You are on page 1of 90

APPLICATION OF THE FLOTO FILTER UNIT

FOR CONTACT FLOCCULATION FILTRATION


OF SURFACE WATERS



by


D.R.Induka B.Werellagama



A thesis submitted in partial fulfillment of the requirement for the degree of Master of Engineering.







Examination Committee: Dr C.Visvanathan (Chairman)
Dr S.Fujii
Mrs Samorn Muttamara




D.R.Induka B.Werellagama

Nationality : Sri Lankan
Previous Degree : BSc.(Eng) Hons.
University of Peradeniya, Sri Lanka
Scholarship Donor : Swedish International Development Agency (SIDA)




Asian Institute of Technology
Bangkok, Thailand
April 1993
-2-



ACKNOWLEDGEMENT



The author wishes to express his sincere appreciation and gratitude to his advisor
Dr.C.Visvanathan for all the advice and guidance given throughout the thesis study period. Sincere
thanks are offered to Mrs. Samorn Muttamara and Dr. S. Fujii for sharing their interest in this study
and serving in the examination committee. Special thanks are also due to Professor R.Ben Aim and
Dr. S. Vigneswaran for their valuable suggestions and help during this study.


The Calgon Corporation of U.S.A. which provided the Catfloc T2 used in the experiments,
Prof. R. Ben Aim and HMC Polymers Ltd of Bangkok who provided synthetic media and the US
Environmental Protection Agency which provided technical information are gratefully
acknowledged.


Thanks are also extended to Udeni for helping during endless sieving sessions to isolate the
correct size of media, Deepa and Thayalan for helping by taking readings when the filter runs
extended beyond 40 hours, Mr. Uttam Manandhar, Jayaweera, Gemunu, DN, Jeyaranie and all other
friends for helping in one way or another, Mr. Varine of the Ambient Laboratory who built the set up
and helped with the repairs whenever necessary, and to the laboratory staff of the Environmental
Engineering Division for their support.


The author wishes to express his gratitude to the people of Sweden and The Swedish
International Development Agency (SIDA) for providing him the scholarship for study at the Asian
Institute of Technology.


Finally the author gratefully dedicates this work to his beloved parents for their affection,
concern and encouragement throughout his career.
-3-



ABSTRACT



Laboratory scale experiments were carried out in contact flocculation filtration using a dual
media filter. The objective of the research was to find an optimum synthetic media combination
which would give acceptable quality water under varying conditions while maintaining a low
headloss. The media was lighter than water hence the bed was floating. Since coarse media remained
at the bottom the flow direction was upflow which had the added advantage that the flow was in the
direction of grain compression. Polypropylene and Polystyrene were selected as the optimum
combination of media which due to their large density difference did not intermix even under severe
agitation. Spherical fine media performed better than angular fine media giving lower headloss and
better effluent quality. The dual media combination and the higher rates of filtration are in line with
the current trends in the water industry. The influent concentrations were kept constant and the flow
velocity and filter media size were varied. The headloss variation along the filter, the influent quality
to the two filter layers and the effluent quality were studied.


The filter performed for over 40 hours producing acceptable quality water at conventional
rapid sand filtration rates, also having low headloss development. Another major advantage was the
ease and economy in backwashing. The filter media did not mix even during or after backwashing,
thereby eliminating the most common problem encountered in conventional multi media filters.


The Floto Filter operation was compared with upflow sand filtration. The headloss
development curves for Floto filter showed a characteristic shape easily identifiable from that for
sand.


The existing mathematical model was able to predict the headloss profile but the
concentration profile was not predicted. The necessity of particle size data for mathematical
modelling of filtration of heterodisperse suspensions is emphasized.
-4-



TABLE OF CONTENTS

CHAPTER PAGE
Title Page i
Acknowledgement ii
Abstract iii
Table of Contents iv
Abbreviations v

1 INTRODUCTION 1

2 LITERATURE REVIEW 4
2.1 Introduction 4
2.2 Theory of Filtration 4
2.3 Contact Flocculation - Filtration 6
2.4 Headloss Development 8
2.5 Contact Times for Polymers 12
2.6 Rate Control Patterns and Methods 12
2.7 Some Parameters Affecting Turbidity 13
2.8 Filter Backwashing 14
2.9 Floating Bed Filters 16
2.10 Mathematical Models for Deep Granular Filters 26

3 EXPERIMENTAL INVESTIGATION 36
3.1 Experimental Set Up 36
3.2 Experimental Runs 41
3.3 Materials 43
3.4 Measurements 48

4 PRESENTATION AND CRITICAL DISCUSSION OF RESULTS 51
4.1 Introduction 51
4.2 Experiments with Floating Media 51
4.3 Effect of Physical Parameters on Filter Runs 69
4.4 Polymer Dosage and Mixing Time of Polymers 74
4.5 Effects of Controls and Processes 75
4.6 Experiments with Sand Medium 79
4.7 Comparison of Experimental Runs and Concluding Remarks 82
4.8 Mathematical Modelling 88

5 CONCLUSIONS 95

6 RECOMMENDATIONS FOR FURTHER STUDY 98


REFERENCES 102

APPENDIX 1 106
APPENDIX 2 138
APPENDIX 3 140

-5-






ABBREVIATIONS



A
c
- area of collectors in a unit volume of filter (m
2
)

A
p
- area of retained particles in a unit volume of filter (m
2
)

C
i
- concentration in ith time step (mg/l)

C
o
- influent concentration (mg/l)

d
c
- diameter of collector (m)

d
p
- diameter of particles (m)

f - porosity of bed at time t

f
o
- porosity of clean bed

g - acceleration due to gravity (m/s
2
)

GAC - Granular Activated Carbon

h
f
- headloss through a clogged bed (m)

H
o
- initial headloss (m)

J - head gradient

JTU - Jackson Turbidity Unit

K - Kozeny's constant

K
w
- Kuwabara's constant

L - filter depth (m)

L - filter depth increment (m)

-6-



N - number of particles directly attached on the filter grain

N
p
- number of particle collectors in unit volume (m
-3
)

n - particle concentration at a given time and depth (m
-3
)

n
i
- particle concentration at i
th
time step (m
-3
)

n
o
- influent particle concentration (m
-3
)

NTU - Nephelometric Turbidity Units

OTV - Omnium de Traitments et de Valorization of France

PAC - Powdered Activated Carbon

PACEFILT - PAC Embedding Filtration

P
e
- Peclet Number

PP - Polypropylene

PS - Polystyrene

REFIFLOC - Refiltration Flocculation Process

RSF - Rapid Sand Filter

S
1
- shape factor of suspended particle

S
2
- shape factor of filter grain

SSF - Slow Sand Filter

t - filtration time (s)

US EPA - United States Environmental Protection Agency

V - filtration velocity (m/s)

V
p
- volume of particles deposited (m
3
)

-7-



- single collector efficiency of a clean collector

c
- combined removal efficiency of a single collector

D
- single collector contact efficiency by diffusion

I
- single collector contact efficiency by interception

Lo
- single collector contact efficiency by London Van der Waals
force

p
- contact efficiency of a retained particle

r
- single collector removal efficiency of a filter grain and its
associated retained particles

s
- single collector contact efficiency by sedimentation

- particle to filter grain attachment coefficient

p
- particle to particle attachment coefficient

- fraction of retained particles acting as particle collectors

' - the fraction of retained particles that contribute to the
additional surface area

1
- fraction of filter grain surface, which is exposed for the
particle deposition

2
- detachment coefficient

d
- porosity of deposit

- density of water (kg/m
3
)

p
- density of particles in suspension (kg/m
3
)

- viscosity of suspension (Ns/m
2
)

- specific deposit

-8-



- fraction of coarse particles which can act as particle
collectors to remove finer particles in the suspension
-1-

CHAPTER 1



INTRODUCTION



1.1 General

There is an increasing need to treat low quality surface water to produce drinking quality
water complying with consent conditions. Water percolating through a bed of granular media
(filtration) is widely used in municipal water treatment for clarifying dilute suspensions with particles
of a wide range of sizes.

In surface water filtration, slow sand filters(SSF) and rapid sand filters(RSF) are widely used
for removal of solids present in surface waters, precipitated hardness from lime softened water and
precipitated iron and manganese. Both these types (SSF and RSF) are deep granular filters and often
the filter media is graded silica sand. Another type of filter is precoat filters, which use diatomaceous
earth, perlite, powdered activated carbon etc; as the filter media.

Rapid sand filters are more popular for municipal applications due to their lower space
requirement, higher production capacity, and higher flexibility of treating waters of different
turbidities. Lower space requirement means lower capital cost to achieve the water of desired
quality. The flow rate in a conventional rapid filter is in the range of 2.4 - 10 m
3
/m
2
.h. The rapid
filters conventionally treat water that are passed through several pretreatment steps like screening,
primary sedimentation, rapid mixing, coagulation and flocculation and secondary sedimentation.

After some period of operation the filter media gets clogged and the production rate declines.
When this happens the filter has to be taken out of operation and its flow direction is reversed in
order to remove the clogging particles. This process is called backwashing. As the production
capacity of a rapid sand filter is increased, the clogging rate also increases, resulting in more frequent
backwashing. In addition to the loss of production time, product water has to be utilized for the
backwashing operation. Rapid gravity sand filtration would typically consume 2 - 5 % of throughput
for backwashing. In view of saving this water as well as saving operator time required for frequent
backwashing operation (which entails valve operation), researchers in the past three decades have
focused their attention on a wide range of process modification such as upflow/ biflow filtration and
mobile bed filtration for enhancing the filter performance.

One notable research work on modifying the filter media itself was the application of the dual
or multi media filtration which utilized various types of sand, crushed anthracite coal, diatomaceous
earth, perlite and powdered or granular activated carbon etc; as the filter media. Even these modified
arrangements had the major drawback of frequent intermixing of filter media after backwashing in
-2-

the conventional manner.

In this research work, the focus was on the modification of filter media. Here the classic sand
media was replaced by synthetic polypropylene and polystyrene beads, whose density is less than
that of water. Therefore these beads floated in water. During filtration the artificially made turbid
water was dosed with the flocculent just before entering the bed, effecting contact flocculation. As
the turbid water moved up the floating bed, suspended solids in the form of floc were captured within
the filter media. The purified water was collected at the top. As the media itself was always kept in
suspension, it facilitated easy backwashing. The energy required to agitate the floating bed to
resuspend floc was much less than for a conventional sand bed.


1.2 Objectives of the Study

The objectives of this research were

(1) Finding a combination of floating filter media able to produce an acceptable effluent under a wide
variety of conditions and also having a uniform floc and headloss distribution over the depth of bed.

(2) Determination of the effectiveness of contact flocculation filtration on deep bed dual media filters
operating in the upflow mode.

(3) Optimization of the filter performance for different filtration rates, different influent turbidities
and for different media types and sizes. The selected operating conditions reflected the current trend
in water treatment plants towards higher filtration rates and dual media. The analysis was by studying
the resulting filter effluent quality patterns and the head variation along the filter.


(4) Comparison of the Floto filter performance with an upflow sand filter and identifying the
advantages/ disadvantages of floating bed filtration over conventional filtration.

(5) Verifying an existing mathematical model using the pilot plant results.


1.3 Scope of the Study

The study was basically a series of experiments utilizing a dual media filter with floating
media. These were followed by analysis, interpretation and limited computer modelling.

The filter operation was studied for the constant rate flow mode under a constant head. The
experiments were carried out in upflow direction. The usage of floating filter media in dual
arrangement with coarse to fine arrangement was verified as the best option by initial hydraulic
studies.
-3-


The final experimental results were headloss variations along the two media layers as the
filtration progressed more than 40 hours, headloss and filtrate quality relations with time, water
quality variation for two media layers, and the breakthrough behavior at different filtration velocities.

Since only a limited selection of floating media were available, the available media was
crushed, sieved and heat treated in order to prepare smaller sized media fractions.

O'MELIA and ALI's (1978) filtration model which had been modified by VIGNESWARAN
and BEN AIM (1985), VIGNESWARAN and CHANG (1986) and MANANDHAR (1990) was
used for simulation and verification of Floto - Filter results.



1.4 Limitations of the Study

The modelling was limited to O'MELIA and ALI's model only. i.e Attention was given only
to microscopic parameters of filtration.

Artificial suspension of Kaolin clay, the turbidity of which was kept constant was the
influent.

Only contact flocculation - filtration mode was studied which, by definition, allows very
small contact times for coagulation and flocculation.

Due to the limitations of the available dosing pumps very dilute stock solutions had to be
used for dosing of flocculents.

The inability to get smaller sized floating media necessitated preparation of them by methods
available at hand. Some of the resulting media therefore had a density variation which might have
had an effect on experimental results.

The absence of particle size data necessitated the use of simplifying assumptions during the
mathematical modelling.
-4-

CHAPTER 2



LITERATURE REVIEW



2.1 Introduction

The design of a granular medium filter system involves the consideration and specification of

* The type of medium, size and depth
* Filtration Rate
* Pressure available and driving force
* Method of filter operation (including cleaning)

The deep granular filters presently in operation (particularly Rapid Sand Filters and Slow
Sand Filters) generally consist of 0.45-0.75 m of filter medium supported on an underdrain system.
The filter may be open to atmosphere (gravity filter) or enclosed completely in a pressure tank
(pressure filter). Filtered water collected in the underdrain is discharged to a reservoir or to the
distribution system. The underdrain system is also used to reverse the flow to backwash the filter.

Most of the results presented here in the literature review were derived from filtration
experiments utilizing graded silica sand as the principal media. Some researchers have utilized other
media like garnet sand, anthracite coal, perlite, and diatomaceous earth. All these media are of higher
density than water. Experiments utilizing filter media of density less than water (e.g. polystyrene,
polyethylene, polypropylene, filter ag and paraffin) are described in section 2.9.




2.2 Theory of Filtration

2.2.1 Coagulation and Flocculation

Coagulation is the destabilization and initial aggregation of colloidal and finely divided
suspended matter by the addition of a floc forming chemical or by biological processes. Coagulation
involves the charge neutralization and destabilization of colloids and formation of microflocs.


Flocculation is the agglomeration of colloidal and finely divided suspended matter after
coagulation by gently stirring using either mechanical or hydraulic means. During flocculation the
-5-

microfloc is agglomerated into macrofloc.



2.2.2 Principal Mechanisms of Filtration

According to the filtration model formulated by O'MELIA and STUMM (1967), the
dominant mechanisms of removal of suspended solids in a filter depend on the physical and chemical
characteristics of the suspension and the medium, the rate of filtration and the chemical
characteristics of the water.


In deep granular filters of coarse material, the removal is primarily within the filter bed
(Depth Filtration). The removal efficiency depends on a number of mechanisms. Some solids are
removed by the simple mechanical process of interstitial straining. Removal of other solids
(particularly micro particles) depend on two types of mechanisms.


(i). A transport mechanism brings the micro particle from the bulk of the fluid within the
interstices, close to the surfaces of the media. Transport mechanisms for depth filtration may include
gravitational settling, diffusion, interception and hydrodynamic forces. These are affected by
physical characteristics such as size of the filter medium, the size of the suspended particles and the
ratio of suspended particle size to media size, filtration rate, fluid temperature (viscosity) and the
density .


ii). As the particle approaches the surface of the medium, or previously deposited solids on
the filter medium, an attachment mechanism is required to retain the particle. The attachment
mechanism may involve :
* Electrostatic interactions
* Chemical bridging or
* Specific adsorption


The efficiency of the filtration process for a given set of hydraulic conditions depend on the
attachment forces. The pores get clogged due to accumulation of material as the run progresses. As
the approach velocity is kept constant in the high rate filtration, the hydraulic gradients increase due
to this accumulation of material. Increase in hydraulic gradient increases the shear forces. The
filtration efficiency is effectively set by the relationship between the attachment and shear forces.
Breakthrough occurs when the hydrodynamic shear forces become greater than attachment forces. A
flocculent causing strong attachment forces in accumulating material will prolong the filtration cycle
but also cause high head loss. To decrease the head loss the diameter of the grains has to be
increased. The other solution of decreasing the flow is not practical (ADIN & REBHUN, 1974).
-6-



2.2.3 Water Pretreatment

Water can be pretreated to improve the performance of a filter. The pretreatment may:
1) Decrease the filtration resistance of the suspended solids
2) Increase the ability of the filter to remove and retain suspended solids that are too
small to be removed solely by straining.


When the suspended solids concentration of a water is not high, small doses of coagulants
(alum, ferric chloride, polyelectrolytes) addition before filtration helps to increase the permeability of
the solids that deposit in the filter. Generally this pretreatment reduces the resistance by no more than
50 %. This type of pretreatment changes the filterability of the solids. It does not decrease the total
amount of solids delivered to the filter. In fact, the total amount of solids collected on the filter
increases. Such pretreatment can reduce the breakthrough tendency of the solids by improving the
ability of the filter to retain them.


ADIN & REBHUN (1974) found out that the alum floc are too weak to withstand high shear
forces while the polyelectrolytes formed strong floc.


2.3 Contact Flocculation - Filtration

2.3.1 Introduction

Sometimes when the turbidity of the raw water is low some of the pretreatment steps of the
conventional filtration process can be taken off. In direct filtration the water is applied directly to the
filter after only screening, coagulant addition, rapid mixing and flocculation. Contact flocculation
filtration is a further development of direct filtration. Only pretreatment is chemical (coagulant &
flocculent) addition (clarification steps by flocculation and sedimentation are omitted). To combine
flocculation and coagulation in a single rapid process, a porous bed is required. The flocculation
occurs during the contact of raw water and flocculent within the filter media and the whole solids
separation process occurs in the filter bed. The process of Contact Flocculation Filtration differs from
the volume flocculation due to the high rate of the flocculation.


Raw Water Flocculent



Filter Bed

-7-




Filtered Water



FIG 2.1 Contact Flocculation Filtration

According to ADIN and REBHUN (1974) who experimented using both hydrolytic and
polyelectrolytic flocculents, the removal mechanism of the contact flocculation filtration process has
three stages. A working in stage, a working stage and a breakthrough stage . The working in stage is
characterized by a rapid decrease in effluent turbidity with time, reaching a low, stable value. The
working stage is the effective (main) stage of filtration giving satisfactory effluent quality. If the run
is not terminated due to head loss, the breakthrough stage can be identified by the deterioration of the
effluent quality. The functioning of the bed can be described by the frontal advancement of the
working layer in which effective filtration is taking place.


ADIN and REBHUN (1974) found experimentally that contact flocculation filtration with
alum alone was not efficient at high rates with coarse media. Upto 0.62 mm grain size, alum alone
gave efficient filtration at filtration rate of 5 - 10 m
3
/m
2
.h. The cationic polyelectrolytes made higher
filtration rates (20 m
3
/m
2
.h) possible. They noted that for higher output per cycle, filtration with
polymer must be done through coarse media.


In contact flocculation filtration, flocculation occurs within the filter bed. If alum is used as
the flocculent it will contribute to a large fraction of the sludge produced due to the aluminum
hydroxide precipitate. (raw water containing 10 mg/L of suspended solids may require an alum dose
of about 25 mg/L ). When polyelectrolytes are used as the sole flocculents, they are applied in
smaller doses (usually in the 1 mg/L range), and the sludge produced is composed almost entirely of
solids which originated in the raw water (SHEA et al. 1971). Therefore, polyelectrolytes are more
suitable flocculents for contact - flocculation and are added just ahead of the filter.


Since the entire solids removal takes place within the filter itself, waters with low turbidity
range are more suited for contact flocculation filtration. This is because in the conventional fine to
coarse arrangement the majority of the particles are removed at the top layer of the filter bed,
resulting in rapid clogging. SHEA et al.(1971) report that among various types of media used for
Contact Flocculation Filtration, the best results were given by coarse and uniform dual media, used
in coarse to fine media arrangement.


2.3.2 Advantages and Disadvantages
-8-


The major advantage in contact flocculation filtration is that it eliminates the sedimentation
and flocculation processes required in conventional filtration. This results in large operational and
capital savings. Another advantage is that the sludge is produced only in the filter backwash process,
hence causing less handling problems. The relatively high cost of polyelectrolytes is offset due to the
low volumes needed, resulting in a net saving in the chemical cost. Operational costs are also
reduced due to lower handling and storage requirements.

The disadvantages are the shorter filter runs resulting from the entire solids removal being
within the filter itself. The shorter filter run would increase the frequency and the degree of
backwashing [VIGNESWARAN et al,1983].


2.4 Head Loss Development

2.4.1 In Contact Flocculation Filtration

ADIN & REBHUN (1974) report that when alum was used as the flocculent, the head loss
increased linearly for a greater part of the cycle. Toward the end of the cycle a nonlinear headloss has
been observed. With the polymer the headloss developed exponentially. Hence for the same
hydraulic and operating conditions the increase in the rate of head loss buildup with polymer was
higher than with alum. They therefore state that the smaller the attachment forces and deeper the
penetration, the slower the development of the head loss.

Head Losses developed at a lower rate when the grain diameter was increased or the flow
velocity was decreased. Grain size strongly affected the headloss, while increasing the filtration rate
(for a given bed) did not. The main effect of flow rate for a specific grain size was related to the
initial head loss only.


SHEA et al. (1971) had observed that the initial headloss was not related linearly to the flow
rate. They report that for the same filter, the initial headloss was 2.3 cm for a flow rate of 7.3 m/h and
14 cm for a flow rate of 22.0 m/h.


2.4.2 Headloss development in upflow filtration

As reported by HAMANN & McKINNY (1968), IVES (1967) found that upflow filter runs
carried to give the same head loss were longer than downflow filter runs. Therefore for identical
lengths of run, head loss through the upflow filter was appreciably lower than through the downflow
filter. The filter used in this case had a sand bed of 1.2 m and the sand was graded from 0.6 mm at
the top to 1.2 mm at the bottom. HAMANN & McKINNY (1968) also report about the work of
MINZ (1962) in Russia. His upflow filters had contained gravel and sand to a depth of 2.3 to 2.6 m.
-9-

Increase in head loss during filtration had been slow. He had attributed this to the removal of a
substantial amount of suspended matter in the coarse portion of the filter where it has less influence
on head loss.


HAMANN & McKINNY (1968) also carried out a series of upflow filtration experiments
for filter beds of depth 60 cm and 120 cm, using alum and polymer as coagulant and flocculent. The
deeper the bed, the initial head loss was higher but the head loss development was lower for the
deeper bed. After 3 to 4 hours the head loss in the 60 cm bed exceeded the head loss in the 120 cm
bed.

They also report that in all cases the filter runs were terminated either by bed lifting or by
fluidization of the finer sand. Both of these phenomena resulted in the escape of the solids
(previously accumulated in the filter bed) with the filtrate. They noted that bed lifting or
breakthrough may occur in an upflow filter when the weight of the bed above a given level becomes
equal to the head loss developed above that level.


ADIN and REBHUN (1974) reported that the main effect of flow rate for a specific grain
size is related only to the initial head loss. Experiments on upflow filtration by PERERA (1982) also
show headloss development similar to ADIN & REBHUN (1974).


DANIEL & GARTON (1969) studied various combinations of sand, coal, glass beads,
walnut shells and pelleted paraffin wax as media for upflow filtration. Their model upflow filter was
10 cm in diameter and 1.83 m high. They added varying amounts of coagulant aid to the waters and
noted that both high turbid and low turbid waters required approximately the same amounts of
coagulant aid. The flow rate through the model filters was 2.44 m/h. Six runs were reported for each
media. Each run was of 23 hours duration. The head values before and after backwash are given in
figure 2.2. This figure also gives the filtrate quality for a particular run.













-10-



















Fig 2.2 Effluent and Headloss values for different media
(DANIEL & GARTON 1969)


ODIRA (1985) studied the headloss development patterns in upflow filtration utilizing sand
media in six different filter bed configurations. The coagulant used had been alum. He reports that


The increase of headloss with time varied with each filter design at the same filtration rate and coagulant
dosage. The finer filter media sustained the highest increase in headloss with time while the
effluent quality was substantially the same for all the filter designs. For a particular filter design,
the rate of headloss development and the terminal headloss were also affected by the coagulant
dosage and the influent turbidity; with the low dosage - low filtration rate - low influent
turbidity combination exhibiting the lowest headloss patterns. The normal values for the
headloss at breakthrough also varied for the various filter designs.


Results of ODIRA (1985) show a linear variation of headloss with time. For filter rates
above 10 m/h the headloss development had been very rapid resulting in very short filter runs.


2.4.3 Head Loss Development in Reverse Graded Filters

Reverse graded dual media and multimedia filters use coarser material on the raw water side
of the filter and finer material on the filtered water side. This accomplishes a much more uniform
-11-

distribution of solids throughout the depth of the filter, with much of the suspended matter being
removed in the coarser material. The head loss in these filters is generated at a much lower rate than
that of a conventional filter. Filtration through reverse graded media provides a filter run that is 2 - 5
times greater than obtained with the conventional filter, other conditions being equal. (SHEA et al.
1971)

















FIG 2.3 (WEBER, 1972)

The figure 2.3 shows the typical relationship of head loss to volume of flow for a reverse
graded filter.


Since the water is filtered through media of increasing fineness, this type of filter is less
subject to passage of solids due to filtration rate changes. Some breakthrough of solids occur, but not
to the same extent as with a conventional filter.


2.5 Contact Time for Polymers

According to TREWEEK (1979), who conducted direct filtration experiments utilizing 3.0
mg/l alum and 0.25 mg/l Catfloc T for treatment of surface water from a reservoir, a flocculation
time shorter than 7 minutes was not sufficient to produce the floc for removal in the filter media. His
filter column was a 30 cm bed of sieved sand (E.S. =0.5 mm and U.C. =1.3) and the flow rate was
11.5 m/h. Flocculation times exceeding 7 minutes produced large visible floc but the effluent quality
did not improve further. ADIN & REBHUN (1974) state that for their experiments in contact
flocculation filtration, the total contact time of the flocculent and the suspension before reaching the
bed was few minutes. LO (1984) added polymer to the influent point of the filter in his experiments.
-12-

The contact time in his case had been less than a minute.


The Calgon Product Bulletin 12-42b (1989) states that organic cationic polymers have a
relatively slow destabilization time (time required for adsorption, charge neutralization and initial
floc formation) as compared to inorganic flocculents like alum. It recommends the use of inorganic
coagulants (about 40% to 60% of the amount previously used) to speed up the total destabilization
time when short mixing times are encountered in practical applications.


It also states that in several water supply systems it was possible to feed the cationic
polymers to the raw water line far enough upstream from the plant to obtain several hours of
additional mixing time in the line. In such cases, as little as 0.5 ppm of the organic polymer has
improved clarification considerably and has eliminated the use of alum, activated silica and lime too.
When cationic polymer is used with alum, it is usually better to feed it into the raw water line ahead
of the rapid mix to obtain a maximum mixing time. Sometimes it is advantageous to premix the
polymer solution with the inorganic coagulant solution.


2.6 Rate Control Patterns and Methods

There are two basic methods of operating filters that differ primarily in the way pressure drop
(driving force) is applied across the filter. These methods are:

1) Constant rate filtration
2) Declining rate filtration

In constant rate filtration, the total operating head on the filter is fixed and the flow through
the filter is controlled at a constant rate by means of a flow control valve. As filtration proceeds, the
filter gets clogged with solids resulting in loss of head and declining flow rate. The flow control
valve is opened slowly to maintain a constant flow rate.

In declining rate filtration the incoming flow is supplied to a group of filters on a free flow
basis to meet their individual operating rates. There are no effluent controllers. The only control is
the effluent overflow level in the clear well.

In water treatment practice, the constant rate filtration is the most popular due to its proven
performance and higher operational control (KAWAMURA, 1991).


In operating pilot plants DANIEL & GARTON (1969) observed that preflocculated water
was difficult to control at a constant rate through the small rotameter flow meters. The unsettled floc
disturbed the rotameter floats.
-13-



According to CLEASBY & BAUMANN (1962) and TUEPKER & BUESCHER (1969), if
the filtration rate on a filter which contains deposited solids is suddenly increased, the hydraulic
shearing forces also suddenly increase. This disturbs the equilibrium existing between the deposited
solids and the water, and some solids will be dislodged to pass out with the effluent. Depending on
the type of solids, and the magnitude of the suddenness of the rate change, the effect can be quite
drastic. All sources of sudden rate change should be avoided in the design of filters.



2.7 Some parameters affecting effluent turbidity

HAMANN & McKINNY (1968) report that the effluent turbidity increased exponentially
with the increasing flow rate. They also noted that the turbidity decreased as the depth of the bed
increased. Increasing the bed depth resulted in better stability of operation and less trouble with bed
fluidization.


In analyzing the results obtained for several types of media DANIEL & GARTON (1969)
suggest that the major influencing factor in finished water turbidity was proper coagulation of the
raw water just prior to filtering. Their turbidity results and corresponding head losses were given in
figure 2.2 in section 2.4.2.



2.8 Filter Backwashing

During filtration, as the water containing suspended matter percolates through the bed, the
material accumulates within the interstices of the granular medium. The head loss builds up beyond
the initial value. Also as the granular medium becomes filled with removed particles, the suspended
matter in the filter effluent starts to increase. When the head loss or the effluent turbidity reaches
some predetermined value, the filter must be cleaned. Ideally, the time required for the head loss
build up to reach the preselected terminal value should correspond to the time when the suspended
matter in the effluent reaches the preselected terminal value for acceptable quality.


Most granular filters are cleaned by reversing the flow through the filter bed. Filtered water is
pumped through the bed at a rate sufficient to expand the bed. The suspended matter arrested within
the filter are removed by the shear forces created by the backwash water as it moves through the bed.
The wash water is drained off in washwater troughs.


-14-

The typical method for backwashing the granular media filters is air scour, water scour and
surface wash. Scour can be achieved by stepping up the velocity or rate of backwash per unit area
sufficiently (high velocity wash). As given by TCHOBANOGLOUS & SCHROEDER (1985) the
typical backwash rates for single medium, dual media and multimedia granular filters are 30 m/h, 48
m/h and 48 m/h respectively.


By directing jets of water into the fluidized bed surface scour is achieved. Surface wash
should start 1 - 2 minutes before backwashing begins. The surface wash is continued while the filter
is being backwashed, until the backwash water begins to clear. Scour of the bed can also be
intensified by stirring the fluidized bed mechanically.


Air scour serves to break up accumulated deposits. Air is blown upward through the bed
before or after fluidization. CLEASBY & BAUMANN (1977) state that the best backwash is
achieved with simultaneous air scour and water wash (as compared to air followed by water
backwash or to surface & subsurface wash). AMIRTHARAJAH (1978) has concluded that
backwashing with water alone is an inherently weak cleaning process due to the limitations in
particle collisions. Air scour and surface wash that promote interparticle abrasions during backwash
are indispensable for effective cleaning.


Dirty filters are commonly backwashed with filtered water. If the filters in water filtration
plants are not cleaned properly, fine material will accumulate in the form of mud balls. Mud balls
should be broken up and washed out or problems will quickly develop. A high pressure jet stream
directed into the expanded bed throughout the wash is required for this.


AMIRTHARAJAH (1988) states that for ordinary solids (with low adhesive forces), wash
water alone, which expands the bed by 30% to 40% to give expanded porosities around 0.65 - 0.70
in the top layers of the backwashed filter, provides optimum cleaning. Simultaneous air scour (54 -
90 m/h) and subfluidization water wash (14.7 - 19.6 m/h) provide the best cleaning for solids with
higher adhesive forces (polyelectrolytes and washwater solids). He also observes that dual media and
multimedia filters have the danger of loss of media with air and surface water wash and that they
need a fluidization wash at the end of backwash cycle to restratify the media.


ADDICKS (1991) reports that the scouring action of the simultaneous backwash is superior
to the water fluidization wash only in the transition zone from the packed bed to the fully three phase
(air,water & filter medium) fluidized bed. He notes that most of the particle abrasion is completed in
the first 1-2 minutes and thereby suggests that a backwash cycle of:

(1) Simultaneous air-water
-15-

(2) Water fluidization
(3) Repeating 1 & 2

will achieve a quicker and better result than prolonged wash cycles.


HAMANN & McKINNY (1968) report that nearly all the early upflow filters made a
"mistake" by sending the backwash water in the reverse (i.e. downward) flow direction. This had
been ineffective as it did not expand the media as occurs in washing the filter by upward flow, hence
suspended matter that had penetrated deep into the media was not completely removed. They report
that the Russian upflow filters of MINZ (1962) were washed with a cocurrent flow of approximately
30 m/h. (normal operating flow rate was kept below 6 m/h to prevent sand expansion).


QUAYE (1991) had used optimal upflow wash rates of 65, 54 and 43 m/h for summer, spring
and autumn, and winter respectively to backwash his dual media filter bed with water only.


SMET and GALVIS (1989) who conducted upflow roughing filtration experiments suggest
shock loading the filter as an efficient method of backwashing, when the backwash is done
downflow. They surged the filter by quickly opening the valves in the underdrain system, keeping it
open for one minute and then rapidly closing and reopening the valves.


2.9 Floating Bed Filters

2.9.1 The Biostyr

The Biostyr or the Upflow Floating Aerated Biofilter incorporates the features of the classical
biological aerated filter with the requirements of the upflow filtration. The filter bed consists of
submerged and floating granular medium (polystyrene).


1 Raw water inlet channel 7 Process air
2 Filter feed and sludge discharge 8 Aerated filtering zone
3 Wash water valve 9 Media retention with nozzles
4 Filter media 10 Treated water storage & discharge
5 Backwash air 11 Recirculating pump
6 Non aerated zone




-16-

Fig 2.4 The Biostyr (OTV, 1991)

The Biostyr is a trademarked wastewater treatment process marketed by the Omnium de
Traitments et de Valorisation (OTV) of France. Water circulates up through the floating media. The
media is not fluidized, but filtration is carried out in the direction of grain compression.


Polystyrene, the filter medium, is lightweight, and is easy to backwash. The size and the
density of the filter grains is controlled to suit the required treatment objective. The fine and regular
medium gives a large specific surface and efficient filtration.

Retaining the filter media at the top of the unit reverses the classical gravity filter system.
Upflow filtration enables feeding of the influent without obstructing the distribution devices. Flow in
the direction of grain compression favors the retention of the suspended solids. Backwash is
facilitated through a simple gravity flush. The lightweight beads facilitate the backwashing through
counter current flushing, rinsing most intensely the filtration zone in contact with the heavily loaded
influent. Sludge can be removed in the direction of gravity by the shortest way (ROGOLLA et al,
1992).


The backwash water is stored on top of the filter, therefore the Biostyr does not need a
backwash pump or a separate clean water reservoir. The backwash water flow rate is about 50 m/h.


2.9.2 Upflow Filter using Filter-ag medium.

RICE et. al.(1980) report about an upflow filter using Filter-ag as the filter medium. Filter-ag
is a commercially manufactured non hydrous aluminum silicate. The properties of this material are
given in table no 2.1.


Table 2.1 Physical properties of Filter-ag (RICE et al, 1980)

Property Description
Color Light grey to off white
Density 385-417 kg/m
3

Effective Size 0.57 mm
Uniformity Coefficient 1.66


-17-


































Fig 2.5 Configuration of filter using filter-ag media.
Note: The dimensions are in cm.


They had used two filter units as shown in figure 2.5. The size of the larger unit was
0.914 m in diameter and 2.134 meters in height. The height of the bed was 0.152 m. The flow
rate in this unit was 5.5 m/h. For raw water turbidity of 48 NTU it had produced effluent of 2.0
NTU to 1.5 NTU. When the raw water turbidity was 13 NTU the effluent turbidity had been 1.4
NTU.

-18-


The filter media, located at about the midheight of the tank, was held in place with 50 mesh
(i.e. 0.3 mm) stainless steel screens, above and below the material. The distance from the water entry
pipe to the filter media was 92 cm. A correct alum concentration was added and was needed for the
proper and efficient operation of the units.

At the design flow rate it took about 22 minutes travel time from the water inlet to the filter
media. Jar tests had shown, for low & moderate turbidity water with correct alum addition, this was
sufficient time for the floc to settle out. Downflow backflushing removed the collected sediment out
of the bottom of the unit. The space below the filter media thus served as a settling basin as well as
the mixing basin for coagulation, flocculation & disinfection (with chlorine).


The smaller unit was of 0.3m * 0.3m cross section and 1.52 m tall. Distance from the inlet to
the filter media was 1.07 m. Bed was similar to the large unit. The flow rate was 2.5 m/h. This unit
was not effective at higher flow rates. (5 to 15 m/h).


2.9.3 Filter Using Pelleted Paraffin Wax Media.

DANIEL & GARTON (1969) experimented with various types of media one of which was
pelleted paraffin wax. The results of this experiment are given in section 2.4.2 and figure 2.2. As this
material has a specific gravity less than 1.0, they required a screen both above and below the filter
media. After 25 hours of operation the paraffin media had given turbidity less than 5 ppm.


2.9.4 The Haberer Process

Development of the Haberer process began as a search for an improved upflow filter design
in which backwashing of the filter would be accomplished in a downward direction (downwash)
rather than upward, as practiced in other upflow filter designs. Downwashing allows downward
movement, with the force of gravity, of the dense floc formed in the upflow filter and therefore
ensures rapid removal of solids from the filter bed. Conventional backwashing in an upward
direction must remove the solids from the filter bed against gravity and therefore requires a
considerable amount of time and water to clean the filter. Whereas conventional backwashing of a
filter may take upto 8 minutes, downwashing cleans the Haberer filter to the same degree in about 2
minutes.


This filter used 1-3 mm foamed polystyrene beads (Styrofoam) as the filter medium. Since
the specific gravity of the medium was less than 0.1 the filter was fitted with a constraint above the
medium rather than with a conventional filter underdrain. A schematic diagram of the filter is given
in Fig 2.6.
-19-



















Fig 2.6 Haberer Process (STUKENBURG & HESBY, 1991)

According to STUKENBURG and HESBY(1991), the filter medium can be of any depth, but
a depth of 1.2m is often used. Typical filtration rate is 10 m/h. Downwash rates for the filter vary
with application. Expansion of the medium occurs at 100 m/h.


This filter is also used as a means to contact water with powdered activated carbon(PAC).
Many types of PAC will adsorb on the polystyrene medium so that, in effect, a carbon column can be
formed with PAC. Variations of this process have been patented as REFIFLOC( Refiltration
Flocculation) and PACEFILT( Powdered Activated Carbon Embedding Filtration). With the Haberer
filter, PAC can be used as efficiently as granular activated carbon(GAC) and can be used
intermittently if desired. A 36,000 m
3
/day plant in Wiesbaden, Germany employs the Haberer PAC
contactor as a second stage process to remove iron and manganese, to nitrify ammonia, and to
remove organic pollutants in water.


STUKENBERG and HESBY (1991) carried out a series of experiments using a Haberer
PAC contactor to treat alum coagulated water. They suggest that it is ideal for package plants, for
which the goal is to provide the maximum water treatment capacity possible in a given volume
and otherwise concluded that it has no apparent economic advantage over conventional
treatment. In their study polymer was added as a filter aid only for the conventional dual media
filter and not for the floating filter. They repeated their study for alum coagulated water (45 to 54
mg/L) without using PAC in the column. Even then the Haberer column gave good turbidity removal
comparable with the conventional dual media filter, until the breakthrough occurred after seven
-20-

hours. They also concluded that the presence of carbon had little effect on the rate of head loss build
up in the column.


In their trial run, the chemical floc produced by the alum contact, accumulated in the free
space below the polystyrene medium. The volume of this space was approximately 30% of the
volume occupied by the medium.

HABERER and SCHMIDT (1991), point out that resin beads made of foamed polystyrene
are better suited for an upflow filter than either polyethylene or polypropylene because of their lower
density and substantially greater buoyancy in water. The polystyrene is inert and poses no health
hazard. The beads should be as homogeneous as possible, and the optimal size depends on the
application. The compact filter bed can be changed to a fluidized bed without additional expenditure
of energy by simply directing an intense rinsing stream downward through the bed.

In this upflow filter, the backwash water is stored above the nozzle plate on top of the
floating filter bed. To initiate backwashing, the inlet to the filter is closed and the discharge valve
opened. Thus, the high backwash velocities required to fluidize the bed are obtained without a
backwash water pump, and air rinsing is not used.




















Fig 2.7 REFIFLOC process with polishing filter
(HABERER & SCHMIDT, 1991)

The upflow filter has a high capacity for the storage of captured solids and will act as a
-21-

flocculator. The Refiltration Flocculation (REFIFLOC) process for wastewater treatment is based on
this principle. Here the effluent is recycled and refiltered, possibly several times. REFIFLOC process
can be used successfully even for the treatment of highly polluted waters, for which large dosages of
flocculents are needed to remove turbidity, algae and other contaminants. HABERER and
SCHMIDT (1991) recommend a REFIFLOC filter followed by a polishing multimedia filter for this
purpose. This process is illustrated in figure 2.7.

PACEFILT process (Powdered Activated Carbon Embedded Filtration) was developed based
on the REFIFLOC process in order to meet the requirements of the adsorption stages of water
treatment. Here as a special pretreatment step, a slurry of PAC is distributed over the entire filter bed
in a high velocity closed recycle stream by using the refiltration pump of the REFIFLOC unit. This
produces an adsorption layer of PAC on the polystyrene beads. PACEFILT combines the advantages
of the PAC, i.e., high reactivity, with those of the filter, i.e., the efficient utilization of adsorption
capacity. After it becomes exhausted, the carbon is removed by an intense backwash stream that is
directed downwards. This is similar to the floc removal in the REFIFLOC process.























Fig 2.8: The 3 Steps of the PACEFILT process
(HABERER & SCHMIDT, 1991)

HABERER and SCHMIDT (1991), give the effect of backwash velocity on bed expansion
for polystyrene beads of two different densities. This is given in figure 2.9.
-22-















Fig 2.9 Bed Expansion due to Backwash Velocity (HABERER & SCHMIDT, 1991)

According to their experiment they specify a backwash velocity of between 70 to 110 m/h.
This value is consistent with that given by STUKENBURG and HESBY.


In their upflow filter HABERER and SCHMIDT used foamed polystyrene beads of 1 - 2 mm
size. The filter bed height was 1.0 m - 1.5 m. They did not use coagulant aid. Their experiments on a
pilot filter with PAC coated polystyrene
media showed that the smaller beads (size 0.9 - 1.3 mm) gave consistently better organics removal
than the larger beads (size 1.6 - 2.5 mm).



TABLE 2.2 MAXIMUM OPERATING PRESSURE IN RELATION TO FOAMED
POLYSTYRENE BEAD DENSITY (HABERER & SCHMIDT, 1991)

Density(kg/m
3
) Permitted Pressure(kPa/cm
2
)
15-25 0.8-1.8
60 5.5
100 8.0

The effect of coagulant aid on iron content of a REFIFLOC test filter after coagulation with
30 g FeCl
3
/m
3
is shown in figure 2.10. In this case the bed depth was 1.35 m and the filter diameter
was 0.19 m. Size of the polystyrene beads was 1.5 - 2.2 mm. Filter velocity was 6 m/h and the
maximum filter run was 8 h.
-23-

























Fig 2.10 Effect of Coagulant Aid


HABERER and SCHMIDT (1991) have carried out experiments to combine REFIFLOC and
PACEFILT processes. Experience to date shows that cylindrical filters are preferred. The nozzle
plate at the top of the filter bed should be designed to withstand the strong buoyancy of the filter
material and to effect a minor head loss to result in equal distribution of backwash water. They
specify that the height of the cylindrical filter should be at least 1.5 times that of the filter bed. If
contact flocculation is used in the REFIFLOC process the filter inlet chamber should be designed to
allow sufficient hydraulic detention time for floc formation. In order to attain complete removal of
the solids during backwashing the lower part of the cylindrical filter should be conical. This
configuration also helped in distribution of PAC to the filter. They also say that due to the buoyancy
of the filter medium, a sieve or screen to prevent its loss through the backwash water outlet is
unnecessary and would impair operation. HABERER and SCHMIDT (1991) also noted that in both
REFIFLOC and PACEFILT processes the smaller grains of the medium remained in the bottom
filtering layer. Therefore the grain composition in the lower reaches of the filter determined the
filtration effect and the required filter bed depth.


-24-

If the removal efficiency of the filter is inadequate, the filtrate quality can be improved by
refiltration. The concentrations of iron manganese and ammonium in the filtrate of the Weisbaden
water utility were considerably reduced in this way. Also the startup period for the plant was
reduced. After maturation of the filter, the refiltration ratio was reduced to the minimum value of 1.6.
This minor but constant refiltration guarantees uniform filtrate quality despite major fluctuations of
flow through the plant. The filtration velocity is kept at a constant level. If the flocculation processes
do not operate optimally, refiltration at a ratio between 2 and 3 may help improve the filtrate quality.
(HABERER and SCHMIDT, 1991). The upper limit of the refiltration rate is determined by the
velocity at which the increased shearing forces begin to destroy the floc and, thus, cause
breakthrough.



















Fig 2.11 Organics Removal vs FeCl
3
dosage
(HABERER & SCHMIDT, 1991)


A pilot filter at Biebeshiem, Germany treated the water of the river Rhine to remove turbidity
and organics using a combined REFIFLOC and PACEFILT plant. In this plant the efficiency of
organics removal with various flocculent dosages was studied. Before the filter, the water was
pretreated by ozonization (1g/m
3
) and FeCl
3
addition (7, 10, 24 g/m
3
). The result is shown in figure
2.11.


In this case the filter material was foamed polystyrene with 1.5-2.0 mm diameter. The
optimal filtration velocity was 6 m/h. The filter run was between 6 to 8 hours. The filter bed was
loaded with 700 g Fe(OH)
3
/m
2
before the breakthrough occurred.
-25-



HABERER and SCHMIDT (1991), report that about five filter bed volumes of backwash
water are typically required to remove the saturated powdered carbon once the bed was exhausted.
At backwashing velocities between 80 and 100 m/h, the corresponding backwash time was 3 - 5 min.
They also specified that backwash should effect 25 to 30 percent filter bed expansion.


2.9.5 Horizontal Filter Using Plastic Media

TANUMIHARJA (1981) used a bench scale horizontal filter to study the effectiveness of
coarse plastic media (25 mm diameter;and specific gravity 0.26) as a filter media. But this media was
prevented from floating by a 10 cm layer of crushed stone of effective size 20 - 25 mm.


This experiment showed that the coarse plastic media have suspended solids removal
efficiencies in the range of 30% to 60%. The flow rates utilized in this experimental study were from
0.5 - 1.5 m/h and the influent turbidity levels were 50 - 100 NTU.


2.10 Mathematical Models for Deep Granular Filters

To predict the behavior of filtration in deep granular filters, several researchers have
developed mathematical models. These can be categorized into two major groups, viz.

* Macroscopic models
* Microscopic models

The common macroscopic models (IVE's approach) relate filtrate quality and head loss with
time, incorporating measurable macroscopic variables of filtration such as filtration rate, grain size
and water viscosity.


Models based on microscopic parameters (O'MELIA's approach) consider microscopic
parameters such as individual particle and filter grain sizes, number of retained particles etc, in
addition to operating parameters of filtration. In this







-26-

Table 2.3 Contact Efficiencies of Clean Bed Filter ( TULACHAN, 1987)

Model Used Contact Efficiencies for Different Mechanisms
STOKES LEVICH 1962 :
D
=P
e
-2/3
; P
e
=Peclet number

YAO 1971 :
I
=1.5 (d
p
/d
c
)
2



s
=(
p
-)gd
p
2
/18v

=
I
+
s
+0.9 P
e
-2/3


HAPPEL HAPPEL 1958: A
S
=Happel's model constant


I
=1.5 A
S
(d
p
/d
c
)
2


COOKSON 1970:

D
=0.9 A
S

1/3
P
e

-2/3


SPIELMAN et al. 1973 :
(1) For small particles (d
p
<1 m)

=2.498 A
s
1/3
(vd
c
/40)
-2/3
{"/("+1)} S(")

(ii) For big particles (d
p
>1 m)

=
I
(4/3) { (9/5) N
Ad
}
1/3


N
Ad
=Adhesion group =-4 H
c
d
c
2
/9 A
s
d
p
4


KUWABARA RAJ AGOPALAN and TIEN 1976 :

=0.72 A
s
N
Lo
1/8
(d
p
/d
c
)
15/8
+

+2.4*10
-3
A
s
N
G
6/5
(d
p
/d
c
)
-2/5
+(4/3) A
s
1/3
Pe
-2/3


N
G
=Sedimentation group
LEE and GIESEKE 1979 :


D
=3.54 { (1 - ')/K
w
}
1/3
Pe
-2/3
; ' =1 - f


I
=1.5 { (1 - ')/K
w
} (d
p
/d
c
)
2
/ [(d
p
/d
c
)+1]
2


Here
D
,
s
,
I
,
Lo
are the contact efficiencies of single collector by diffusion sedimentation, interception, London-
Vander Waals force respectively.
-27-


research the model utilized had been developed using the microscopic approach. Hence it is
discussed in detail.


In 1978 O'MELIA and ALI presented their model to predict the development of headloss and
removal efficiency during filtration by packed beds. Many equations were developed during the
preceding two decades to calculate the clean bed filter efficiency from the fundamental forces
responsible for particle retention within the filter bed. These were given in the table 2.3 .


2.10.1 O'MELIA - ALI Model

O'Melia and Ali (1978) formulated their model based on the postulate that some retained
particles can act as filter media and thereby improve the filtration efficiency. Since retained particles
modify the performance of a filter as filtration proceeds, the equation governing the clean bed filter
becomes inapplicable. They considered the particle transport and attachment mechanisms within the
filter, analyzed it microscopically and proposed a set of equations (Equations 2.1 to 2.5 and 2.8) to
predict the variation of filter efficiency and head loss during filtration. They defined the single
collector efficiency as


Rate at which particles strike the collector
=--------------------------------------------------------------- (2.1)
Rate at which particles approach the collector



The actual collector consists of a filter grain and a number of particles attached to it which
also act as collectors. The combined removal efficiency
c
of a single collector is given by


(Rate at which particles +N * (Rate at which particles strike
a retained particle
strike the filter grain) acting as a collector)

c
=------------------------------------------------------------------------------------------------------- (2.2)
Rate at which particles approach the collector



nv (/4) d
c
2
is rate at which particles approach the collector where n is the bulk particle
concentration and v is the fluid velocity.

-28-


Equation 2.2 can be rearranged as

c
= +N
p
(d
p
/d
c
)
2
(2.3)

where
p
is the contact efficiency of retained particles which is defined as the rate at which
particles strike a retained particle acting as a collector divided by nv(/4) d
p
2
, where d
p
is the
diameter of the suspended particles.


O'MELIA and ALI considered the removal in the packed bed is dependent upon the transport
and attachment. Transport is defined by ,
r
and
p
. They introduced two empirical coefficients
and
p
to define attachment and rewrote the equation 2.3 as

r
= + N
p

p
(d
p
/ d
c
)
2
(2.4)



where, d
c
=diameter of the filter grain
d
p
=diameter of particles in the suspension
N =number of particle collectors attached to a filter grain
=particle - filter grain attachment coefficient

p
=particle - particle attachment coefficient
=contact efficiency of a single collector

c
=combined removal efficiency of a single collector

p
=contact efficiency of a retained particle

r
=single collector removal efficiency of a filter grain and its
associated retained particles.


The number of retained particles that can act as collectors, N, at a depth L and at time t can
be calculated using the equation 2.5.


where n =local concentration (number of particles per unit volume)
=fraction of retained particles which act as particle
collectors


Considering the mass balance of suspended particles,

ACCUMULATION =INPUT - OUTPUT - REMOVAL (2.6)


(2.5)

-29-


where L =filter depth
t =filtration time
f = porosity of the filter
bed


Simplifying


Equations 2.4, 2.5, and 2.8 comprise the mathematical model to
describe the removal efficiency of a filter bed in time and space.


With time taken discontinuously n/t =0

O'MELIA assumed
r
to be constant over some depth L and gave the solution of the
above equation 2.8 for t> L/v for i
th
time step as


where n
io
is the particle concentration entering the media
layer at i
th
time step.


For the case of clean bed filter it is given as


Again similarly for i
th
step equation 2.5 can be given as




Substituting equations 2.9 and 2.11 in 2.4 we can get




Substituting equation
2.12 in 2.9 we can get the
fraction of remaining particles
at any time t, provided that the values of and
p
are known.


(2.7)


(2.8)


(2.9)

(2.10)

(2.11)


(2.12)

-30-

The values of can be calculated theoretically from the equations given in table 2.3 or by
using equation 2.10.


This model assumed that the porosity of a filter bed remains constant throughout the filter
ripening stage.


VIGNESWARAN and CHANG (1986) developed a mathematical model based on
O'MELIA and ALI (1978) and considering the particle detachment coefficient
2
as suggested by
ADIN and REBHUN (1977). They assumed the decrease of filter efficiency during the post ripening
period to be caused by the hydraulic gradient and number of particles already retained on the filter
grain. The final model obtained by them is given as equation 2.13 where detachment is represented
by the last term.




t

ri
= [1 +
p

p
V (/4) d
p
2
n
i-1
t exp{(-3/2)(1-f)
ri-1
(L/d
c
)}]
i=1
t
- (
2
/n) J
i-1

r
n
i-1
(2.13)

i=1


where J
i-1
=hydraulic gradient at (i-1)
th
time step



2.10.2 Headloss Models

KOZENY defined the headloss caused by the clean filter bed in equation 2.14.
where,

(h
f
/L)
o
=hydraulic gradient of clean filter bed
f
0
=porosity of clean bed filter
k =Kozeny's constant
=density of fluid
=dynamic viscosity
s
o
=specific surface of clean bed
v
0
=superficial velocity
(2.14)

-31-



According to CARMAN, k ranges between 4 to 6 for a variety of liquids.

But as the filter run progresses specific surface and porosity change due to deposits of
particles inside the pores. Hence the equation 2.14 is modified to 2.15.



where
h
f
/L =hydraulic gradient of deposited media
s =specific surface area of the deposited media
f =true porosity after deposition
=(f
0
- )

There are several models relating headloss, h
f
, and specific deposit, , as given in Table
2.4 .

Table 2.4 Summary of the Headloss Models (TULACHAN 1987)

Researcher Model
Shekhtman, 1955 H =H
0
(1 + /f
0
)
-3

Mackrle, 1961 H =H
0
(1 +a/f
0
)
3
(1 - /f
0
)
-3

Camp, 1964 H =H
0
(1 + /(1 - f
0
)
4/3
(1 - /f
0
)
-3

Sakthivadivel, 1969 H =H
0
(1 - f
0
+ )
2
(1 - /f
0
)
-3
(1 - f
0
)
-3

Herzig, 1970 H =H
0
(1 +k)
Letterman, 1976


i
=k log H/H
0


O'MELIA gave the development of headloss based on KOZENY's equation for clean bed
filter as

where

=dynamic viscosity, kg/m.s or g/cm.s
=density of the fluid, kg/m
3
or g/cm
3

k =empirical constant
v =undisturbed superficial velocity above the bed , m/s
(2.15)


(2.16)


-32-

or cm/s
A
c
=Surface area of the filter grain, m
2
or cm
2

V
c
=volume of the filter, m
3
or cm
3


O'MELIA and ALI (1978) considered the change in the surface area as dendrites develop while
neglecting the porosity changes during the progress of the filter run and presented the equation 2.17.


where V
p
and A
p
are volume and area of particle deposited.


The headloss equations presented by O'MELIA and ALI
(1978 ) can be written in the following form.




where
N
p
=number of retained particles in
the unit volume of bed

N
c
= number of filter grains
(collectors) in unit volume

' =fraction of total number of retained particles that contribute to additional
surface area.

Since deposition is non uniform with depth, the bed is subdivided to several layers and h
f
/L
was calculated for each layer.


VIGNESWARAN and CHANG (1986) developed the headloss equation for clogged bed,
based on Kozeny's equation for clean bed. They give the decrease of porosity during filter run as


f =1 - (/6) [ N
c
d
c
3
+( N
p
d
p
3
)/(1 -
d
) ] / { A L} (2.19)


Where N
p
is the number of deposited particles of size d
p
. The porosity of the deposit,
d
is
included in calculating the volume of deposited particles. Shape factors S
1
and S
2
of particles and
filter grains respectively have been introduced in the headloss equations.
(2.17)


(2.18)



-33-


Equation 2.20 gives the
modified head loss model.


VIGNESWARAN and BEN
AIM (1985) had shown the influence of
particles of different sizes on the
removal efficiency of a particle of
another size. They showed that the ratio of coarse particles to finer particles influences the removal
of finer particles. A constant is defined by them which is the fraction of coarser particles which act
as particle collectors to remove fine particles in suspension.


After extensive investigations CHANG (1989) and MANANDHAR (1990) had developed
the computer programs to simulate the whole cycle of filtration based on the equations 2.13 and 2.20.
CHANG (1985) cites previous investigators in assigning a value of 8.5 for S
1
. MANANDHAR
(1990) has simulated the operation of graded bed filters and multi media filters, for varying influent
concentration, varying particle size distribution and for constant and declining rate filtration.
(2.20)



-36-

CHAPTER 3



EXPERIMENTAL INVESTIGATION



3.1 Experimental Set Up

The schematic diagram of the experimental set up is shown in figure 3.1. The main
components of the setup are:

1) Raw water feeder system
2) Chemical dosing system
3) Filter column and the Filter bed
4) Backwash air/water feeder system

The relevant design details of the system are given in the sections 3.1.1 to 3.1.4.


3.1.1 Raw water feeder system

(a) Solution preparation tank

The artificial suspension of Kaoline clay was prepared in a 50 l tank (No. 1 in Fig. 3.1). Once
the required quality was ensured, the suspension was pumped from this tank to the raw water tank
using a small centrifugal pump of 16 l/min capacity.


(b) Raw Water Tank

The artificial suspension of raw water was stored in this tank (No. 3 in Fig. 3.1). The tank
capacity was 290 l which facilitated a 5 hour filter run without replenishment at the maximum flow
velocity used (15 m/h). Since most of the filter runs were much longer than this, the raw water tank
was periodically replenished from the solution preparation tank. The raw water tank had a stirrer
arrangement which continuously stirred the suspension to prevent the suspended solids from settling.


c) Constant Head Tank

A centrifugal pump with 45 l/min capacity fed raw water from the raw water tank to the
constant head tank (No. 4 in Fig.3.1). The level was kept constant by an overflow arrangement which
-37-

recycled the excess water from the overhead tank back to the raw water tank.


The constant water level in the overhead tank was 2.05 m above the datum level. (i.e 2.65 m
above the ground level or 2.15 m above the filter bottom)


3.1.2 Chemical Dosing system

The coagulant was Aluminum Sulphate (alum). 20 mg/l of alum was added direct to the stock
solution preparation tank.


The required dosage of Catfloc-T2 (Cationic polymer) was introduced close to the influent
end of the filter directly and continuously to achieve contact flocculation. The Catfloc dosage for 30
NTU influent turbidity was established as 0.5 mg/l in jar test experiments. This confirmed the result
obtained by LIANG (1982). The dosages for 60 and 90 NTU influent turbidity too were found using
Jar Test experiments.


Using a magnetic stirrer the flocculent was thoroughly mixed with water for few minutes,
and kept in a storage bottle of volume 2.7 l (No. 5 in Fig. 3.1) to be fed to the system through a
dosing pump of capacity 2 - 80 ml/min (No.6 in Fig. 3.1). The speed of the dosing pump was kept
constant (at 6.67 ml/min) and the concentration of the flocculent stock solution was varied to suit the
required dosage and the flow rate. The total contact time between the flocculent and suspension
before they reached the filter bed was 1.6 minutes for 15 m/h flow rate and 4.8 minutes for the 5 m/h
flow rate.


3.1.3 Filter Column

(a) Filter Column

The filter column was a 1 m high acrylic column of 6.4 cm internal diameter. The transparent
column allowed observation of the media as the filtration process was in progress. Initially there
were a total of 5 ports for head measurement with 3 of them in the top 60 cm where the floating bed
was. After the filter run #18 the column was modified with 5 new ports being introduced. After this
modification the column had 8 piezometer taps in the top 60 cm with 4 piezometer ports for each
media layer. These ports are placed at 7.5 cm intervals. The details of the filter column with ports for
sampling and headloss measurement are shown in figure 3.2. Of the 5 sampling ports only one was
utilized after run #18.


-40-

The filter column was fixed vertically with raw turbid water entering from the bottom, for the
upflow mode of operation.


(b) Filter Bed

The filter bed was a 0.6 m deep packed bed w granular media. The filter bed had a 30 cm fine
media layer over a 30 cm coarse media layer for experiments with floating media. For the
experiments with sand medium the bed consisted of a 60 cm height of uniform sand. The media sizes
are given in section 3.3.2.


(c) Sampling Arrangement

The sampling and piezometer ports were copper tubes of 6 mm diameter. The mouths of
these tubes were covered with #50 mesh (i.e. 0.3mm mesh) to prevent loss of media. After the initial
experiments showed that continuous sampling along the filter bed disturbs it, only one sample was
taken from the center (at the port number 4 for the floating bed and at the port number 9 for the sand
bed). This location was selected so that in the dual media filter the center sample represented influent
quality to the fine media layer. The sample flow rate was controlled by the use of a pinchock clamp.
The sample flow rate was adjusted prior to the experiment to give the sample flow velocity equal to
the filtration velocity. This ensured that scouring did not occur and that the sample was
representative. Effluent too was sampled.


(d) Headloss measurement

The piezometer tubes were connected to the pressure taps using flexible plastic tubes of 6
mm internal diameter. The topmost pressure tap was connected to the piezometer No 1, and the other
taps in the descending order of height were connected to piezometers 2 to 9, the bottommost tap
being connected to piezometer no 10. The datum level for head measurements was 60 cm above the
ground level (i.e. 10 cm above the filter bottom).


The piezometer tubes were of 215 cm height. The static head (corresponding to the water
level in the constant head tank ) was 205 cm. This was the maximum value of head loss allowable in
this series of filtration experiments.


(e) Flow Pattern and Control

All experiments were conducted in the upward flow direction with constant rate. A pinchock
clamp (valve 13) was used to control the flow rate. The valve was adjusted frequently to maintain the
-41-

required flow rate. The flow rates were 15, 12.5, 10, 7.5 & 5 m/h. For the lower flow rates (10, 7.5 &
5 m/h) the flow rate was measured using a stop watch and a measuring cylinder.


The flow entered the filter from the bottom and exited from valve 12 (in fig 3.1) and was
wasted to the drain through the rotameter (10-60 l/h).


3.1.4 Backwash System.

The backwash procedure consisted of sending compressed air in upflow direction at 100 m/h
for 2 minutes followed by water wash for 3 minutes at 50 m/h. For floating media the water
backwash was downwards achieving 65% expansion of the filter bed ( 30% expansion in polystyrene
and 100% expansion in polypropylene).There was no intermixing of media during or after the
backwash. For sand medium the water rinse too was upflow. Cocurrent wash for upflow sand filter
had been recommended by HAMANN & McKINNY (1968) too.

The backwash air and water was obtained from the tap water supply in the Ambient
Laboratory. Both backwash air and water flow rates were measured using rate indicators. The flow
rate meter for air had the range 0 - 25 l/min. The backwash water rate was measured from a rotameter
(Metric 14 XS, 2 - 200 l/h). A simple scale fitted along the filter column served as the media
expansion gauge. Visual observation of the bed expansion was complemented in some instances by
photographic recording too.


After a filter run, the valve 7 (see fig 3.1) was closed and remained so until the backwash
operation was complete. The backwash air entered from the valve 8 and exited from valve 12. The
valves 15 and 16 served either as inlet or outlet for the backwash water depending on the direction of
backwash water flow.


3.2 Experimental Runs

The filter runs #1 to #18 were conducted to determine an optimum media arrangement as
well as hydraulic conditions, chemical dosing etc. The data of these runs are presented in Appendix
2. After reviewing the results of these runs the filter column which initially had 5 piezometer ports,
was modified by addition of 5 more. Also after run #18 the center sampling was limited to sampling
the influent to the fine media layer only. Alum was added as primary coagulant from run #15. The
summary of the filter runs (Runs #19 to #48; different combinations) is shown in the Table 3.1. The
filtration rates were adjusted frequently to the given value by adjusting the control valve.

Table 3.1 Summary of the Filter Runs

-42-

Media NTU Surface Loading m
3
/m
2
.h
15 12.5 10 7.5 5
Dual Media
PP 3.66 mm
PS 1.54 mm spherical
30 #19 #20 #21 #22 #30
60 #23 #24 #25 #26 #27
90 #28
#29
Dual Media
PP 3.66 mm
PS 1.54 mm angular
30 #31 #32 #33 #34 #35
60 #36
90 #37
Dual Media
PP 2.57 mm
PS 1.54 mm spherical
30 #38 #39 #40 #41 #42
60 #44
90
Dual Media
PP 2.57 mm
PS 1.54 mm angular
30 #43
60
90
Single Medium
Silica Sand 1 mm
30 #45 #46
60 #48
90 #47

#denotes a filter run
* PP =polypropylene, PS =polystyrene
* All runs upflow
* Run #28 with 0.5 mg/l Catfloc T2, Run #29 with 0.2 mg/l Catfloc T2
* Data on Runs #1 to #18 presented in Appendix 2.

3.3 Materials

3.3.1 Artificial suspension

The turbid raw water utilized in this experimental study was an artificial suspension of
kaoline clay dissolved in AIT tap water. The turbidity of the suspension was kept at 30, 60 or 90
NTU. The usage of an artificial suspension facilitated the comparison of various parameters while
-43-

keeping the other conditions uniform.


The mean particle size (D
50
) of the Kaoline clay was 3.3 microns.


Before pouring into the raw water tank Kaoline clay was thoroughly mixed in a beaker, using
a magnetic stirrer. This prevented the sticking of kaoline clay and made it uniformly distributed.
Continuous stirring arrangement was provided in the raw water tank to prevent the settling of the
suspension and thereby keep the influent water quality constant.

The concentration of Kaoline clay required to get a given turbidity varied linearly with the
NTU value and confirmed the result obtained by LO (1984). This relationship is given in figure 3.3.




















Fig 3.3 Turbidity vs Dosage of Kaoline Clay

3.3.2 Filter Media

Figures 3.4 to 3.9 show the 6 types of media utilized in these experiments. The data about
these filter media are summarized in Table 3.2. For dual media runs polystyrene was the fine medium
and polypropylene was the coarse medium.

Table 3.2 Filter Media Characteristics

-44-

Media Size
(mm)
Specific
Gravity
Source / Method of Preparation
Polypropylene 4.36 0.903 Obtained from France
Polypropylene
**
3.66 0.903 Commercially available in Bangkok
Polypropylene 2.54 0.903 Crushing the 3.66 mm beads and sieving.
Polystyrene
(spherical)
1.54 0.05 Obtained from France.
Polystyrene
(angular)
1.54 0.24 Crushed the raw material
*
.
Boiled in water for 1 minute at 100
0
C. The
resulting polystyrene was dried in the sun and
sieved to obtain the required fractions.
Silica Sand 1.00 2.70 Washed sand, dried and sieved. The sand utilized
was passing the #16 (1.18mm) sieve and was
retained in the #20 (0.85mm) sieve.
Geometric Mean Size =1 mm

*
The Raw material was Singlite expandable polystyrene grade 143 S, Manufactured by Thai
Petrochemical Industry Co. Ltd. of Bangkok, Thailand. This is a high strength polystyrene. The
density before expansion is 1040 kg/m
3
. Composition is 93% - 97% polystyrene and 7% - 3%
blowing agent. This material is recyclable. The cell structure is uniform after expansion. Polystyrene
does not dissolve in hot water, sea water, weak acids or inorganic hydroxides. Hence it's chemical
composition is not affected in a drinking water treatment operation.


**
Polypropylene commercially available was PRO-FAX 6331 manufactured by the HMC polymers
Ltd. of Bangkok, a licensee of HIMONT Inc. of USA. This polypropylene is a general purpose
homopolymer resin. It meets all requirements of the US Food and Drugs Administration as specified
in the code of federal regulations, Title 21, Section 177, 1500, covering safe use of polyolefin articles
and components of articles intended for food contact uses. The hardness of the media is 97 in the
Rockwell R scale. The deflection temperature at 455 kPa is 96
0
C.









-45-












Fig. 3.4 Polypropylene 4.36 mm






















Fig. 3.5 Polypropylene 3.66 mm








-46-












Fig. 3.6 Polypropylene 2.57 mm























Fig. 3.7 Polystyrene 1.54 mm (spherical)







-47-













Fig. 3.8 Polystyrene 1.54 mm (angular)























Fig. 3.9 Silica Sand 1mm

3.3.3 Coagulant

20 mg/l of alum was added to the solution preparation tank. The alum for the experiments
was obtained in the liquid form (50% solution).

-48-


3.3.4 Flocculent

The flocculent utilized was CatFloc T 2 manufactured by Calgon Corporation of Pittsburgh,
Pennsylvania, USA. This has a molecular weight of 250,000. The US EPA potable water approval
limit for Catfloc T is 30 ppm as product. Hence the dosage of 0.5 ppm was within the specified
standards.


The dosages of CatFloc T2 (as determined by Jar Tests) were 0.5, 0.5, and 0.2 mg/l for 30, 60
and 90 NTU respectively. Since the runs for 60 NTU were not very successful a CatFloc T2 dosage
of 0.03 mg/l was used for the last two runs with 60 NTU.


During the Jar tests to determine the optimum CatFloc T2 dosages the alum dosages were
kept constant at 20 mg/l.


The normal sequence of a Jar Test was
5 minutes @ 100 rpm
5 minutes @ 50 rpm
10 minutes @ 20 rpm
1.5 minutes settling



3.4 Measurements

3.4.1 Turbidity Measurement:

The influent turbidity was kept constant at 30 60 or 90 NTU. The influent, center and
effluent turbidity values were measured using a HACH model 2100 A turbidimeter. In the first one
hour grab samples were obtained at 5
th
, 10
th
, 15
th
, 30
th
, 45
th
and 60
th
minutes. After one hour the
samples were obtained hourly. Most of the filter runs lasted 30 to 40 hours (to reach terminal head
loss or turbidity breakthrough).


Due to the moderate hardness of the AIT tap water, the sampling cell of the turbidimeter had
to be cleaned every 3 hours to prevent errors being introduced in the reading due to deposited solids.
The cell was cleaned with dilute HCl acid and soap. The cleaning of the cell was most important
when the effluent turbidity was less than 1 NTU.


-49-

3.4.2 Headloss

The values of the available head were read from the piezometers every time just after
sampling. Later the headloss was calculated as the difference between this value and the initial head
loss.


3.4.3 Flow Measurement

Filtration rate range was 5 - 15 m/h. A rotameter (Metric 7p, 0- 60 l/h ) was used to measure
the filtration rate. For lower filtration rates the rate was verified using a stop watch and a measuring
cylinder. The backwash flow was through a rotameter (Metric 14 X).


3.4.4 Porosity Measurement

The porosity was measured using the simple relation

Volume of voids
Porosity =--------------------------
Total Volume


500 ml dry sample of the floating media was put in to a 1000 ml graduated cylinder

Table 3.3 Porosity values for different media

Media Porosity
Polypropylene (4.36 mm) 0.33
Polypropylene (3.67 mm) 0.34
Polypropylene (2.57 mm) 0.44
Polystyrene (1.54 mm, spherical) 0.34
Polystyrene (1.54 mm, angular) 0.35
Silica Sand (1mm,angular) 0.41
and this was restrained by a #50 mesh. Then a volume of water was added using 50 ml and 10 ml
pipettes until the water level reached the mesh (at 500 ml mark). This volume of water gives the
initial volume of voids. The water was added slowly, along the wall of the cylinder so as to allow all
air to escape. To help ascertain the level of water, blue ink was added to the water used. Since all the
synthetic media was white in color the water meniscus was clearly identifiable.
-50-


The values of porosity obtained were given in Table 3.3


3.4.5 Particle Size Analysis

The particle size of the Kaoline clay was analyzed using the Shimadzu centrifugal particle
size analyzer Model SA - CP2. This result is given in Table 3.4.

TABLE 3.4 Particle Sizes of Kaoline Clay

Diameter (micron) Weight Percentage
0 - 2 25.1
2 - 3 19.6
3 - 4 16.2
4 - 5 13.0
5 - 6 10.9
6 - 8 15.2

-51-

CHAPTER 4



PRESENTATION AND CRITICAL DISCUSSION OF RESULTS



4.1 Introduction

The Floto Filter was primarily different from other filters due to its floating bed. Several
combinations of synthetic plastic media whose specific gravity was <1 were used for experimental
runs in this filter. Four runs with sand medium were used for comparison. The first part of this
chapter will focus on the experimental analysis while the discussion on model study is given in the
latter part in section 4.8.


4.2 Experiments with Floating Media

4.2.1 Media Selection

The experiments to select an optimum media combination was carried out using media like
polyethylene, polypropylene and polystyrene of various sizes. Polyethylene was rejected as it
intermixed with both polypropylene and polystyrene. The combination of polypropylene and
polystyrene was selected as these media did not intermix even when agitated using compressed air
and water. The polypropylene (larger particles), being denser, stayed at the bottom of the floating
bed. This enabled the operation of the filter in the coarse to fine arrangement. With coarse medium at
the bottom and the fine medium on top, the flow direction had to be upflow. Upflow filtration with
floating media has the added advantage of the filtration being in the direction of grain compression.


4.2.2 Media Preparation

The method of preparation of the materials that were prepared at the AIT was given in
section 3.3.2. In addition to the 1.54 mm angular polystyrene a smaller polystyrene of geometric
mean size 1 mm (passing #16 mesh and retained in #20) was also obtained. When this was used in
the dual media filter (coarse medium being 3.66 mm polypropylene) the experiment was not
successful as the media intermixed due to non uniform density of polystyrene. The proportion of
fraction heavier than polypropylene, fraction intermixing with polypropylene and fraction lighter
than polypropylene was approximately 25% : 25% : 50%. This experimental result is shown in figure
4.1.


-52-


































Fig 4.1 Experiments With Media
Intermixing of 1 mm Polystyrene With Polypropylene



4.2.3 Summaries of Experimental Runs

The experiments were carried out for three values of influent turbidity, namely 30, 60 and 90
NTU. The results are presented in the tables 4.1, 4.2 and 4.3. Majority of the experiments were
carried out for 30 NTU influent turbidity.
-53-


Table 4.1 Summary on 30 NTU runs

Media Run# Break through
time
(hours)
Lowest NTU
achieved
Maximum
headloss at
break
through
Reason for termination of the run
PP 3.66 mm
PS 1.54 mm(spherical)
#19 8.5 7 NTU 35 cm Turbidity Breakthrough
#20 22.8 4.5 NTU 110 cm Turbidity Breakthrough, run continued till
terminal headloss
#21 27.5 4.7 NTU 105 cm -do-
#22 22.1 5.5 NTU 65 cm Turbidity Breakthrough
#30 More Than
40 hours
2.7 NTU Terminal Head Loss (200 cm)
PP 3.66 mm
PS 1.54 mm(angular)
#31 11 NTU Poor Removal
#32 3 NTU Poor Removal
#33 12 4.8 NTU 55 cm Turbidity Breakthrough
#34 More Than 28
hours
2.9 NTU Terminal Head Loss

#35 34 hours 2.0 NTU 150 cm Turbidity Breakthrough (Terminal
headloss)
PP 2.57 mm
PS 1.54 mm(spherical)
#38 5 2.6 NTU 10 cm Turbidity Breakthrough
#39 7 3.1 NTU 20 cm -do-
#40 15 1.4 NTU 45 cm -do-
#41 28 2.5 NTU 74 cm -do-
#42 More Than
41.5 hours
1.5 NTU Head loss at 34 hrs was only 60 cm.
Run was terminated as the bed was
disturbed due to flow adjustment at
41.5 hours.
PP 2.57 mm
PS 1.54 mm(angular)
#43 28 hours 0.81 NTU 55 cm Head loss at 45 hrs was 144 cm. Run
terminated at 45 hrs as the bed was
disturbed due to flow adjustment
Silica Sand 1 mm #45 13 hrs 0.55 NTU Terminal Head Loss at 13hrs
#46 More Than 23 0.15 NTU Terminal Head Loss at 23hrs
-54-

hours

* PP =polypropylene * PS =polystyrene







Table 4.2 Summary of 60 NTU runs

Media & (Polymer Dosage) Run Break
through
Time
(hours)
Lowest
NTU
achieved
Maximum
headloss at
break
through
Reason for termination of the run
PP 3.66 mm
PS 1.54 mm(Spherical)

(Cat floc T2 0.5 mg/l)
#23 11 NTU Poor Removal
#24 7 NTU Poor Removal
#25 9.5 NTU Poor Removal
#26 18 NTU Poor Removal
#27 29 3.8 NTU 81.5 cm Turbidity Breakthrough
PP 3.66 mm
PS 1.54 mm(Angular)
(Cat Floc T2 0.5 mg/l)
#36 5.5 1.3 NTU 9 cm Turbidity Breakthrough
PP 2.57 mm
PS 1.54 mm(Spherical)
(Cat floc T2 0.03 mg/l)
#44 3.5 1.9 NTU 5 cm Turbidity Breakthrough
Silica Sand 1 mm
(Cat floc T2 0.03 mg/l)
#48 3.75 0.22 NTU 24.5 cm Turbidity Breakthrough

PS =polystyrene PP =polypropylene









-55-

Table 4.3 Summary of 90 NTU runs

Media Run Break
through
Time
(hours)
Lowest
NTU
achieved
Maximum
headloss at
break
through
Reason for termination of the run
PP 3.66 mm
PS 1.54 mm(spherical)
#28 28 Poor removal
#29 6 5.5 8 cm Turbidity Breakthrough
PP 3.66 mm
PS 1.54 mm(angular)
#37 4.5 1.5 10 cm Turbidity Breakthrough
Silica Sand 1 mm #47 3.5 0.35 22.5 cm Turbidity Breakthrough.

*PS =Polystyrene
*PP =Polypropylene
*Catfloc T2 dosage for run#28 was 0.5 mg/l while for run#29 was 0.2 mg/l.




4.2.4 Graphical Presentation - Concentration and Headloss Profiles

The headloss development along the filter is presented in the figures 4.2 to 4.13. The
ordinates show the position of the piezometer tapping points in the filter and the abscissas show the
head loss. The curves are for a (stated) specific time. The curves for dual media (floating bed) filter
show a characteristic curve which is easily identifiable from the curve for sand. The curves for dual
media filter runs show that as the interface of the coarse and fine media layers is passed the headloss
increases rapidly. While figures 4.2 to 4.12 depict longer, successful runs, figure 4.13 is for run#33
which did not perform well. Analyzing this figure, we can see the lower headloss development,
which indicates that the failure was due to media not capturing the floc. This may be due to higher
flow velocity (10 m/h) or due to flocculation not taking place. Also when fig.4.13 is compared with
another figure showing lower total headloss development (say fig.4.9) we can notice a higher
headloss within the coarse media layer for the successful run.


The concentration profiles together with relevant headloss profiles for runs #19 to #48 are
presented in Appendix 1.


4.2.5 Headloss Development and Breakthrough Behavior

The head measurements were piezometric head measured from a common datum. The total
-56-

head loss at any given time was the sum of the head in the inlet system and the head loss in the bed.
As the clean bed head loss is a function of flow velocity, higher initial headlosses were observed for
higher filtration velocities. The initial headlosses for the five filtration velocities used in this series of
experiments are shown in table 4.4.

Table 4.4 Initial Headlosses for Different Filtration Velocities.

Flow Rate (m/h) Average Value of Initial
Headloss (cm)
15.0 25
12.5 17
10.0 12
7.5 7
5.0 5

For a threefold increase in the flow rate, the initial headloss in the floating bed filter has
increased fivefold. Therefore the initial headloss is not linearly related to the rate of filtration. SHEA
et al (1971) had also reported a similar phenomenon for sand filters.


Since the entry losses remain constant for constant filtration velocity the subsequent increase
of head is a direct result of particle capture. The particles captured by the media may aid in
subsequent collection of particles. CLARK et al (1992) and ADIN and REBHUN (1974) have
reported that increased flow velocity result in lower additional headloss. But this result should be
seen only in the view of lower particle retention with higher filtration velocity.














-69-














Fig.4.14 Headloss and Effluent Quality for run #30

For Floto Filter the headloss vs time graphs (given in figure 4.14 and appendix 1) show an
exponential type of headloss development. As seen in fig.4.14, in experiment #30, the head loss in
first 20 hours of operation was only 52 cm whereas it was 150 cm in the next 20 hours of operation.
This exponential headloss curve may be due to the partially fluidized media matrix allowing the
deposited material to be compressed. A slight shortening of the filter bed (about 1.5%) was observed
when the filter runs exceeded 30 hours of operation.


The breakthrough was generally accompanied by a sudden sharp increase of head loss as can
be seen in headloss profiles for runs #20, 21, 22, 35, 41 and 43. These headloss profiles are shown in
Appendix 1.


For most of the (dual media) filter runs, the head development within the coarse media layer
(polypropylene) was linear whereas the fine media layer showed exponential headloss development.
From the higher head loss in the fine media we can assume that most of the removal occurs in the
fine media layer. SHEA et al (1971) had also noted that media zones that exhibit large slope in
piezometric head profiles are zones where concentrated floc accumulation has occurred. The
headloss development in the coarse media is linear even for 40 hour filter runs.


In the runs that failed or showed an early floc breakthrough, the headloss development was
minimal. This may be due to the flocculation not taking place properly and the formed floc particles
not being strong enough to form aggregates that could be retained within the pores of the filter
media.


The headloss development for floating media takes a much slower rate than for sand. The
-70-

possible causes for this are discussed in section 4.6.


4.3 Effect of Physical Parameters on Filter Runs

4.3.1 Influent Turbidity

As shown in Table 4.3, the filter did not perform well for 90 NTU influent turbidity even
though lowest (5 m/h) velocity was utilized for all the runs. The runs yielded good results only for
short durations (generally less than 5 hours). For 60 NTU influent turbidity, only run #27 (flow
velocity 5 m/h, PP 3.66mm, PS 1.54mm, spherical) gave good results.


4.3.2 Filtration Rate

Both sand and the floating media performed well at filtration rates lower than 7.5 m/h with 5
m/h being the best. This filtration rate is comparable with the filtration rate of a conventional rapid
sand filter. Only one media combination (PP 3.66 mm and PS 1.54mm spherical media) gave
reasonably good filtrate for filtration rates 10 m/h and 12.5 m/h.


Since the higher filtration velocities push the particles deeper into the bed, the bed height of
60 cm utilized in this series of experiments may not be sufficient to retain the particles at higher
velocities.


When the results of successful runs with different filtration velocities were compared for the
same mass balance, the turbidity and headloss results showed reasonable agreement. This is
especially evident when comparing run #34 with #35 and #41 with #42. After 30 hours of operation
the headloss for run 41 (7.5 m/h) was 90 cm and for run #42 (5 m/h) was 60 cm. When comparing
runs #22 and #30, run #30 (5 m/h) reached terminal headloss at 42 hours while still giving good
quality effluent whereas run #20 (7.5 m/h) reached terminal headloss at approximately the same time
after giving good quality filtrate only upto 22 hours.


4.3.3 Media Size

As expected the smaller size coarse media gave better results than the larger size. This is
evident when we compare the runs #22 & #41 (fig 4.15) and #30 & #42 (fig 4.16). Experimental run
#41 performed well for a longer period than #22 before breakthrough occurred. The headloss
development in the fine media was also less in run #41, indicating that some of the floc particles had
been successfully removed by the 2.57 mm coarse media. Runs #30 and #42 gave almost identical
quality but for run #42 the maximum headloss, which represents particle capture by fine media, was
-71-

much less. This confirms that the 2.57 mm coarse media had been successful in particle capture than
3.66 mm media. The result that capturing a particle with coarse media causes lower additional head
than capturing it with fine media, confirm the findings by CLARK et al (1992).



4.3.4 Media Shape (Angular vs Spherical Media)

Angular fine media performed comparably with the spherical fine media but the head loss
build up was more rapid for angular media. Figure 4.17 compares headloss and effluent quality of
runs #42 and #43. The combination of spherical polystyrene 1.54 mm with polypropylene 3.66 mm
was the only combination to give long filter runs at 12.5 m/h and 10 m/h. A comparison of filter runs
using spherical and angular fine media is given in Table 4.5.


Table 4.5 Comparison of Results for Angular and Spherical Media

Compared Filter Runs Lower headloss
given by
Acceptable quality given
for a longer period by
spherical angular
#22 #34 Spherical Media Angular Media
#30 #35 Spherical Media Spherical Media
#42 #43 Spherical Media Spherical Media
#27 #36 Spherical Media Spherical Media




Depending on this result we can safely conclude that spherical polystyrene performs better
than angular polystyrene for conditions simulated in this series of experiments. However it should be
mentioned that the headloss curves for both shapes of media followed the same pattern of
development.


4.3.5 Terminal Head Loss

The terminal headloss of the system was 205 cm which facilitated longer runs before
reaching the breakthrough. This value is comparable with a conventional water treatment plant.



-75-

4.4 Polymer Dosage and Mixing Time of Polymers

ADIN and REBHUN (1974) had shown that for equal hydraulic conditions cationic polymer
is more efficient in removal ability than a conventional hydrolytic flocculent like alum. In these
experiments the cationic polymer CatFloc T2 was used as a filtration aid in the contact flocculation
filtration. The optimum dosages as determined by the Jar Test were used for polymer dosing.


The dosage required for 60 NTU runs was 0.5 mg/l but generally this did not give good
results. For the last two runs with 60 NTU influent turbidity (i.e runs #44 and #48), a polymer dosage
of only 0.03 mg/l was applied. This lower dosage was selected by conducting Jar tests using a 0.1%
alum solution as given in section 4.5.3. A very dilute CatFloc T2 stock solution had to be utilized due
to the limitations in the available dosing pumps. (The manufacturer recommended a stock solution
concentration of 1%). In the preparation of the polymer stock solution, the degree of mixing the
polymer with water seemed to have a direct bearing on the results, the longer the mixing period, the
better the results.( The manufacturer's literature says that the polymer readily mixes with water).


Jar tests carried out using 10 hour old stock solution of Catfloc T2 gave results comparable to
that with fresh solution. Since the stock solution was replenished every 6 hours any effects due to
hydrolysis of the polymer were not anticipated.


4.5 Effect of Controls and Processes

4.5.1 Flow Rate Controllers

The approach velocity was kept constant throughout the filter run by means of a pinchock
clamp. Filter effluent quality was adversely affected by changes in the rate of flow, even though care
was taken to open the valve slowly. The disturbances mostly occurred when the filter had been in
operation for a long time (i.e. when the bed was loaded with removed floc). When the bed was
disturbed, a large amount of accumulated floc were released into the effluent stream resulting in very
high effluent turbidity. After such a bed disturbance the filtrate quality returned back to the earlier
quality (or a better quality) within 10 minutes. This may be due to the fine media again capturing the
floc, within the free space created due to the floc release after the bed disturbance.

4.5.2 Backwashing

A valve arrangement was utilized to effect compressed air backwashing which was followed
by the water backwash. The air was supplied in the upflow (cocurrent) direction at a velocity of 100
m/h for a period of 2 minutes. The water was then sent downwards (countercurrent) for a period of 3
minutes, achieving 65% expansion of the filter bed ( 30% expansion in polystyrene and 100%
expansion in polypropylene). The air wash duration is similar to that used in a conventional rapid
-76-

gravity filter. The water wash velocity of 50 m/h used here is less than the value of 80 - 100 m/h
specified (for polystyrene) by HABERER and SCHMIDT (1991) but achieved the same (30%) bed
expansion as specified by them. But it has to be kept in mind that the floto filter bed was shorter and
of a different composition than their filter bed.


The backwash process is shown in Figures 4.18 to 4.21. As evident from these figures the
media interface was clearly defined even during backwashing. Also there was no intermixing of
media after the backwash. This was the most prominent advantage of the















Fig. 4.18, Floto
Filter at the End
of a Run














Fig 4.19, Backwash With
Air. Note the color
-77-

Change in Fine Media.
























Fig.4.20, Backwash
With Water
















-78-



Fig.4.21, Filter
After the Backwash








floating media arrangement as it eliminates the frequent problem of intermixing, faced by other types
of dual media arrangements. In dual media filters using other media like sand and anthracite, there is
some intermixing in the interface zone. This zone (which can be about 15 cm) has a smaller porosity
and therefore promotes the accumulation of floc in the interface, necessitating an auxiliary scouring
system like subsurface washing.



In the floating bed filter, the air flow agitated the coarser medium (polypropylene) very well
while the polystyrene beads did not agitate. But the floc being pushed away from the fine media due
to the air flow was apparent from the change in the color of polystyrene bed (from brown to white,
see Fig. 4.19). Using this phenomenon, a backwash system which provides a small amount of water
simultaneously with the air backwash would be able to achieve complete cleaning of the bed, without
using a large amount of water. This backwash water can be sent from the bottom or laterally from the
walls.


Both floating media agitated very well in the countercurrent water rinse. Even without an
airflow, the water rinse was sufficient to clean the bed. In a practical situation where there is some
storage of filtered water above the filter bed, it can be utilized for backwashing.


In this experimental unit, countercurrent (i.e downflow) backwashing was impossible
immediately after the air wash, as the air pressure built up inside the filter due to the absence of an air
release mechanism. This high pressure air pocket prevented the flow of water into the filter. Sending
water into the filter against this high pressure would have damaged the apparatus. This problem was
overcome by filling the bed slowly with water flowing upwards, releasing the air through the valves
on top of the filter, and then sending the backwash water. This fault has to be rectified in a future
design by introducing a device to release air from the filter. Also a future design should have a valve
arrangement which should facilitate simultaneous air and water wash as simultaneous air/water
backwash was impossible in the filter used for these experiments.
-79-



4.5.3 Effectiveness of Jar Tests

CLARK et al.(1992) who did contact flocculation filtration experiments using CatFloc T
have verified that the turbidity measurement after jar tests (or 0.45 membrane filtration) gave the
best indication of the optimal dosage for particle destabilization. They also cite previous investigators
to show the correspondence between the optimal polymer dosage in Jar Tests and that in filtration.

Yet in this set of experiments, the optimum polymer dosages determined in the jar tests
sometimes did not give very good results especially for runs with higher influent turbidity (See
Tables 4.2 and 4.3). This may be due to the Cat Floc T2 stock solution being very much more dilute
than the manufacturer's recommendation.


The 50% strength alum solution utilized for alum dosing was directly used in the jar test
experiments too. KAWAMURA (1991) has stated that a freshly made 0.1% alum solution is more
suited for jar tests. Also the feeding conditions of chemicals (intermittent vs continuous feeding) and
the differences in the actual mixing conditions might have affected the correlation between jar test
results and the plant performance.


4.5.4 Effectiveness of Turbidity Measurement for Quantifying Filter Performance.

The influent turbidity was kept constant at 30, 60 or 90 NTU. The efficiency of the filter in
removing the suspended solids was basically assessed on the turbidity of the effluent. The
experiments by CLARK et al. (1992) show that there exists a good correlation between the
concentration of suspended matter and the turbidity of a suspension.



4.6 Experiments With Sand Medium

The summary of the four filter runs using silica sand were shown in tables 4.1, 4.2, and 4.3.


These comparative runs with a monomedium bed of silica sand gave very good filtrate
quality as expected but the runs were much shorter, reaching the terminal head loss quickly. For 15
m/h flow rate even the sand did not perform well with partial fluidization occurring in the bottom 5
cm, after 6 hours of operation. This is shown in figure 4.22.


For sand medium cocurrent (i.e. upflow) backwash with water (preceded by air) seemed the
-80-

better method of backwashing. This obtained fluidization of bed easily. This also had the additional
advantage that the bed naturally graded itself to coarse to fine arrangement after the water flow was
stopped. Also since the backwash direction was upflow, backwash water could be sent immediately
into the filter after the air wash, as the two valves on the top of the filter facilitated air release during
the water entry.






































-81-

Fig 4.22 Fluidization of Sand Bed During the Run


The best run with the floating media is compared with the best run with sand media in the
figure 4.23. The headloss shown is the maximum headloss for both cases. The filter runs with sand
gave relatively better product water quality but the headloss development was much more rapid
reaching the terminal headloss much faster than floating media. One reason for this may be the larger
size of the floating media. Also the dual media arrangement and the partially fluidized nature of the
floating bed may have allowed the fine floc to penetrate deeper into the bed and thereby utilize the
filter bed more efficiently, resulting in lower headloss. Another reason may be that the sand bed was
in contact with the water inlet as compared to the bottom of the floating bed which was 40 cm above
the water inlet. The additional contact time thus available for the floating filter might have ensured
the formation of larger floc which could be retained in the coarse floating media. Also some larger
flocs settled without reaching the floating bed as could be observed from the deposition on the lower
mesh.


4.7 Comparison of Experimental Runs and Concluding Remarks

The comparison of Floto Filter results with the results for sand runs is given for the 30, 60
and 90 NTU influent turbidities in Tables 4.1, 4.2 and 4.3 respectively.


In contact flocculation filtration using sand media, one of the disadvantages is the rapid
clogging of the filter which results in shorter filter runs and higher frequency of backwash. The Floto
filter is advantageous in both these aspects.


Both sand and the floating media performed well at surface loadings lower than 7.5 m
3
/m
2.
h
with 5 m
3
/m
2.
h being the best. The major advantage of the floating media was evident from the fact
that they could run even beyond 42 hours without reaching terminal quality or terminal head loss.
The fluidization of the bed in contact with the underdrain as seen in sand (for run#45) will not occur
in a Floto Filter as it has a buffer zone of water between the water inlet and the bottom of the filter
bed.


The advancement of the filtration front was visually observable for runs with sand media. For
synthetic media, the advance of a front was not noticeable. This may be due to floc being retained
throughout the filter bed. For synthetic media bed a change of color from white to brown was
noticeable as the run progressed but again this color was even for the whole bed. This result shows
that the removal mechanisms are different for the sand filter and the floating filter.


-83-

The sand medium used the same amount of backwash air as floating media but needed a
relatively larger amount of water for backwashing. While the floating media could be backwashed
with water alone, the sand always needed air, to break up the piston effect. The result of attempting
to backwash the sand with water alone can be seen from figure 4.24. DIAPER and IVES (1965) have
attributed the piston effect seen in upflow wash of sand, to the arching effect due to the clogged
media forming an impervious barrier. Considering the buoyant nature and the low affinity of the
floating media particles to each other, formation of such an impervious front in the Floating Filter is
not possible.






























Fig 4.24 Attempt to Backwash the Sand Filter With Air
Note the Impervious Front


A typical dual media filter needs a fluidization wash at the end of the backwash cycle, to
-84-

restratify the media. Due to the large difference in the densities of two media used in this experiment,
such a restratifying wash is not required.


When considering filter ripening (the period over which filtrate quality generally improves),
the sand filter runs for 5 m/h flow velocity ripened within the first 15 minutes of operation.

Quick filter ripening for synthetic media was observed in runs #41, #34 and #18 (all with 7.5
m/h flow velocity). For other long runs with synthetic media (e.g. #20, #21, #27, #30, #42,and #43)
the filter ripening took from 30 minutes to 2 hours. In a practical operation the product during this
period can either be recycled or wasted, but it is not an attractive solution. Hence the filter operation
has to be optimized to have ripening periods less than 15 minutes. Since the filter ripened quickly for
the higher (7.5 m/h) flow rate, operating first at a high rate and then gradually coming back to the 5
m/h rate may bring in faster ripening.


CLARK et al (1992) reported that distilled water was pumped with polymer injection, for a
minimum of one hour prior to their filtration experiment. This has been done to ensure that any
ripening observed in the experiments were due to particles captured in the media and not to changes
in the surface chemistry of the media, caused by the polymer after the experiment began. Coating the
filter media by polymer had been reported by HANEY and STEIMLE (1974) and O'MELIA and
ALI (1978) too. The durations employed by them had been 30 minutes and 15 minutes respectively.
ADIN and REBHUN (1974) also observed that coating the media with flocculent increases the
chances of efficient attachment. SHEA et al (1971) had indicated that addition of 20 mg/l of alum
would provide the necessary coating to the media in making it effective in collecting the incoming
floc. Since 20 mg/l of alum was added, coating the media with polymer was not attempted in this
series of experiments. Yet coating of media with polymer may provide an answer to the problem of
hastening of the filter ripening period.


The floto filter gave an increased length of run to a given headloss. The coarse to fine
arrangement allowed the removal of floc along the depth of the filter. This lower head loss and
higher capacity for retention of suspended solids is an inherent advantage of dual media filters.


Most of the runs utilizing the floating media had to be terminated (often after 30 or 40 hours
of operation) as the media beds were disturbed when the flow was adjusted. This happened despite
precautions being taken to minimize rapid changes in flow rate (Screw type flow control valve was
used and the valve was opened slowly). Hence a flow control device which does not introduce shock
loads into the filter should be incorporated into future experiments.


An upper retention grid is needed in this type of filter to prevent the loss of filter media. This
-85-

grid should be strong enough to withstand the buoyant force exerted by the floating polystyrene.


In the course of experiments, some floc deposition was observed on the upper retention grid
too. This showed that flocculation occurred even after the water left the media bed. Perhaps a much
deeper filter bed would have been more effective in capturing these floc and thereby improving the
filtrate quality further.


Also during some experiments, some large floc particles were observed in the conical inlet to
the filter, below the bottom mesh. Since this was a #50 mesh (i.e. 0.3 mm) it is obvious that floc
particles larger than 0.3mm can form at the inlet. In a practical operation the lower retention grid can
be omitted if necessary. This would allow these floc particles to go in. Also as discussed in section
4.6 if some large floc particles settle, absence of the grid will allow them to settle into the conical
bottom. In such a case water inlet can be placed somewhere at the base of the cone. If the bottom
mesh is removed, the available head will be improved due to the absence of a floc layer clogging
either side of the bottom mesh. However care should be taken to prevent excessive downward
expansion of media during the downflow backwash.


In a practical operating filter low operating costs will result from higher water production per
run and by using a low percentage of product water in backwashing. The experiments have shown
that the floating filter can provide acceptable quality water for more than 40 hours, at conventional
rapid sand filtration rates, without backwashing. The Floto Filter is also more attractive in that
backwash can be achieved with a lower water consumption.


Of the media combinations considered in this series of experiments the most desirable filter
bed consisted of 2.57 mm Polypropylene as coarse medium and spherical polystyrene beads of 1.54
mm as fine medium, both in terms of quality and head loss. For higher filtration rates (10 and 12.5
m/h) better runs were obtained with the combination of 3.66 mm Polypropylene and spherical 1.54
mm polystyrene. Since better results were expected with 2.57 mm polypropylene combinations, this
result has to be verified by further experiments.


For measuring the porosity of the floating media a method had to be improvised as a standard
reference for measuring the porosity of floating granular media was not available. Still the results
obtained show a good correlation with experimental results by CARMAN (1937) for sand media
which is given as Appendix 3. Accordingly, the crushed polypropylene showed the highest porosity
among the synthetic media. All the porosity values were given in section 3.4.4.



-88-

4.8 Mathematical Modelling

4.8.1 Introduction

The theoretical framework for the mathematical model utilized was given in section 2.10.
The computer program prepared by MANANDHAR (1990) based on this model was utilized in this
study to predict the Floto Filter performance. This program could simulate the performance of a
graded bed filter and a multimedia filter for varying influent concentration, varying particle size
distribution and constant and declining rate filtration. Due to the time constraints the model was used
as it is, without major modification. The results of this study are presented in section 4.8.2.


4.8.2 Model Calibration and Verification

This model considers the filter divided into infinitesimal depth increments with time
increments covering the whole run. The model was directly used for upflow filtration mode. Using
the equations 2.13 and 2.20 to describe the filter performance needs the assessment of the terms ,
,
p
, ,
2
and '. The values of
p
and are given as the product
p
. and are given
by the product .


According to CHANG (1989) and MANANDHAR (1990), the ripening period of filtration is
affected by and
p
. The breakthrough period is affected by
2
and headloss is affected by '.


Using the method utilized by O'MELIA and ALI (1978) and subsequent investigators, the
single collector removal efficiency , was calculated from the Floto filter data. The C
i
/C
0
curve
was extended to meet the time =0 axis (i.e clean filter condition) and this value was utilized to
calculate the n
i
/n
0
value. C
i
and C
0
are the particle concentration measured in terms of turbidity. n
i

and n
0
are the particle concentration in number per unit volume. This value substituted in equation
2.10, considering L as the depth of bed, gives the value. The C
i
/C
0
curves for the four filter
runs considered for simulation of the model are presented in the figures 4.25 to 4.28.


Once was established the values of
p
,
2
and ' were varied to obtain the profiles of
the experimental headloss and concentration profiles. The model was able to predict both the
headloss and concentration profiles for upflow filtration in sand media for 5 m/h filtration rate. This
result is shown in figures 4.29(a) and 4.29(b).





-89-
















Fig. 4.29(a), Run#46, Experimental and Calculated Headloss Profiles.
v =5 m/h, =.00106,
p
=.32,
2
=.008, ' =.007






















Fig. 4.29(b), Run#46, Experimental and Simulated Concentration Profiles
v =5 m/h, =.00106,
p
=.32,
2
=.008, ' =.007


-90-
























Fig 4.29(c), Run#45, Experimental and Simulated Headloss Profiles
v =15 m/h, =.00304,
p
=.0035,
2
=.02, ' =.039


Figure 4.29 (c) shows the headloss profiles for the run #45 which used sand media with 15
m/h filtration velocity. The model was unable to predict the concentration profile. The predicted
values were much lower than the actual.


The calculation of the value for the dual media filter posed a problem, as there are two
sets of values for porosity, media grain size and C
i
/C
0
, resulting in two values. The model could
not accept two values. Also simulation using two different values necessitated the actual
particle size of the influent to the fine media layer. Unfortunately this data was not available. As a
preliminary step an attempt was made to predict the filter behavior using the average value of the two
values. This assumption may be unrealistic but helped in identifying the other parameters
necessary to do a realistic modelling of the Floto filter. The simulation could give the headloss
profile but the predicted concentration profile differed from the actual. The values for two media
layers are shown in the Table 4.6. The headloss profiles calculated by the model are shown in
Figures 4.30 and 4.31.

-91-



















Fig. 4.30 Experimental and Simulated headloss profiles for run #30
v =5 m/h, =.00323,
p
=.25,
2
=.025, ' =.007





















Fig 4.31 Experimental and Computed Headloss Profiles for run #42.
v =5 m/h, =.00368,
p
=.28,
2
=.04, ' =.001
-92-



Table 4.6 values for coarse and fine media
Run #30 Run #42
for coarse media 0.00371 0.00504
for fine media 0.00274 0.00231


An attempt was made to simulate the behavior of coarse and fine media separately. The aim
was to predict the headloss and concentration profiles for the coarse media layer and then using this
as an influent value for simulation of the performance of the fine media layer. The results are shown
in figures 4.32 and 4.33.


























Fig. 4.32 Headloss Profiles for coarse Media, Run#30
v =5 m/h, =.00371,
p
=.30,
2
=.09, ' =.0001
-93-



























Fig. 4.33 Headloss Profiles for Coarse Media, Run#42
v =5 m/h, =.00504,
p
=.28,
2
=.2, ' =.001

Again the simulation was good for headloss prediction only, concentration predictions being
lower than the experimental values. The performance of the fine media layer could not be simulated
due to the absence of the particle size entering that layer. A future experiment should analyze the
particle size at the entry point to the fine media layer, in addition to the influent and effluent particle
sizes.


In all the simulations listed above, the influent particle size was considered as 3.3 microns,
which is the average size (50% size on weight basis) of the Kaoline clay particles. The model utilized
for the simulation (MANANDHAR, 1990) had the facility to individually simulate the removal of
different sized particles. But for each particle size the removal rate is different and each size has a
different C
i
/C
0
curve. The resulting n
i
/n
0
values mean that there is a particular value for each
particle size considered. Again a particle size analysis was necessary to get this data. Even in the
presence of such data, the effects of particle agglomeration due to flocculation makes the formulation
-94-

of a model a complex task. Also in the case of analyzing the removal of different particle sizes
individually, the model required individual
p
and
2
values but only one value of '.

In these simulations the shape factor of sand was taken as 7.5 while for 3.66mm PP, 2.57mm
PP and Spherical 1.54mm PS the shape factors were 6.1, 8.5 and 6.0 respectively. The media sizes
utilized were the geometric mean sizes.


Commenting on the deficiencies of data (viz. porosity, shape factors, densities etc.) for
modelling, MANANDHAR (1990) recommends the remedy of selecting realistic values depending
on past work. While for a sand filter this is possible, this remedy is inapplicable to the Floto filter.


MANANDHAR (1990) notes that the model parameters will also be affected by filtration
velocity. i.e. the same filter will have different model parameters for each individual velocity. He had
also noted that the predicted curves for smaller particle sizes were giving a much closer fit than for
the larger particles.


The values of
p
(contact efficiency of a retained particles) and (the fraction of coarser
particles that can act as particle collectors) were taken as unity due to the absence of better
information. These assumptions might have introduced errors to the simulations. Also the calculation
of C
i
/C
0
for the clean bed is approximate. If the C
i
/C
0
is not used, the initial contact efficiencies can
be calculated from the models (YAO, HAPPEL, RAJAGOPALAN and TIEN, LEE and GIESEKE)
given in table 2.3 too. MANANDHAR (1990) gives comparison of initial collection efficiency
values generated from these models with the experimentally calculated value, which only show
approximate agreement.


In these simulations the model showed a headloss increase with increasing '. Increasing the

2
only marginally increased the concentration profile but heavily reduced the headloss values.
Increasing
p
resulted in increase in headloss values, with a marginal change of the concentration
profile. The concentration profile showed a decrease with decreasing ' but this rate of decrease was
very low.



-95-

CHAPTER 5



CONCLUSIONS




The best combination of floating media for the Floto filter are polypropylene and
polystyrene. Due to the large density difference in these two media, they do not intermix.


Of the media combinations considered in this series of experiments, the most desirable filter
bed consisted of 2.57 mm Polypropylene as coarse medium and spherical polystyrene beads of 1.54
mm as fine medium, both in terms of better quality and lower head loss. For higher filtration rates (10
and 12.5 m/h) better runs were obtained with the combination of 3.66 mm Polypropylene and
spherical 1.54 mm polystyrene.


The method of preparation employed to make smaller size polystyrene has to be improved to
ensure that all grains have the same density. If a smaller size polystyrene is commercially available, it
should be utilized for future runs.


The floto filter gave an increased length of run to a given headloss. The coarse to fine
arrangement allowed the removal of floc along the depth of the filter. This lower head loss and
higher capacity for retention of suspended solids is an inherent advantage of dual media filters.


For a Floto Filter, the best flow direction is upflow as it is the direction of grain compression,
which aids in better removal of suspended solids.


The initial head loss for the floto filter show a non linear increase with increasing velocity.
For a threefold increase of flow velocity, the initial headloss has increased fivefold.


The partially fluidized media matrix of the floto filter allows compression of deposited floc.


In the Floto filter most of the floc removal occurs in the fine (polystyrene) media layer. The
smaller the particle size of coarse medium, the better the floc removal. The particle capture by the
-96-

coarser medium results in less additional head than for particle capture by fine medium.


The filter was very successful in treating the water of 30 NTU turbidity. For higher turbidities
it did not perform well, probably due to inadequate flocculation.


Both sand and the floating media performed well at surface loadings lower than 7.5 m
3
/m
2.
h
with 5 m
3
/m
2.
h being the best. This velocity compares with the filtration rate of a conventional rapid
sand filter.


The headloss development curves for floating filter takes a characteristic shape easily
identifiable from the curve for sand filter.


The sand retained the flocs in an advancing front, whereas the floc removal in the floating
media was spread throughout the bed.


In contact flocculation filtration using sand media, one of the disadvantages is the rapid
clogging of the filter which results in shorter filter runs and higher frequency of backwash. High
operating costs will result from lower water production per run and usage of a high percentage of
product water in backwashing. These experiments have shown that the floating filter can provide
acceptable quality water for more than 40 hours, at conventional rapid sand filtration rates, without
backwashing. As the backwash can be achieved with a lower water consumption this process is an
economically attractive alternative to rapid sand filtration.


The energy required to resuspend the solids in the floating bed is much less than the energy
required for a sand bed. Therefore the Floto filter can be backwashed with compressed air and a little
amount of water. Alternatively a downflow water rinse can backwash the filter to the same degree.
The sand medium used the same amount of backwash air as floating media but needed a relatively
larger amount of water for backwashing. While the floating media could be backwashed with water
alone, the sand always needed air, to break up the piston effect. Considering the buoyant nature and
the low affinity of the floating media particles to each other, formation of an impervious barrier due
to clogged media in the Floating Filter is not possible.


In Floto filter, the media interface was clearly defined even during backwashing. Also there
was no intermixing of media after the backwash. This was the most prominent advantage of the
floating media arrangement as it eliminates the frequent problem of intermixing in the media
interface zone which results in rapid clogging, faced by other types of dual media arrangements.
-97-



A typical dual media filter needs a fluidization wash at the end of the backwash cycle, to
restratify the media. Due to the large difference in the densities of two media used in this experiment,
such a restratifying wash was not required.


Since the Floto filter has a buffer zone between the water inlet and the bottom of the bed, bed
disturbances due to inflow conditions will not occur. Also this buffer zone eliminates the need for
elaborate underdrain systems, thereby reducing capital and operation and maintenance costs.


The mathematical model was able to predict only the headloss profiles for upflow sand filter
and the Floto filter. The concentration profiles were not predicted.


This mathematical model has to be modified to accept different values for different
media layers in the multimedia filter. Simplifying assumptions had to be made to compensate for the
absence of particle size distributions at the entry point to the fine media layer and in the effluent.


During the simulations, the model parameter ' had an effect on headloss profile as
expected. The effect of model parameters
p
and
2
which were respectively expected to affect
the ripening and breakthrough stages of the concentration profiles showed only a marginal effect.
Higher values of the detachment coefficient
2
reduced the predicted headloss values significantly.
-98-

CHAPTER 6


RECOMMENDATIONS FOR FURTHER STUDY



Filter effluent quality was adversely affected by changes in the rate of flow, even though care
was taken to open the valve slowly. Hence an improved flow control device which does not
introduce shock loads into the filter should be incorporated into future experiments. The simple flow
control device shown in figure 6.1 can be utilized for this purpose.

























Fig 6.1 Float Activated Flow Control Device (ADIN et al. 1979)


In this experimental unit, countercurrent (i.e downflow) backwashing was impossible
immediately after the air wash, as the air pressure built up inside the filter due to the absence of an air
release mechanism. This high pressure air pocket prevented the flow of water into the filter. Sending
water into the filter against this high pressure would have damaged the apparatus. This fault has to be
-99-

rectified in a future design by introducing a device to release air from the filter. Also a future design
should have a valve arrangement which should facilitate simultaneous air and water wash as
simultaneous air/water backwash was impossible in the filter used for these experiments.


Removal of the bottom retention grid is desirable if the downflow water rinse is to be
employed as a backwash method. Yet in case of accidental dewatering of the bed the media may be
lost if the bottom grid is not there. Hence this grid size should be enlarged according to the size of
the coarse media. Also the backwash water outlet should be replaced by a larger diameter pipe, (say
1.5 cm diameter).


If the bottom grid is removed or if a larger mesh is used, the filter bottom should be modified.
It can be shaped like a funnel with a valve arrangement allowing periodic removal of accumulated
floc from the funnel. The upflow water and air inlets should be positioned above the funnel.


Freshly made 0.1% alum solution should be used for jar tests to obtain better correlation
between jar test results and the plant performance. The jar test results can be supplemented by 45
membrane filtration.


The filter operation has to be optimized to have ripening periods less than 15 minutes. Since
the filter ripened quickly for the higher (7.5 m/h) flow rate, operating first at a high rate and then
gradually coming back to the 5 m/h rate should be attempted to bring in faster ripening.


Coating of media with polymer may provide an answer to the problem of hastening of the
filter ripening period. This should be attempted for varying lengths of time before start of filter run to
ensure that any ripening observed in the experiments are due to particles captured in the media and
not to changes in the surface chemistry of the media, caused by the polymer after the experiment
began. Perhaps this would allow elimination of using alum altogether. If alum is not used the
effective polymer dosage can be determined by trial runs in a smaller filter unit.


For higher filtration rates (10 and 12.5 m/h) better runs were obtained with the combination
of 3.66 mm Polypropylene and spherical 1.54 mm polystyrene. Since better results were expected
with 2.57 mm polypropylene combinations, this result should be verified by further experiments.


The method of preparation employed to make smaller size polystyrene has to be improved to
ensure that all grains have the same density. If a smaller size polystyrene is commercially available, it
should be utilized for future runs.
-100-


A backwash system which provides a small amount of water simultaneously with the air
backwash would be able to achieve complete cleaning of the bed, without using a large amount of
water. Such a backwash system should be attempted with the backwash water sent from the bottom
or laterally from the walls.

The exact amount of water consumed during the backwash operation should be analyzed.


Modification of the media chemically to have different surface properties should be
attempted.


Application of the Floto filter for treatment of natural suspensions should be attempted. The
removal should be assessed for removal of turbidity, color, algae, coliform and iron and manganese.


The coagulant and flocculent used in this research were alum and CatFloc T2. The usage of
other coagulants and flocculents (e.g. Magnafloc cationic coagulants and flocculents, poly aluminum
chloride, or poly aluminum silicate sulphate) should be attempted.

Continuous monitoring of pressure head using pressure transducers and continuous recording
of effluent quality should be attempted as this would bring in a better picture of the filter operation.

Polymer dosing pumps capable of delivering the required dosage at the recommended stock
solution strength should be utilized. Ability of using a drip bottle/ roller grip arrangement for
polymer dosing should be analyzed.


A pilot scale study should be made to further assess the performance of the Floto Filter and
consolidate the results gained in the laboratory study.


Extensive work is required to upgrade the existing model to predict the Floto filter behavior.
The model should be upgraded to accept different values for different layers in the multimedia
filter.


Particle size distributions of the influent to the filter, influent to the fine media layer and of
the effluent should be obtained in future experiments to facilitate accurate prediction of model
parameters for mathematical modelling. A particle size analyzer capable of giving size analysis of
dilute suspensions is required for this. Also the samples should be analyzed immediately after
sampling considering possible effects of flocculation and agglomeration.





APPENDIX 1

FILTER RUNS # 16 TO # 48

CONCENTRATION AND HEADLOSS PROFILES











































APPENDIX 1

The headloss and concentration profiles for runs # 15 to # 48 are given
in the following pages. This summary gives the operational and hydraulic
conditions.

Summary of filter Runs # 15 - # 48
Media NTU Surface Loading m
3
/m
2
.h
15 12.5 10 7.5 5
Dual Media
PP 3.66 mm
PS 1.54 mm spherical
30 #15 #16 #17 #18
60
90
Dual Media
PP 3.66 mm
PS 1.54 mm spherical
30 #19 #20 #21 #22 #30
60 #23 #24 #25 #26 #27
90 #28
#29
Dual Media
PP 3.66 mm
PS 1.54 mm angular
30 #31 #32 #33 #34 #35
60 #36
90 #37
Dual Media
PP 2.57 mm
PS 1.54 mm spherical
30 #38 #39 #40 #41 #42
60 #44
90
Dual Media
PP 2.57 mm
PS 1.54 mm angular
30 #43
60
90
Single Medium
Silica Sand 1 mm
30 #45 #46
60 #48
90 #47

# denotes a filter run
* PP = polypropylene, PS = polystyrene
* All runs upflow
* Center sampling from run #19
* Run #28 with 0.5 mg/l Catfloc T2, Run # 29 with 0.2 mg/l Catfloc T2


























APPENDIX 2

FILTER RUNS #1 TO #18

SUMMARY AND COMMENT



















-139-

APPENDIX 2

Previous Filter runs:

The filter runs #1 to #18 were conducted to determine an optimum media arrangement as
well as to verify hydraulic conditions, chemical dosing, optimum bed height, media mixing
properties etc. No alum was added up to run #15. Only chemical dosing was CatFloc T 0.5 mg/l. In
run #1 no flocculent was added. Runs #15 to #18 used alum 20 mg/l and Catfloc T2 0.5 mg/l.

Media NTU Surface Loading (M
3
/m
2
.h)
15 12.5 10 7.5 5
Polypropylene 4.36 mm 30 #4 #5
50 #3
Silica Sand 1 mm

30 #6 #7 #8 #9 #10
Dual Media
Polypropylene 4.36 mm
Polystyrene 1.54 mm (Spherical)
30 #12 #13 #14
Dual Media
Polypropylene 3.66 mm
Polystyrene 1.54 mm (Spherical)
30 #15 #16 #17 #18


Downflow Runs

#1 - PP 5mm, 30 NTU, 15 m/h, No flocculent addition
#2 - PP 5mm, 30 NTU, 15 m/h, With Cat Floc T
#11 - Silica Sand, 30 NTU, 15 m/h, With CatFloc T

Comment:

First 14 runs were generally unsatisfactory as the flocculent did not act due to the absence of
alum. Also since sampling was carried out from all five ports, the bed was disturbed after few hours
of run. For runs #15 to #18 only the effluent was sampled. The filter column was modified after run
#18 with the addition of 5 more piezometer points.

You might also like