 &
GARNET M-36 0
0.
0.01 072
99.rW.99
PERCENT PASSING BY WEIGHT
Fig. 62. Sieve analysis of garnet sand media,
coefficients for the graded sands and coals have been presented pre-
viously in Table 29.
In two cases, 10% of a medium was removed by skimming the upper-
most layer of the filter bed through the use of a siphoning hose
after the medium had been stratified by backwashing. Unskimmed
graded media were alphabetically designated, for example, Sand A,
whereas skimmed media were additionally labelled with a subscript,
for example, Sand A«. Figures 60 and 61 demonstrate that a slightly
larger effective size and a smaller uniformity coefficient resulted
from skimming as the finest media particles were removed.
The graded media, as shown in Figs. 60, 61, and 62, followed a near
log-normal distribution.
The results of the sieve analyses of the uniform silica sands and
uniform coals are presented in Appendix A. The results indicate
that at least 70% (generally greater than 80%) of the total sample
was retained between the adjacent sieves indicated in the size de-
signation (e.g., -20+25).
252
-------
Average Particle Size Determinations
The average sizes of the particles for the uniform media were deter-
mined from the sieve analysis presented previously and the equivalent
diameter of a spherical particle. The data for the equivalent diameter
of a spherical particle is presented in Table 34. A summary of the
results for the average particle size for the various methods is pre-
sented in Table 35 . The table includes average diameter by the
arithmetic mean defined by Eq. (39), the inverse diameter defined by
Eq. (38), the diameter of the 60% finer size from the sieve analyses
plotted in Figs. 60, 61, and 62, and the equivalent diameter of a
spherical particle by Eq. (40).
It is apparent from Table 34 that the equivalent spherical diameter
of the various media is larger than the mean of adjacent sieve size
for the uniform media. This is to be expected because of the non-
sphericity of the media.
Table 35 indicates that the inverse definition for the graded media
gave a somewhat lower average diameter than the arithmetic mean
diameter. The diameter corresponding to the 60% finer size was
closer to the arithmetic mean diameter.
Densities
The results of the density determinations for the garnet sand, silica
sands, and coal are given in Table 36. Several replicates are reported
for the garnet sand to illustrate the precision of the analysis.
The garnet sand used for density determinations was from the uniform
-14+16 and -25+30 sizes (Table 36). The density of the -25+30 garnet
was somewhat lower than that of the -14+16 garnet sand. The pycno-
meter bottles were shaken vigorously to remove air attached to the
particles and allowed to cool to room temperature. Therefore, it was
assumed that the inclusion of air or the temperature effects on the
density of water were not the source of the difference. The probable
source of difference was that the -25+30 garnet contained a small
amount of the less-dense, grayish-colored material that was assumed
to be removed. The difference was quite small; therefore, the average
of all six determinations was used in all necessary calculations in-
volving garnet sand media.
Porosities
The fixed-bed porosities for the media used in the experiments were
determined two ways; (1) the graduate cylinder technique and (2)
the column technique (using both the 2-in. and 6-in. diameter fluidi-
zation columns). The results for the porosity determinations by the
graduate cylinder technique and by the column technique are given in
Table 37. In some cases, data were collected by both investigators
(Woods [143], Boss [14]). Both values are reported to indicate the
spread between measurements made by different investigators.
253
-------
Table 34. Average diameter of uniform media by two methods, mean of
adjacent sieve sizes and mean equivalent spherical diameter
by the count and weigh method [Eq. (40)].
Particles
Mean of
adjacent sieve
Mean equivalent
spherical
Uniform media
mesh range
Sand A (-10+12)
Sand A (-12+14)
Sand A (-14+16)
Sand A (-16+18)
Sand A (-18+20)
Sand A (-20+25)
Sand A (-25+30)
Sand A (-30+35)
Sand A (-35+40)
Sand A (-40+45)
Coal F (+4)
Coal F (-4+7)
Coal F (-7+8)
Coal F (-8+10)
Coal F (-10+12)
Coal A (-12+14)
Coal A (-14+16)
Coal A (-16+18)
Coal A ( -18+20)
Garnet (-14+16)
Garnet (-14+16)
Garnet (-25+30)
Garnet (-25+30)
Garnet (-50+60)
Garnet (-50+60)
Number
counted
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
110
110
110
110
98
91
Weight,
g
1.0754
0.5752
0.3532
0.2190
0.1514
0.0434
0.0406
0.0231
0.0174
0.0095
12.8546
3.5898
1.8299
1.3413
0.8490
0.3410
0.3040
0.1331
0.0791
0.5798
0.5968
0.0729
0.0685
0.0042
0.0042
sizes, mm
(1)
1.840
1.545
1.300
1.095
0.920
0.775
0.650
0.545
0.460
0.385
—
3.790
2.595
2.180
1.840
1.545
1.300
1.095
0.920
1.300
—
0.650
—
0.273
—
diameter, mm
(2)
1.98
1.61
1.36
1.16
1.02
0.78
0.66
0.55
0.46
0.41
Sand mean
5.22
3.40
2.72
2.46
2.10
1.57
1.51
1.15
0.92
Coal mean
1.347
1.360
0.675
0.661
0.270
0.277
Garnet mean
(2) /(I)
1.08
1.04
1.04
1.06
1.11
1.07
1.02
1.01
1.00
1.06
1.05
—
0.90
1.05
1.13
1.14
1.02
1.16
1.05
1.00
1.08
1.036
1.046
1.039
1.017
0.989
1.015
1.024
Settling Velocities
The experimentally determined settling velocities for the three dif-
ferent uniform garnet sand media at the two different temperatures are
given in Table 38. The mean time, the standard deviation, and mean
velocity of the time measurements are included in Table 38.
A comparison of these experimental values of Vs with the velocity
intercept (Vj_) at a porosity equal to one on the log V vs log e plots
254
-------
Table 35. Summary of the average diameters of the media - d (by
several methods).
Size
Sand A
Sand C
Coal A
Coal B
Coal C
Coal D
Coal E
Coal F
Garnet (-14+16)
Garnet (-25+30)
Garnet (-50+60)
Garnet (M-60-80)
Garnet (M-36)
Source :
«%•
nun
0.711
1.008
1.285
1.305
1.572
1.988
2.094
3.228
1.275
0.639
0.265
0.272
0.642
Eq. (39)
"inverse'
HKQ
0.685
0.946
1.205
1.246
1.480
1.800
1.912
2.941
1.271
0.637
0.263
0.249
0.615
Eq. (38)
d60% finer,
nun
0.74
1.03
1.35
1.35
1.65
1.90
2.00
3.30
1.27
0.635
0.267
0.29
0.67
Figs. 60, 61, and 62
presented later in Table 41 shows that the experimental Vs was sub-
stantially lower than the Vj^.
These results are consistent with the findings of Carvalho for
crushed coal [28] previously discussed and can be attributed to the
nonspherical shape of the particles.
255
-------
Table 36. Densities of media — p .
s
Media
Garnet sand (-14+16)
Garnet sand (-25+30)
Silica sand A
Silica sand C
Coal A
Coal B
Coal C
Coal D
Coal E
Coal F
Density of sample, g/ml Densitvf
123 g/ml
4.1337 4.1384 4.1359
4.1228 4.1243 4.1201 Avg 4.13
2.65
2.65
1.70
1.70
1.71
1.72
1.74
1.73
The average settling velocities for the -25+30 and -50+60 garnet sands
were slightly faster at 17 °C than at 25 °C, contrary to expectations.
However, the differences are not statistically significant.
Fixed Bed Hydraulic Profiles in Dual-Media Filters -
Coal and Sand
Downflow Observations of Single-Media Filters
To observe the effect of various degrees of intermixing on the per-
meability of dual-media filters, head loss vs depth was observed for
the individual graded sand and graded coal media. These hydraulic
256
-------
Table 37. Fixed-bed porosities of the three media determined by the
two techniques (c ) and two investigators.
o
Porosities
Graduate
cylinder
Media technique
Sand A (-10+12)
Sand A (-12+14)
Sand A (-14+16)
Sand A (-16+18)
Sand A (-18+20)
Sand A (-20+25)
Sand A (-20+25)
Sand A (-25+30)
Sand A (-30+35)
Sand A (-30+35)
Sand A (-35+40)
Sand A (-35+40)
Sand A (-40+45)
Sand A (-40+45)
Garnet sand (-14+16) 0.454
Garnet sand (-25+30) 0.504
Garnet sand (-50+60) 0.550
Garnet sand (M-60-80) 0.557
Garnet sand (M-36) 0.508
Coal F (-4+7)
Coal F (-7+8)
Coal F (-8+10)
Coal F (-10+12)
Coal A (-10+12)
Coal A (-12+14)
Coal A (-14+16)
Coal A (-16+18)
Coal A (-18+20)
Coal A (-20+25)
Coal A (-25+30)
Column Column
technique technique
(2 in.) (6 in.)
0.440
0.442
0.443
0.448
0.440
0.439
0.436
0.444
0.451
0.441
0.456
0.458
0.457
0.452
0.453
0.505
0.554 0.580
0.536
0.495
0.554
0.560
0.573
0.573
0.551
0.555
0.552
0.554
0.554
0.558
0.572
Investigator
*>..}£]
Boss JJ
Boss "
Bo.. "
Bosslri,J01
tt , [143]
Woods
Boss [14]
Boss [14]
Woods']
Boss [14]
Woods [143]
Boss [14]
Boss [14]
Woods [143]
Woods |jJ3j
Woods Jg
Woods J14,J
Woods |"J
Woods [U3J
[14]
Boss|l4j
Boss 4]
Boss £
Boss 14
Boss 141
Boss 14
Boss 14
Boss[l4]
Boss 14]
Boss 4
Boss1 J
profiles for graded media of 6 to 12-in. depth were studied in the
6-in. fluidization column. The hydraulic profiles of the media
were then to be compared to the hydraulic profiles of the correspond-
ing dual-media filters, comprised of the same components, to observe
the effects of intermixing.
257
-------
Table 38. Settling velocities of uniform garnet sand media.
Garnet sand media
-14+16 -25+30 -50+60
Temperature,°C 16.5 25.0 17.0 25.0 17.0 25.0
Number of
particles
observed 100 100 100 100 60 60
a
Mean time,
sec 3.56 3.55 6.92 7.00 17.86 18.53
Standard
deviation,
sec 0.553 0.615 0.993 0.991 2.953 3.151
Mean velocity,
fps 0.703 0.703 0.362 0.357 0.140 0.135
a
Distance of timed free fall = 2.5 ft.
Several variables affecting head loss in a filter are correlated
in Eq. (17) and are evident in Figs. 63 and 64. The variables have
previously been studied and correlated in equations by several in-
vestigators. However, the figures are included here to help the
reader visualize the nature of the raw downflow data collected.
A fluid temperature of 18 °C, the lowest consistently attained temper-
ature with the laboratory water system, and an 8 gpm/sq ft flow rate
were chosen for the downflow measurements. At high filtration rates,
the head losses were more substantial and could thus be measured with
greater relative precision. Thus, the high flow rate of 8 gpm/sq ft
was chosen for the experiments.
Given the same flow rate and temperature, the difference in cumulative
head losses between two plotted points in Figs. 63 or 64 is the dif-
ference in pressure between two adjacent piezometers on the large
column. From this pressure difference, the experimental values of
head Ipss per 1-1/2-in. depth were taken to plot the hydraulic profile.
Bed depth vs the head loss per 1-1/2-in. depth was then plotted in
Figs. 65 and 66 to provide downflow hydraulic gradient profiles of
all graded media used. Analysis of Figs. 65 and 66 demonstrates that
the top sand layer had a head loss of 0.35 ft, while the bottom coal
layers had a head loss of 0.01 to 0.04 ft. Since precision of piezo-
258
-------
vO
1.00
o.ao —
0.40 —
4:
*
r
SAND A
FLOW RATE 9
30.0 °C O
26.5 °C
22.0
18.0 O
0.40 —
0.20 —
0.00
0.50 0.75 1.00 1.25
CUMULATIVE H£Af> LOSS, fc, ft
1.50
1.75
Fig. 63. Fixed-bed head loss for graded sand A at 9 gpm/sq ft for various
temperatures.
-------
1.00
N>
0>
o
0.80
0.60 h~
o
Q
SAND A
TEMPERATURE
9 gpm/*q ft o
8 gpnviqft
O
0.00 0.25 0.50 0.75 1.00 1.25
CUMULATIVE HEAD LOSS, h, ft
Fig. 64. Fixed-bed head loss for graded sand A at 26.5 C for various flow rates.
-------
meter readings was limited to 0.01 ft, the small head losses between
adjacent piezometers resulted in larger relative errors in the coal
head loss readings.
Study of Fig. 65, the hydraulic profile of the basic sand, reveals
that a great head loss change occurs in the top half of the filter
bed depth. Head loss throughout the graded coal media filter, as
shown in the hydraulic profile, Fig. 66, remained fairly constant
with depth. The surface head loss in the graded coal media filter
was only slightly larger than that of the lower layer. Therefore,
skimming of the sand media is expected to have more impact on filter
performance than skimming of the coal media.
Downflov Observations of Dual-Media Filters
To determine the effect of intermixing on the fixed-bed hydraulic
profiles of dual-media filters is one of three specific aims of this
research, and Figs. 67 through 72 illustrate typical related experi-
mental findings. Figure 67 is a typical example of cumulative head
loss vs bed height for dual media M. The gradual bend of the plotted
lines in Fig. 67 exemplifies the effect of modest intermixing on head
loss.
Since pressure readings could only be made to the nearest 0.01 ft,
some of the data points were relatively inaccurate. Such points on
figures such as Fig. 67 were adjusted by drawing the best fitting,
curved line for the cumulative head loss. Values from this curved
line were then chosen at 1-1/2-in. increments of the filter bed.
These values were used to plot the solid-line hydraulic profile of
the dual-media filters, as in Figs. 68 through 72. The adjusted points
did not vary by more than ± 0.02 ft from the original experimental
points.
Figures 68-72 also provide combined, individual hydraulic profiles
of the dual-media filter components. These dashed-line hydraulic
profiles show the head loss which would result at the interface if
intermixing did not occur. To obtain these theoretical profiles, the
hydraulic profiles for the appropriate sand and coal from Figs. 65
and 66 were combined. Coal was plotted above sand, and the two
hydraulic profiles were connected with a horizontal line. The hori-
zontal connecting line was plotted at the bed height corresponding
to the sand depth indicated in Table 32. Thus, the hydraulic profiles
of a hypothetical nonintermixed filter and actual intermixed filter
are shown in Figs. 68 through 72. In Figs. 69 and 72, where the
sand or coal depth was not the same as in Figs. 65 and 66, the gradient
curve for the sand and coal was compressed vertically to fit the ex-
perimental sand depth. The actual observed location of the inter-
mixing zone is shown in Figs. 68 through 72. The upper limit of
the intermixing zone corresponds to the lowest filter level where only
coal particles were evident, and the lower limit corresponds to the
261
-------
1.00
NJ
0*
NJ
£
LU
Q
0.80
0.60
0.40
0.20
0.00
GRADED SAND
FLOW RATE 8 gpm/«q ft
TEMPERATURE 18 °C
SAND A O
SAND A2 A
SAND C D
SAND C2 •
0.00 0.05 0.10 0.15 0.20 0.25 0.30
HEAD LOSS PER 1-1/2-m. DEPTH, PLOTTED AT TOP OF THE
DEPTH INCREMENTS, ft
Fig. 65. Head loss for individual graded media in 1-1/2-in. unit filter sections.
0.35
-------
CO
COM.
COAL
COAL
COAL
COAL
COAL
COAL
COAL
COAL
COAL
0.00 0.01 0.02 0.03 0.04 0.05 0.06
HEAD LOSS PER 1-1/2-in. DEPTH, PLOTTED AT TOP OF THE
DEPTH INCREMENTS, ft
Fig. 66. Head loss for individual graded media in 1-1/2-in. unit filter sections.
0.07
-------
2.00
*
%
JE
LU
Q
Q
to
^
4s
DUAL MEDIA AA
o
18"C4gpin/sqft O
0.50 -
0.00
0.00
0.25
0.50
0.75
1.25
CUMULATIVE HEAD LOSS, ft
Pig. 67. Fixed-bed head loss of dual Media AA at various temperatures and flow rates.
-------
10
£
X
LU
Q
2
CD
85
2.00
1.60
1.20
0.80
0.40 —
0.00
0.00
TEMPERATURE 18 °C
FLOW RATE 8 gpm/iq ft
INTERFACE RANGE
UPPER LIMIT T~
AVERAGE LIMIT
LOWER LIMIT
0.05
0.10
0.15
0.20
0.25
0.30
0.35
HEAD LOSS PER 1-1/2-in. DEPTH, PLOTTED AT TOP OF THE
DEPTH INCREMENTS, ft
Fig. 68. Head loss per 1-1/2-in. unit depth in dual media AA and head loss for the
two-component media if unmixed.
-------
NJ
2.00
1.60
1.20
04
Q
0.80
0.40
0.00
0.00
COAL C
DUAL MEDIA AC
INTERFACE RANGE
UPPER LIMIT ~T
AVERAGE LIMIT —
LOWER LIMIT _L
TEMPERATURE 18 °C
FLOW RATE 8 gpm/sq ft
I
I
0.05
0.10
0.15
0.20
0.25
0.30
0.35
HEAD LOSS PER 1-1/2-in. DEPTH, PLOTTED AT TOP OF THE
DEPTH INCREMENTS, ft
Fig. 69. Head loss per 1-1/2-in. unit depth in dual media AC and head loss for the
two-component media if unmixed.
-------
2.00
1.60
-------
N>
ON
00
2.00
1.60
1.20
x
£ o.so
Q
9
oo
55
5 0.40
0.00
INTERFACE RANGE
UPPER LIMIT
AVERAGE LIMIT
LOWER LIMIT
TEMPERATURE 18°C
FLOW RATE 8 gpm/sq ft
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
HEAD LOSS PER 1-1/2-in. DEPTH, PLOTTED AT TOP OF THE
DEPTH INCREMENTS, ft
Fig. 71. Head loss per 1-1/2-in. unit depth in dual media A~E and head loss for the
two-component media if unmixed.
-------
a*
VO
1.60
^
£ 1.20
t
LU
0
a
CD
2j 0.80
^
0.40
0.00
0.
/ M T
1 V
-1- COAL F ^s.
L
1 1
00 0.05 0.10
INTERFACE RANGE
UPPER LIMIT T~
AVERAGE LIMIT —
LOWER LIMIT J_
DUAL MEDIA AF
^L_ -----jr^-^^-^
^^-a^- — . — — — — ^^
TEMPERATURE 18°C
FLOW RATE 8 gpm/sq ft
SAND A
1 1 1 1
0.15 0.20 0.25 0.30 0.
35
HEAD LOSS PER 1-1/2-in. DEPTH, PLOTTED AT TOP OF THE
DEPTH INCREMENTS, ft
Fig. 72. Head loss per 1-1/2-in. unit depth in dual media AF and head loss for the
two-component media if unmixed.
-------
highest filter level where only sand particles were evident. The
average interface was judged as the filter level at which the coal
and sand particles are approximately equal in number. Of course,
these intermixing zone distinctions are based on visual observations
and are subject to errors in judgment.
Figure 68 shows a typical example of head loss reduction at the inter-
face, which results when intermixing occurs. Without intermixing, the
head loss gradient in the sand layer at the interface, as shown in
Figure 68, would be expected to be 0.35 ft. Intermixing of the coal
and sand reduced this head loss gradient to 0.17 ft, a 50% reduction
in the head loss. This reduction was attributed to intermixing, be-
cause the zone of intermixing was visually observed to occur when-
ever the experimental hydraulic profile differed from the individual
component hydraulic profiles. As the very fine sand mixed up into
the coarser coal layer, a redistribution of the hydraulic profile
was observed, causing a greater head loss in the coarser coal layer,
with a correspondingly decreased head loss in the sand layer. The
enclosed areas of Figs. 68 through 72, which illustrate this decrease
in head loss, labelled I, in the sand layer and increase in head loss,
labelled II, in the coal layer, are approximately equal in area, ex-
cept in cases where excessive intermixing occurred.
The installation of various graded coal media with the same graded
sand medium, as illustrated in Figs. 68 through 70, did not signi-
ficantly change the 0.17>~ft maximum head loss at the interface.
Therefore, the graded sand medium was deemed as controlling the
interface head loss, regardless of the coal media involved in the
intermixing process. The effect of using increasingly coarser coals
with the same sand is also evident in Figs. 70 through 72. The
portion of bed showing intermixing increases with the coarser coals.
The general effect of intermixing in Figs. 68 through 72 is to cause
a gradual decrease in permeability with depth. This decrease would be
due to decreased average pore dimension and decreased porosity. The
influences of these variables on head loss (permeability) are evident
in Eq. (17).
The general effect of intermixing is to provide gradual coarse to
fine filtration in the direction of flow. The desirability of this
coarse to fine filtration is accepted by all researchers studying
filter performance. The fact that the peak hydraulic gradient of
the dual-media filter is less than the peak gradient of the sand
filter alone would indicate that the effectiveness of the sand layer
in filtration would be somewhat diminished. However, the diminishment
would be offset by the improved effectiveness of the lower coal layers
where intermixing is present. The prediction of relative filter per-
formance, in light of these observations, cannot be made at this time.
Figures 70 through 72 demonstrate excessive intermixing. Sand
particles actually reached the surface of the coal layer and caused
270
-------
an increase in the head loss throughout the coal layer as the coal
bed permeability decreased due to the presence of the sand. The head
loss increase in the coal layer was especially significant, as shown
in Fig. 72. The capacity of the dual-media filter could be expected
to be decreased due to the finer pore size in the surface layers of
the filter. Thus, Fig. 72 gives an example of too much intermixing.
The water shutdown procedure following backwashing was also considered
important to downflow hydraulic profiles and thus to filter performance,
Preliminary work was done with dual media A2C2 to determine the effect
of shutdown on intermixing. Two shutdown procedures, a slow 1-min
valve closure and a fast 5-sec valve closure, were tested. The bed
height was 12.9 in. for all test measurements. Given high fluidizing
flow rates, a large amount of intermixing was achieved before shut-
down. A fast shutdown procedure allowed for a greater amount of the
intermixing to be retained as the coal and sand particles settled
than did a slower shutdown procedure. The slower, 1-min procedure
allowed the particles of the two media to separate and stratify.
Expansion — Flow Rate Observations
Garnet Sand
The raw data collected during expansion of garnet sands in the 6-in.
column were; (1) flow rate, (2) bed height, (3) manometer readings
at 3-in. increments of the bed, and (4) temperature.
The following analyses of these data were made.
1. Flow meter reading was corrected by the appropriate calibration
equation and changed to velocity (in gpm/sq ft and fps) based
on the open cross-sectional area of the column.
2. Bed height reading was corrected by adding 1 in. The average
porosity ratio of the expanded bed was calculated from the
expanded height of the bed by Eq. (2) using the previously
determined e and the observed 1 .
o o
3. Pressure loss through the bed was calculated by subtracting
a piezometer reading above the expanded bed from the piezometer
reading at the bottom of the bed column.
A typical example of the data (run 1, Series A-13) is given in
Table 39.
The computed values of the data for each expansion run were plotted
as pressure vs flow rate, e.g., Figs. 73 and 74, and expanded bed
height vs flow rate, e.g., Figures 75 through 84.
An analysis of the pressure loss vs flow rate figures leads to the
following observations:
271
-------
Table 39. Expansion — flow rate data of run 1, Series A-13 (-14+16
garnet sand media).
Flow rate
gpm/sq ft
110.0
99.8
93.0
82.3
73.7
68.2
61.3
56.4
51.8
48.8
45.4
42.6
38.6
35.4
32.5
27.5
23.6
18.6
12.2
Corrected
(V)
fps
0.246
0.222
0.207
0.184
0.164
0.152
0.137
0.127
0.116
0.109
0.101
0.0950
0.0859
0.0788
0.0725
0.0613
0.0525
0.0415
0.0273
Bed height
U),
in.
20.50
19.50
18.75
18.00
17.75
17.25
16.63
16.25
15.75
15.63
15.50
15.25
14.88
14.50
14.25
13.80
13.70
13.65
13.65
Average
porosity
ratio
(e)
0.636
0.618
0.603
0.586
0.580
0.568
0.552
0.541
0.527
0.523
0.519
0.511
0.499
0.486
0.477
0.460
0.456
0.454
0.454
Pressure loss
through bed,
ft
1.83
1.87
1.87
1.90
1.90
1.90
1.88
1.88
1.87
1.86
1.86
1.85
1.85
1.86
1.86
1.73
1.41
1.04
0.64
for this run collected during contracting of fluidized bed only,
Water temperature = 16.5°C.
272
-------
2.0
RUN 1
TEMPERATURE 16.5 °C
000 0000 OOOOO 00 O
1.0
c 0.0.
8
UJ
at
to
e
2.0
0.05
0.10
0.15
0.20
RUN 2
TEMPERATURE 25.0 °C
OOOQO oo o o
O O O
0.25
1.0
0.0
0.0
OJO 0.15
FLOW RATE, V, fps
0.20
0.25
Fig. 73. Pressure loss - flow rate diagram for garnet sand
media (-14+16).
273
-------
2.0
1.0
c o.o ^
„ 0.0
8
LU
2.0|
1.1
RUN 7
TEMPERATURE 17.0 °C
«••• •
OT02
0.04
0.06
0.08
RUN 8
TEMPERATURE 25.0 °C
OA4 0.06
FLOW RATE, V, fps
0.08
0.10
0:10
Fig. 74. Pressure loss - flow rate diagram for garnet sand
media (M-60-80).
274
-------
WEN AND YITS MOOR
AUTHOrS MCCB. fe
to
•vl
Ui
If
17 —
13
O O
MEDIA - 14 + 16
TEMPEIATUKE 16 °C
0.04
0.08
0.20
0.12 0.16
FLOW RATE, V, fps
Fig. 75. Expansion - flow rate characteristics (garnet sand, run 1).
0.24
0.28
-------
K>
••4
21
— 17
m •'
x
15
13
WEN AND YU'S MODa
AUTHOR'S MODEL 60
O O O
0.04
MEDIA -14 + 16
TEMPERATURE 25.0 °C
0.08
0.20
0.12 0.16
FLOW RATE, V, fps
Fig. 76. Expansion - flow rate characteristics (garnet sand, run 2).
0.24
0.28
-------
ro
WEN AND YU'S MODEL
AUTHOR'S MODEL 6a
MEDIA -25+30
TEMPERATURE 16.0 °C
0.04 0.06
FLOW RATE, V, fps
0.10
0.12
0.14
Fig. 77. Expansion - flow rate characteristics (garnet sand, run 3).
-------
NJ
~J
oo
23
, 19
O
UJ
a
15
13
WEN AND YU'S MODEL
AUTHOR'S MODEL 60
0
tf
00,0 00)
X
X
MEDIA -25+30
X TEMPERATURE 25.0 °C
X
0.02
0.04
0.10
0.06 0.08
FLOW RATE, V, fps
Fig. 78. Expansion - flow rate characteristics (garnet sand, run 4).
0.12
O.U
-------
40
38 —
36
34
32
30
!E
o
26
24
22
20
18
16
14
WEN AND YU'S MODEL /
AUTHOR'S MODEL 60 /
AUTHOR'S MODEL 6b /
MEDIA -50 + 60
TEMPERATURE 17.0 °C
0.02 0.04 0.06
FLOW RATE, V, fps
0.08
0.10
Fig. 79. Expansion - flow rate characteristics (garnet sand,
run 5).
279
-------
WEN AND YU'S MODEL
AUTHOR'S MODEL 6a
AUTHOR'S MODEL 6b
MEDIA -50 + 60
TEMPERATURE 25.0 °C
0.02 0.04 0.06
FLOW RATE, V, fps
Fig. 80. Expansion - flow rate characteristics (garnet sand,
run 6).
280
-------
WEN AND YU'S MODEL /
AUTHOR'S MODEL 60
- AUTHOR'S MODEL
MEDIA M-60-80
TEMPERATURE 17.0 °C
FLOW RATE, V, fps
Fig. 81. Expansion - flow rate characteristics (garnet sand,
run 7).
281
-------
WEN AND YU'S MODEL /
AUTHOR'S MODEL 6a /
AUTHOR'S MODEL 6b 6
MEDIA M-60-80
TEMPERATURE 25.0 °C
0.02
0.04 0.06
FLOW RATE, V, fps
0.08
0.10
Fig. 82. Expansion - flow rate characteristics (garnet sand,
run 8).
282
-------
to
00
32i
30
28
, 26
g 24
S
CO
16
22
20 —
18
WEN AND YU'S MODEL
AUTHOR'S MODEL 60
O
MEDIA M-36
TEMPERATURE 17.0 °C
0.02
0.04
0.06 0.08
FLOW RATE, V, fps
0.10
0.12
0.14
Fig. 83. Expansion - flow rate characteristics (garnet sand, run 9).
-------
32
30
.E28
^26
*
§24
22
LU
r
18
16
WEN AND YU'S MODEL
AUTHOR'S MODEL 60
20 I rv
f^ '" MEDIA M-36
TEMPERATURE 25.0 °C
doo oPi r^ I I
0 0.02 0.04 0.06 0.08 0.10 0.12 0.14
FLOW RATE, V, fps
Fig. 84. Expansion - flow rate characteristics (garnet sand,
run 10).
1. The apparent minimum fluidization can readily be determined
as the intersection of the two linear portions of the curve.
These values as well as the Vmf values determined from Wen
and Yu's equation [Eq. (53)], Leva's equations [Eqs. (51) and
(52)], and Frantz's equation [Eq. (54)] are listed in Table 40.
2. The effect of channeling is greater at about Vp and decreases
as expansion increases. The evidence of channeling is the
lower pressure drop observed for flow values just above minimum
fluidization. Channeling is especially noticeable for the finer
media (Figs. 77 through 82) and was visually observed and noted
during the actual fluidization. The channeling is attributed
to poor flow distribution into the expansion column.
The bed height vs flow rate figures (Figs. 75 through 84) also
present the results of calculated bed height for garnet sand media
by various expansion models. The models and discussion will be
presented later.
284
-------
Table 40. Summary of minimum fluidization velocities of garnet sand
media - V
mf
Minimum fluidization velocities, fps
From head loss Wen and
Garnet vs flow rate, e.g., Yu's Leva's Frantz's
Run media Figs. 73 and 74 Eq. (53) Eqs. (51) and (52) Eq. (54)
1
2
3
4
5
6
7
8
9
10
-14+16
-14+16
-25+30
-25+30
-50+60
-50+60
M-60-80
M-60-80
M-36
M-36
0.
0.
0.
0.
0.
0.
0.
0.
0.
067
074
027
033
0071
0078
0067
0076
031
0.036
0
0
0
0
0
0
0
0
0
0
.067
.075
.021
.026
.0040
.0048
.0042
.0051
.022
.026
0
0
0
0
0
0
0
0
0
0
.0059
.064
.024
.029
.0049
.0058
.0051
.0061
.025
.029
0
0
0
0
0
0
0
0
0
.158
.196
.040
.049
.0070
.0085
.0074
.0089
.041
0.049
The analyses of the log V vs log e plots for the various media are
the most important analyses of the expansion-flow rate experiments.
One typical plot, log V vs log e (run 1, Series A-13), is shown on
Fig. 85. The important characteristics of this figure are the
slope of the line or n slope and the velocity intercept at porosity
ratio equal to one. To remove the bias of fitting a straight line
to this plot, a linear regression analysis was performed. The V
and e data that were used for the linear regression analysis were
all of the data points above the intersection of the two straight line
portions of the pressure loss vs flow rate plot, e.g., Figs. 73 and
74. The results of this analysis for all the expansion flow rate runs
on garnet sand are given in Table 41. Included in this table are V±
in gpm/sq ft and fps, n slope, number of points used in the regression
analysis, the correlation coefficient of the log-log line, the
Reynold's number based on the arithmetic mean diameter and Vt, and
Richardson and Zaki's n slope calculated from Eq. (30), neglecting
d/D.
285
-------
-0.4
Fig. 85.
-0.3 -0.2 -0.1
LOG POROSITY RATIO, c
Log plot of V vs e for garnet sand media (-14+16)
(run 1, Series A-13.
The experimentally determined n slope values are higher than the n
slope values from Richardson and Zaki's Eq. (30). This was expected
from the literature and is attributed to the particle shape. Because
the garnet sand media has a very irregular shape, no attempt will be
made to correlate the n slopes for the data with the calculated n
slopes by Richardson and Zaki's equations. Rather, a unique equation
for the solution of n slope for garnet sand will be proposed.
286
-------
Following the same approach as Richardson and Zaki, log n vs log
Re^ was plotted (Fig. 86). The linear regression correlation
equation was determined using all the 10 sets of points from column
1 and column 7 of Table 41. The resulting equation is,
n = 5.758 Re
-0.0541
(69)
with a coefficient of determination for the log-log line, r = 73.32%.
For the evaluation of n slope, the Re. must be defined. The prop-
erties of the water and the linear dimensions of particle size are
readily available or can be determined. The available methods for
evaluating the velocity intercept, which equals the settling velocity
of a discrete particle for spherical particles, are discussed in
the section on prediction of settling velocities, beginning on page 229,
1.00
0.75
UJ
Q.
2
to
o
Q
0.50
0.25
o o
I
I
0 0.5 1.0 1.5 2.0 2.5 3.0
LOG REYNOLD'S NUMBER, Re. (BASED ON VELOCITY INTERCEPT, V.)
Fig. 86. Log plot of n slope vs Reynold's number - Re^ (for
garnet sand media, runs 1 through 10, Series A-13
through A-17).
287
-------
The weaknesses in three methods for determining settling velocity are
discussed as follows:
A comparison of the experimental settling velocities of discrete
particles (Table 38) with V^ (Table 41) does not show any reasonable
agreement. This observation is consistent with expectations from
the literature as previously described (pages 216 through 228) .
Graphical correlations of the ratio of Re^Re^ vs Ga have the in-
herent weakness of requiring a family of curves for varying e^ ,
and the graphs presented did not fit the data points presented in
the literature as well as would have been desired.
The graphical solution from a log- log plot of ReQ vs Ga would be
a fast and simple method of determining settling velocities (pages 229-
234 ) . An attempt was made to use the correlation of log Re0 vs
2/3 Ga of Coulson and Richardson [33, p. 147]. However, values
of Vg thus determined did not compare well with the experimental
V^ values in Table 41. But following the same method of plotting
the log Re^ vs log Ga (dropping the constant 2/3), the experimental
values plotted closely as a straight: line (Fig. 87). The linear
regression equation of this line is,
R6i = 0.0702 Ga0'823 (70)
2
with a coefficient of determination for the log- log line, r = 99.38%.
The values used for this determination and the resulting values from
use of Eq. (70) are listed in Table 42. The agreement of the Re.
values from the above equation when compared with the experimental
Re. result in an error of - 14.7 to + 16.7%.
Presentation of Garnet Sand Expansion Model
The modified model for the expansion of garnet sand is proposed based
on the empirical -correlations presented previously [Eqs. (69) and (70)]
It is outlined in the following steps:
1. Experimentally determine arithmetic mean diameter (djj), the
density of the particles (ps), and the fixed-bed porosity (e f)
2. Calculate Ga from the fluid and particle properties.
3. With Ga, calculate Re. from Eq. (70).
4. Calculate V. from Re and n slope from Eq. (69).
5. The porosity at any desired flow rate can then be determined
from Vi, n slope, and Eq. (28),
288
-------
Table 41. Results of the log V vs log e relationship for garnet sand media.
ts>
00
Run
1
2
3
4
5
6
7
8
9
10
n
slope
4.326
4.089
4.437
4.420
5.321
5.197
4.932
4.926
4.139
4.173
Velocity
log VL
2.90380
2.87682
2.42580
2.47664
1.8961
1.94649
1.81071
1.87366
2.33674
2.39184
intercept
gpm/sq ft
801.3
753.0
266.6
299.7
78.73
88.41
64.67
74.76
217.1
246.5
-------
o
o
4.0
3.0
0>
a:
eo
Z
Q 2.0
O
Z
fc
1.0
1.0 2.0 3.0 4.0
LOG GALILEO NUMBER, Go
5.0
6.0
Fig. 87. Log plot of Reynold's number, Re^ vs Galileo number,
Ga (for garnet sand media, runs 1 through 10, Series
A-13 through A-17).
e =
1/n
6a. Using the experimentally determined value of e ~ and selecting
any desired initial bed height (£Q), the expanded height of
the bed can be determined from Eq. (2),
I =
(1 - e)
6b. As an alternative approach, which eliminates the need to mea-
sure e^f in step 1, the Vmf can be calculated from a minimum
fluidization velocity equation. For this study, Leva's
equation [Eq. (51)] with Zabrodsky's correction factor [Eq. (52)]
for Re > 10 gave the most consistent results (Table 40) with the
290
-------
Table 42. Values of Reynold's numbers and Galileo's number for the
garnet sand media.
Run
and
series
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
A-13
A-13
A-14
A-14
A-15
A-15
A-16
A-16
A-17
A-17
Garnet
media
-14+16
-14+16
-25+30
-25+30
-50+60
-50+60
M-60-80
M-60-80
M-36
M-36
Ga
51,876
80,370
6,530
10,117
490
722
530
780
6,969
10,261
Re^ based
on V^ and
d_
m
625
730
104
146
13.1
17.8
11.0
15.5
87.5
120
Re^ from
Eq. (70)
533
764
96.8
139
11.5
15.8
12.3
16.9
102
140
7, error
- 14.7
4.6
- 6.9
- 4.8
- 12.2
- 11.2
11.8
9.0
16.6
16.7
V £ values determined from the intersection of the two straight
l?ne portions of the head loss vs flow rate plots. The emf is
then calculated from Eq. (28),
/.v \ 1/n
€mf = (~vf )
using the V^ and n slope from step 4 above. The expanded height
is then calculated in the same way with Eq. (2).
The values of Re.^, n, and V^, calculated from Eqs. (70) and (69)
and the definition of Reynold's number, are listed in Table 43.
The emf values calculated from Leva's Vmf values of step 6b are also
listed in Table 43.
The results of expanded bed height calculated using the above model,
including step 6a (designated authors' model 6a), are presented in
Figs. 75 through 84 for runs 1 through 10. Similar results for the
291
-------
Table 43. V,, n. « . to be used in author's expansion models.
l tnr
Run
1
2
3
4
5
6
7
8
9
10
Garnet
media
-14+16
-14+16
-25+30
-25+30
-50+60
-50+60
M-60-80
M-60-80
M-36
M-36
Ga
51,876
80,371
6,530
10,117
490
722
530
780
6,969
10,260
Re^ from
Eq. (70)
533
764
96.8
139
11.5
15.8
12.3
16.9
102
140
n from
Eq. (69)
4.10
4.02
4.50
4.41
5.05
4.96
5.03
4.95
4.49
4.41
Vi (fps)
from Rei
definition
1.52
1.77
0.552
0.637
0.154
0.175
0.160
0.182
0.565
0.641
'./
0.453
0.443
0.497
0.496
0.505
0.504
0.505
0.503
0.497
0.496
Calculated from emf = (Vmf/Vi) where V^ determined by Leva's
equations [Eqs. (51) and (52)] and V^ and n from this table.
model using step 6b (designated authors' model 6b) are presented for
runs 5, 6, 7, and 8. The results of using step 6b for runs 1 through
4, 9, and 10 do not differ more than 1/4 in. from the results using
step 6a and, therefore, are not shown in the figures.
The garnet expansion model using step 6b requires a minimum of experi-
mental work. The average diameter from the sieve analysis and a
density determination for the particles are all that is required ex-
perimentally .
For comparison purposes, the expansion equation of Wen and Yu
[Eq. (36)] was also used to calculate expanded bed height, and the
results are presented in Figs. 75 through 84. It is apparent that
the Wen and Yu equation does not provide good predicted expansions.
Silica Sand and Coal
The expansion vs flow rate data for the uniform silica sands and
coals were analyzed in the same manner as the garnet sands. This
292
-------
was done to develop a model which could be used to predict expansion
of silica sand or coal using the same general approach previously
described for garnet.
The expansion vs flow rate data for the uniform media were first
analyzed to determine the porosity at each flow rate. Log porosity
was then plotted against log superficial velocity such as Fig. 85
presented previously. The slope of these curves n and the inter-
cept velocity (Vi) at e = 1.0 were determined by regression analysis,
The results are presented in Table 44.
The relationship between the n slope and Reynold's number based on
Vi was then determined for all the sand data and separately for all
Table 44. Results of log V vs log e regression analyses for uniform
sized silica sands and coals.
Media
designation
n
slope
Velocity
intercept
Vi,(fps)
Number of
values
used
Coefficient
of
determination
r2, 7.
based on
V and da
Galileo
no.
Silica Sand
A-10+12
A-12+14
A-14+16
A-16+18
A-18+20
A-20+25
A-25+30
A-30+35
A-35+40
A-40+45
Coal
2.573
2.665
2.900
2.909
3.153
3.296
3.504
3.704
4.126
4.206
0.557
0.500
0.492
0.420
0.383
0.320
0.281
0.258
0.236
0.187
11
13
13
14
15
16
14
13
14
11
99.76
99.95
99.75
99.82
99.88
99.37
99.47
99.54
99.47
99.55
347
263
217
157
119
84.7
61.9
47.9
36.9
26.1
125,296
74,939
44,054
26,803
15,662
9,553
5,539
3,327
1,958
1,404
F-4+7
F-7+8
F-8+10
F-10+12
A-10+12
A-12+14
A-14+16
A-16+18
A-18+20
A-20+25
A-25+30
3.016
2.832
2.920
2.939
3.254
3.448
3.667
3.479
3.523
3.988
3.739
ad * arithmetic mean
0.565
0.506
0.465
0.388
0.321
0.291
0.271
0.220
0.186
0.171
0.139
of adjacent
7
7
9
10
11
12
9
10
10
8
7
sieve
99.49
99.68
99.53
99.27
99.64
99.79
99.85
99.86
99.68
99.67
99.43
sizes.
726
447
344
242
201
153
119
82.0
58.1
45.1
30.6
484,270
157,652
92,145
55,558
53,284
31,869
18,734
11,398
6,660
4,062
2,356
293
-------
the coal data in Table 44. The results of the regression analysis
yielded the following relationships:
For silica sand A, all sizes in Table 44,
n = 7.973 Re^0'1947 (71)
2
with coefficient of determination, r = 98.16%.
For coal F and coal A, all sizes in Table 44,
n = 5.517 Re.'0'1015 (72)
2
with coefficient of determination, r = 78.36%.
Following the approach of the garnet sand expansion model, the rela-
tionships between Re. and the corresponding Galileo number (Ga) were
determined for the uniform silica sands and uniform coals. The values
of Ga are presented in Table 44. The regression analysis of log Re.
vs log Ga gave the following relationships:
For silica sand A, all sizes in Table 44,
Re± = 0.5321 Ga°'5554 (73)
2
with coefficient of determination, r = 99.74%.
For coals A and F, all sizes in Table 44,
Re± = 0.2723 Ga°*6133 (74)
2
with coefficient of determination, r = 99.39%.
The above relationships were then tested for validity by calculating
the expansion vs flow rate data for all uniform sands and coals.
The procedure was exactly as described previously for garnet sand
(authors' model 6a) using the appropriate n vs Re. and Re^ vs Ga
relationship for the media under analysis.
The results for the prediction of uniform sand expansions were very
good as expected from the high coefficients of determination for
Eqs. (71) and (73). The predicted values of n and V± for the various
sands are presented in Table 45, along with the maximum percent error
in the predicted expanded bed depth up to a total expansion of 70%.
It is apparent from Table 45 that the uniform sand expansions can be
predicted acceptably, with errors not exceeding about 7% of actual
observed values.
294
-------
Table 45. Predicted values of n, V^ with errors of prediction and
maximum error pf prediction of expanded bed depth for
uniform sands.
Sand
designation
A(-10+12)
AC-12+14)
A (-14+16)
A(-16+18)
A(-18+20)
A(-20+25)
A(-25+30)
A(-30+35)
A(-35+40)
A( -40+45)
Vi
from
Eq. (73),
fps
0.559
0.499
0.443
0.397
0.352
0.316
0.280
0.250
0.222
0.206
%
error
in
V
- 3.64
0.25
'9.51
5.38
7.17
8.42
1.94
0.94
3.60
- 0.45
n
from
Eq. (71)
2.533
2.678
2.836
2.993
3.172
3.346
3.549
3.751
3.972
4.117
%
error
in
na
1.52
- 0.51
2.19
- 2.90
- 0.60
- 1.53
- 1.29
- 1.26
3.73
2.11
Maximum
7, error in
expansion'3
+ 2.2
- 1.1
- 7.5
- 7.1
- 6.5
- 2.3
- 2.1
- 3.3
- 3.6
+ 5.9
aCompared to values in Table 44.
[(Observed depth - predicted depth) 100/observed depth]. Maximum
observed error in predicted expansion up to a total bed expansion
of 70% above fixed bed depth.
On the other hand, the prediction of coal expansions were not as good
as shown in Table 46. It appears that the model provides an acceptable
prediction for the uniform sizes from coal A, but is unable to do as
well for the uniform sizes from coal F. This may be due, in part, to
different sphericity for the two coals. It may also be due to less
well-defined uniform sizes due to the angularity of the crushed coals
and resulting sieving difficulties. More work should be done on the
crushed coals to attempt to improve the prediction accuracy.
The difficulties encountered with the coal expansion prediction empha-
size one important weakness in the expansion models: they do not
incorporate any direct measure of sphericity. They are empirical
295
-------
Table 46. Predicted values of n, V^ with errors of prediction and
maximum error of prediction of expanded bed height for
uniform coals.
Coal
designation
F-4+7
F-7+8
F-8+10
F-10+12
A-10+12
A- 12+14
A-14+16
A-16+18
A-18+20
A- 20+25
A-25+30
Vi
from
Eq. (74)
fps
0.629
0.459
0.395
0.343
0.334
0.289
0.249
0.217
0.187
0.163
0.140
%
error
in Via
- 11.10
9.20
15.13
11.69
- 3.96
0.39
7.85
1.28
- 0.19
4.73
- 0.62
n
from
Eq. (72)
2.787
2.989
3.090
3.189
3.197
3.301
3.413
3.520
3.639
3.753
3.883
7.
error
in
na
7.58
- 5.52
- 5.85
- 8.49
1.74
4.24
6.94
- 1.16
- 3.30
5.89
- 3.83
Maximum
% error in
expansion"
13.7
- 15.0
- 26.0
- 24.8
4.7
5.0
1.9
- 1.6
- 3.8
4.1
- 2.7
Compared to values in Table 44.
b[(Observed depth - predicted depth) 100/observed depth]. Maximum
observed error in predicted expansion up to a total bed expansion
of 70% above fixed bed depth.
models appropriate to media of about the same sphericity as that used
in their development. Thus, they should be used with caution on
media from other sources with potentially different sphericity. Future
work should include development of simple direct measures of sphericity
and collection of additional data on the effect of sphericity on the
drag coefficient and on the expansion models.
Since the prediction of expansion of the uniform sands was considered
acceptable, the models were used to predict the expansion of the
three graded (A, A-2, and C) sands previously described in Fig. 60
296
-------
and for graded sand observations reported by Amlrtharajah [4]. The
results of the prediction are summarized in Table 47. The procedure
consisted of the following steps: (1) calculation of the average
grain diameter by the inverse definition [Eq. (38)], (2) calculation
of the Galileo number from the properties of the media and fluid,
(3) calculation of Re from Eq. (73) and V± from Re^ (4) calculation
of n from Eq. (71), (5) calculation of e at each superficial flow
rate, V from Eq. (28) using V. and n calculated above, and (6) cal-
culation of expanded bed depth from Eq. (2).
It is evident from Table 47 that the predicted bed depths are all
higher than observed, and the prediction is not too good. An attempt
was made to determine the reasons for the consistent over-prediction.
One cause is the choice of e0 used in Eq. (2). Values of eo of 0.42
were used for the graded sands of Boss and 0.41 for those of
Amirtharajah. These choices were based on porosities by the
graduate cylinder technique for graded sands described previously
and values reported by two investigators. If a value of 0.44 had
been selected as determined by the column technique (Table 37),
the prediction would have improved. The arithmetic mean diameter
[Eq. (39)] consistently yields a larger diameter than the inverse
diameter [Eq. (38)], as evidenced by Table 35. If the larger
diameter defined by Eq. (39) had been used, the prediction would
be improved. A spot check of the effect of these two factors on
prediction for three graded sands in Table 47 indicates that the maxi-
mum percent errors reported would be reduced about 5% (e.g., from
-15 to -10%).
From this analysis, it is evident that the prediction adequacy is
sensitive to choice of eo and mean diameter. Acceptable predictions
seem possible if the column technique is used to evaluate e and the
arithmetic mean definition [Eq. (39)1 is used to calculate mean grain
diameter.
The expansion of the graded sands was also calculated using the
incremental approach presented in some engineering textbooks[46].
This approach calculates the expansion of increments of the media
between adjacent sieves and sums the expanded depth of the increments
to determine the total expanded depth. The results of the incremental
approach were no better than, and in some cases worse than, those
presented in Table 47.
In view of the relative inaccuracy of prediction of expansion of the
uniform coals, no attempt was made at this time to test the prediction
accuracy of the models for graded coals.
Minimum Fluidization Velocity of all Media
The minimum fluldization velocity (Vmf) can be defined in a number of
ways. For uniform media which fluidizes sharply at a particular flow
297
-------
Table 47. Prediction of expanded bed depths for graded sands using
models developed for uniform sands and average diameter
based on the inverse definition [Eq. (38)].
Sand
From
A
A-2
C
From
A
A
A
A
A
B
B
B
B
C
C
Avg
dia,
nun
Boss [14]
0.686
0.716
0.947
Water
temp»
°C
17
26
22
Ga
4,383
7,736
14,838
R6i
56
76.9
110.4
Vi,
fps
0.2926
0.3092
0.3353
n
3.640
3.424
3.191
Maximum
% error
in expanded
depth3
- 7.10
- 6.6
- 17.3
Amirtharajah [4]
0.612
0.612
0.612
0.612
0.612
0.919
0.919
0.919
0.919
0.688
0.688
15
13.5
30
30
30
15
15
30
30
22
35.5
2,824
2,613
5,773
5,773
5,773
9,599
9,599
19,623
19,623
5,725
10,401
43.9
42.0
65.3
65.3
65.3
86.6
86.6
129
129
65.0
90.6
0.270
0.269
0.281
0.281
0.281
0.354
0.354
0.370
0.370
0.298
0.309
3.818
3.850
3.534
3.534
3.534
3.347
3.347
3.096
3.096
3.537
3.316
- 9.5
- 11.5
- 14.9
- 16.2
- 16.2
- 15.8
- 17.8
- 15.8
- 16.8
- 17.4
- 17.7
a[(Observed depth - predicted depth) 100/observed depth], over the
full range investigated, generally to an expanded depth of about
50% of fixed bed depth.
298
-------
rate, it is frequently defined as the point of intersection obtained
through extrapolation of the two linear sections of the envelope curve
of head loss vs superficial velocity such as Figs 73 and 74. The
values of Vmf based on this definition for garnet sands have been pre-
sented in Table 40. However, this definition is not meaningful for
graded media because the coarser sizes of media comprising the bed
are not fluidized at the Vmf defined in the above manner.
It is also difficult to define Vmf on the basis of first visual ap-
pearance of complete fluidization because it is subject to the ob-
server's visual definition of complete fluidization. In view of these
difficulties and the fact that all media studied were graded in size
to some extent (even the uniform media), the minimum fluidization
velocity was defined as that flow rate required to achieve 10% bed
expansion. This expansion is close to the minimum rate at which the
bed first appears fluidized. The Vmf values thus determined for the
three media studied are presented in Fig. 88. These data are based
on the expansion studies for single uniform media only in Table 33
(Series C-l and 3 and D-l) and in Table 31 (Series A-13, 14, and 15).
This type of data is believed to be useful in determining minimum
backwash rates to ensure fluidization of all media comprising the bed.
If the sieve analysis of each media is available or specified, the
backwash rate needed to achieve fluidization of the coarse sizes in
each media should be provided. Furthermore, it would appear that
media should be specified so that the coarse sizes of each media
comprising the bed are fluidized at roughly the same minimum
fluidization velocity. Thus, the entire bed will become fluidized
simultaneously. The total bed expansion can then be determined by
summing up the calculated expansions of the individual media.
Unfortunately, the experiments depicted in Fig. 88 were not conducted
over a broad enough range of temperatures to present similar empirical
data for other temperatures. The effect of temperature on Vmf can be
judged from various models for V £ such as Eqs. (51) through (54).
Data were collected on graded sand A and graded coal A at four dif-
ferent temperatures from 16 to 30 °C. To illustrate the effect of
temperature on V^, the values of V ^ to achieve 10% expansion of
these two media were determined. The results are plotted in Fig. 89.
It is evident that temperature does have a distinct effect on Vmf
as would be expected in the transitional or laminar fluid regimes.
These curves could be used to select a rough temperature correction
factor to be applied to the data of Fig. 88 in selecting Vmf for
other temperatures. Inspection of Fig. 89 shows that change in Vmf
is not a linear inverse function of viscosity as would be expected
in the transitional fluid regime. Only in the laminar regime would
the relation be linear.
299
-------
COAL p = 1.7 o
SILICA SAND p = 2.65 D
GARNET SAND p« 4.13
.0 0.25 0.5 0.75 1 1.5 2
MEAN SIEVE SIZE, mm
2.5
Fig. 88. Minimum fluidization velocity, V f, to achieve 10% bed
expansion at 25 °C. m
0.04
0.03
0.02
0.01 h
10
20
TEMP, °C
30
1.0 1
0.5
§
u
i/>
O
u
i/>
£
Fig. 89» Effect of temperature on V - for sand and coal and en
-• - -* viscosity of water.
300
-------
Comparison of Fig. 89 with Table 40 shows reasonable agreement with
change in Vmf with temperature. For example, the average ratio of
Vmf values for garnet at 25 °C and 17 °C is 1.14 in Table 40. The
same ratio for sand and coal in Fig. 89 is 1.17.
It should be noted that the coarse sizes in a given media control
the point of complete minimum fluidization, and skimming the fines
will not alter the V c.
mi
Intermixing Observations
Garnet and Silica Sand
The objective of the Intermixing experiments was to test the validity
of the bulk density approach for the prediction of intermixing of
the small-sized dense garnet sand and larger-sized less dense silica
sand. A uniform-sized garnet sand and various uniform-sized silica
sands were first fluidized individually, and then the two-component
mixtures of the silica sands and garnet sands were fluidized to ob-
serve their intermixing behavior.
The data collected for the single-component fluidization experiments
included flow rate and bed height. From the individual component
fluidization data the following values were calculated:
1. Flow meter readings were corrected to give corrected flow rate
using the appropriate calibration equation.
2. From the bed height reading, the average porosity of the bed
at various flow rates was calculated from the known weight of
media in the column, the column cross-sectional area, and the
particle density.
3. The average bulk density was then calculated at the same flow
rates from the above porosity values and the solid and fluid
density by Eq. (41).
The computed values of flow rate and bulk density from single-component
fluidization data for each media studied are shown in Fig. 90. An
interesting and important fact can be observed on Fig. 90. The slope
of the garnet sand curve is steeper than any of the silica sand curves.
Thus, maximum bulk density differences (garnet — silica sand) occur at
the lowest flow rates.
The data collected for the two-component fluidization experiments
were the bed height, flowmeter reading, the visual observations of
relative media concentrations at various depth in the bed. The
results of these two-component fluidization experiments are shown in
Figs. 91 through 94.
301
-------
- 2.4
o GARNET SAND-50+ 60
D SILICA SAND-20+ 25
SILICA SAND-30+ 35
• SILICA SAND -35 + 40
A SILICA SAND-40+ 45
- 1.0
10 20 30 40
FLOW RATE, V, gpm/sq
Fig. 90. Bulk density vs flow rate for garnet sand and silica sand.
Before the intermixing observations were made, the two-component mix-
ture was fluidized and contracted to a fixed-bed state very slowly.
The data presented in Figs. 91 through 94 were then collected during
the expansion of the two-component mixtures. After each incremental
increase in flow, sufficient time was allowed to reach equilibrium
conditions before the intermixing observations were recorded. The
expansion was carried up to about 200%. The major problem encountered
in collecting and presenting this type of data was that it was difficult
302
-------
OJ
o
60
50
40
g30
x
Q
GARNET SAND S
SILICA SAND 0
INTERMIXING D
120
10
0 10 20 30 40 NO FLOW
FLOW RATE, V , gpm/sq ft
Fig. 91. Intermixing of -50+60 garnet sand
and -2OI-25 silica sand.
60
50
c 40
**
x
§ 30
x
o
LU
CO
20
10
GARNET SAND S
SILICA SAND 0
INTERMIXING D
Fig. 92.
10 20 30 40 NO FLOW
FLOW RATE, V , gpm/sq ft
Intermixing of -50+60 garnet sand
and -30+35 silica sand.
-------
U)
GARNET SAND C3
SILICA SAND CZ3
INTERMIXING O
Fig. 93.
10 20 30 40 NO FLOW
FLOW RATE, V , gpm/sq ft
Intermixing of -50+60 garnet sand
and -35+40 silica sand.
60
50
40
O30
i
Q
20
10
GARNET SAND E3
— SILICA SAND E3
INTERMIXING I—I
10 20 30 40 NO FLOW
FLOW RATE, V , gpm/sq ft
Fig. 94. Intermixing of -50+60 garnet sand
and -40+45 silica sand.
-------
at times to decide in which regions of the expanded bed one component
was the sole component and in which the components were intermixed.
Three regions usually existed — one region at the top and one at the
bottom of the bed where the components were clearly or closely the
sole component of the layer, and one between these two separate layers
which was a region of intermixing. Within this region of intermixing,
the concentration of each component decreased with distance away from
the adjacent region where it was the major component. The interfaces
between the three different regions are distinguished on the figures
by a solid line which indicates a sharp interface or a dashed line
representing a diffuse interface over a bed depth of 2 to 4 in.
After the two components were expanded to the maximum expanded height,
the flow of water was quickly shut off, and the bed was also allowed
to settle. The stratification of the bed after this settling is also
included in Figs. 91 through 94.
The analysis of Figs. 91 through 94 indicates a trend for the garnet
sand to occupy the lower layer of the bed at low flow rates, then as
the flow rate increases, the two-component mixtures are intermixed.
At still higher flow rates, the garnet sand occupies the upper portion
of the bed, and the silica sand occupies the bottom of the bed. These
trends would be expected from the bulk density plots of the single
component data as shown in Fig. 90, because the bulk density of the
garnet sand decreased more rapidly than the silica sand as the flow
rate increased.
An exception to this trend was noticed in the fluidization of the
-20+25 silica sand and -50+60 garnet sand mixture where the silica
sand occupied the lower portion of the bed at all flow rates. This
can be explained by considering the minimum fluidization velocities
of the silica sand and the garnet sand. The minimum fluidization for
the -50+60 garnet sand was observed to be 3.0 to 3.5 gpm/sq ft (0.007
to 0.008 fps). The silica sand minimum fluidization velocity was
approximately 7.5 gpm/sq ft (0.017 fps) for the -20+25 size. There-
fore, when the fluidized mixture of -20+25 silica sand and garnet
sand was contracted slowly, the -20+25 silica sand reached the fixed
bed state while the garnet sand was still fluidized and before the
bulk density of the garnet sand was sufficiently greater than the
silica sand to occupy the bottom layer of the bed. Just below mini-
mum fluidization for the -20+25 silica sand, it was noticed that the
garnet sand, which was still fluidized, displaced a small portion of
the silica sand that was in a fixed state up into the intermixed
layer.
The garnet sand was below the silica sand at all flow rates for the
-40+45 silica sand and -50+60 garnet sand two-component mixture.
The maximum flow rate of this two-component mixture was 40 gpm/sq ft
(0.089 fps). At this flow rate, the garnet s-and still had a higher
bulk density than the -40+45 silica sand (from single-component data,
305
-------
Thus, as predicted by the bulk density approach, the garnet sand
should occupy the lower portion of the bed.
The sensitivity of results was hampered by the following factors.
1. The uniform sands used were not actually unisized and of con-
sistent shape. Because of this, there would be a tendency for
stratification within each individual media, and a bulk density
gradient would exist in each individual media. This was sup-
ported by the observation for all of the uniform media when ex-
panded and then contracted slowly, or when expanded and allowed
to settle after the fluid flow was stopped. There was a slight
but noticeable difference in particle size between the top and
bottom layers.
2. Because of the physical properties of the fluid and solids, the
fluidization of the components occurred in the transitional
regime of flow, and mixing and circulation patterns existed in
the bed during fluidization. This would tend to diminish the
bulk density gradients within the individual and two-component
mixtures.
3. The distribution of flow into the fluidizing column from the
calming section was not perfectly uniform. However, it was
quite good with short circuiting of upward flow usually limited
to 2 to 3 in. above the entrance and rarely extending 6 in. up
the column.
There are three major conditions of interest in the relative location
of the garnet sand component:
1. stratification of the garnet sand component in the bottom of the
bed with or without intermixing of garnet sand and silica sand
in layers above.
2. maximum intermixing of the garnet sand with silica sand, and
3. stratification of garnet sand in the top layer with or without
intermixing in layers below.
The following bulk density differences, Table 48, were obtained from
Fig. 90 and the appropriate two-component intermixing figures (Figs.
91 through 94) for the three conditions. The ratio of the diameters
of garnet sand to silica sand is also given below. From Table 48
or from Figs. 90 and 91 through 94, maximum intermixing of the two
components (Condition 2) does not occur at zero bulk density difference
as would be expected, but at a slightly positive bulk density differ-
ence, about 3 to 8 Ib/cu ft. These observations are in agreement
with the experimental results of Le Clair [74] who found that the
velocity for a homogeneous mixture as predicted from single-component
data was greater than the observed velocity where homogeneous mixing
occurred.
306
-------
Table 48. Bulk density difference, Ib/cu ft (garnet-silica sand).
Silica sand
media
Condition
(1)
Condition
(2)
Condition
(3)
, Silica
m sand
d_ Garnet
m ,
sand
-20+25
Did not
occur
8 to 3
< 12
2.84
-30+35
-35+40
-40+45
> 20
> 10
> 5
15 to 5
8 to 4
10 to =" 0
< 9
< 6
Did not
occur
2.04
1.70
1.47
It is evident from Table 48 that a single value of bulk density
difference cannot be readily selected which could be used by the
design engineer to ensure the desired degree of stratification or
intermixing between garnet sand and silica sand. For example, the
bulk density differences (garnet-silica sand), which ensure that
Condition 1 or 3 will exist, decreased with lower diameter ratio
(silica sand/garnet sand). The lower bulk density differences, indi-
cated in the table for Condition 1 and 3, occur at higher flow rates
and higher bed porosities.
Based on the data in Table 48, if one wanted to ensure that some
garnet sand would always remain on the bottom regardless of the back-
wash rate of the filter bed, one would need to select the media so
that the ratio of the bottom silica sand size to the bottom garnet sand
size is not more than 1.47. The next larger ratio could be used (1.70)
providing that the backwash rate was limited to 15 to 20 gpm/sq ft
(0.033 to 0.045 fps) or the operator would need to allow for a slow
contraction of the fluidized bed after backvashing to achieve restrati-
fication.
Attention should be drawn to the converging nature of the garnet sand
and silica sand curves in Fig. 90. Because of this converging nature,
excessive backwash rates result in increased tendencies to inter-
mixing and bed inversion. This fact should be considered in selecting
the media and backwash rate for dual- and multi-media filters.
In view of the apparent inadequacies, or insensitivity of the equal
bulk density approach to the prediction of intermixing, an attempt was
made to utilize the intermixing theory proposed by Camp et al. [26].
307
-------
This theory as stated in Eq. (49) suggests that mixing will occur if
the bulk density of the lower bed of smaller, more dense garnet grains
is less than the density of the larger, less dense upper particles
minus a drag force term for the upper grains. Figure 95 shows the
results of the calculations.
Figure 95-A shows that the -204-25 mesh silica sand should intermix
at all flow rates since its particle density less the drag term is
still greater than the bulk density of the garnet, top or bottom
layer. This intermixing was evident in Fig. 91.
Similarly, Fig. 95-B for -40+45 silica sand would predict complete
intermixing if the sand grains were spherical. However, if they were
cubical in shape, the drag term would be increased, and intermixing
would be only partial. The sand grains used in this study are rounded
in shape and would not be cubical in sphericity. Thus, intermixing
should be complete. However, Fig. 94 shows that most of the garnet
sand remained on the bottom of the bed at all flow rates for this
sand and garnet combination.
These observations do not prove or disprove the validity of the inter-
mixing model of Camp et al. [26]. An adequate measure of sphericity
would be needed to test the model. The unavailability of such a
measure is the same weakness preventing good prediction of bed expan-
sion for nonspherical particles discussed previously. Until this
weakness is resolved, the intermixing model of Camp et al. [26] is
no more useful than the equal bulk density model of Le Clair [74].
Silica Sand and Coal
Bulk density data and intermixing observations for the silica sand
and coal were collected in the same manner previously described for
garnet and silica sand.
A full range of uniform media was tested in a series of tests using
the 2-in. fluidization column. The uniform sizes tested ranged from
-10+12 mesh to -40+45 mesh for sand and from -4+7 mesh to -25+30 mesh
for coal. A known dry weight of a particular uniform filter medium
was placed in the column. The expanded bed height vs flow rate was
observed. Porosity at any flow rate was calculated from the known
weight of the medium, particle density, and expanded bed height. Bulk
density for each flow rate was calculated by Eq. (41) and then plotted
in Fig. 96. Bulk density, as defined in this study, is actually the
average fluid and particle composite density value within a filter
bed cross section. At zero flow rate, the sand bulk density was
approximately 1.9 g/cc while the coal bulk density was approximately
1.3 g/cc.
Figure 96 illustrates that, with an increasing flow rate, sand bulk
densities decrease much more rapidly than the coal bulk densities.
An extreme bulk density decrease of from 1.90 g/cc at zero flow rate
308
-------
to
UJ
O
u
i
UJ
2.5
2.0
1.5
1.0
0.5
2.5
2.0
1.5
1.0
0.5
p SILICA SAND
DRAG FORCE ON SAND
ASSUMING SPHERICAL
1HAPE
DRAG FORCE ON SAND -
ASSUMING CUBICAL SHAPE
GARNET - BOTTOM OF BED
«.
pb
GARNET - TOP OF BED
[A) -20 + 25 SILICA SAND AND -50 + 60 GARNET SAND
AT 25 °C
p$ SILICA SAND
DRAG FORCE ON SAND
ASSUMING SPHERICAL
SHAPE
DRAG FORCE ON SAND -
ASSUMING CUBICAL SHAPE
P. GARNET - BOTTOM
OF BED
GARNET - TOP OF BED
(I) -40 + 45 SILICA SAND AND -50+60 GARNET SAND
^—- _ ft _
AT 25 "C
0.00 0.02 0.04 0.06 0.08 0.10
SUPERFICIAL VELOCITY, fps
Fig. 95. Intermixing of silica sand and coal according to the
model of Camp et al. [26].
309
-------
-25+30
-30+35
-35+40
-40+45
F-7+8
F-8+10
F-10+12
65 —
u
u
o>
l/l
z
LU
Q
Fig. 96.
20 40 60 80
FLOW RATE, V , gpm/sq ft
Bulk density vs flow rate for coal and silica sands
(data points not shown on all curves for drafting con-
venience) .
310
-------
to 1.15 g/cc at 50 gpm/sq ft was exhibited by the -40+45 mesh sand.
The coal bulk densities showed less dependence on flow rate, especially
coarser coals down to 12 mesh. The general pattern is similar to that
for garnet and silica sand presented previously in Fig. 90.
A bulk density comparison between uniform media of the same size
range but different origin (i.e., -10+12 mesh coal A and -10+12
mesh coal F) is also plotted in Fig. 96. The bulk density difference
of these identically-sized uniform coals was attributed to dif-
ferences in specific gravity and particle shape of the two source
samples. Limited data restricted such comparisons of uniform
media from different sources to the two uniform media above.
A large bulk density difference also exists between the majority
of the sand and the majority of the coal media* If the back-
washing flow rate is limited to 30 gpm/sq ft, only the fine -30
+45 mesh sand approaches the bulk density of the coarse coal.
Intermixing observations were limited to -4+7 coal F and the various
uniform sands shown in Fig. 96. Again, it was desired to test the
validity and sensitivity of the equal bulk density theory of inter-
mixing of Le Glair [74]. In retrospect, it would have been better
to use several narrow, size ranges for the coal. Use of only the
broad -4+7 mesh coal in the intermixing observations led to incon-
clusive results regarding the bulk density difference associated
with partial or complete intermixing. Additional work in this area
should be conducted.
Expansion of Graded Dual-Media Filters
This section presents some general characteristics associated with
the expansion of dual-media filters.
(1) An example of the expansion plot for a graded dual-media filter
bed is found in Fig. 97. The bed height ordinate shows that approxi-
mately equal 12-in. quantities of sand A and coal A were systematically
expanded. The similarity shown between the expansion rate of sand A
and coal A resulted, because the coal average particle size was such
that the V , values were almost the same for sand A and coal A.
ml
Figure 97 illustrates that the predicted expansion, obtained by
adding bed expansion of the individual components, closely approxi-
mated the experimentally determined expansion of the dual-media
filter. The results of Fig. 97 support the work of Le Clair [74],
who, in restating the cell theory, suggested that in all cases each
fluidized particle within a medium has a definite, surrounding cell
of fluid at a particular flow rate. Introducing particles of dif-
ferent densities does not influence the size of the fluid cells for
the particular flow rate. Therefore, the expansion of the individual
components can be summed to predict dual-media filter expansion.
311
-------
42
36
30
24
i
o
a is
12
SUM OF TWO
SINGLE MEDIA
O SAND A
O COAL A
A DUAL MEDIA AA
10 15 20 25
FLOW RATE, V, gpm/sq ft
30
Fig. 97. Expansion vs flow rate of dual media AA and the two-
component media at 22 °C.
312
-------
(2) The expansion of dual media A-2^2 an<* fc^e tw£>-component media are
presented in Fig. 98. Because sand A2 became fluidized first, ex-
pansion of the dual-media filter at lower flow rates was close to
the amount of sand expansion. At higher flow rates, above fluidi-
zation for coal C2, expansion of dual media A2C2 was close to the
sum of the expanded depths of the two single media.
It is apparent from Fig. 98 that sand A£ begins to expand at about
7 to 8 gpm/sq ft and reaches 107. expansion at about 12 gpm/sq ft.
Coal C2 on the other hand begins expanding at about 15 to 16 gpm/sq ft
and reaches 10% expansion at about 22 gpm/sq ft.
The most important fact to be concluded from Fig. 98 is that if the
sand and coal of a dual-media filter are not selected to fluidize
at about the same flow rate, one media may remain fixed (or nearly
fixed) while the other is fluidized. The fixed bed portion may not
be cleaned adequately during the backwashing. For example, if dual
media h-2^2 were expanded 20%, the minimum normally required for
adequate backwashing, the coal would be expanded only about 5%. At
5% expansion, the coal is essentially in a fixed-bed condition and
would not clean well. Furthermore, to achieve 20% expansion of the
coal in this dual media would require a flow rate of 27 gpm/sq ft,
and the total bed expansion would be about 35%. Thus, observation
of a total bed expansion of 20% for this dual media would not
necessarily mean that both media would receive an adequate backwash.
313
-------
35
30
25
SUM OF TWO
SINGLE MEDIA
O SAND A2
COAL C2
a DUAL MIDIA
10 20 30
FLOW RATE, V, gpm/sq ft
40
Fig. 98. Expansion vs flow rate of dual media &2C2 and the two~
component media at 22 °C.
314
-------
XI. EFFECT OF MEDIA INTERMIXING ON
DUAL-MEDIA FILTRATION
Introduction
Dual-media filters composed of anthracite coal over silica sand are
widely used in water and wastewater filtration because they achieve
greater water production per filter run than sand filters if other
filtration conditions are the same. Furthermore, because of greater
potential production per filter cycle, the percentage of filtered
water used in backwashing is less for dual-media filters than for
sand filters.
The coal and sand of a dual media filter will sometimes intermix at
the interface, depending on the size, shape, and density of the two
media at the interface, the rate of backwash, and the backwash valve
closure rate. The most important parameter affecting the amount of
intermixing is the relative size ratio of the sand and coal at the
interface. The media may be so graded as to maintain a sharp inter-
face (coal size to sand size ratio at the interface of about 2 to 1)
or to allow a substantial zone of intermixing (coal size to sand size
of about 4 to 1).
There are two conflicting ideas concerning the desirable amount of
intermixing. Some researchers feel there should be no mixing at the
interface, with the sand acting as a polishing filter after the coal
roughing filter. Others feel that intermixing results in a more
uniform decrease in grain size with depth and allows more efficient
use of the storage space in the media and, thus, longer filter runs.
An intermixed bed is a closer approximation of the ideal coarse to
fine filter bed and thereby eliminates an impervious layer that might
build up at a sharp interface.
Objectives and Scope of This Study
Because of the differences of opinion concerning the desirability of
intermixing in a dual-media filter, this study was undertaken to show
what effect, if any, the intermixing has on the performance of dual-
media filters. Performance differences, if they exist, will be shown
by measurements of head loss development and effluent quality versus
time during filtration.
It must be kept in mind that this study was not meant to be a com-
parison of dual-media and single-media filtration nor a study to
determine the optimum amount of intermixing; rather, it is meant to
show how the media intermixing or non-intermixing in dual-media fil-
tration affects performance.
315
-------
Experimental Investigation
Apparatus and General Approach
The system used in this study consisted of three, 4-in. inside diam-
eter plexiglass filter columns as shown in Fig. 99. In one column,
the coal and sand were placed together and allowed to mix as they
might. In two other columns, operated in series, identical coal and
sand were placed separately. Two gradations of coal and sand media
were used, one which resulted in a mixed interface, dual-media bed
and one which resulted in a sharp interface dual media.
Prior to further detailed description of the experimental investiga-
tion some terms must be defined for clarity. The term "mixed inter-
face media" refers to the media which exhibited intermixing of the
anthracite and sand at the interface when both media were placed in
column 1. The term "sharp interface media" refers to the media which
exhibited no intermixing of the anthracite and sand at the interface
when both were placed in column 1. The term "combined media" refers
to the sand and anthracite media when placed together in column 1.
The term "separate media" refers to the same amount and gradation of
sand and anthracite placed in separate columns. The same anthracite
as in column 1 was placed by itself in column 2. The same sand as
in column 1 was placed by itself in column 3.
A filter run consisted of measuring the head loss buildup and efflu-
ent quality for the combined media and the head loss buildup and
effluent quality for the separate media when the separate and combined
media filters were operated concurrently. Comparisons between the
results obtained with mixed and with sharp interface media in subse-
quent filter runs were then made to evaluate the effects of inter-
mixing on filter performance.
The filters were operated as pressure type filters with the pump
applying approximately 25 psig pressure to the top of filter column
1 and column 2 (Fig. 99). The suspensions of particulates were
pumped from the supply to columns 1 and 2. From the separate coal
(column 2), the water flowed to column 3, containing the separate
sand. The filtered water coming from column 1 and column 3 flowed
through a pressure regulator, a needle valve, a rotameter, and to
waste.
Constant rate filtration was achieved by use of Fisher, type 95L
pressure regulators (Fisher controls Company, Marshalltown, Iowa).
These regulators were used to reduce any incoming pressure to a con-
stant exit pressure of 4 psig. This constant 4 psig pressure leaving
the regulator was applied to a fixed effluent needle valve to achieve
a constant flow.
This system provided a reasonably constant flow, although slight
variations in flow were observable on the effluent rotameter which
316
-------
PUMP
FILTER INFLUENT
BACKWASH
EFFLUENT
_BACKWASH_
EFFLUENT
I
NEEDLE VALVE
COLUMN
NO. 2
BACKWASH
COLUMN]
NO. 3
BACKWASH LINE
NEEDLE VALVE
Fig. 99. Schematic diagram of apparatus.
-------
had a full-scale capacity of 0.78 gpm. The rotameter discharged to
waste.
Tap water at normal main pressure was used for backwash. The back-
wash system is shown in the plumbing diagram, Fig. 99.
Head loss buildup was measured using a multiple tube manometer con-
taining four single leg manometers (Model 33KB35, Multiple Tube Ma-
nometer, Merian Instrument Division, The Scott and Fetzer Company,
Cleveland, Ohio). A 1/4-in. OD copper tube was connected from the
top of a filter housing to the bottom end of one manometer, while
another 1/4-in. OD copper tube was connected from the bottom of a
filter housing to the top end of the same manometer. The manometer
operates on the principle of a U-tube with one leg of water and one
leg of mercury.
Head loss buildup was observed by the differences in the readings of
the mercury level at successive time intervals during a filter run.
Initial head loss through each filter could not be readily measured,
but head loss increase with time was readily and precisely observed.
Before and after every run, the filters were backwashed. The filters
were expanded to 50% during the backwash and allowed to wash thor-
oughly. When a cake of solids formed on the surface of the media,
air scour was sometimes used. After a thorough backwash, the back-
wash valve was closed rapidly. Then the bed was shocked momentarily
by rapidly opening and closing the backwash valve several times in
succession. This procedure was used to insure an equally dense bed
for every filter run. The bed depth was recorded after every wash
for each run and did not vary.
Because the patterns of head loss buildup and effluent quality were
of more importance than absolute quality, the filters were operated
at 7 gpm/sq ft. This high filtration rate was also selected to en-
courage deep penetration of some solids into the bed, hopefully
through the interfacial region to allow the effects of the interface
to be observed.
Filter Media
Two different dual media were used for this study. One dual media
was selected in order that intermixing would occur at the interface.
The second dual media was selected in order to produce a sharp, well
defined interface. The sharp interfaced dual media was intended to
act as a control, to show what differences in head loss and filtrate
quality could be attributed to the experimental apparatus and opera-
tion. However, the sharp interface media was not a control in the
strict sense of the word, because the sharp interface media and the
mixed interface media were not operated at the same time.
318
-------
Any effects on head loss development and effluent quality produced by
the method of operation or the equipment (presenting two surfaces to
the flow, flowing through an underdrain and plumbing, etc.) had to
be determined. The filtration with the sharp interface media was
done to illustrate these effects. As there was no mixing of the
media in column 1 (Fig. 99), there would be no effect of intermixing
on the performance of the filter in column 1. There was obviously
no intermixing between the coal in column 2 and the sand in column 3
and no effect due to intermixing. A comparison of the results of the
combined media filtration with the separate media filtration using
the sharp interface media would show what effect the operating proce-
dure and equipment had on the results. This observation would then
be useful in interpreting the data collected using the mixed inter-
face media in subsequent filtration runs.
The dual media, selected to have a well intermixed interface, was
similar to a media that is commonly specified for water treatment
plant filters.
The coal was "Philterkol" from the Reading Anthracite Coal Company,
Pottsville, Pennsylvania. The sand was from the Northern Gravel Com-
pany, Muscatine, Iowa. Fines from the coal and sand were skimmed
from the media after hydraulic gradation of the media by backwashing.
This was done because it was felt these fines might cause problems
and because skimming is commonly practiced in filter plant construc-
tion.
The detailed skimming procedure was as follows. Approximately 30 in.
of sand or coal was placed in the plexiglass column. The media was
fluidized to approximately 50% expansion and allowed to stabilize.
The backwash valve was closed rapidly, and the media was allowed to
subside. The top inch or so was skimmed off by a siphon. The pro-
cedure was repeated until 10 to 20% of the finer media was removed.
The media was then removed from the column and dried in an oven. A
sieve analysis was run on the skimmed coal and sand. The effective
sizes and uniformity coefficients of the skimmed media were as
follows;
Sharp interface media
sand ES = 0.85 mm
UC - 1.29
coal ES = 0.91 mm
UC = 1.45
Mixed interface media
sand ES = 0.46 mm
UC = 1.49
coal ES = 0.92 mm
UC = 1.60
The size ratio at the interface is the most important factor deter-
mining the amount of intermixing of coal and sand. In this study,
319
-------
the 99% finer size by weight (dgg^) was used to approximate the
coarsest size coal, and the d^o size was used to approximate the
finest sand. The size ratio at the mixed interface was 5.34. The
size ratio at the sharp interface was 2.32. The size ratios based
on the dgo% for the coal and the dio% for the sand were 4.05 for the
mixed interface media and 1.93 for the sharp interface media.
The specific gravity of "Philterkol" ranged from 1.65 to 1.7 on
various shipments whereas the sand has consistently been 2.65.
Source of Suspensions
Several different suspensions of participates commonly encountered in
filtration were filtered in an attempt to ascertain what effect, if
any, intermixing of the media in dual-media filtration has on filter
performance. It was felt that different types of suspensions might
have different transport and attachment mechanisms for the removal
of solids, and the effect of intermixing for various typical filtra-
tion situations might be different.
Five different suspensions were filtered. They were an iron floe,
lime-soda ash softening precipitate, aluminum sulfate coagulated and
settled trickling filter effluent, activated sludge settled effluent,
and trickling filter settled effluent.
Iron floe. Ferrous sulfate was mixed with Iowa State University tap
water to prepare an influent suspension containing precipitated iron
floe for the filtration study. A stock feed solution of 0.2M fer-
rous sulfate in an acid solution of approximately 0.1 N HCL was made
up in sufficient quantity to fill a 20-liter feed bottle.
To achieve a mixing tank effluent of 9 to 9.5 mg/1 of iron. 17.4
ml/min of the stock ferrous sulfate solution was fed into 5.76 gpm
of tap water. The iron solution was dripped from a constant head
capillary feeder into a mixing tank to achieve the desired iron con-
centration. The iron solution was mixed by a paddle mixer in the
reaction tank.
The type of precipitate formed by the addition of ferrous sulfate to
water depends on the pH, alkalinity, and temperature of the water
and upon the time allowed for reaction. It may consist of Fe(OH)o,
FeC03, or Fe(OH>2. In this research, air was not used in mixing,
and the mixing time was short, about 27 min. Because of the hard
alkaline nature of the tap water and the conditions of mixing, one
would expect the precipitate to be mainly FeCOs; however, the exact
nature was not determined.
Trickling filter effluent. Final effluent from the Ames Water Pollu-
tion Control Plant was filtered. The Ames sewage treatment plant
consists of comminutors, pumping, aerated grit chambers, primary
320
-------
settling, and standard rate trickling filters followed by final
clarifiers.
Secondary effluent from the final clarifiers was pumped from the
final collection chamber to the filters. Most of this filtration of
trickling filter effluent was done during the month of July 1973.
The raw wastewater during July averaged 130 mg/1 BOD5, 161 mg/1 sus-
pended solids, and 337 mg/1 COD. The secondary effluent for July
averaged 16 mg/1 BOD5, 20 mg/1 suspended solids, and 68 mg/1 COD.
Aluminum sulfate coagulated trickling filter effluent. A pilot plant
was in operation at the Ames Water Pollution Control Plant that was
using alum to coagulate and flocculate secondary effluent for phos-
phorous removal. The secondary effluent was pumped to an erdlator
(upflow solids contact unit) where approximately 200 mg/1 of alum
were added for precipitation of the phosphorous.
Average performance for the effluent from the erdlator during the
summer of 1973 was as follows [118]:
Influent Effluent
Average suspended solids (mg/1) 25.4 10.32
Average turbidity (units) 10.9 2.35
Average BOD5 (mg/1) 38.6 10.80
Average TOG (mg/1) 14.0 7.59
Average total P04 (mg/1) 19.3 3.91
Average ortho P04 (mg/1) 18.7 2.39
Average total Kjehldahl N (mg/1) 4.9 4.93
The settled effluent described above was used as the filter influent
for this study.
Activated sludge effluent. An activated sludge pilot plant was oper-
ated so as to produce an effluent to filter. The pilot plant was a
Smith and Loveless "Oxigest" (Smith and Loveless Model "C" Oxigest,
Smith and Loveless, Division-Union Tank Car Company, Lenexa, Kansas)
unit located at the Ames Water Pollution Control Plant. The aeration
tank volume is 2000 gal., and the settling tank volume is 672 gal.
The "Oxigest11 unit was operated at a constant rate of 5 gpm. This
gave a detention time of 6.67 hr. The influent to the unit was pri-
mary effluent from the Ames Water Pollution Control Plant. The aver-
age characteristics of influent to the "Oxigest" for the month of
July 1973 were 82 mg/1 of 5 day BOD, 65 mg/1 suspended solids, and
204 mg/1 COD. Typical characteristics of effluent from the "Oxigest"
unit were 5 to 15 mg/1 BOD5 and 9 to 10 mg/1 suspended solids. The
mixed liquor suspended solids was maintained at around 2400 mg/1.
Effluent from this activated sludge pilot plant was pumped from the
effluent trough t6 the filters.
321
-------
Lime-soda ash softening precipitate. The water for the City of Ames
comes from ground water. This water is softened in a modified, split
treatment, lime-soda ash system. The water from wells is aerated,
slaked lime is added and mixed, and without intermediate settling,
the split flow and soda ash are added and mixed to precipitate addi-
tional hardness. The water is then allowed to settle, and some of
the sludge formed from the settling precipitants is returned to aid
in the precipitation reactions in the mixing step. After settling,
approximately 2 mg/1 of polyphosphate is added to stabilize the water
and stop the reaction of chemical precipitation so that further pre-
cipitate will not form on the filter media, piping, or in the clear
wells. Chlorine and fluoride are also added.
The pilot filter plant was placed in the pipe gallery at the Ames
plant, and water was taken out of the influent pipe leading to one
of the city filters. The average turbidity of the influent during
this study was 6.4 FTU, with a range of 4.3 to 12 FTU. The turbidity
is due mostly to presence of calcium carbonate particles.
Sampling and Measurement
Filter performance was monitored by observing head loss development
and filter influent and effluent quality. Head loss was measured,
as previously discussed, by head loss buildup using mercury manom-
eters. Quality was measured by periodically collecting a grab
sample and analyzing it. The period between samples varied with
the rate of head loss buildup. If head loss increased rapidly, the
interval between samples was decreased. Samples were collected
every 20 min for iron filtration and up to every 3 hr for softening
precipitate filtration.
The quality of the influent and effluent was measured in various
ways. For the iron filtration series, the total iron in the sample
was measured. For the softening precipitate, trickling filter efflu-
ent, activated sludge effluent, and the alum flocculated secondary
effluent, turbidity was used as one measure of quality. Since sus-
pended solids are frequently specified as an effluent quality param-
eter for sewage treatment plants, grab samples were taken and com-
posited over several time intervals for suspended solids analysis.
When an adequate amount of sample had been collected, a suspended
solids analysis was done. Researchers have found that, within
limits, suspended solids concentrations found in treated wastewater
can be roughly correlated to turbidity measurements [59,128]. Never-
theless, suspended solids analyses were also run in this study be-
cause it was felt that the turbidity measurement might not measure
the larger particles in the filter influent and effluent. For each
of the treated wastewaters filtered, after the suspended solids were
filtered from the sample, a sample of the filtrate was taken and the
turbidity determined. This result was considered to be background
color or colloidal material.
322
-------
Iron measurement. Iron was measured using 1,10 phenanthroline. This
method is a spectrophotometric method in which the complex formed
with the ferrous iron produces an orange-red color that obeys Beer's
law. The method was developed so that the analysis of iron samples
could be automated by using the Technicon Auto Analyzer II (Technicon
Industrial Systems, A division of Technicon Instruments Corporation,
Tarrytown, New York). Forty samples could be analyzed per hour.
After the sample was taken, a few drops of hydrochloric acid and some
hydroxylamine hydrochloride were added to reduce the ferric iron to
ferrous iron and to acidify the mixture to keep the iron in solution.
By doing this, the samples were preserved for later analysis on the
automatic analyzer.
Turbidity measurement. Various turbidimeters were used to determine
turbidity for this study. As the quality differences are based on
the relative turbidity in the filter influent and filtrate, the ef-
fect of using different machines was negligible. The instruments
used were all manufactured by Hach (Hach Chemical Company, Ames,
Iowa) and included the Model 2100 and Model 2100A laboratory turbidi-
meters, the "Surface Scatter 3," and the Low Range Turbidimeter,
model 7120. The latter two instruments are continuous flow and
reading turbidimeters. All the turbidimeters were calibrated against
prepared standards of fonnazin polymer, prepared by Hach. The unit
of turbidity measurement was the Formazin Turbidity Unit (FTU).
The continuous flow turbidimeters were checked against the laboratory
turbidimeters. Very close agreement was found. The continuous flow
turbidimeters were used to measure the influent to the filters when
filtering trickling filter effluent, the alum floe, and the softening
precipitate. The effluent from the filters was monitored by grab
samples analyzed on the laboratory turbidimeters.
Suspended solids measurement. The procedure adopted for the deter-
mination of total suspended matter was a slight modification of the
procedure given in Standard Methods [117]. Whatman GF/C glass fiber
filter paper was used. The filter disks were not prewashed and
dried, as it was known from prior experience with this type of paper
that the effect of not washing would be negligible.
Results
Quality and Head Loss
In this empirical study, filter performance was measured by head loss
buildup and effluent quality.
The head loss buildup was measured in inches of mercury. Quality
was measured by various parameters, as previously described. The
effluent concentration divided by the influent concentration, C/C ,
was then calculated. Both head loss buildup and C/CO were plotted
against the total volume of filtrate. Total volume of filtrate
323
-------
during a filter run was determined by taking the flow rate, multi-
plying it by the time, and summing to yield the total volume of fil-
tered water. Total volume of filtrate was used to attempt to elimi-
nate the influence of any small difference in flow rate that existed
between the filters during a filter run.
As previously described, five different suspensions were filtered by
two different dual media; thus there were ten series of filter runs.
The series are identified in Table 49. Several filter runs were made
in each series so that a consistent trend in the data was observed.
Some conditioning runs were undertaken at the beginning of each
series in which the suspension was filtered without close control
to acclimate the filter media to the suspension.
Table 49. Identification of experimental series.
Series Dual media used Suspension filtered
I S Sharp interface Iron floe
I M Mixed interface Iron floe
II S Sharp interface Activated sludge effluent
II M Mixed interface Activated sludge effluent
III S Sharp interface Alum coagulated trickling filter eff
III M Mixed interface Alum coagulated trickling filter eff
IV S Sharp interface Trickling filter effluent
IV M Mixed interface Trickling filter effluent
V S Sharp interface Lime-soda ash softening ppt
V M Mixed interface Line-soda ash softening ppt
A typical plot of head loss buildup versus volume of filtrate and
C/C0 versus volume of filtrate for each series is presented in Figs.
100 through 109.
There are three different parameters presented in the graphs of head
loss versus volume of filtrate. They are (1) the head loss buildup
in the combined media (that media which was in column 1, Fig. 99)
labeled "COMBINED TOTAL HL," (2) the head loss buildup in the sepa-
rate coal (that media which was in column 2, Fig. 99) labeled
"SEPARATE COAL HL," and (3) the sum of the head loss buildup in the
separate coal and separate sand (that media which was in column 2
324
-------
O)
Q
0.80-
0.60
0.40
0.20
0.00
4.00
3.00
2.00
1.00
0.00
QUALITY
O COMBINED C0
A SEPARATE C/C0
HEAD LOSS (HL)
O COMBINED TOTAL HL
A SEPARATE COAL HL
Q SEPARATE TOTAL HL
0.00 8.00 16.00 24.00 32.00 40.00 48.00 56.00
VOLUME FILTRATE, gal. (x 101)
'Fig. 100. Head loss and filtrate quality vs volume of filtrate,
Series I S, run 2, sharp Interface, filtration of
iron with C - 8.68 to 9.64 mg/1 Fe, Avg 9.07 mg/1.
o
325
-------
0.16
0.12
o
u
0.08
0.04
0.00
8.00
o>
I
.c'6.00
H
O
o4.00
x
2.00
0.00
QUALITY
o COMBINED C0
A SEPARATE C/C0
HEAD LOSS (HL)
O COMBINED TOTAL HL
a SEPARATE COAL HL
D SEPARATE TOTAL HL
I
0.00 4.00 8.00 12.00 16.00 20.00 24.00 28.00
VOLUME FILTRATE, gal. (x 101)
Fig. 101. Hetd lomm and filtrate quality v« volua* of filtrate,
S*ri«« I M, run 11, mixed interface, filtration of
iron with C0 - 9.2 to 9.7 mg/1 Fe, Avg 9.42 mg/1.
326
-------
0.80
0.60
0.40
0.20
0.00
8.00
o>
x
.c'6.00
s
9
Q 4.00
x
2.00
0.00
QUALITY
o COMBINEDC/CO
A SEPARATE C/C0
HEAD LOSS (HL)
O COMBINED TOTAL HL
a SEPARATE COAL HL
D SEPARATE TOTAL HL
0.00 8.00 16.00 24.00 32.00 40.00 48.00 56.00
VOLUME FILTRATE, gal. (x 101)
Fig. 102. Head Loss and filtrate quality vs volume of filtrate,
Series II S, run 1, sharp interface, filtration of
activated sludge effluent with Co = 4.5 to 8.5 FTU,
Avg 6.16 FTU.
327
-------
QUALITY
O COMBINED C/Co
A SEPARATE C/Co
HEAD LOSS (HL)
o COMBINED TOTAL HL
A SEPARATE COAL HL
D SEPARATE TOTAL HL
0.80
0.60
0.40
0.20
0.00
8.00
£6.00-
2
Q 4.00
2.00
O.OOLjQ,
0.00 8.00 16.00 24.00 32.00 40.00 48.00 56.00
VOLUME FILTRATE, gal. (x 101)
Fig. 103. Head loss and filtrate quality vs volume of filtrate,
Series II M, run 1, mixed interface, filtration of
activated sludge effluent with Co - 2.6 to 8.6 FTU,
Avg 4.15 FTU.
328
-------
QUALITY
o COMBINED C/Co
SEPARATE C/C0
HEAD LOSS (HL)
O COMBINED TOTAL HL
SEPARATE COAL HL
D SEPARATE TOTAL HL
0.00
0.00 8.00 16.00 24.00 32.00 40.00 48.00 56.00
VOLUME FILTRATE, gal. (x 101)
Fig. 104. Head loss and filtrate quality vs volume of filtrate,
Series III S, run 3, sharp interface, filtration of
alum coagulated trickling filter effluent with C0 -
3.8 to 10 FTU, Avg 5.21 FTU.
329
-------
QUALITY
o COMBINEDC/CO
A SEPARATE C/C0
HEAD LOSS (HL)
o COMBINED TOTAL HL
A SEPARATE COAL HL
D SEPARATE TOTAL HL
0.80
0.60
0.40
0.20
0.00
8.00
.E 6-°°
s
2
Q 4.00
i
2.00
0.00
0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00
VOLUME FILTRATE, gal. (x 101)
Fig. 105. Head loss and filtrate quality vs volume of filtrate,
Series III M, run 5, mixed interface, filtration of
alum coagulated trickling filter effluent with Co -
1.7 to 6.2 FTU, Avg 2.95 FTU.
330
-------
o
00
o
•o
QUALITY
O COMBINED C/C
A SEPARATE C/C °
O !
I
]
si
• i
o
8
•
00
oo
q s
8
CM*
8
•
o
HEAD LOSS (HL)
O COMBINED TOTAL HL
A SEPARATE COAL HL
D SEPARATE TOTAL HL
0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00
VOLUME FILTRATE, gal. (xlO1)
Fig. 106. Head loss and filtrate quality vs volume of filtrate,
Series IV S, run 3, sharp interface, filtration of
trickling filter effluent with Co = 4.7 to 8.5 FTU,
Avg 6.29 FTU.
331
-------
0.80
U
0.60
o
0.40
0.20
0.00
8.00
o>
X
c 6.00
|,«,
i
2.00
0.00
QUALITY
O COMBINED C/C
A SEPARATE C/C °
HEAD LOSS (HL)
O COMBINED TOTAL HL
A SEPARATE COAL HL
D SEPARATE TOTAL HL
0.00 4.00 8.00 12.00 16.00 20.00 24.00 28.00
VOLUME FILTRATE, gal. (xlO1)
Fig. 107. Head loss and filtrate quality vs volume of filtrate,
Series IV M, run 3, mixed interface, filtration of
trickling filter effluent with CQ - 12 to 15 FTU,
Avg 6.14 FTU.
332
-------
QUALITY
o COMBINED C/C0
A SEPARATE C/C0
HEAD LOSS (HL)
O COMBINED TOTAL HL
£ SEPARATE COAL HL
o SEPARATE TOTAL HL
0.00
I
I
8.00 12.00 16.00 20.00 24.00 28.00
VOLUME FILTRATE, gal. (x 10*)
Fig. 108. Head !••• and filtrate quality YB veluae of filtrate,
Sorie* Y S, run 4, •harp laterfact, filtration of lina-
•«4a aah aoftaming pvaclpitata with C0 • 5.6 to 6.5
RU, *rg 6.14 FTU.
333
-------
0.00
QUALITY
o COMBINED C/Co
A SEPARATE C/Co
HEAD LOSS (HL)
O COMBINED TOTAL HL
A SEPARATE COAL HL
a SEPARATE TOTAL HL
A A
1
0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00
VOLUME FfLTRATE, gal. (x 102)
Fig. 109. Head loss end filtrate quality vs volume of filtrate,
series V M, run 3, mixed Interface, filtration of lime-
soda ash softening precipitate vlth Co - 6.8 to 8.3 FTU,
Avg 7.5 FTU.
334
-------
and column 3, Fig. 99, respectively) labeled "SEPARATE TOTAL HL."
The graphs of C/C0 versus volume of filtrate compare the effluent
from the combined media (column 1, Fig. 99) labeled "COMBINED C/CO"
to the effluent from the separate media (the effluent from column 3,
Fig. 99) labeled "SEPARATE C/CO."
Suspended Solids
Suspended solids tests were run on the three treated wastewater
streams, both influent to and effluent from the filters. The range
of values and the average value for the influent, combined media
effluent, and the separate media effluent are given in Table 50 for
all suspended solids data from the indicated series.
Table 50. Influent and effluent suspended solids data, average and
range, for wastewater series.
Influent Series
suspension no.
II S
Activated
sludge
effluent II M
Alum m S
coag
TF
effluent IIZ M
IV S
Trickling
filter
effluent IV M
Range
Avg
Range
Avg
Range
Avg
Range
Avg
Range
Avg
Range
Avg
Influent,
mg/1
4.8-8.0
6.0
3.8-14.6
9.53
5.0-10.2
8.05
5.6-11.2
7.61
10.0-30.8
19.37
18.0-34.8
23.30
Combined eff,
mg/1
0.9-2.6
1.38
0.3-1.2
0.60
1.0-3.9
2.05
0.5-2.9
1.49
3.0-9.6
5.50
1.9-6.2
4.48
Separate eff,
mg/1
0.9-2.3
1.26
0.2-1.2
0.61
1.0-3.2
1.91
0.4-2.8
1.36
3.0-9.0
5.57
2.0-6.2
4.13
Background Turbidity
The background turbidity (turbidity of filtrate remaining from the
suspended solids tests) of the three treated wastewater streams was
determined. The average values of the background turbidity are
given in Table 51. From the data presented, it can be seen that the
background turbidity was affected very little by filtration (the
background turbidity of the influents was approximately equal to the
335
-------
Table 51. Average values of background (bg) turbidity for the
treated wastewaters.
Influent
suspension
Influent bg
turbidity,
FTU
Combined eff
bg turbidity,
FTU
Separate eff
bg turbidity,
FTU
Activated sludge
Alum coagulated
TF effluent
Trickling filter
effluent
0.56
0.53
1.60
0.59
0.59
1.70
0.59
0.58
1.80
background of the effluents). It may also be noted that the back-
ground turbidity was a significant percentage of each total effluent
turbidity for these treated wastewater streams. Furthermore, this
colloidal background matter is not measured by the suspended solids
analysis.
Discussion
The results presented in Figs. 100 through 109 are summarized in
Table 52. Where there was no difference or only a slight difference,
"no difference" is indicated. Where the separate media had a better
performance (i.e., lower head loss buildup or better filtered water
quality), "separate better" is indicated. In the series where the
combined media had a better performance, "combined better" is indi-
cated. In Series II S and Series IV S, no comparison could be made,
as will be explained later; therefore, "no comparison" is indicated.
Table 52. Summary of results comparing filter performance for sharp
and mixed interface media.
Series Filtering Interface
Head loss
Quality
I S
I M
II S
II M
III S
III M
IV S
IV M
V S
V M
Fe
Fe
AS
AS
Alum
Alum
TF
TF
Soft
Soft
Sharp
Mixed
Sharp
Mixed
Sharp
Mixed
Sharp
Mixed
Sharp
Mixed
Separate better
Separate better
No comparison
No difference
Combined better
Separate better
No comparison
Separate better
No difference
No difference
No difference
No difference
No comparison
No difference
No difference
No difference
No comparison
No difference
Separate better
Separate better
336
-------
The sharp interface media was used as a control to establish what
differences in performance could be attributed to the equipment and
operating procedure. Thus, if the sharp interface media presented a
significant trend in one direction and the mixed interface media
demonstrated the same trend in greater magnitude or the opposite
trend this difference could be attributed to the mixing of the media
at the interface.
Upon examination of the data and plots for the filtration of the iron
floe, it can be seen that there is no difference in the quality that
can be attributed to media mixing at the interface. The head loss
data and plots reveal that the mixing at the interface resulted in
slightly larger head loss.
No valid conclusions can be drawn from the filtration of the acti-
vated sludge effluent or the trickling filter effluent. This is be-
cause all the removal was in the coal during sharp interface filtra-
tion. Thus, the control offered by the sharp interface data was
lost, and the effect of the equipment and operating procedure was
not known.
The data and plots for the filtration of the alum coagulated trick-
ling filter effluent by the sharp interface media showed the equip-
ment and operating differences caused the combined media to have
less head loss buildup than the separate media. However, the data
and plots of the filtration of this suspension by the mixed interface
dual media show that the separate media have less head loss buildup.
Thus, the effect of the intermixing at the interface was a detrimen-
tal one because there was a greater head loss in the combined mixed
interface media. This greater head loss is attributable to the
intermixing. The quality data and plots could show no difference in
quality attributed to intermixing.
The filtration data for the lime-soda ash softening precipitate re-
vealed no difference in head loss buildup between the combined and
separate media for both the mixed and sharp interface dual media.
The quality data and plots for the sharp interface media showed that
the separate media had a better effluent. This observed difference
can be attributed to the equipment and operating procedure. The
quality data and plots for the mixed interface media showed that the
separate media also produced a better quality effluent. Since both
the sharp and mixed interface media had the same trend, no difference
in quality can be attributed to the mixing of the dual media at the
interface. A visual observation of the upper surface of the separate
sand revealed a layer of white precipitate. This blanket of precipi-
tate on the upper surface of the separate sand in both the sharp and
mixed interface media may have acted like a filter itself, thereby
improving the effluent quality from the separate media.
337
-------
From the results, it can be seen that the mixing of the dual media
at the interface causes little or no effect on either head loss
buildup or effluent quality of a dual-media filter. Some slight in-
crease in head loss buildup may be attributed to media mixing at the
interface for some suspended solids, but this is not true of all
solids, and the magnitude is slight.
Filtrate quality as measured by turbidity or iron concentration
showed no benefit or detriment due to intermixing. The suspended
solids data in Table 50 also show no apparent effect of the media
intermixing at the interface.
A point must be made concerning the effluent quality. As can be
seen from the data, all of the suspended material was not completely
removed (C/CO greater than zero). Thus, some of the suspended matter
must have reached the interface if some passed through the entire
dual-media filter. In the case of Series II S and IV S, little of
the suspended matter that reached the interface or the sand was re-
tained due to the size of the sand.
It would be of interest to compare these observations with theoreti-
cal models of filter performance. However, it is difficult to draw
firm conclusions because the models have been derived and verified only
for single-media filters. Nevertheless, an attempt will be made to
apply filtration theory to support the observations. Boyd and Ghosh
[18] recently summarized the various models for filtrate quality.
They restated the now commonly accepted idea that removal of sus-
pended solids by a granular filter involves a transport step and an
attachment step.
All of the parameters in the various models for both transport and
attachment are properties of the particles in suspension or the
carrying fluid except for the diameter of the filter grains. Two of
the transport models, namely, interception and Brownian diffusion,
indicate that removal effectiveness is an inverse function of grain
diameter. Therefore, one would expect better filter efficiency with
finer grain size, if all other parameters were unchanged.
Most models for prediction of head loss in granular media filtration
formulate head loss to be a function of solids capture [64,84], Thus
if the solids capture is the same for the combined or separate media,
the total head loss would be expected to be about the same.
The models do not deal with the effect of grain diameter in the in-
termixed zone on the efficiency of a dual-media filter. If the same
grains are present they should yield the same removal efficiency
whether they are intermixed in a dual-media bed, or operated sepa-
rately in series. This idea would support the experimental results
reported herein which showed no difference between the filtrate for
the combined and separate media. However, it is probable that the
grain diameter in the models is really important as an indirect
338
-------
measure of the size and shape of the void spaces in the filter bed.
When the small sand grains move up into the coarser coal layers,
they tend to fill the larger voids of the coal, causing a pore size
and permeability between that associated with the sand and the coal
as shown previously in Chapter X. Similarly, the coal grains that
subside into the upper sand layer create a pore size and permeability
between the sand and coal. Intermixing resulted in a total bed depth
slightly less than the sum of the depths of the two media, measured
individually in this work. A similar observation had also been re-
ported previously in Chapter X. Since the overall depth of the dual-
media bed is less, it would have a lower clean bed porosity and per-
meability in the intermixed zone. Because of this, one would expect
better filtrate quality but higher head loss from the combined media
than the separate media. The higher head loss for the combined media
was observed for three of the solids reported herein, but the better
quality was not observed.
Substantial differences in performance were observed between the
sharp and mixed interface media. These differences would be expected
from the foregoing theories. The mixed interface media produced a
better filtrate quality than the sharp interface media, but also
generated higher head loss. Substantially the same size coal was
used in the two media, but a much finer sand was used in the mixed
interface media than in the sharp interface media. From the theory
presented, the finer sand should result in more efficient solids re-
moval, which in turn would generate more head loss. These media
sizes were purposely selected to achieve the desired degree of inter-
facial intermixing or separation. The coarse sand in the sharp in-
terface media was required to produce a sharp interface with the ex-
pectation that it would cause poorer solids capture and lower head
loss.
The degree of interfacial intermixing is dependent on the size, den-
sity, and shape of adjacent media, as well as the backwash flow ve-
locity and valve shut-off procedure as shown previously in this re-
port in Chapter X. If a normal coal size and gradation are selected,
one cannot achieve a sharp interface except by sacrificing expected
filtrate quality.
The controversy as to the desirability or detriment of interfacial
intermixing was started by Conley [32] and Camp [23], Camp [23,24]
stated that he felt a sharp interface acts as a safeguard against
breakthrough. He discouraged intermixing. Ives [63] and Gregory
[54] have also implied that one should discourage interfacial inter-
mixing. However, other researchers and actual case studies have
shown that filters perform quite adequately when intermixing occurs
at the interface either by accident or by plan.
The results of this study showed substantially better filtrate qual-
ity for the mixed interface media than for the sharp interface media.
339
-------
Thus the results are in opposition to the design recommendation of
Camp. The better filtrate observed for the mixed interface media is
due to the use of finer sand and not to the intermixing itself.
However, intermixing is the inevitable result of selecting the dual
media to provide substantially finer top sand size than top coal
size.
None of the researchers have presented proof to back up claims that
the intermixing itself is good or bad. Brosman and Malina [19] are
the only ones who have attempted to measure the effect of inter-
mixing on both head loss and effluent quality. The writers feel that
their study had some experimental weaknesses [109] which led them
to the false conclusions that interfacial intermixing, in and of it-
self, provided benefits to both filtrate quality and head loss. That
conclusion is not supported by the results reported herein.
Conclusions
The following points are concluded from this study:
1. Interfacial intermixing does not in itself affect filter per-
formance as measured by both head loss development and effluent
quality of dual-media filters.
2. The mixed interface filter media produced better filtrate qual-
ity than did the sharp interface filter media for all suspen-
sions filtered in this study; this was due to substantially
finer sand in the mixed interface media.
3. The mixed interface filter media produced higher head losses
than the sharp interface filter media for all suspensions due
to greater suspended solids removal.
4. Intermixing at the interface of dual-media filters is an un-
avoidable phenomenon which results when United States anthracite
(sp. gr. about 1.7) and sand are used and the sizes are selected
to achieve coarse to fine filter media in the direction of flow.
340
-------
XII. ABRASIVE LOSS OF COAL DURING AIR SCOUR
The effect of air scour on anthracite coal filter media became of
concern because of the possible widespread use of air scoured fil-
ters using anthracite coal as one of the filter media. Widespread
use seems probable for the following reasons: (1) the apparent bene-
fits of air scour as an adjunct to water backwash, (2) the apparent
necessity of auxiliary agitation in wastewater filtration, (3) the
necessity of using dual media in wastewater filtration to achieve
adequate filter run length.
The visual appearance of coal filter media when subjected to air
scour was observed in plastic filter housings. The coal appears to
be in rather violent motion, especially the coal near the top surface
of the bed. The granules of the coal are in contact with each other
and rubbing against each other, unlike those in a fluidized bed where
collisions and rubbing are absent. Thus the abrasive loss of coal
was of concern due to the softer nature of coal as compared to other
common filter media such as silica sand and garnet sand.
An experiment was conducted to gain some view of the potential seri-
ousness of the abrasive loss problem. The experiment consisted of
subjecting a filter bed of anthracite coal to essentially continuous
air scour for two weeks. Two weeks of such exposure would be com-
parable to about 20 years of normal filter operation assuming 24-hr
filter cycles and 3 min of air scour per cycle. The filter was
water backwashed twice a day to observe the visual loss of coal dust.
Furthermore, changes in the depth of bed and pressure loss profile
through the bed in both upflow and downflow were observed to obtain
evidence of changes in the filter media.
Experimental Procedure
The details of the experiment were as follows:
The starting coal was a standard crushed anthracite coal "Philterkol"
obtained from the Reading Anthracite Coal Company, Pottsville, Penn-
sylvania. The suppliers specified effective size was 0.6 to 0.79 mm,
the uniformity coefficient was less than 1.8, and the reported hard-
ness was 3 on the Moh Scratch Test comparison scale.
The coal was split to obtain a representative sample, and 16 in. of
coal was placed in a 6-in. ID pilot filter. The coal was backwashed
with water and allowed to stratify. A total of about 2 in. of the
fine coal was skimmed from the surface of the stratified bed in two
increments during backwashing. The coal was then removed from the
bed, air dried and oven dried, and two separate representative sam-
ples were collected for sieve analysis. A riffle type splitter was
used to collect the representative sample.
341
-------
The sieve analyses were performed on a nest of United States Standard
sieves (12, 14, 16, 18, 20, 25, and 30 mesh) using a Tyler portable
shaker for 5 min. The sample size was 356.6 and 407.9 g for the two
samples.
The remainder of the coal was weighed and was placed back in the fil-
ter column, fluidized to permit stratification and the pressure loss
profile was observed from the piezometer tubes located at 3-in. depth
increments as shown in Fig. 4. A water temperature of 22 °C was used.
After the upflow and downflow pressure profile observations, the
water level was lowered to about 6 in. above the coal surface, and
the air scour was applied at a rate of 4 cfm/sq ft. The air was kept
on continuously except for two short periods each day when the air
was shut off and the filter was subjected to backwash with 50% bed
expansion, and to upflow and downflow pressure profile observations.
The amount of coal dust in the initial backwash water was observed
qualitatively as one measure of abrasive loss. The downflow pressure
drop between piezometer taps at 7 gpm/sq ft and 22 °C was observed to
see if there was any progressive increase in pressure drop due to the
formation of finer media size as a result of abrasion. The pressure
drop during upflow at a rate which would provide 50% bed expansion at
22 °C was measured to see if there were any progressive changes due
to abrasive loss. The upflow rate was selected to provide an ex-
panded bed depth of 21 in. The development of a finer media would
result in greater expansion at a given upflow rate and less pressure
drop between the piezometer taps. The depth of the bed after each
backwash was observed after carefully and slowly closing the backwash
valve to ensure a consistent degree of packing of the bed. Any major
loss of media or change in sphericity due to abrasion would manifest
itself by a decrease in bed depth.
At the end of two weeks of the above procedure, the coal was removed
from the filter, dried, weighed, and again two representative samples
were taken and subjected to sieve analysis.
Results
Some abrasive loss was evident from the color of the coal dust in the
water at the end of each period of air scour. The initial backwash
water was black with coal dust but cleared up in the first 2 min of
backwash.
There was no substantial loss in bed depth or bed weight as shown in
Table 53. Furthermore, there was no substantial change in head loss
either upflow or downflow, indicating no substantial change in size
of the media.
The sieve analysis before and after the two-week period of air scour
is presented in Fig. 110. Each curve represents the average of two
342
-------
Table 53. Changes in coal bed over a two-week period of air-scour
exposure equivalent to about 20 years of normal service.
Bed depth,
Bed weight
in.
dry, g
Initial
value
14-1/8
5019
After
two weeks
13-7/8
4775
Total downflow head loss, ft
Upflow rate to achieve 21 in.
0.30
0.29
bed depth, gpm
Total upflow head loss, ft H20
Effective size of media, mm
Uniformity coefficient
5.1
0.35
0.80
1.59
5.3
0.34
0.78
1.60
sieve analyses. It is evident that the coal decreased in size
slightly in this air scour exposure, about 0.02 mm in effective size,
which is a negligible change.
Conclusions
A typical crushed anthracite coal filter media was subjected to air
scour for two weeks to simulate the abrasive effects of 20 years of
typical filter service. From this study, it is concluded that some
abrasive loss of coal does occur, but it is a negligible amount. The
total loss of media and changes in media head loss and size were less
than 3%. This is considered negligible because coal filter media
suppliers usually request and are granted about a 107o tolerance on
the effective size of the media which they supply.
343
-------
N
CO
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
7-12-72
BEFORE
AIR SCOUR
CD
AFTER
AIR SCOUR
12 5 10 20 30 40 50 60 70 80 90 95
PERCENT PASSING BY WEIGHT, % finer
Fig. 110. Sieve analysis of coal before and after 14 days of
continuous air scour.
344
-------
XIV. REFERENCES
1. Adler, I. L., and Happel, J. The fluidization of uniform smooth
spheres in liquid media. Chem. Eng. Prog., Symp. Ser., 58(38):
98-105, 1962.
2. American Society for Testing Materials. 1968 Book of ASTM
Standards. Part 10. Concrete and mineral aggregates. ASTM,
Philadelphia, 1972.
3. American Water Works Association, Inc. Water Quality and Treatment.
New York, McGraw-Hill Book Co., Inc., 1971.
4. Amirtharajah, A. Expansion of graded sand filters during backwash-
ing. Unpublished M.S. thesis. Ames, Iowa, Library, Iowa State
University of Science and Technology, 1970.
5. Amirtharajah, A. Optimum expansion of sand filters during back-
wash. Unpublished Ph.D. thesis. Ames, Iowa, Library, Iowa
State University of Science and Technology, 1971.
6. Amirtharajah, A. and Cleasby, J. L. Predicting expansion of
filters during backwashing. J. Am. Water Works Assoc., 64:52-59,
1972.
7. Anderson, T. B., and Jackson, R. The nature of aggregative and
particulate fluidization. Chem. Eng. Sci., 19:509-511, 1964.
8. Babbitt, H. E., Doland, J. J., and Cleasby, J. L. Water Supply
Engineering. 6th ed. New York, McGraw-Hill Book Co., Inc., J.yb2.
9. Barret, A. D. Immedium filter in tertiary treatment. Process
Biochem., 6_(1): 33-35, 1971.
10. Baylis, J. R. Nature and effects of filter backwashing. J. Am.
Waterworks Assoc., 51:126-156, 1959.
11. BayUs, J. R. Review of filter design and methods of washing. J. Am.
Water Works Assoc., 51:1433-1454, 1959.
12. Becker, H. A. An investigation of the laws governing the spouting
of coarse particles. Chem. Eng. Sci., 13:245-262, 1961.
13. Boby, W. M. T., and Alpe, G. Practical experiences using upward
flow filtration. Proc. Soc. Water Treat. Exam., 16:215-229, 1967.
14. Boss, R. R. Hydraulics and intermixing of sand and coal filters.
Unpublished M.S. thesis. Ames, Iowa, Library, Iowa State University
of Science and Technology, 1973.
15. Botterill, J. S. M. Progress in fluidization. Br. Chem. Eng.,
10:26-30, 1965.
345
-------
16. Botterill, J. S. M. Progress in fluidization. Br. Chem. Eng.,
13:1121-1126, 1968.
17. Bourgeois, P., and Grenier, P. The ratio of terminal velocity to
minimum fluidizing velocity for spherical particles. Can. J. Chem.
Eng., 46:325-328, 1968.
18. Boyd, R. H., and Ghosh, M. M. An investigation of the influences
of some physiochemical variables on porous-media filtration. J.
Am. Water Works Assoc., .66:94-98, 1974.
19. Brosman, D. R., and Malina, J. E. Intermixing of dual media filters
and effects on performance. Technical Report EHE-72-4 CBWR-86.
Center for Research in Water Resources. The University of Texas
at Austin, 1972.
20. Buevich, Yu. A., and Markov, V. G. Pseudo-turbulent diffusion of
particles in homogeneous suspensions. Int. Chem. Eng., 10:570-
575, 1970.
21. Cairns, E. J., and Prausnitz, J. M. Longitudinal mixing in
f luidization. AIChE J., ji:400-405, I960;
22. Cairns, E. J., and Prausnitz, J. M. Macroscopic mixing in fluidi-
zation. AIChE J., i:554-560,1960.
23. Camp, T. R. (Discussion of) Experience with anthracite-sand
filters. J. Am. Water Works Assoc., 5.3:1478-1483, 1961.
24. Camp, T. R. Theory of water filtration. ASCE J. Sanitary Eng. Div.,
90:1-30, 1964.
25. Camp, T. R. (Discussion of) Theory of water filtration. ASCE J.
Sanitary Eng. Div., £1:55-69, October 1965.
26. Camp, T. R., Graber, S. D., and Conklin, G. F. Backwashing of
granular water filters. ASCE J. Sanitary Eng. Div., No. SA6,
pp. 903-925, December 1971.
27. Camp, T. R., and Stein, P. C. Velocity gradients and internal
work in fluid motion. J. Boston Soc. Civ. Eng., 30:219-237, 1943.
28. Carvalho, W. J. Fluidization of crushed anthracite coal. Unpub-
lished M.S. thesis. Ames, Iowa, Library, Iowa State University of
Science and Technology, 1970.
29. Clark, J. W., Viessman, W., and Hammer, M. J. Water Supply and
Pollution Control. Scranton, Pennsylvania, International Textbook
Company,1971.
346
-------
30. Cleasby, J. L. Iron removal, a case study. J. Am. Water Works
Assoc., £7(3):147-149. March 1975.
31. Cleasby, J. L., and Baumann, E. R. Selection of sand filtration
rates. J. Am. Waterworks Assoc., 54(5):579-602, 1962.
32. Conley, W. R. Experience with anthracite-sand filters. J. Am.
Waterworks Assoc., 53:1473-1478, 1961.
33* Coulson, J. M., and Richardson, J. F. Chemical Engineering. Vol.
2, 2nd ed. New York, Pergamon Press Inc., 1968.
34. Gulp, R. L., and Gulp, G. L. Advanced Wastewater Treatment. New
York, Van Nostrand Reinhold Company, 1971.
35. Davidson, J. F., and Harrison, D. Fluidized Particles. New York,
The Syndics of the Cambridge University Press, 1963.
36. Davies, L., and Richardson, J. F. Gas interchange between bubbles
and the continuous phase in a fluidized bed. Trans. Inst. Chem.
Eng., 44:T293-T305, 1966.
37. Degremont, S. A. Water Treatment Handbook. 3rd English ed.
Kensington, London W14, England, Hugh K. Elliot, Ltd., 1965.
38. Dvorin, R. Advances make wastewater filtration economical and
effective solution. Water Sewage Works 116.(9) -.1W/12-1W/16, 1969.
39. Efremor, G. I. and Vakhrusher, I. A. A study of the hydrodynamics
of three-phase fluidized beds. Int. Chem. Eng., 1():37-41, 1970.
40. Ergun, S. Fluid flow through packed columns. Chem. Eng. Prog.,
48:89-94, 1952.
41. Evans, S. C., and Roberts, F. W. Twelve months' operation of
sand filtration and microstraining at Luton. Inst. Sewage
Purif., J. Proc., 4:333-341, 1952.
42. Evans, S. C., and Roberts, F. W. Recent developments in sewage
treatment at Luton. Inst. Sewage Purif., J. Proc., 3_:225-236, 1955.
43. Evers, R. H. Tool up with mixed-media filters. Water Wastes Eng.,
8(5):C14-C16, 1971.
44. Ewing, G. W. Instrumental Methods of Chemical Analysis. 3rd ed.
New York, McGraw-Hill Book Co., Inc., 1969.
45. Fair, G. M., and Geyer, J. C. Water Supply an«j W,as,t;q Water Dis-
posal. 8th ed. New York, John Wiley and Sons, Inc., 1958.
46. Fair, G. M., Geyer, J. C., and Okun, D. A. Water and Wastewater
Engineering. Vol. 2. New York, John Wiley and Sons, Inc., 1968.
347
-------
47. Fair, G. M., and Hatch, L. P. Fundamental factors governing the
streamlike flow of water through sand. J. Am. Water Works Assoc.,
25:1551-1565, 1933.
48. Frantz, J. F. Minimum fluidization velocities and pressure drop
in fluidized beds. 1 Chem. Eng., Symp. Series, 62(62):2l-31, 1966.
49. Friedlander, S. K., and Topper, L. Turbulence. Classic Papers
on Statistical Theory. New York, Interscience Publishers Inc.,
1961.
50. Furukawa, J., and Ohmae, T. Liquidlike properties of fluidized
systems. Ind. Eng. Chem., 50:821-828, 1958.
51. Galloway, T. R., and Sage, B. H. A model of the mechanism of
transport in packed, distended and fluidized beds, Chem. Eng. Sci.,
25:495-516, 1970.
52. Ghosh, M. M., O'Connor, J. T., and Engelbrecht, R. S. Precipitation
of iron in aerated ground waters. ASCE J. Sanitary Eng. Div.,
90(SA1):199-213, 1966.
53. Godard, K., and Richardson, J. F. Correlation of data for minimum
fluidizing velocity and bed expansion in particulately fluidized
systems. Chem. Eng. Sci., 24; 363-367, 1969.
54. Gregory, R. Media specification and backwashing - W.R.A. Paper
presented at Anthracite-Sand Filtration, A Working Conference,
University of Reading, England, January 3-5, 1972.
55. Handley, D., Doraiswamy, A., Butcher, K. L., and Franklin, N. L.
A study of the fluid and particle mechanics in liquid-fluidized
beds. Trans. Inst. Chem. Eng., 44:T260-T273, 1966.
56. Hanratty, T. J., Latinen, G. and Wilhelm, R. H. Turbulent
diffusion in particulately fluidized beds of particles. AIChE J
1:372-380, 1956.
57. Harrison, D., Davidson, J. F., and Kock, J. W. On the nature of
aggregative and particulate fluidization. Trans. Inst. Chem.
Eng., .39:202-211, 1961.
58. Holding, J. C. Polishing sewage effluents, Effluent Water Treat.
J., .12:665-667, 669-672, 1972.
59. Huang, J. Y. C. Granular filters for tertiary wastewater treat-
ment. Unpublished Ph.D. dissertation. Ames, Iowa, Library, Iowa
State University of Science and Technology, 1972.
60. Hudson, H. E. Filter backwashing experiments at the Chicago
experimental plant. J. Am. Waterworks Assoc., 27_:l5b7-l56b, 1935.
348
-------
61. Huisman, L. Rapid filtration, Part 1. Mimeographed book of
lectures in English, Delft University of Technology, Depart-
ment of Civil Engineering, 1974.
62. Hulbert, R., and Herring, F. W. Studies on the washing of rapid
filters. J. Am. Water Works Assoc., 21:1445-1513, 1929.
63. Ives, K. J. Progress in filtration. J. Am. Water Works Assoc.,
16(9):1225-1232, 1964.
64. Ives, K. J. Theory of filtration. Special Subject No. 7. Proc.
Int. Water Supply Cong. Exhib., Vienna, 1969. London, International
Water Supply Association, 1969.
65. Ives, K. J. Personal communication, September 24, 1974.
66. Jackson, R. The mechanics of fluidized beds. Part I: The stabil-
ity of the state of uniform fluidization. Trans. Inst. Chem.
Eng., 41:13-21, 1963.
67. Johnson, R. L., and Cleasby, J. L. Effect of backwash on filter
effluent quality. ASCE J. Sanitary Eng. Div., 92:215-228, 1966.
68. Joslin, J. R., and Greene, 6. Sand filter experiments at Derby.
Water Pollut. Control, 69:611-622, 1970.
69. Jottrand, R. An experimental study of the mechanism of fluidiza-
tion. J. Appl. Chem., ^(Supplementary Issue,1):5l7-526, 1952.
70. Jung, H., and Savage, E. S. Deep-bed filtration. J. Am. Water
Works Assoc., 66:73-78, 1974.
71. Kada, H., and Hanratty, T. J. Effects of solids on turbulence in
a fluid. AIChE J., <>:624-630. 1960.
72. Reefer, C. E. Tertiary sewage treatment. Public Works, 93(11):
109-112, 1962.
73. Kramers, I. H., Westermann, M. D., de Groot, J. H., and Dupont,
F. A. A. The longitudinal dispersion of liquid in a fluidized
bed. Symposium on the Interaction Between Fluids and Particles.
London, Institution of Chemical Engineers, pp. 114-119, 1962.
74. Le'Clair, Brian P. Two component fluidization. Unpublished M.A.
Sc. thesis. British Columbia, Canada, Library, University of
British Columbia, 1964.
75. Lemlich, R., and Caldas, I. Heat transfer to a liquid fluidized
bed. AIChE J., 4:376-380, 1958.
76. Leva, Max. Fluidization. New York, McGraw-Hill Book Co., Inc.,
1959.
349
-------
77. Leva, M., Grummer, M., Weintraub, M., and Pollchik, M. Intro-
duction to fluidization. Chem. Eng. Prog., 44:511-520, 1948.
78. Leva, M., Grummer, M., Weintraub, M., and Pollchik, M. Fluidi-
zation of solid nonvesicular particles. Chem. Eng. Prog., 44;
619-626, 1948.
79. Lewis, E. W., and Bowerman, Ernest W. Fluidization of solid
particles in liquids^ Chem. Eng. Prog., 48:603-609, 1952.
80. Malik, A. M. Air and water backwashing of granular filters.
Unpublished M.S. thesis. Ames, Iowa, Library, Iowa State Univer-
sity of Science and Technology, 1972.
81. McCune, L. K., and Wilhelm, R. H. Mass and momentum transfer in
solid-liquid system. Ind. Eng. Chem., 41:1124-1134, 1949.
82. Metcalf and Eddy, Inc. Wastewater Engineering. New York,
McGraw-Hill Book Co., Inc., 1972.
83. Michaelson, A. P. Under the solids limit at Ashton-under-Lyne.
Water Pollut. Control, 70(5)-.533-535, 1971.
84. Mohanka, S. S. Theory of multilayer filtration. ASCE J. Sanitary
Eng. Div., 95:1079-1095, 1969.
85. Murray, J. D. Mathematical aspects of bubble motion in fluidized
beds. Chem. Eng. Prog. Symp. Ser., £2(62):71-82, 1966.
86. Naylor, A. E., Evans, S. C., and Dunscombe, K. M. Recent develop-
ments on the rapid sand filters at Luton, Water Pollut. Control,
66(4):309-320, 1967.
87. Nicolle, N. P. Humus tank performance, microstraining and sand
filtration. Inst. Sewage Purifi., J. Proc., .1:19, 1955.
88. Nicolle, N. P. Operational aspects of control of rapid gravity
sand filters. Inst. Sewage Purif., J. Proc., 3:273-275, 1957.
89. Ostergaard, K. The effect of particle size and bed height on the
expansion of mixed phase (gas-liquid) fluidized beds. Chem. Eng.
Sci., .21:413-417, 1966.
90. Ostergaard, K. On the growth of air bubbles formed at a single
orifice in a waterfluidized bed. Chem. Eng. Sci., 21:470-472.
1966.
91. Othmer, D. F. Fluidization. New York, Reinhold Publishing Corpor-
ation, 1956.
92. Pettet, A. E. J., Collett, B. A., and Summers, T. H. Mechanical
filtration of sewage effluents. Part 1, Removal of humus. Inst.
Sewage Purif., J. Proc., 4:399-410. 1949.
350
-------
93. Pettet, A. E. J., Collett, W. F., and Waddington, J. I. Mechan-
ical filtration of sewage effluents. III. Removal of humus:
Further experiments at Luton. Inst. Sewage Purif., J. Proc.,
4:195-202, 1951.
94. Pettet, A. E., Collett, W. F., and Waddington, J. I. Rapid filtra-
tion of sewage effluents through sand and anthracite. Sewage Ind.
Wastes, 24(7):835-843, 1952.
95. Pinchbeck, P. H., and Popper, F. Critical and terminal velocities
in fluidization. Chem. Eng. Sci., 6.:57-64, 1956.
96. Pruden, B. B. Particle size segregation in particulately fluid-
ized beds. Unpublished M. A. Sc. thesis. British Columbia,
Canada, Library, University of British Columbia, 1964.
97. Pruden, B. B., and Epstein, N. Stratification by size in particu-
late fluidization and in hindered settling. Chem. Eng. Sci.,
^9:696-700, 1964.
98. Reuter, H. On the nature of bubbles in gas and liquid fluidized
beds. Chem. Eng. Prog., Symp. Ser., 6i2(62):92-99, 1966.
99. Rice, G. A. Direct filtration of secondary effluent, backwashing
and performance. Unpublished M.S. thesis. Ames, Iowa, Library,
Iowa State University of Science and Technology, 1973.
100. Richardson, J. F., and Zaki, W. N. Sedimentation and fluidization:
Part I. Trans. Inst. Chem. Eng., .32:35-53, 1954.
101. Rigby, G. R., Blockland, G. P. V., Park, W. H., and Capes, C. E.
Properties of bubbles in three phase fluidized beds as measured by
an electro-resistivity probe. Chem. Eng. Sci., 25:1729-1741, 1970.
102. Ripley, P. G., and Lamb, G. L. Filtration of effluent from a
biological-chemical system. Water Sewage Works, 120(2); 66-69,
1973.
103. Robeck, G. G., and Kreissl, J. F. Multi-media filtration: princi-
ples and pilot experiments. Transactions of 17th Annual Conference
of Sanitary Engineering. Univ. Kans. Public Bull. Eng. Arch.,
5_7:lO-22, 1967.
104. Romero, J. B., and Johanson, L. N. Factors affecting fluidized
bed quality. Chem. Eng. Prog., Symp. Ser., 58(38):28-37, 1962.
105. Rowe, P. N. Drag forces in a hydraulic model of a fluidized bed.
Part II. Trans. Inst. Chem. Eng., 39.: 175-180, 1961.
106. Rowe, P. N., and Henwood, G. A. Drag forces in a hydraulic
model of a fluidized bed. Part I. Trans. Inst. Chem. Eng., 39;
43-54, 1961.
351
-------
107. Ruckenstein, E. Homogeneous f lutdizatlon. Ind. Eng. Chem. Fund.,
3:260-268, 1964.
108. Savage, E. S., Personal communication, February 19, 1975, describ-
ing media and backwash provisions used by Pintsch Bamag (Germany)
and Dravo (U.S.) .
109. Sejkora, G. D. Effect of Media Intermixing on Dual Media Filtra-
tion. Unpublished M. S. thesis. Ames, Iowa, Library, Iowa State
University of Science and Technology, 1974.
110. Shannon, P. T. Fluid dynamics of gas fluidized batch systems.
Microfilm copy. Unpublished Ph.D. thesis. Chicago, Illinois,
Library, Illinois Institute of Technology, 1961.
111. Shi reman, C. Filtration boosts tertiary treatment. Water Wastes
., £(4):34-37, 1972.
112. Siusnonds, M. A. Public water supplies of Queensland, Australia, J.
Am. Waterworks Assoc., 55_: 1044 -1080, 1963.
113. Simpson, H. C., and Rodger, B. W. The fluidization of light solids
by gases under pressure and heavy solids by water. Chem. Eng.,
Sci., 16_:153-180, 1961.
114. Singer, P. C., and Stumm, W. The solubility of ferrous iron in
carbonate-bearing waters. J. Am. Water Works Assoc., 62; 198-202.
1970.
115. Slis, P. L., and Willemse, T. W., and Kramers, H. The response of
the level of a liquid fluidized bed to a sudden change in the
fluidizing velocity, Appl. Sci. Res., 8:209-218, 1959.
116. Sosewitz, B., and Bacon, V. W. Chicago's first tertiary treatment
plant. Water Wastes Eng., 5_:52-55, 1968.
117. Standard Methods for the Examination of Water and Wastewater-
13th ed. New York, American Public Health Association, 1971.
118. Stangl, E. W. Air and water backwash of wastewater filters. Un-
published M.S. thesis. Ames, Iowa, Library, Iowa State University
of Science and Technology, 1973.
119. Steward, P. S., and Davidson, J. F. Three-phase fluidization:
Water, particles and air. Chem. Eng. Sci., 12:319-322, 1964.
120. S too key, L. L. Ferrozine - A new spectrophotometric reagent for
iron. Anal. Chem., 42:779-781, 1970.
121. Streander, B. Mechanical filtration of sewage. Water Works
Sewerage, 82(7) : 25 2- 257, 1935.
352
-------
122. Streander, B. Sewage filtration with silica sand filters I. Water
Works Sewerage, 87(8)-.351-357, 1940.
123. Streander, B. Sewage filtration with silica sand filters II.
Waterworks Sewarage, 87(9)-.422-429, 1940.
124. Streander, B. Sewage filtration with silica sand filters III.
Waterworks Sewerage, 87(10)-.493-495, 1940.
125. Taylor, G. I. Statistical theory of turbulence, Parts I-IV.
Proc. R. Soc.,London, Ser. A., l5JL:421-478, 1935.
126. Taylor, G. I. The spectrum of turbulence. Proc. R. Soc. London,
Ser. A, 164:476-490, 1938.
127. Tchobanoglous, G. Filtration techniques in tertiary treatment.
J. Water Poll. Control Fed., 42:604-623, 1970.
128. Tchobanoglous, G., and Eliassen, R. Filtration of treated sewage
effluent. ASCE J. Sanitary Eng. Div., 96:243-265, 1970.
129. Tebbutt, T. H. Y. An investigation into tertiary treatment by
rapid filtration. Water Res., J. Intern. Assoc. Water Pollution
Research, 5_:81-92, 1971.
130. Tesarik, I. (Discussion of) Theory of water filtration. ASCE J.
Sanitary Eng. Div., 90:67-69, 1965.
131. Truesdale, G. A., and Birkbeck, A. E. Tertiary treatment of
activated-sludge effluent. Water Pollut. Control, 67(5):483-494,
1968.
132. Ulug, S. E. The backwaahing of rapid sand filters. Unpublished
thesis for the Diploma of membership of Imperial College of Science
and Technology. London, United Kingdom, Library, Imperial College
of Science and Technology, University of London, 1967.
133. United States Environmental Protection Agency. Process Design
Manual for Suspended Solids Removal. U.S. EPA Technology Transfer,
1975.
134. Van Heerden, C., Novel, A. P. P., and Van Krevelen, D. W. Studies
on fluidization - The critical mass velocity. Chem. Eng. Sci.,
l.:37-49, 1951.
135. Volpicelli, G., Massimilla, L. and Zenz, F. A. Non-homogeneities in
solid-liquid fluidization. Chem. Eng. Frog., Symp. Ser., 62(67):
42-50, 1966.
136. Vosloo, P. B. B. Some experiments on rapid sand filtration of
sewage works effluents without coagulation. Inst. Sewage Purif., J.
Proc., 1:204-209, 1947.
353
-------
137. Walker, J. D., Aurora, Illinois. Dual media tertiary filters
design parameters. Private communication with Walker Process,
August 1, 1972.
!38. wen, C. Y., and Yu, Y. H. Mechanics of fluidization. Chem. Eng.
Prog., Symp. Ser., .62(62) : 100-111, 1966.
139. Whitmore, R. L. The sedimentation of suspensions of spheres. Br.
J. Appl. Phys., 6:239-245, 1955.
140. Whitmore, R. L. The relationship of the viscosity to the settling
rate of slurries. J. Inst. Fuel, 30:328-342, 1957.
141. Wilhelm, R. H., and Kwauk, M. Fluidization of solid particles.
Chem. Eng. Prog., 44:201-218, 1948.
142. Wood, R., Smith, W. S., and Murray, J. K. An investigation into
upward flow filtration. Water Pollut. Control, 6_7(4) :42l-428,
1968.
143. Woods, C. F. Expansion and intermixing of garnet and silica sand
during backwashing of granular filters. Unpublished M.S. thesis.
Ames, Iowa, Library, Iowa State University of Science and Tech-
nology, 1973.
144. Zabrodsky, S. S. Hydrodynamics and Heat Transfer in Fluidized
Beds. Cambridge, Massachusetts, The M.I.T. Press, 1966.
145. Zenz, F. A. and Othmer, D. F. Fluidization and Flufil Particle
Systems. New York, Rheinhold Publishing Corporation, 1960.
354
-------
Sieve analyses of uniform media.
U>
Designation
F(-4+7)
F(-7+8)
F(-8+10)
F( -10+12)
A(-12+14)
A( -14+16)
A(-16+18)
A(-18+20)
AC -20+25)
Coals
U.S.
sieve
4
7
Pan
7
8
Pan
8
10
Pan
10
12
Pan
12
14
Pan
14
16
Pan
16
18
Pan
18
20
Pan
20
14
Coals (continued)
retained
3.4
79.7
16.9
3.3
85.0
11.7
4.6
81.0
14.4
4.8
84.2
11.0
4.7
80.5
14.8
2.2
83.6
14.2
7.0
77.8
15.2
3.8
77.8
18.4
10.2
Designation U.S. %
sieve retained
A(-25+30) 25
30
Pan
A( -30+35) 30
35
Pan
A(-35+40) 35
40
Pan
A(-40+45) 40
45
Pan
14
Silica sands
A(-10+12) 10
12
Pan
A(-12+14) 12
14
Pan
A(-14+16) 14
16
Pan
A(-16+18) 16
4.9
84.1
11.0
3.2
83.3
13.5
8.8
81.7
9.5
4.2
83.6
12.2
0.1
83.7
16.2
9.2
80.1
10.7
5.9
70.1
24.0
2.7
14
Silica sands (continued)
Designation U.S.
sieve
A(-18+20) 18
20
Pan
A(-20+25) 20
25
Pan
A(-25+30) 25
30
Pan
A (-30+35) 30
35
Pan
A(-35+40) 35
40
Pan
A (-40+45) 40
45
Pan
Garnet sands
A(-14+16) 14
16
18
Pan
retained
4.3
87.0
8.7
0.6
74.9
24.5
2.1
77.6
20.3
1.3
70.9
26.8
3.6
72.6
23.8
6.2
74.8
19.0
143
i^Tn^
0.24
87.72
12.00
0.04
-------
Sieve analysis of uniform media (continued).
tn
Designation
A(-25+30)
A(-50460)
14
Coals
U.S.
sieve
25
Pan
25
30
35
Pan
50
60
70
Pan
1A
Coals'""' (continued)
%
retained
78.3
11.5
0.11
88.91
10.46
0.53
0.14
81.81
16.56
1.49
Designation U.S.
sieve
18
Pan
A(-20+25) 20
25
30
Pan
A(-30+35) 30
35
40
Pan
%
retained
74.5
22.8
0.22
78.40
21.33
0.05
0.54
88.76
10.58
0.12
14
Silica sands (continued)
Designation U.S.
sieve
A(-35-»40) 35
40
45
Pan
A (-40445) 40
45
50
Pan
%
retained
0.42
85.17
14.33
0.08
0.70
96.12
3.15
0.03
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/2-77-016
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
BACKWASH OF GRANULAR FILTERS USED IN WASTEWATER
FILTRATION
5. REPORT DATE
April 1977 (Issuing Date)
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
J..L. Cleasby and E. R. Baumann
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Iowa State University
Ames, Iowa 50011
10. PROGRAM ELEMENT NO.
1BB043.ROAP 21-ASO.Task 15
11. CONTRACT/GRANT NO.
R802140
12. SPONSORING AGENCY NAME AND ADDRESS
Municipal Environmental Research Laboratory--Cin.,OH
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, Ohio 45268
13. TYPE OF REPORT AND PERIOD COVERED
Final, 9/1/71-5/31/76
14. SPONSORING AGENCY CODE
EPA/600/14
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The use of deep granular filters in waste treatment is of growing importance. The
key to long-term operating success of such filters is proper bed design and adequate
bed cleaning during backwashing. A number of questions related to adequate backwash-
ing of granular filters are investigated which lead to the following conclusions:
Cleaning granular filters by water backwash alone to fluidize the filter bed is
inherently a weak cleaning method because particle collisions do not occur in a
fluidized bed and thus abrasion between the filter grains is negligible.
Due to the inherent weakness of water backwashing cited above, auxiliary means of
improving filter bed cleaning are essential for wastewater filters. Three auxiliary
methods were compared in a wastewater pilot filtration study. The most effective
backwash was provided by air scour and water backwash simultaneously at subfluidiza-
tion velocities. The other two methods, surface and subsurface wash auxiliary or
air scour prior to water fluidization wash were about comparable in effectiveness.
The performance of coarse sand, dual-, and triple-media filters was compared, and
the backwashing routines appropriate for each media are discussed. A number of
investigations concerning the design and backwashing of dual media filters are
presented.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
COS AT I Field/Group
Filtration , Sewage Filtration ,
Sewage Treatment, Waste Treatment
Water Pollution, Backwashing*,
Clarification
Filter Backwash, Sus-
pended Solids Removal,
Filter Media, Filter
Cleaning, Filter Design
13B
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
ZO. SECURITY CLASS (This page)
UNCLASSIFIED
21. NO. OF PAGES
381
22. PRICE
EPA Form 2220-1 (9-73)
357
-1977-757-056/5573 Region No. 5-11
------- |