Untitled
Untitled
Untitled
.
PERFORMANCE EVALUATION OF DYNAMIC
ROUGHING FILTRATION
Examination Committee:
Supervisor:
Mentors:
Ir. T. Ahsan
Jr. J. P. Buiteman
1~J j~4O
LO: ~ ~
The findings, interpretation and conclusions expressed in this study do
neither necessarily reflect the veiws of the International Institute for
Hydraulic and Environmental Engineering, nor of the individual members of the
NSc—committee nor of their respective employers.
TABLE OF CONTENTS
Acknowledgements i
Abstract ii
List of Figures iii
List of Tables iv
Notations V
Abbreviations Vi
CHAPTER 1: INTRODUCTION
1.1. Background 1
1.2. Objectives 2
1.3. Approach and Methodology 2
5.1. Introduction 32
5.2. Description of the pilot plants 32
5.3. Description of the experimental set-up 33
5.4. Overview of experimental parameters 35
6.1. Introduction 38
6.2. Presentation of results 39
6.2.1. Raw water characterization 39
6.2.2. Column settling test results 41
6.2.3. Treated water quality 50
6.2.4. Scouring of particles 53
6.2.5. Other measurements 55
6.2.5.1. Mass balance 55
6.2.5.2. Flow measurements 56
6.2.5.3. Headloss 57
CHAPTER 7: DISCUSSION OF RESULTS
8.1 Conclusions 65
8.2 Recommendations 66
REFERENCES 67
APPENDIXES 70
Appendix 2
Linear regression model between suspended solids and raw water
turbidity based on column settling test results.
Table 2-1 Relationship between raw water turbidity and
suspended solids. 77
Appendix 3
Table 3-1 Real removal efficiency variations over the filter
runs for different Q1’s and V = 2.0 rn/h. Descriptive Statistics
for physical, chemical and bacteriological parameters. 78
Appendix A
Graphical representation of Column Settling Tests.
Cumulative frequency distribution of settling velocities for
different Q1’s.
Figure A-i a) Nominal Q = 1.01/s. 80
Figure A-i b) Nominal Q, = 2.0 1/s. 80
Figure A-i c) Nominal Q, = 2.65 1/s. 80
Appendix B
Typical variations in water flows, Q1, Q and Q0 over the
filter runs.
Figure B-i a) Nominal Q = 0.42 1/s and V = 2.0 m/h. 80
Figure B-i b) Nominal Q = 0.70 1/s and V = 2.0 rn/h. 80
Figure B-i c) Nominal Q1 = 1.501/sand V = 2.0 rn/h. 80
Figure B-i d) Nominal Q = 2.65 1/s and V = 2.0 rn/h. 81
Figure B-i e) Nominal Q, = 1.50 1/s and V = 3.0 rn/h. 81
Figure B-i f) Nominal Q~= 1.50 1/s and V = 4.0 m/h. 81
Appendix C
Typical variations of headloss for different Q1’s and
V = 2.0 rn/h.
Figure C-i a) Nominal Q, = 0.42 I/s. 82
Figure C-i b) Nominal Q, = 0.70 1/s. 82
Figure C-i c) Nominal Q1 = i.001/s. 82
Figure C-i d) Nominal Q, 1.50 1/s. 83
Figure C-i e) Nominal Q, 2.0 1/s. 83
Figure C-i f) Nominal Q, 2.65 1/s. 83
ACKNOWLEDGEMENTS
The author would like to thank his supervisor Prof. Dr. Jr. G.J.F.R. Alearts, and his mentors
Jr. T. Ahsan and Jr. J.P. Buiteman of the International Institute for Infrastructural Hydraulic
and Environmental Engineering (IHE), Deift, to their valuable advice and suggestions and
continuous guidance and inspiration during the research period. He also like to extend his
profound thank to Jr. J.T. Visscher of the International Water Supply Center (IRC), The
Hague, who reviewed his drafts.
He is grateful to all CINARA and UNIVALLE families for their co-operation and assistance
during carrying out the research project.
Finally he wishes to express deep appreciation to his wife and mother who encouraged and
inspired him during the period of study.
August, 1994
Deift
The Netherlands Jorge Latorre Montero.
I
Performance Evaluation of Dynamic Roughing Filtration
ABSTRACT
Dynamic Roughing Filtration (DyRF) has been proposed as a first pretreatment step before
roughing filters and/or slow sand filter for turbid rivers. DyRF comprises a thin layer of fine
gravel on top of a shallow bed of a coarse gravel with a system of underdrains. The influent
is distributed into two directions: part of the flow (effluent) passes downward through the
filter medium and then to the subsequent treatment units; and the other part (overflow) flows
over the gravel bed and is normally returns to the raw water source.
Research was carried out in the Research and Technoiogy Transfer Station of CINARA (Inter
regional Center in Water Supply and Sanitation) in Cali, Colombia, where two pilot DyRF
plants with declining rate filtration treated raw water from Cauca river. This research was set
out to achieve a better understanding of DyRF processes. The specific objectives are:
1. To characterise the raw water quality
2. To study the particle removal process
3. To assess the impact of the surface overflow on surface particle scouring and on
treated water quality
4. To suggest design guidelines
2. In DyRF sedimentation is the main particle removal process which occurs in two different
locations: i) Plain sedimentation onto the exposed surface of the gravel bed. Here the removal
efficiency is < 10%. ii) Sedimentation in the gravel bed.
3. The surface overflow did not have much impact on treated water quality. It may have a
negative impact on the hydraulic perfornance of the DyRF.
4. The following removal efficiencies were obtained for DyRF units operating at filtration
rates between 2.0 and 4.0 rn/h and surface flow velocities ranging from 5 cm/s to 18 cm/s:
turbidity from 50% to 52%, suspended solids from 83% to 87%, true color from 13% to
24%, total iron 55% to 84% and faecal coliform from 0.4 to 1.0 log.
5. Due to the limited and possibly negative impact of the overflow it is better to design the
DyRF as a declining rate down flow roughing filter with maximum filtration rate of 4.0 mlh
and gravel sizes between ~ 6 to 13 mm in the upper layer, 4 13 to 19 mm in the middle and
4 19 to 25 mm at the bottom. The thickness of each layer can be taken as 0.20 m. It needs
further research to explore if constant rate filtration would provide similar results.
11
LIST OF FIGURES
111
Figure 6-7b) Treated water suspended solids for different Q1’s. 51
Figure 6-7c) Treated water volatile solids for different Q1’s. 51
Figure 6-7 51
d) Treated water true color for different Q1’s.
Figure 6-7 e) Treated water total iron for different Q’s. 52
Figure 6-7 f) Treated water faecal coliform counts for different Q~’s. 52
Figure 6-8 Suspended solid removal efficiencies for different filtration velocities. 53
Figure 6-9 Graphical representation of theoretical relationship between non
cohesive particles and scouring velocities. 54
Figure 6-10 Graphical representation of theoretical relationship between cohesive
particles and scouring velocities. 55
Figure 6-11 Bar chart of sludge accumulation distribution in different layers of
the DyRF for different Q,’s under V = 2.0 rn/h. 55
Figure 6-12 Bar chart of storaged sludge per layer under different ifitation rates. 56
Figure 6-13 Bar chart of sludge loading for different filtration rates. 56
Figure 6-14 Typical flow variations over the filter run for Q,, Q0 and Q~. 57
Figure 6-15 Typical variation of headloss through filter depth over the filter run. 57
LIST OF TABLES
iv
NOTATIONS
The following symbols and units are used in this thesis:
A = surface area of DyRF
d (m2)
1 = particle diameter (~m= ixl0~m)
d50 = median particle diameter of bed material (m)
FCU = forming colony unit
F~ = Froude number (—)
= friction factor 2)
(—)
g = acceleration of gravity (9.80 m/s
(m)
h = water depth (m/m)
I = hydraulic gradient (m)
= effective bed roughness (m)
= gravel layer thicknesses
NTU = nephelometric turbidity unit (%)
P = remaining concentration
(—)
p = porosity (1/s)
Q. = treated water, effluent (L’s)
Q, raw water, in fluent (us)
Q0 = overflow (%)
R = removal ratio
(—)
Reynols number (crn/sx 10~)
S = overflow rate (m2)
S~ = available surface area for sedimentation in gravel beds (cm/sx 10~)
S1 = surface loading (mg/i)
SS~ = effluent suspended solid concentration,treated water (mg/i)
~~1 = influent suspended solid concentration, raw water (mg/l)
SS0 = overflow suspended solid concentration
STD = standard deviation (cm/s)
U = scouring velocity (rn/h)
V = filtration rate (cm/sx 1 0~)
V~ = settling velocity
x (cm/s)
= flow velocity
(%)
= significance level (m2/s)
-y = kinematic viscosity
0~~ (—)
= dimensionless critical shear stress (kg/rn3)
p = water density (kg/rn3)
p’ particle density
1 = (N/rn2)
roe = critical shear stress (—)
(I = roughness coefficient
V
ABBREVIATIONS
vi
CHAPTER 1
Introduction -
1.1 Background
Experience in several countries has shown that for many communities Slow Sand Filtration
(SSF) is a very appropiate alternative for drinking water treatment (Visscher et al. 1987).
Nevertheless, the performance of such systems may not be sufficient to cope with the level
ofcontamination of raw water sources. For instance, high level of faecal contamination (over
500 CFU/100 ml), Lloyd (1991), or medium to high turbidity levels (over 10 mg/I S1O2 for
longer periods than a few weeks), Huisman and Wood (1974).
Under this situation, pretreatment is required to lower the influent turbidity and faecal
contamination to the SSF to an acceptable level. Obviously the pretretment process should be
such that its level of complexity is compatible with that of the SSF.
The pretreatment research carried out in Colombia on water from a highly polluted lowland
river and from other Andean rivers, clearly shows the potential of combining two-stage
roughing filtration with SSF, CINARA-IRC (1992).
The research reveals that roughing filtration as a pretreatment alternative helps to reduce the
load in suspended solids and creates an essencial additional barrier against the transmission
of disease carrying organisms and other harmful substances in the water, Galvis (1992). Also,
preliminary studies by CINARA indicate that two-stage roughing ifitration is a feasible
alternative to lower the chemical consumption in conventional treatment plants.
In the two-stage roughing filtration, the Dynamic Roughing Filter (DyRF) provides a very
good first stage in the treatment process. The results in the investigations in Cali-Colombia
show that:
i) Costs are rather low (on average less than 5% of the capital investment in the
treatment plant).
ii) DyRF contributes to the removal of suspended solids between 57% and 80% of
average suspended solids loads in raw water in the range of 60 to 190 mg/i.
iii) Faecal coliform counts are reduced between 33% and 78% for raw water in the
average range of 8476 and 73182 CFU/100 ml.
iv) Turbidity is being removed between 36% to 45 % for average turbidity levels in the
raw water between 25.8 and 238 NTU.
1
vi) Color removal is between 11% and 17%.
According to the available reviewed literature and taking into account the optimization done
by CINARA, DyRF comprises a thin layer of fine gravel on top of a shallow bed of coarse
gravel with a system of underdrains. The water entering the unit (Influent) passes over the
gravel bed. Part of it is drained through the bed to the next treatment unit (Effluent) and the
other part is returned to the river (Overflow). Under normal conditions the unit will gradually
clog and will need to be cleaned every week or twice a week. When peak loads in suspended
solids are being received clogging goes much faster. The clogging will reduce the flow to the
subsequent treatment units, thus protecting the total treatment plant from peak loads, Galvis
et al. (1992).
In spite of the good findings resulting from the DyRF performance, further improvements of
this stage seem very well possible. Up to now DyRF systems have been basically studied as
part of multi-stage filtration plants. Only initial research on DyRF itself is carried out by
CINARA in collaboration with IRC and mCWD.
A particular point for further study is the Impact of the surface overflow on the DyRF
performance. Although the available reports do recognize the impact of the overflow on the
behaviour of the units, there is not any systematic study of this parameter to support design
or operational criteria. Data mentioned in the literature shows a wide variation in the ratio
Overflow/ Effluent ranging between 0 and 10. This thus requires further research in order to
achieve a better understanding of DyRF processes and also to develop adequate guidelines for
design and operation and maintenance.
1.2 Objectives
The objectives of this study are:
ii) To assess the impact of the surface overflow on surface particle scouring and on
treated water quality
i) The raw water quality of Cauca river will be analysed and results obtained from the
bacteriological and phisical-chemical analysis will be compared with, existing water
2
quality standards to assess the sanitary risk of the water.
ii) Sedimentation has been mentioned as one of the most important mechanisms in the
removal of particles in roughing filtration. This process will be studied in DyRF by
comparing the real efficiencies (established on the basis of the experimental data for
suspended solids) and theoretical removal efficiencies established with column settling
tests.
iii) The statistical comparison of removal efficiencies for each parameter and for different
overflow will allow to analyse the impact of the overflow on the quality of treated water.
iv) The impact of the overflow on scouring of particles settled on the surface of the
gravel will be estimated by comparing the remaining concentration of particles for
different flow velocities.
v) On the basis of the results and data available from literature possible changes in the
design will be presented to optimize the system.
To stablish this research pilot plants will be established in the Research Station of CINARA
in Cali, Colombia, where water will be treated from Cauca river. The plants will operate
under declining rate filtration and will be designed on the basis of the preliminary design
criteria presented by CINARA, IRC (1993).
3
CHAPTER 2
Review on Pre-treatment Alternatives
2.1 General
In Europe the limitations presented in slow sand filtration application resulted in the
development ofpre-treatment techniques which initially were rather simple such as, long term
storage and micro straining. Gradually, more complicated systems were put in place prior to
slow sand filtration including coagulation, using chemicals, and flocculation followed by
sedimentation and rapid sand filtration. These processes however hold little promise
considering the conditions in most less developed countries. This situation has revitalized
research in other pre-treatment alternatives that do not require the addition of chemicals and
that are simple to operate and maintain. Most of these experiments involve rather small scale
pilot plants and in fact only limited data are presented in literature and no evidence was found
of comparative research of the different techniques. According to CJNARA, IRC (1993)
experiments reported in the literature focus particularly on the removal of suspended solids -
some using kaoline suspensions - and that the performance of the systems is mostly being
explained on the basis of sedimentation theory.
Pre-treatment Alternatives are categorized below and will be discussed in that order.
1. Infiltration Wells
2. Infiltration Galleries
3. Storage
4. Plain Sedimentation
5. Roughing Filters
Dynamic Roughing Filters (DyRF)
Downflow Coarse Sand Filters
Downflow Roughing Filters in Series
Upflow Roughing Filters
Horizontal Roughing Filters (HRF)
Infiltration wells are wells dug or drilled in the banks of rivers. Depending on the water
quality in the river and the soil conditions, water drawn from infiltration wells can be directly
put into supply after disinfection or bringing it to a slow sand filter plant. Engels et al.(1989)
have reported problems with the resuspension of iron and manganese oxides when levels of
oxygen in the ground and river drop below 1 mg/i.
Another disadvantage of filtration wells is that changes may occur underground which may
4
result in reduction of the water flow and which can not be remedied by maintenance activities.
Infiltration galleries basically consist of perforated pipes placed in the river bed. If the natural
permeability of the river bed is low, the material can be removed and replaced partially by
other material such as gravel and sand. Figure 2-1 illustrates two posibilities to install the
filter material. Flow velocities applied in river bed filtration have been reported in the range
between 0.25 and 1.5 rn/h depending on the turbidity and the requirements to improve the
water quality. Removal efficiencies for the system indicated in Figure 2-1 (b) have been
reported by Salazar (1980) as 98% for turbidity removal from rivers with turbidity levels
ranging from 48 to 200 NTU. A study carried out by CINARA, IRCWD (1988) in Colombia
indicated however that the real efficiency of the systems is rather low and may reach only
some 20%. In 1984, Nagarkar et.al. indicated that filtration galleries may not be suitable as
a pretreatment method for treating raw water with colloidal turbidity greater than 100 NTU
and may be considerably more costly for construction and maintenance than other systems.
The periodic blockage of the infiltration zone makes cleaning or repositioning of the material
needed. In practice this maintenance is extremely complicated as the material is located under
water in the rivers. Galvis et al.(1993) reported that these types of limitations have motivated
the development of a modified system now being known as dynamic roughing filters.
(a) (b)
Figure 2-1 a) Infiltration Gallery without interfering in the flow gradient (Smet
et. al. 1989) and (b) Infiltration Gallery Abstraction System with a weir Salazar
(1980).
2.2.3 Storage
Schultz and Okun (1984) have reported that storage serves the following purposes:
5
4. It can improve the reliability of the water supply system
5. It can be used to overcome periods of excess turbidity.
Long term storage also may have draw backs. Ellis (1985) stated that in the case of storage
the problem of eutrophication may occur and monitoring of nutrient should be employed.
Plain sedimentation can very much contribute to reducing the level of suspended solids in the
water source but has a limited impact for water sources with a turbidity of colloidal nature.
Cleasby (1991) reported that in Cincinnati, plain sedimentation of water from the Ohio river
reduced the suspended solids content from 170 to 100 mg/I, after a retention time of six days.
Two different applications can be identified: a) a system using a short retention time, less than
one day, and b) a system with a very long retention time in the order of several days or
weeks. For systems with a short retention time tests with sedimentation columns are
recommended to explore the expected quality improvement. Cleasby (1991) experimentally
recognized that these tests however are not suitable for estimating the effect of long term
storage, as the process in the sedimentation column will not reflect the normal situation where
other factors such as stratification because of temperature, and the influence of algae can be
very important.
Long term storage is very common in England. WRC (1977) has found that in London
turbidity reduction in large storage basins have reduced turbidity levels from 30 NTU to
values below 4. Long term storage may also have a significant effect on the bacteriological
quality. In 1974 Taylor confirmed that in the period from 1961 to 1970 the average faecal
coliform counts of 6,680 per 100 ml were reduced to 249 per 100 ml. However the periodic
blooms of algae made it necessary to introduce micro-screeners or rapid filters before the
slow sand filters treating the stored water. The potential of long term storage for tropical
countries has to be evaluated carefully before large scale application can be promoted.
Tilted plated settlers or tube settlers can be applied as sedimentation units with short retention
times and may reduce the required surface area. Tilted plate settlers have been applied with
good results in chemical coagulated water but hardly any experience exists in its application
for non-coagulated water.
Roughing filters are basically boxes filled with gravel or coarse sand. The efficiency of
roughing filtration is primarily based on the large surface area available in the gravel bed
which facilitates the available mechanisms to remove impurities from the water. These
mechanisms are of physical, chemical and biological nature. In the following section different
types of roughing filters are described and are classified according to their characteristics and
direction of flow.
6
2.2.5.1 Dynamic Roughing Filters (DyRF)
CINARA, IRC (1993) described the DyRF as a layer of fine gravel (3 to 6 mm) of some 0.2
to 0.3 m heigth placed over a layer of coarser gravel (12 to 25 mm) of some 0.2 to 0.4 m of
height. In the bottom layer perforated pipes are placed as drainage system. Figure 2-2 shows
a longitudinal section.
The water flowing into the unit partly infiltrates into the gravel bed to the drainage system
from where it will flow to the next unit. The other part flows over the gravel and usually
returns to the raw water source. These units operate at a filtration rate which ranges from 1
to 9 rn/h. Under normal operation conditions the fine gravel layer will gradually clogg as a
result of the retention of suspended solids. When higher levels of suspended solids are being
received the rate of clogging may be higher and - depending on characteristics of the particles
- may lead to a complete blockage. Once or twice a week the gravel bed has to be cleaned
by raking the fine gravel layer. Galvis et al.(1992) reported that every six to twelve months
the filter material has to be removed, washed and re-installed in the unit to maintain the
filtration capacity of the system.
WATER
Since the begining of this century downflow rapid sand filters have been applied prior to slow
sand filters in Europe and the U.S.A.. Cleasby (1991) reported that although these alternatives
are somewhat different from conventional rapid sand filtration, as somewhat coarser sand is
being used and no addition of chemical products is required, still the ifiters need to be
backwashed frequently to clean the sand. This complicates the application of the technology
and limits its use to areas where backwashing may be readily applied.
Rajapakse et al.(1989) have found another system comprising pebbles and sand called pebbled
matrix filtration which is currently being investigated as pre-treatment to be combined with
slow sand filtration. The stones have a size of some 50 mm and are surrounded by coarse
sand. In this way rather a large quantity of suspended solids can be retained with a relative
limited headloss development. An experimental pilot unit with a sand bed depth of 1.3 m
7
operating at flow velocities between 0.5 and 1.5 rn/h treating water with kaoline suspension
ranging between 100 to 5000 mg/l of suspended solids produced effluents with suspended
solid concentrations below 25 mg/l. On the basis of these results Rajapakse et.al.(1989)
consider that this technology can be used as a pre-treatment method for surface water sources
with a suspended solid content below 2000 mg/l, applying a filtration rate of 0.7 rn/h.
Thus, downflow Coarse Sand Filters seem to have good potential for pre-treatment of surface
water with high suspended solid content. However, its operation and maintenence
requirements which include backwashing under considerable pressure may restrict its
aplication.
Perez et al.(1985), Pardon (1989), and Galvis et al. (1987) carried out studies in Peru and
Colombia respectively. The findings stimulated new interest in this pre-treatment option. The
figure 2-3 indicates a schematic design of a Downflow Roughing Filter in Series with three
units. The system has a moderate capacity to store sludge which makes periodic cleaning
necessary. This is done by draining the filter with the help of a fast drainage valve connected
to the drainage system.
Ir,tlow -~
Outflow
A study was made in Peru on a system with three units of 15 cm in diameter filled with 0.6
m of gravel. Gravel was used ranging from 50 to 12 mm diameter and flow velocities were
applied ranging from 0.1 to 0.8 rn/h. The turbidity of the Rimac river during the study period
was approximately 50 NTU. Removal efficiencies obtained were 45 % with respect to NTU
for the highest filtration rate and 55 % for the lowest rate. When turbidity levels were
8
increased to 200 and 300 NTIS through sludge dosing, efficiencies incresed to approximately
70% for the highest filtration rate and 90% for the lowest. On the basis of the experiments
a filtration rate of 0.3 rn/h was recommended to ensure an enfluent below 20 NTU, for
surface water with turbidity levels below 300 NTU. It was also found that the flow velocity
needed to clean the system was very high and had to be in the order of 90 mlh to transport
the deposited material to the underdrain system.
In Colombia studies were also realized with pilot scale units similar to those utilized in Peru
but with layers of graded gravel with grain sizes ranging from 6 to 18 mm, and filtration rates
of 0.7 mlh were used. Quiroga (1988) and Galvis et al.(1989) recognized that the studies were
complicated by the fact that the small diameter of the units (15 cm) and the low flow
velocities made operation and maintenance of the units rather complicated. To obtain reliable
results of such experiments very close monitoring is required and therefore it is better to use
larger units. Quiroga (1988) and Galvis et al.(1989) reported that the first results of the
studies with water from the Cauca river in Cali were obtained with raw water turbidities
ranging from 20 to 100 NTLJ, apparent color levels ranging from 49 to 200 units, and faecal
coliform levels in the order of 100,000 MPN/100 ml. The removal efficiency obtained for
turbidity ranged between 50 and 90% for apparent color between 45 and 85% and for faecal
coliforms between 70 and 99%. In subsequent investigations with technical scale units of 2m
diameter the potential of this pre-treatment alternative was confirmed, but very little
experience on full scale plants is available.
Frankel (1974) and Wolters et al. (1989) have reported that in Asia Downflow Roughing
Filters were also used but instead of applying gravel other filter material was used such as
coconut fiber. Raw water turbidities ranging between 25 and 130 JTU (Jackson Turbidity
Units) were reduced to below 1 JTU by applying coconut fibre. However, this filter medium
needs to be replaced every time when the filter needs to be cleaned which in this case was
every month.
In upflow roughing filters the water flows upwards through a series of different gravel layers
which are decreasing in size. Galvis et al. (1993) distinguish two alternatives namely, a)
upflow gravel filtration in layers (URFL) when the gravel layers are installed in the same
unit, as is shown in Figure 2-4 and b) upflow roughing filtration in series (IJRFS) when the
gravel layers are installed in two or three different units, as is presented in Figure 2-5.
Galvis et al.(1987) have expressed that the units have a moderate sludge storage capacity and
therefore require periodic cleaning. This cleaning is done by draining the units by opening
a fast drainage valve. The cleaning effect of draining can be increased by rapidly opening and
closing the fast drainage valve.
Galvis et al. (1989) have pointed out that both alternatives of upflow roughing filtration have
been evalualed in Cali, Colombia, using water from the River Cauca. For the first trials with
9
the URFS, filter columns were used each 15 cm in
diameter filled with gravel ranging from 18 to 6
mm. The filtration rate which was applied was 0.7
Inflow- m/h and raw water quality with a turbidity level
ranging from 20 to 100 NTU, color from 50 to 200
TCU and faecal coliforms in the order of 100,000
MPN per 100 ml. Galvis et al. (1989) and Wolters
et al.(1989) have reported that the results obtained
Drains for filter CleaninQ indicated removal efficiencies for turbidity between
75 and 90% for apparent color between 50 and 70%
. u
Figui re 2-4 Overview of an p flOWand
alternative of coliforms
for faecal between for
URFL evaluated 70 and
the 99.9%.
same water
The
Roughmg Filter in Layers. quality showed lower removal efficiencies.
CINARA, IRC (1993) reported that between 1990 and 1993 both systems were evaluated
using water from the Cauca river but using systems of 2 m diameter and gravel ranging from
25 to 1.6 mm. Filtration rates were applied ranging from 0.3 rn/h to 0.75 rn/h. Over the test
period the raw water source had turbidity values ranging from 52 to 106 NTU, true color
levels from 35 to 73 TCU, suspended solids levels ranging from 61 to 187 mg/i and faecal
coliforms counts between 30,185 and 148,575 MPN per 100 ml. Removal efficiencies for the
URFL were reported by CINARA, IRC (1993) ranging between 46 to 71 % for turbidity,
between 10 and 46% for true color, between 49 and 94% for suspended solids and between
73.3 and 98.4% for faecal coliforms. The efficiencies for the IJRFS ranged from 69 to 83%
for turbidity, between 29 and 68% for true color, between 92 and 97% for suspended solids,
and between 97.7 and 99.7% for faecal coliforms.
Inflow —
10
al.(198l) and by Mbwette (1983).
The International Reference Center for Waste Disposal (IRCWD) has carried out laboratory
studies in Switzerland with experimental pilot units using different kaoline suspensions. This
study continued with a monitoring of full scale units constructed in different countries
including Tanzania, Peru, Sudan and Colombia, Wegelin (1986).
CINARA, IRCWD (1988) has reported that in Colombia one experimental unit and three full
scale plants have been constructed and monitored. The first experience with this alternative
was obtained with an HRF of 7.14 m length which included a drainage system for cleaning
purposes. The gravel size utilized ranged from 25.4 to 2mm and the filtration rate ranged
from 0.3 to 0.6 rn/h.
Figure 2-6 presents the schematic lay-out of a typical HRF which basically consists of the
inlet, the filter bed, the outlet, and the underdrainage structure.
Inflow iuttlow
More recently the International Institute for Infrastructural, Hydraulic and Environmental
Engineering (IHE) based in The Netherlands carried out research to enhance the
understanding of Direct Horizontal-Flow Roughing Filtration’s (DHRF) behaviour and to
further develop it, Ahsan et al.(1991). In DHRF coagulants are added prior to filtration. Lab
scale pilot plant experiments systematically showed better removal efficiency of DHRF even
at higher filtration rates (e.g. 5 rn/h) when compared to HRF.
Chowdhury (1993) so far concluded that the feasibility of DHRF, proves to be an attractive
and promising pre-treatment process on bench scale. Its application on full plant scale prior
slow sand filtration and rapid sand filtration in semi-urban areas and the small towns of
11
developing countries needs to be investigated and Ahsan et.al.(1991 ) added that the
construction cost of DHRF is estimated to be about two times lower than HRF or a
conventional flocculation-sedimentation system.
CINARA,IRC (1993) pointed out that the pre-treatment research carried out in Colombia on
water from a highly polluted lowland river and from other Andean rivers, clearly revealed
the potential of combining two-stage roughing filtration with slow sand filtration; Galvis
(1992) reported that roughing filtration as a pre-treatment alternative helps to reduce the load
in suspended solids and creates an essential additional barrier against the transmission of
disease carrying organisms and other harmful substances in the water. Also, a preliminary
study done by CINARA, IRC (1993) indicated that two-stage roughing filtration is a feasible
alternative to lower the chemical consumption in conventional treatment plants.
In two-stage roughing filtration, the Dynamic Roughing Filter (DyRF) provides a very good
first stage in the treatment process. Figure 2-7 provides data from the evaluation brought
about in Colombia, showing that the DyRF gave better results in suspended solids removal
than the plain sedimentation units and the tilted plate settlers.
1. The DyRF construction costs are rather low (on average less than 5% of the capital
investment in the treatment plant).
2. DyRF contributes to the removal of suspended solids between 57 and 80% of average
suspended solid loads in raw water in the range of 60 to 190 mg/i.
3. Faecal coliform counts are reduced between 33 and 78% for raw water in the average
range of 8476 and 73182 FC/100 ml.
4. Turbidity is being removed between 36 to 45% for the average turbidity levels in the
raw water between 25.8 and 238 NTU. - - -
12
— l1.~ —
200
•
Penod =
=
~° 200 —
• Period II D~BFA
D~iiFB
• —
1P5
DW1~ — — 0yR~C
100 ~ 100
31,50
1F~i~r
ScUd,
: huT
PS -
TurDlthCy TrL* Colour
Figure 2-7 Average levels of different contaminants in raw water and effluent
of dynamic roughing filters and other conditioning processes. (CINARA-IRC,
1993)
A general overview of a DyRF is presented in Figure 2-8. The water flow Q, entering the
filter is distributed into two flows. The first flow Q passes through the filter medium and on
to the subsequent treatment units; the other flow Q0 normally returns to the raw water source.
It is important to note that the filter medium grades from fine at top to coarse at the bottom
where it is placed over the drainage system. In this way the system is designed to accumulate
the suspended solids basically at the surface which very much facilitates its cleaning.
Due to the relatively coarse gravel which is being used, the headloss over the unit is very
small. However, if the valve which controls the flow towards the other units of the system
is not being opened, the flow through the filter will gradually reduce as a result of the small
increase in headloss due to gradual clogging of the surface area. After some time too little
water will flow through the units and then cleaning will be needed. In case of peak loads of
suspended solids clogging will go much faster and the water flow through the other treatment
units may be blocked completely. In this way the other units are being protected from
excessive loads of suspended solids. Galvis et aL(1993) reported that the potential to react to
increased loads in suspended solids is the reason for the name Dynamic Roughing Filtration.
Galvis et al.(1994) added that the filter bed is the most important element and requires special
attention as it is crucial to the functioning of the system. The grading of the gravel (a layer
with fine grains on the top and a layer with coarser grains at the bottom) differs from other
types of roughing filters where grain size reduces and not increases in the direction of flow.
Just raking the surface provokes the resuspension of the retained material which is easily
13
carried away with the overflowing water. It is very important to keep this grading of the
gravel as otherwise suspended solids will be carried deeper into the bed which would make
cleaning much more difficult. A simple raking of the surface level would then not be
sufficient to restore the filtration capacity.
•r.v.i b.d.
Two types of DyRF can be designed depending on the type of raw water source. The first
type of system is designed to reduce the quantity of suspended solids in the raw water, thus
minimizing operation and maintenance problems in the other parts of the treatment plant.
Galvis et al.(1992) state that this type of DyRF is a clear first sanitary barrier to improve the
water quality because important reductions in suspended solids and also turbidity and faecal
coliforms can be obtained.
A second design possibility for DyRF is used for rivers which normally transport limited
quantities of suspended solids but occasionally show sharp peak loads of short duration. In
this case the filter is designed to quickly block whenever the suspended solid load in the river
is showing a rapid increase. This type of system can be seen as an automatic valve which
blocks partially or totally the inflow to the other treatment units when the river is carrying
too much suspended solids.
The velocity of the surface flow over the filter needs to be controlled because too high a
velocity would carry away the fine gravel. Based on the preliminary experience Galvis et
al.(1992) recommended that a surface flow between 0.05 and 0.15 rn/s is a good range when
the DyRF is designed to improve the water quality. A lower velocity below 0.05 mIs is
proposed when the system has to protect against peak loads in suspended solids. In the second
alternative the effect of sedimentation on the filter medium will help to block the surface
quicly when peaks arrive at the unit, whereareas in the first alternative automatic cleaning of
14
the surface is being strived for. A higher flow will ensure that less material will sediment on
the surface of the gravel.
During a filter run the DyRF will operate as a declining rate filter and flow through the filter
bed will gradually reduce with time. This reduction may be between 20 to 40% during an
operation period of one week provided that no peaks in suspended solids are being received.
Galvis et. al. (1993) reported that the design capacity needs to be at least 1.4 times the
requiered capacity of the system.
The preliminary design criteria differentiating the two design alternatives and preliminary
specifications of the filter medium are been indicated in Tables 2-1 and 2-2 as recommended
by CINARA, IRC (1993).
2.3.4. Limitations
In spite of the promising findings resulting from the DyRF perfomance, further improvements -
of this stage seem very well possible. Up to now DyRF syitems have been basically studied
as a part of multi-stage filtration plants. Only preliminary research on DyRF itself has been
carried out in Colombia.
A particular point for further study is the Impact of the Surface Overflow on the DyRF
Perfomance. Although the available reports do recognize the impact of the Overflow on the
behavior of the units, there is not any systematic study of this parameter to support design or
operational criteria. This thus requires further research in order to achieve a better
understanding of DyRF processes and also to develop adequate guidelines for design.
Table 2-1 Preliminary Design Criteria for two DyRF design alternatives. CINARA, IRC
(1993).
PRINCIPAL ROLE
PARAMETER First barrier to Protection
improve quality against peak
of water loads
Filtration Velocity (rn/h) 0.5 - 3.0 3.0 -5.0
Range size of gravel in the upper layer (mm) 3.0 - 5.0 < 3.0
Surface flow (mIs) 0.05 - 0.15 < 0.05
Surface wash velocity (m/s) 0.2 - 0.4 0.1 - 0.3
Depth of bed(m) 0.6 0.4 —.
15
Table 2-2 Specification of Filter Media for DyRF. CINARA, IRC (1993).
Position of First barrier to improve qulity Barrier against peak loads
layer . .
Depth of layer Diameter (mm) Depth of layer Diameter
(m) (m) (mm)
Top 0.2 3.0 - 5.0 0.20 1.5 - 3.0
Middle 0.2 5.0 - 15.0 0.10 3.0- 5.0
Bottom 0.2 15.0 - 25.0 0.10 5.0 - 15.0
16
CHAPTER 3
Review on Roughing Filtration and Sedimentation
Wegelin (1991) reported that Roughing filtration using coarse and fine gravel is a simple and
efficient method for the removal of suspended solids due to the large surface area available
for sedimentation, absorption and biological and bio-chemical activities in the ifiters.
Roughing filters usually comprise filter material which gradually decreases in size from coarse
in the first part of the system to relatively fine in last part. The filter media is often relatively
coarse and of much larger size than material used in slow sand filtration or rapid sand
filtration as is indicated in the following comparation given by Schulz and Okun (1984).
The filtration rate normally used ranges from 0.3 to 3 rn/h. Schulz and Okun (1984) and
Galvis (1992) reported that the applied rate will depend very much on the type of filter, the
sanitary risk of the water source and the required treatment efficiency. Galvis et.al (1993)
indicate that the treatment mechanisms in roughing filters can be classified as mechanisms of
transport, attraction and purification.
This is the process which brings the particles into contact with the gravel and includes:
3.1.2.1. Screening
This process removes particles of larger diameter than the pores between the gravel grains.
During the filtration process the diameter of the pores will gradually decrease and so smaller
particles will be retained. Huisman (1984) recognized that this mechanism however, has only
limited importance in roughing filtration due to the large diameter of the gravel
(approximately 2mm), which corresponds with a pore of 500 tim. Wolters (1988) indicates
that only at the end of the filter run when the size of the pores has decreased screening of
suspended solids may be of some importance.
3.1.2.2. SedImentation
17
reported by Ives (1975) and Wegelin (1986). The difference is that in sedimentation tank only
the bottom is available as a precipitation surface whereas in roughing filters the total surface
area of the grains is available.
Flocculation of smaller particles creates larger flocs which then also can precipitate on the
gravel surface. Colloidal material however, is not being removed by sedimentation. Wegelin
(1986) and Siripatrachai (1987) have reported that sedimentation is the main mechanism for
particle removal in horizontal roughing filters. This mechanism will be discussed for DyRF
according to specific data collected during the experimental phase.
Huisman (1986) reported that sedimentation removes particulate suspended matter of finer
sizes than pore openings by precipitation upon the surface of the grains. A gravel bed with
a pore space p, one m3 of spherical filter grains with a diameter d1 has a gross surface area
2.
of S~= (6/dJ(1 - p) m
Lebcir (1992) proposed that the available surface area for sedimentation in gravel beds (Se)
can be determined as:
(3-1)
In which, 1/6 = reduction factor for available upward surface, 1/2 = reduction factor for
contact of adjacent grains and 2/3 = reduction factor for high flow which prevent deposition.
Therefore, the equation (3-1) can be expressed for any volume of gravel bed (V
1) as follows:
Se~(1P)Vg (32)
In gravel beds, the Surface loading (S1) taken as quotient of the amount of water to be treated
(Q~) and the area of deposition will now be very small and hence the removal efficiency will
be greater than for a conventional settling tank. For S1 the following equation can be written:
3d8Q~ (3—3)
Se (1P)Vg
3.1.2.3. DiffusIon
The brown movement or molecular diffusion is caused by the collision of particles with water
molecules. Huisman (1982) reported that this movement is important for the removal of
colloidal material and does not affect particles above 2 ~m. Removal efficiency by diffusion
increases with the increase in size of the suspended particles and temperature and with the
reduction of the flow velocity and the grain size.
18
3.1.2.4. InertIal and Centrifugal Forces
During the passage of the water through the filter the flowlines curve around the grains. Due
to inertial and centrifugal forces particles may be forced to leave the flowlines and come into
contact with the gravel grains. The removal efficiency increases with the increase of particle
density and flow velocity and is reduced if larger grains are being used. Galvis et.al (1993)
indicate that particle removal through this mechanisms is limited in roughing filtration because
low filtration velocities are being applied and the gravel grains are relatively large.
3.1.2.5. Interception
Part of the particles in the water will stick to the sides of the grains and in doing so gradually
reduce the diameter of the pores. Initially these particles will stick to the grains where they
entered the filter but gradually part of the material deposited in this area will be transported
further into the filter bed.
Yao et.al (1971) indicate that the removal efficiency through interception is independent of
operational factors such as flow velocity. In 1967 O’Melia and Stum reported that the
efficiency increases with increasing particle size and decreases with increasing gravel grains.
Because of the large size of grains used in roughing filtration interception does not play an
important role in the removal of impurities.
The main forces that hold particles in place once they have made contact with the gravel are:
electrostactic attraction, and mass attraction. A combination of these forces is frequently
referred to as absorption.
Mass attraction between particles (van der Waals force), is always present but very much
decreases with the distance between the two. The impact of this force is therefore very limited
beyond the distance of 0.01 ~zm.The attraction between opposite electrical charged particles
(Coulomb force) is inversely proportional to the square of the distance between the particles.
Like the van der Waals force it may supplement other transport mechanisms when these have
brought a particle into the near vicinity of gravel grains having an opposite electical charge.
As a result of the attachment of materials to the gravel the electrical charge of the gravel
grains will change constantly thus attracting alternately particles with a positive or negative
charge. It appears that active absorption is of limited importance in roughing filtration.
Mass attraction and electrostatic attraction although of minor importance to draw particles
from the water, is considerably more important in holding the particles to the grain surface
19
once they have been put in contact.
Particles of organic origen deposited on the surface of the gravel will quickly become the
breading ground of bacteria and other micro-organisms. This will produce a stick slime layer
to which particles from the raw water may easily attach. The organic material is gradually
assimilated to become part of this sticky layer and may form large chains of organic material
which may easily intercept smaller particles.
The purification processes whereby the trapped impurities on the filter grains are broken down
are independent and therefore better described in combination than separately. The two
principal processes are chemical and microbiological oxidation, but other biological processes
may play a significant role as well.
Through biochemical oxidation organic material is being converted into smaller particles and
eventually into water, carbon dioxide and inorganic salts. Iron salts are also being transferred
into several oxides which form a thick layer around the grains. The chemical and biochemical
reaction only takes place on the surface of the grains where the catalic agents as well as large
quantities of bacteria are present. These mechanisms only can take place after such agents
have been attached to the grains. This biochemical oxidation plays an important role in the
removal of color (true and apparent), and the removal of iron and manganese in roughing
filters.
Wolters (1988), Smet et.al (1989), Wegelin (1989) among others, have recognized the
importance of biological processes in roughing filtration. Through all the mechanisms
indicated before, bacteria attach themselves to the surface of the grains. This concerns both
bacteria which are useful for the removal processes as well as pathogene bacteria. To satisfy
the energy requirement for their metabolism, bacteria oxidize part of the organic material they
will encounter. Furthermore, they convert organic material into cell material for theirgrowth.
For certain types of bacteria such as faecal coliforms, the conditions in the filter are not very
good, and these will gradually die off when attached to the grains.
20
3.2. Sedimentation
3.2.1 General
Sedimentation is the separation by gravitational settling of suspended solids from the water
which carrries them. For sedimentation to take place, the density of the suspended solids must
be greater than that of water (otherwise flotation would occur).
The suspended solids may already exist in the raw water and/or may result from a preliminary
procces, such as hydrolysis, precipitation, etc. Generally speaking, there are two types of
suspended solids: 1. discrete particles, which settle separately with a constant settling rate;
and 2. more or less flocculated particles resulting from natural or induced agglomeration of
the colloidal substances in suspension.
Four distinct types of sedimentation which reflect the concentration of the suspension and the
flocculating properties of the particles were described by Fitch in 1958. These are: discrete,
flocculant, hindered zone, and compression settling sedimentation.
In the discrete settling region, the particle moves down independently at a constant rate with
a negligible amount of natural aggragation. With flocculant settling, one particle overtakes a
slower settling particle and in passing collides, aggregates and consequently settles at a higher
rate.
When particles are sufficiently close, interparticle forces are able to hold them in fixed
positions relative to each other. As a result, the particles subside as a large mass rather than
as discrete particles. This type of sedimentation is called zone settling. When the particles
actually contact each other, the resulting structure of the compacting mass acts to restrict
further consolidation. This is called compression.
The terminal velocity of a settling discrete particle is given by Newton’s formula and taken
into account that Weber (1972) introduced the forces F~,Fb, Fd, which are due to gravity,
buoyancy, and drag respectively. The following equation can be written:
dv (3—4)
m.
The impelling force equals the submerged weight of the particle, as: F~- Fb = (p~-
in which, p and p are the particle and fluid density respectively, g the acceleration of gravity
and V~,the particle volume.
21
The drag force is a function of the dynamic viscosity ~, p, V1, and characteristic diameter d
of the particle.
2 I 2,Based
where,on A~
Newton’s drag coefficient
projected particle area(CD), thedirection
in the friction drag equals:
of flow and FD
V =
CD.AP.P.V1 1 =
settling velocity. Incorporating dynamic behaviour of the particle into the above equation and
taking into consideration that after an initial transient period, the acceleration, dV3/dt,
becomes zero and the V1 becomes constant, the general equation for the settling velocity of
sphere particle of diameter d is:
5 (35)
~ 3 CD ( ~ P .d]°
In 1946 Camp and Fair et al. (1968) have presented the coefficient of drag CD as a function
of the Reynolds number, R,. For laminar condition, R~< 1, and the frictional resistance is
only due to viscous force, CD = 24/Re and the corresponding settling velocity is:
Chaudhari and Tare (1993) have developed a simple regression model to predict Re as a
function of the non-dimensional numbers CDRC2 (diameter term) or CDJRC (velocity term). Both
these terms can be computed as follows:
CD/Re=(4/3) [gv(p
5—p)] (1/V5p) (3—8)
22
2-0.00368094(X 3+0.O00273403(X
1ogR~——1.35873+0.987198(X
1)—0.0251154(X2) 1) 1)’
(3—9)
2)
and X
Where X1 = log(C~R~ 2 = log(CDJR~)
2—0.00239434(X 3 (3—10)
logR1.0.7355—0 .570226 (X3) +0.025256986 (X2) 2)
The frequency distribution of the settling velocities for variQus particles is measured directly
in the laboratory. The details of the experimental set up and procedure of the laboratory
testing are explained in Chapter 5 of this report. In this experiment, the samples of
suspension were taken at different depths at pre-determined time intervals and were analyzed
for turbidity and suspended solid concentrations.
Settling velocity is given by the quotient of the depth of the sampling point to the time
elapsing from the begining of the test. A typical cumulative frequency distribution of settling
velocities is shown in Figure 3-1.
Since 1946, Camp found that all particles in the settling zone of an ideal rectangular basin
travel in a straight-line path as shown in Figure 3-2. They are determined by the vector sum
of the settling velocity V1 of the particle and the displacement velocity VH of the liquid.
The efficiency of the basin depends wholly on surface area (A) and the discharge (Oj, which
together constitute the displacement velocity. The detention time is calculated from the length
of the basin (L). The equation for both the velocity and detention time is as follows:
In which, B is the width of the tank. The velocity of the particle which settles a distance equal
to the effective depth (H) of the tank in a travelling time (tv) is the overflow rate (S =. V0)
of the tank; i.e. S = V0 = H/tv. All particles with a settling velocity V1 S are completely
removed, while for particles with a lower settling velocity than S, the removal ratio amounts
to: h/H = V,/S. The overall removal efficiency is given by:
23
3
0 0.5 1.0 m~seC
t5x10
which can be evaluated from the cumulative frequency distribution curve, suggested by Camp
in 1946 and further by Ingersel et.al. in 1956, as ilustrated in Figure 3-3. The integral part
represents the shaded area, which can be found graphically by drawing a horizontal line in
such a way that the two shaded areas are equal.
0
24
.~ ~8O
I60
~4O
20
0 0
0 v~ 03 v~ 0 0.5 V
1 1.0 1.5
25
CHAPTER 4
4.1. General
Scour refers to the removal of material by running water. Galay et al.(1987) have reported
that scour in rivers and canals is generally the result of secondary currents or vortices that
occur in conjunction with river features such as bends, abrupt changes in flow directions,
obstructions, constrictions, confluences, subsurface sills, control structures and piers of
various types and sizes.
H. Chang (1988) pointed out that scour criteria are involved with physical conditions
pertaining to the threshold of motion for the material. Therefore, determination of such
criteria is the prerequisite for scour control and sediment transport.
4.2.1. Introduction
Particle movement will occur when the instantaneous fluid force on a particle is just larger
than the instantaneous resisting force related to the submerged particle weight and the friction
coefficient.
The basic aspects that play an important role in the initiation of motion will be descibed in
the following sections.
The motion of the particle is under the interaction of two opposing forces: the applied force
and the resisting force. The former is caused by the hydrodynamics of flow; the latter is
associated with the submerged weight. The particle will be moved or entrained if the applied
forces overcome the resistance. At the critical condition for entrainment, that is, threshold of
movement, the applied forces are just balanced by the resisting force.
In a flowing stream, the forces acting on a grain of noncohesive sediment lying on the stream
bed consists of the hydrodynamic drag (FD), the hydrodynamic lift (F1), and the submerged
weight (W), as shown schematically in Figure 4-1. The drag FD is in the direction of flow,
and the lift FL is normal to the flow. The lift force is not considered seperately in most
analytical models because of the difficulty in determining its magnitude, Moreover, lift force
is directly related to the drag. It depends on the same variables as the drag force and therefore
the effect of the lift force is automatically taken into account by the empirical coefficients.
At the threshold of movement, the resultant of these two forces is along the direction of the
26
FL friction angle; that is, the ratio of forces on the
Point of supporl grain acting normal to the bed is equal to tan e.
C =ctanO (4—1)
(y3—y)d
-o~~
w This situation can be expressed as a function of
critical shear stress r~. For horizontal bed, 4) is
Figure 4-1 Schematic of forces 0~equalsto zero and the formula becomes:
sediment grain on sloping bed. F÷Wsin~________ (4-2)
tanO= D
Wc os 4,
In which y1 and y are the specific weights
for sediment and water respectively, c =
constant value which must be determined experimentally and d grain diameter.
Major variables that affect the incipient motion of uniform sediment on a level bed include
r~,d, y, - y, p and v. From dimensional analysis, they may be grouped into the following
dimensionless parameters
t Lid
C •c ) (4—3)
(y8-y)d v
Where U.,, = (r~/p)½is the critical friction velocity, and v = kinematic viscosity. The left-
hand side of this equation is the dimensionless critical shear stress and is often referred to as
the critical Shields stress, r.~.The right-hand side called the critical boundary Reynolds
number is denoted by R.. When any bed shear stress T0, other than r~,is used in the two
quantities in equation 4-3, they become the Shields stress and boundary Reynolds number and
are designated as r. and R., respectively.
Figure 4-2 shows the functional relationship of equation 4-3 based on experimental data,
obtained by Shields et. al. in 1936 on flumes with a flat bed. It is generally referred to as
the Shields diagram. Each data point corresponds to the condition of incipient sediment
motion or vanishing bed load.
Three different regions can be identified in Shields diagram based on Reynolds number (R.):
1. Laminar region where R. is less than about 2, the particle size is less than the thickness
of the laminar sublayer and, hence, is enclosed in a thin laminar film. Since the
boundary is hydraulically smooth, the movement is mainly caused by viscous action; the
critical shear stress is inversely related to R.~or r.~ CIR.~,where C is a constant.
27
2. The transition region of intermediate boundary Reynolds numbers where the grain size
has a magnitude similar to that of the laminar sublayer; therefore, the movement is
partially influenced by viscosity. The critical Shields stress has a minimum value of 0.03
at the R. value of about 10.
3. The turbulent region presents large Reynolds numbers (R. > 400), and the laminar
sublayer is interrupted by the grain size. For this hydraulically rough boundary, the
citical Shields stress has a constant value of 0.06, independent of the Reynolds number.
In 1963, Zeller suggested a lower value of 0.047 in this region.
~, 9~”
.r o~
MT~ ~
F~6y~
111111
91b~M~9
‘~y I.0~l;.
II
1h~}— ~
.~ ~ ~
+~euu.~~.n
263
265
~IHII ii XS~eU&WESJ 265
-~ :~ (ás.n97v~*.I
: = : = 261
0!i
— ~ t. ~- ~ s.~ (~~l 210
:— :- — :
-
-I— —- — ~.
- -~•~tA~U~UW1~.P
~ s~a
75
~; :: J.b.~~+Ia1~fI
‘°~‘
I E - jni~.
0 100
I~
0~
-~[O.1(-!.f_1)gd)h/2 (4—4)
V y
which appears as a family of parallel lines in the diagran. From the value of the third
parameter, the value of the critical Shields stress is obtained at an intersection with the Shields
curve from which r~can be calculated.
28
4.2.3 Influence of Criterion
Van Rijn (1989) has reported that the complexity of defining a critical bed-shear stress for
initiation of motion is mainly caused by the stochastical character of the driving fluid forces
and the stabilizing resisting forces and by lack of an unambiguous definition of initiation of
motion.
Since 1936, Shields found that the critical bed-shear stress is the zero transport rate after
extrapolation of measured transport rates, another group of researches Neil et aI.(1969)
indecate that the critical bed-shear stress is the number of particles displaced per unit area
and time and Deift Hydraulics (1972) experimentally recognized that it is the qualitative
transport stage based on visual observation and found measurable quantities of transported
particles at much smaller bed-shear stress than the Shields-values.
Breusers (1988) has reported that experiments with particles of differents shapes show that
the r.~parameter is not much affected by the shape of the particles when the nominal
diameter (diameter that yields the same volume) is used as the characteristic parameter. Very
flat particles have larger r.~values (factor 1.5 to 2).
In 1965, Egiazaroff experimentally recognized that Gradation has an effect when the size
range is rather wide (d~jd~ > 3), because the large particles will be more exposed, while
the smaller particles are shielded by the larger particles. Armouring will occur when the bed-
shear stress is not large enough to move the largest particles of the bed material. When there
is no supply of smaller particles from upstream, all smaller particles will eventually erode,
and the coarser particles will form an armour layer preventing further scour. Based on
experimental results, the median diameter (d50) of the armour layer will approximately be
equal to the d85 of the initial bed material. This phenomenon has been observed downstream
of weirs.
Mantz (1977) studied the initiation of motion offine cohesionless flaky sediments with particle
sizes (d9) in the range of 10 to 100 j~mand the equation 4-5 represents the critical bed- shear
stress (r00):
(4—5)
i) Bed forms determine the roughness of a stream and a change in bed form can give
29
changes in friction factors of 4 or more.
i) Lower flow regime: (Froude Number F1 < 0.7). The bed form is either ripples or
dunes or some combinations of ripples and dunes all of which are triangular shape
elements of irregular shape. The common mode of bed material transport is for the
individual grains to move up the back of the ripple or dune and avalanche down its face.
ii) Upper flow regime: (F1 > 0.7 ± 0.2). The usual bed form are plane bed or
antidunes. The mode of sediment transport is for the individual grains to roll almost
continuosly downstream in sheets a few diameters thick. Simons et al.(1972) believe that
as soon as the sediment transport process is established, ripples and dunes are formed
on the bed.
Sediment mixtures with a fraction of clay particles (d, < 4 ~m) larger than about 10% have
cohesive propierties because electrostatical forces comparable to or higher than the gravity
forces are acting between the particles. Consequently, the sediment particles tend to stick
together forming aggregates known as flocs whose size and settling velocity are much larger
than those of the individual particles.
Fluid-sediment mixtures consisting of water, fine silts, clays and organic materials are
generally called muds. When the bed consists of silty and muddy materials, cohesive forces
between the sediment particles become important and depending on the type of clay minerals,
the effect may be more or less pronounced. Biological activity at the bed may also influence
the critical values for initiation of motion especially in muddy and silty environments as is
reported by Van Rijn (1993).
Parchure and Trimbak (1985) found that critical shear velocities and erosion rates are greatly
variable, depending upon type of mud and consolidation time. Tests with the actual sediment
are necessary to obtain accurate values.
Fresh mud deposits have a very loose texture of mud floes which already have a low density
themselves. The95%
3 of which wet bulk density
or more of such
consists a deposit
of water. may
In this be within
stage the range
the cohesive of in
forces 1050
theto 1100
deposit
kg/m
are still very low and scouring can occur easily. A sand bed with small percentages of silts
and clays (silty or clayed sand) already shows a distinctly increased resistance against erosion.
Van Rijn (1989) reported that most of the clay particles have a negative charge and the
flocculation process for cohesive particles requires particle collisions: The three most
30
flocculation process for cohesive particles requires particle collisions: The three most
important collision mechanisms are: (1) The Brownian motions of particles (d~< 4 ~m) due
to the random bombardment by the thermally agigated water moleculas, (2) turbulent mixing
due to presence of velocity gradients in the fluid and (3) differential settling velocities because
the larger particles have larger settling velocities and may therefore “fall” on the smaller
particles.
Winterwerp et al.(1992) have recognized that the effective settling velocity (V1) in the range
of 0.001 to 1.0 mm/s decreases with increasing critical bed shear stress (r00). Experiments
carried out with cohesive sediments
2), have taken
shownfrom western Scheldtwhich,
the following ratio: the maximum r~
applied amounted to 0.2 (N/rn
¶ 588 (4—6)
0~=o.3X~o-’v;°~
In which V = effective settling velocity (mis) and r~= critical bed shear stress (N/m2)
Scouring velocity (U) can be expressed as a function of critical shear stress as follows:
U=( . . (4—7)
p ~
f~=O.24[1og( (4—8)
Experimental research shows that effective roughness is mainly related to the largest particles
of the top layer of the bed. Van Rijn (1992) proposed that k~can be taken as lc~ = (Id
50
where Il is ranging between I and 2 for gravel beds and d50 = median particle diameter of
bed material.
31
CHAPTER 5
Materials and Methods
5.1. Introduction
The main objective of this research project is to identify the main treatment processes
involved in DyRF and to assess the impact of different overflows (Q0) on treated water quality
as well as in scouring of particles and to suggest design guidelines for DyRF. The research
was carried out in the Research and Technology Transfer Station of CINARA in Cali,
Colombia, where two pilot plants with declining-rate ifitration were arranged to treat raw
water drawn from Cauca river. Two different types Qf experiments were conducted. First a
series of tests in which the initial filtration rate in the DyRF was set at 2.0 rn/h and different
overflow rates were used to assess the influence on the water quality improvement. Thereafter
three runs were made with higher initial filtration rates.
1. Inlet Structure: This is designed for control, measurement and flow distribution.
2. Main Box: This is the most important device in the DyRF structure and contains the
filter bed and underdrain system. The dimensions were taken as: length = 1.50 m; width
0.50 m; and depth = 0.70 m, including free bord taken as 0.10 rn.
For filter beds, three different gravel layers were placed:
Upper layer (thickness = 0.20 m): 4 6 to 13 mm, D10 = 5.1 mm, D~= 6.4 mm
and D~= 8.6 mm.
Middle layer (thickness = 0.20 m): ~ 13 to 19 mm, D10 = 9.0 mm, ~ 10.0
mm and D~= 1.23 mm.
Bottom layer (thickness = 0.20 m): 4 19 to 25 mm, D10 = 13.0 mm, D30 = 15.4
mm and D~= 20.1 mm.
The Underdrain system was formed by manifolds 4 50 mm with 30 orifices 4 12.7 mm.
3. Outlet Structure: To facilitate the sampling and measurements activities two different
chambers were constructed, one for overflow and one for treated water.
4. Drain Valve: One drain valve per each DyRF was installed in order to facilitate the
operation and maintenance activities.
32
Raw water
Mixing tank
Influent: Q~ Influent: Q~
~P~ometer~
Q2,i board 1
DyRF2 DyRF1
Overflow: Overflow: Q~
Q2,3 01,3
Effluent: 0e Effluent: 0e
02,2 Qi,
2
Figure 5-1 Layout of the pilot plants.
33
For the second period Q was kept constant at 1.5 1/s and the initial filtration rates were set
at 3.0, 4.0 and 5.0 rn/h. Two filter runs were tested per each filtration velocity while gravel
sizes and their layer thicknesses remained constant. An overview of the applied initial flow
velocities is indicated in Table 5-1. A list of material and tools used in the research is
presented in Table 5-2.
Table 5-1 Overview of initial flow velocities in the different research periods.
1. Riutangular filters (1.5 m lesigth. 05 m width, 07 m depth) x2 14 Whitman filter paper (934 AH 1.2 tim)
2.Orivel4,ótol3mmipprox.015m’ 15.Bacteriologicalkitx2
4. Gravel 4i 19 to 25mm approx 0 15 in’ 17 Drying oven, for use it 103 ‘Cx 1
34
5.4 Overview of experimental parameters
The parameters and sampling frequency used in the research are summarized in Table 5-3.
These parameters have been selected on the basis of information on roughing filtration in the
literature and available equipment.
Water samples were taken at the same time at the different sampling points after which
analyses were carried out. The parameters were: turbidity, suspended solids, volatile solids,
true color, total iron and faecal coliform Counts. The selection criteria for physical, chemical
and bacteriological parameters are supported below and their experimental tests were
accomplished based on guidelines given by APHA (1986) and APHA (1989).
Turbidity: This is the universal parameter used to evaluate water treatment system. It is
related with particle sizes present in the water as colloids or suspended matter which
frequently is associated to high bacteriological contamination. High levels of turbidity can
protect micro-organisms from the effects of disinfection, stimulate the growth of bacteria and
exert a significant chlorine demand.
Suspended Solids: The amount of suspended matter removed by DyRF can be expressed in
terms of suspended solids. Whatman filter paper of pore 1.2 ~imwas used.
True Color: Colour in drinking water may be due to the presence of coloured organic matter
such as humic substances, metals such as iron and manganese, or highly coloured industrial
waters.
Faecal Coliform Counter: The principal risks to human health associated with community
water supplies are microbiological. Yet this comprises a wide range of bacteria and therefore
faecal coliform counts are commonly used as indicator to determine the degree of
bacteriological contamination.
Furthermore the following physical analyses were made: flow and head loss measurements,
column settling test, mass density and mass balance. A general description of how these
analyses were carried out is given below.
Flow Measurements: The volumetric procedure was accomplished and influent (Q), effluent
(Q) and overflow rate (Q
0) were measured at the same time at three different points, twice
a day.
35
Table 5-3 Overview of selected parameters an-d samnling frequency
Column Settling Test (CST): This indirect method was used to establish the sedimentation
diameter and frequency distribution of settling velocities. Several testings during the research
project (according to raw water quality variations and overflow rates) were accomplished and
water samples to fill the CST were drawn from Q, Q0, Q0,, at the same time. The quiscent
settling column test apparatus used in this experiment consisted of cylindrical columns made
of plexyglass of 0.30 m diameter and 1.0 m depth with four taps located at 0.26, 0.46, 0.61
and 0.71 m below the water surface. - -
Mass Balance: In order to know the particle removal efficiency for each DyRF run, mass
balance was determined and the suspended solids variation for each sampling point (influent
C1, effluent C~, and overflow C0) was calcu]ated. Moreover, the total storaged suspended
solids concentration (TSSS) in each pilot plant was measured and added to mass balance
according to:
36
TSSS=M1- (MR+MO)
(5—3)
Determination of Sediment Density: For different filter-runs three Imhoff cones were
respectively filled with influent, effluent and overflow samples drawn from the DyRFs and
the sedimentation process was given for three hours. After that, three different layers were
/ observed at the Imhoff bottom. Each layer were carefully drawm and its sediment density
I determined. The experimental procedure is given by Van Rijn (1986).
37
CHAPTER 6
Presentations of Results
6.1. Introduction
The results have been organized in five blocks in line with project objectives:
5. Other measurements: The removal efficiency for suspended solids in a DyRF has been
estimated on the basis of a mass balance. For each Q, a mass balance was made as well as
for the experiments with the different filtration rates.
Furthermore headloss was measured at different depths and flow velocities for Q1, Q and Q0
were made per each Q1.
Although during the project it was tried to keep Q, constant this proved not possible due to
operational problems not under control of the project. This has resulted sometimes in flow
variations of 50%. To limit the impact of this variation only the filter runs have been taken
into consideration were variations were less than 15% of the nominal Q~value.
38
6.2 Presentation of results
CINARA’s research station receives water from the highly polluted Cauca river, which
receives water from the highland rivers and untreated sewage from small and large
settlements, as well as waste and runoff of agriculture lands.
During the research period the raw water in Cauca river had average turbidity levels ranging
from 71 to 167 NTU with peaks ranging from 254 to 420 NTU, suspended solids ranging
from 146 to 333 with peaks between 367 to 881 mg/I, volatile solids ranging from 54 to 64
with peaks between 86 and 154 mg/i, true color ranging from 60 to 109 with peaks between
102 to 157 PCU, total iron ranging from 13.37 to 15.76 with peaks between 25.60 to 57.50
mg/l, and faecal coliform counts ranging from 19440 to 64143 with peaks between 29000 and
242000 FCU/100 ml.
Figures 6-1 (a) to 6-1 (1) give an indication of the raw water quality and its variation for each
of the the five test periods. More detailed descriptions are presented in AppendIx 1 Tables
1-1 a) to 1-1 1). This includes minimum, mean and maximum values, standard deviations and
number of samples for the following parameters: turbidity, suspended solids, true color,
volatile solids, total iron and faecal coliform counts. On the basis of the removal efficiency
obtained in the first research period (averages ranging from 83.4% to 87.2%) suspended
solids removal was selected as indicator for the experiments with different ffltration velocities
in the second research period.
ti t
2 t3 t4
— ~
39
1000
900
800
700
~600
500
~400
~300
200
100
0
LI L
2 t3 t4 Es
— Mw
200
150
g 100
> 50
0
tI LI LI
— Minirrui Mean
200
150
~10c
I-
50
LI LI LI
— ~rIrTu~ M~xvyij,
40
100
80
60
40
0
I—
20
0
LI LI LI
— ~r,drru, Mean Maxirr,.r
— M~r~s~ Mean
On the basis of the test results (Appendix A Figures A-i a) to A-i c)) a linear regression
model was stablished which gives the relation between suspended solid concentration and raw
water turbidity. This model is presented in Appendix 2 Table 2-1.
Figure 6-2 illustrates the typical pattern of the remaining concentration (P) and the settling
velocity (V) for samples taken simultaneously from Q, Q~and Q0. The value of P is
calculated as (C,/C,,)x 100 where C,, is the initial suspended solid concentration in Q. at t =
0 and C, are the measured values at different times and depths during the the CST. On the
basis of Figure 6-2 it may be concluded that:
41
0~
I 0 10 20 30 40 50 50 70 80
3. Only some 17% of the initial concentration of suspended solids in the raw water
remains in the treated water thus indicating the impact of DyRF on suspended solid
removal.
Laboratory analysis indicated three types of densities in the sediment of the CST. The samples
were taken from Q~and Q,, and determined at a raw water temperature of 25 °C(±3.1 °C)
using the procedure indicated in section 5.4. The following densities were found: Pi = 2650
kg/rn3, P2 = 2125 kg/m3 and p~= 1365 kg/rn3. Furthermore the average density of the deposit
on top of the gravel in the DyRF was established as 1564 (STD = 78.2) kg/rn3.
With help of the equations presented in 3-8 and 3-10 the diarneters of the particles (d,) present
in the raw water have been calculated for different densities Pi, p
2 and p3 at 25 °C.Results
are presented in Table 6-1 and Figure 6-3. Table 6-1 also includes the values of theP
3 crn/s,
remaining concentration (P) for Q, and Q0 taken from
is about 100% which implies that all particles in Q and Q Figure 6-2. For V3 80x10
3 crnls. This
indicates the presence of particles with d 0 have a V < 80x10
1 < V31, 38 and 64 ~m for densities
3 respectively. For P of 70%, of 2650,
3 cm/s which 2125a
implies
and
high 1365 kg/rn
concentration 3 is7 smaller
of fine particles with 4, < 5, and 13 than 5x10
j~mrespectively.
42
Table 6-1. Particle diameters for different sediment densities and their remaining
concentrations as a function of settling velocities.
80 31 38 64 100 100
60 27 32 57 92 91
20 15 18 32 91 90
5 5 7 13 70 70
1 0.84 1.10 2.70 17 17
...-i -r — A1 ‘Fr - r- —- rL_ -. . —-
In Table 6-2 the results are being presented for both research periods. This concerns:
i) Variation of Q,, Q~and Q throughout the filter runs for different initial Q~’sand for
different filtration rates.
The reduction of Qe and the increase in Q
0 during the ifiter run can well be observed.
ii) Suspended solid concentrations in Q,, Q~and Q0. These figures show a wide variation
due to the variation in the raw water. Table 6-2 does not include data on suspended
solids for the overflow when Q, = 0.42 I/s because when V = 2.0 rn/h Q0 0.
100
90
80
70 — _a
— _.— —
60
— e- —
50
40
30
20
10
0
0 20 40 60 80 100 120
In order to get a better understanding of the results a general description of the normal
performance of the DyRF during one test period is given.
43
Table 6-2 Experimental data for different nominal Q, and over filter run
Nominal O~ 1 01/i V = 2.0 rn/h run: I NOminal Q~ 0.42 Eli V 2.0 rn/h run: 1
Time Flow van~ion(1/.) Suipended ,olidz (mg/I) Flow variation (I/i)_________ Suipended polids (mg/I)
(dayi)
(~ Q. Q. S_S
1 SS• SS• Q Q. Q. SS~ SS~ S-Se
Nominal Q 2.0 I/a V = 2 0 rn/h run 2 Nonunsi Q~ 1.50 I/i V 2.0 rn/h run 2
1 1.40 0.65 0.75 442 165 384 1.20 0.76 0.44 470 205 391
2 1.66 0.24 1 42 597 136 484 1.38 031 1.07 595 169 494
3 1.65 005 160 180 8 176 1.40 0.04 1 36 20* 14 180
NominslQ, = 1.50 1/a V = 5.0 rn/h run 1
44
Under normal conditions the inflowing water Q~is divided into two flows a horizontal flow
(Q
0) and a vertical flow (Q~).
i) The horizontal flow, varies along the length of the filter and is minimum at the
begining of a filter run and maximum when the ifiter is clogged. (see Table 6-2). The
typical flow variation is also shown in Figure 6-14.
During the development of the filter run, gradually a sludge deposit is formed with
a variable depth between 0 at the beginning and 15 mm at the end of a filter run,
with the greatest depth in the first half of the surface area.
ii) Vertical flow. This flow enters into the gravel bed gradually causing the clogging
of the filter. Q~varies and is maximum at the beginning and gradually reduces during
the filter run. (see Table 6-2 and Figure 6-14).
For horizontal flow the difference between the suspended solid concentration in Q and
Q0 determines the horizontal removal efficiency for each overflow rate (S = Q0/A, A
surface area). This efficiency has been calculated for each Q~based on experimental data
shown in Table 6-2.
The theoretical removal efficiencies R (removal ratio) have been calculated for different
Q.’s on the basis of sedimentation theory with help of the cumulative frequency
distribution and based in the equation 3-12. Figures 6-4 a) to 6-4 e) show these theoretical
and the measured suspended solid removal efficiencies as a function ofS for different Q1’s.
2
a
•ff~ciency •ffici~cy
Figure 6—4 a) Removal efficiency comparation for Q1 0.70
1/s.
Figure 6-4 a) shows efficiencies below 5 % in the first days of the run and thereafter the
efficiency increases to 36% and then drops to 20%. These values are not in line with the
data obtained in subsequent runs and as there isiio logical explanation for this behaviour
it is most likely resulting from an experimental mistake. The other Figures 6-4 b) to 6-4
e) show high efficiencies at3the begining
cm/s, of the filter
but thereafter run,
these with a maximum
efficiencies of 29% for
drop to between Q
2 and
= 1.0
9%. Us and
The higher 94.7x10
S =values clearly divert from the theoretical values and they can not be
logically explained on the basis of sedimentation theory. A possible reason could be the
design of the system in combination with cleaning. As a routine maintenance operation the
45
top gravel layer of the DyRF was removed after each run and replaced after washing. It
was observed that in the beginning of each run the gravel bed was slightly higher and
extended some 5 to 10 mm over the overflow weir. After one day the gravel pack settled
and reduced in height till the level of the weir. The initial situation makes that the water
overflowing the weir is actually passing through the gravel bed whereas the second day
this no longer is the case. Therefore it may be concluded that the data obtained for the
second and subsequent days are more in line with normal operation conditions and show
average removal efficiencies of less than 10% which is much more in line with theoretical
sedimentation efficiencies.
I
0 50 100 150 200 250 300 350 400
100
80
I
0 50 100 150 200 250 300 350 ~400
Figure 6-4 e) shows a negative efficiency for Q~= 2.65 1/s this is a results from the high
flow velocity which is able to produce scouring of deposited material.
In Table 6-3 the theoretical and measured efficiencies are being presented for the
beginning of the filter run (situation 1) and for the end (situation 2) for different Q’s.
In situation 1 the difference in theoretical and measured efficiency is considerable with
46
averages of 14 and 4.7% respectively which is probably due to the filtration effect before
the water passes over the weir. For situation 2 and particularly for the experiments with
higher flow velocities the theoretical and measured efficiencies are much closer. The
efficiency measurements of 20% for Q, = 0.7 Ifs and 29% for Q, = 1.0 1/s have been
excluded from the calculations as they are probably resulting from experimental mistakes.
100
80
I 60
40
I 20
0
0 50 100 150 200 250 3(X) 350 400
I
I
0 50 100 150 200 250 300 350 400
Overflow rateS, S (105-3 ci~’sI
—4-’--- Theoretical --8— Real
efficiency efflcler~y
Figure 6-4 •) Removal efficiency comparation for Q1 2.65
1/s.
For the vertical flow (Q) removal efficiency have been calculated on the basis of
difference in suspended solid concentration in the influent (SSJ and effluent (SS~)
presented in Table 6-2.
The theoretical efficiency is calculated on the basis of the cumulative frequency
distribution for each Q1 based on the equation 3-12.
The variation of the efficiency over the filter run is expressed in terms of surface loading
(S1), which is a function of Q~and the available area for sedimentation (Si). These have
been calculated with help of equations 3-1, 3-2 and 3-3. Figures 6-5 a) and 6-5 b) show
the efficiencies for different S1 values for Q, = 0.70 and 2.65 Ifs over the length of the
filter run. Table 6-4 shows the theoretical and measured efficiencies in suspended solid
removal for different values of Q1 at the beginning of the filter run (condition 1) and at the
end (condition 2).
47
Tabla 6-3. Suspended solids removal efficiencies on surface area of a DyRF due to Q0.
Table 6-4 Suspended solids removal efficiencies in gravel beds due to vertical flow (Q~).
The Figures 6-5 a) and b) and Table 6-4 indicate that the measured efficiencies are slightly
higher 6.1 % for condition 1 and 6.2% for condition 2, than the theoretical efficiencies.
48
100
~80
160
2 4 6 8 10 12
A slight improvement in removal efficiency is observed in the course of the filter run.
This is in line with the sedimentation theory as the value of S~decreases due to the
reduction in flow resulting from gradual clogging shown in Figures 6-14 and 6-15
Although the values obtained for Q, = 0.42 1/s have been included in Table 6-4 these have
not been taken into account in the calculations because of irregularities in the performance
of the DyRF. The efficiency for Q~= 0.42 1/s varies strongly during the filter run as a
result of the hydraulic performance of the system. In the beginning no overflow is
produced and part of the DyRF remains dry at the surface. Gradually this pattern changes
until the total bed is under water. This irregular water distribution will no doubt have an
impact on removal efficiency. This different behaviour is also shown in Figure 6-6 were
the relation between suspended solids removal efficiencies and surface loading S1 is given
for different Q1’s. For the other values of Q behaviour is very much
3 cm/s thebeginning
at the same with
of
average efficiencies of 76% (STh
the runs and 95% (STh = 2.4%) with S = 2.3%) with S1 = 10.5x10
1 = 0.105x10~cm/s ending filter run. The
average efficiency for the total filter run for different Q’s was above 83.3% (STh = 9%).
100
—S 80
60
2 4 6 8 10 12
49
S
I
0 2 4 6 8 10 12
&rface Ioa~ 51 (105—3 an/sI
—4— 01 —8— 02 ~-- 03 —e— Q4 —— 05 06
The impact of- the overflow on the quality of treated water is shown for different
parameters in Figures 6-7 a) to 6-7 1’) as function of the efficiency over the filter run. Day
1 corresponds with the start of the filter run whereas the last days depends on the
reduction in the effluent flow Q. When Qe dropped below 0.05 1/s the filter run was
stopped. More details including statistical information is presented- in Appendix 1 Table
1-1 a) to 1-1 1) and Appendix 3, table 3-1.
100
80
~- 60
I 0
2 3 4 5 6
Filtraticn time (daysl -
50
100
80
0
1 2 3 4 5 6
100
I:
2 3 4 5 6
0~
I
1 2 3 4 5 6
51
100
80
0. 60
40
20
0
1 2 3 4 5 6
2.00
1.50
1.00
0.50
000 1 2 3 S 5
S -
The efficiencies showed a small increase in the course of the filter run.
In the second research period the effect of the higher filtration rates (V = 3.0, 4.0 and
5.0 rn/h) with a constant Q, of 1.5 1/s were evaluated and compared with the results of the
second run in the first period Q, = 1.5 1/s and V = 2.0 rn/h. Figure 6-8 shows the
removal efficiencies for suspended solids for the second period. Table 6-5 shows the main
statistical parameters when suspended solid removal efficiencies were evaluated under
different filtration rates and constant Q, = 1.50 1/s. It may be observed that at the start
of the run efficiencies are close to 60% for S1 = 23.3x103 cm/s and increase towards the
end to 95% for S~< 1x103 cm/s.
52
--
80
I 60
40
I- 20
0-
0 5 10 15 20 25
rn/h mTh
Results for V = 2.0, 3.0 and 4.0 rn/h are rather similar with effective removal efficiencies
of 73.8% (STD = 18.5%) for V = 4.0 m/h and 83.8% (STh = 13.4%) for V = 2.0
rn/h. For V = 4.0 rn/h a low removal efficiency was obtained in the beginning, but this
is likely due to insufficient cleaning of the gravel. With a filtration rate V of 5.0 rn/h
efficiencies are considerably lower and vary between 58% at the start to 76% at the end
with an average of 63.6% (STD = 9%) over the filter run.
Tabla 6-5 Main statistical parameters for different filtration rates and constant Q, = 1.50
1/s.
Statistics Filtration rates V (rn/h)
2 rn/h 3 rn/h 4 rn/h 5 m/h
Mean 83.8 78.5 73.8 63.6
Maximum 97.0 95.6 93.3 76.3
Minimum 71.0 62.7 56.4 53.3
STD 13.4 16.5 18.5 9.2
Standard error 6.7 9.5 10.7 3.7
95% confidence 13.2 18.7 21.0 7.3
99% confidence 17.3 24.6 27.6 9.6
Size 4 3 3 6
Table 6-6 shows the hydraulic parameters which have been measured at the end of the
filter runs. Overflow rates have varied between 86.7x103 and 358x103 cm/s and flow
velocities between 10 to 18 cm/s for Q, = 0.7 and 2.65 1/s respectively. In a similar way
the Reynolds numbers have differed between R~= 1074 for Q = 0.70 1/s and laminar
flow conditions and R~3415 for Q~= 2.65 1/s and turbulent flow. The average flow
velocities have been estimated as 0.75 times the velocity of the flow at the surface as is
reported by Yalin (1977).
53
The flow at the surface has been measured at the beginning during and at the end of each
filter run. Flow variation was rather small with velocities starting with 9 cm/s for Q
0.70 1/s and ending with 10 cm/s and starts with 15.5 cm/s for Q~= 2.65 I/s and ending
with 18 cm/s.
Table 6-6 Main hydraulic parameters at the end of filter runs for different nominal Q~
Flow Hydraulic Reynolds Froude
Nominal Overflow Overflow Water Velocity Gradient Number Number
Q (Its) Q0 (its) rate S depth 3 X (cmls) IxiO3 Re F
(cm/s)xlO’ (rn)x10 (rn/rn) 7
A theoretical relation has been stablished for the calculation of the scouring velocity (U)
as function of particle diameter (d1). For the three densities the scouring velocity has been
calculated using the equations 4-5, 4-7 and 4-8 for non cohesive sediments. This
theoretical expression is based on initiation of motion of fine cohesionless flaky sediments
with particle diameters between 10 to 100 ~
In Figure 6-9 is shown the theoretical relationship between noncohesive particles with
d1 and scouring velocities (U) for different sediment densities: 2650, 2125 and
diameters 3.
1365 kg/rn
40
35
~30
3
25
20
> 15
10
0 20 40 60 80 10.0 120.
Particle diameter. da (105—6 ml
—4--— 2650 --8--- -2125 —0— 1365
k~/m3 kg/m3 k~(m3
Based on equation 4-6 to 4-8 and for cohesive sediments, U was calculated as a function
of V and critical shear stess (r~).In Figure 6-10 is shown the U variation versus d
3. 1 for
Pi = 2650 kg/rn
As shown in Table 6-6 for Q
0 = 0.65 1/s the minimum flow velocity X of 10 cm/s is
obtained. Figure 6-9 shows that non cohesive particles -with 4, < 27, 32 and 80 ~m (for
54
2650, 2125 and 1365 kg/rn3) can be scoured. These particles are approximately 91% of
the particles in Q~and Q
0. The
3).same behaviour may be expected for cohesive sediments
with maximum
For d1 >3.2 ~imX =(2650 kg/rn non cohesive particles with d
18 cm/s
1 < 57, 82 and 176 ~m and
cohesive particles with 4,> 0.5 ~m will be scored. Under this maximum X, almost 100%
of existing particles in Q~and Q0 can be scoured.
25
20
3
15
8
10
0
0 5 10 15 20 25
P~rticIe diameter. c~(10E-~&m)
Figure 6-10 Theoretical relationship between cohesive
particles and scouring velocities.
Figure 6-11 indicates the sludge accumulation distribution in the different layers of the
DyRF for different Q1’s with V = 2.0 rn/h. It is clear that the main accumulation takes
place in the upper layer (L1) with an average of 80.5% (STD = 9.5%). Tn the second
layer (L2) 11 % (STD = 4.2%) is retained and in the bottom layer (L3) 8.5% (STD =
6.2%). For Q, = 0.42 1/s a different pattern is found due to the different hydraulic
behaviour as explained in 6.1.
80
60
40
20
0.1 EJ as
02 03 04 05
Figure 6-11 Storaged sludge per layer under V 2.0 in/h and
different Q.
55
Figure 6-12 shows the sludge retention distribution for the filter runs with respectively V
2.0, 3.0, 4.0 and 5.0 rn/h and Q1 = 1.5 1/s. The averages are for L1 = 75.8% (STD
= 14.9%), L2 = 13.5% (STD = 6.7%) and L3 = 10.8% (STD = 8.3%). A different
behaviour was found for V = 5.0 rn/h were the average was 50% for L1, 25 % for L2 and
25% forL3.
Figure 6.13 shows the sludge storage per square meter per day for V = 2.0, 3.0, 4.0 and
5.0 rn/h. The figure shows an increase in sludge storage for V 3.0 rn/h due to the
increase in Qe. For V 4.0 rn/h the stored quantity reduces due to the high flow rate
which causes resuspension of sediment material.
100
80
I 60
40
(I, 20
a
Gravel layers Q 1. L2. L3)
V=2 rrilh~ V3 m/h~ V~4 rri/h~ V~5 mdh
10
~ 6
6
4
Piltrati~rate. V lm!ri)
Vi V2 V3 V4
2mTh 3m/h 4rn/h 5mTh
Figure 6-14 shows the typical variation in water flows Q, Q and Q0 over the filter run.
Q~varies between 0.42 1/s (V = 2 rn/h) to almost 0 1/s after 4 days of filter run. The
reduction of Qe implies an automatic increase in Q0 until almost reaching the value of Q.
56
In Figure 6-14 day 1 corresponds with the start of the filter run and the last day will
depend on the length of the run. In general the filter is stopped when Q < 0.05 ifs. A
similar behaviour was found for V = 3.0, 4.0 and 5.0 mlh as shown in Appendix B
Figures B-i a) to B-i 1).
-
DyRP2, INrLU~NT 01 — 1.1 p.
IILTRATJON RAT! 2.0 rn/h
Q:QocDe~
3 1 2 2 4 1 2 3 4
RiJ!’~~AY
6.2.5.3 Headloss
The difference in headloss over the gravel bed is presented in Figure 6-15. It can be
observed that the main headloss is concentrated in the first 20 cm corresponding with the
finest gravel layer. In Appendix C Figures C-i a) to C-i I) are shown the typical
headloss variation over the filter run for different Q, and V = 2.0 rn/h.
DyRF2 INFLUENT 01 • 1.5 ps
FILTRATION RATE 2.0 rn/h
FILTER RUN 3
DEPTH
cm 0
RUN DAY
10 -4—
-M- 2
20 —e—
30
40
60
60
0 05 1 16 2 25 3 35 4 --
HEAD LOSS (oml
57
CHAPTER 7
Discussion of Results
7.1 Raw water characterization
A large part of the diseases in developing countries are water borne. Many of these are
directly related to faecal contamination of the water sources. The Cauca river which is the
raw water source for several million colombian inhabitants falls in the worst category (E)
(coliform countsll0O ml > 1000) proposed by Lloyd et al. to WHO and UNEP. During
the research period the average faecal coliform count ranged between 19440 and 64143
with peaks between 29000 and 242000. Implying a very high sanitary risk.
Also the physical-chemical quality is very poor. With values for several parameters will
above limitations suggested for the application of Slow Sand Filtration (SSF) as a
treatment alternative. Cleasby (1991), Spencer et al.(1991) and Di Bernardo (1991) have
for example proposed upper limits for SSF application for turbidity between 5 to 10 NTU,
true color between 15 and 25 TCU, total iron between 0.3 and 2.0 mg/i. The raw water
quality of the Cauca river is well above these values thus clearly indicating the need for
pre-treatment alternatives.
A further complicating factor is the composition of the raw water. According to the
research findings 91 % of the particles have a diameter d1 < 27, 32 and 57 ~m and 70%
3
has a diameter < 5, 7 and 13 ~m for densities of 2650, 2125 and 1365 kg/rn
respectively. A large portion of fine material may affect the performance of slow sand
filters (Galvis et al. 1993) but may also influence the behaviour of DyRF because of the
large pore size usually above 500 ~m for gravel sizes above 2mm (Huisman 1986). In
view of the rather special water quality of Cauca river findings for the behaviour of the
DyRF treating this water quality may not be the same for other water qualities.
Sedimentation is the most important process in DyRF as clearly shown by the results
obtained in the research as presented in 6.2.2.
On the basis of the operating conditions two sedimentation areas have been distinguished:
The impact of the sedimentation process is analysed by comparing the theoretical and
measured suspended solid removal efficiencies for each of the two sedimentation areas.
The calculation of the measured efficiency permits the elimination of the effect of the
variation in suspended solid concentration in the influent. The theoretical efficiencies are
based on the removal rate R established with column settling tests (CST).
58
7.2.1. Plain sedimentation on the exposed surface of the gravel bed
This area can be considered as a plain sedimentation tank where (R) depends on two
factors: i) the frequency distribution for settling velocities of the suspended particles and
ii) overflow rate (S). Only the last factor can be influenced by the design or operating
conditions of the DyRF. The efficiency in this area is independ of the depth of the basin
and of the detention time.
In comparing the efficiencies at the beginning of the filter run considerable differences
were found for a significance level a of 5%, the higher measured efficiency however is
probably the result of a the water flowing through the expanded gravel bed to the weir.
The higher efficiency at the start of the filter run is favoured by the cleanliness of the filter
and the low value of Q0, which permits horizontal flocculation to take place in the upper
part of the gravel bed (0 10 mm) such as the formation of the bed forms ripples and
-
dunes. The latter is favoured by the low value of the Froude number (F1 < 0.7).
In the beginning of each filter run X, Re and Fr have the lowest values due to Q~being
highest and Q1 being approximately constant. The flocculation effect can be stimulated by
the high concentration of total iron (average ranging from 13.37 to 15.76 mg/l with peaks
between 25.60 and 57.50 mg/i). This effect is very much reduced after the first day when
the gravel bed is settled and horizontal filtration no longer takes place.
If the results of the first day are excluded, no significant differences exist for a = 5%
between the theoretical and measured efficiencies except for the value of 20% obtained for
= 0.70 1/s and reported as a likely experimental mistake.
Excluding the effect ofhorizontal filtration it may be concluded that the removal efficiency
in the horizontal flow is mainly due to plain sedimentation and is presenting values below
10%.
iv) Negative effects resulting from the formation of a sludge layer on the surface
causing the development of horizontal and vertical flow lines which reduce
sedimentation efficiency.
v) Limitations in the hydraulic design of the system which does not permit a good
flow distribution in the inlet structure or a good abstraction at the outlet.
59
The efficiency of plain sedimentation can also be influenced by the bed forms such as
ripples and dunes which may be formed when F1 < 0.7 and have the common mode of
bed material transport for the individual grains to move up the back of the ripple or dune
and avalanche down its face. Another aspect to take into account is the effect of the
gradation because the large particles will be more exposed while the smaller particles are
shielded by the larger particles.
For a 95% confidence interval it has been established that the horizontal removal
efficiency for Q, values between 0.70 1/s and 2.65 1/s does not depend on the value of Q1.
In the gravel the efficiency depends on the surface loading S1 = QjAI, which Q~reducing
during the ifiter run generating a reduction in S1 and thus an increase in efficiency. In
Table 6-4 average theoretical efficiencies range from 60.4% (STD = 0.65%) at the start
of the filter run and 78.7% (STD = 1.54%) at the end. Nevertheless measured efficiencies
are higher and range from 66.5% (STh = 2.3 %) to 84.9% (STh = 2.16%) respectively.
For a 95% confidence interval differences between theoretical and measured efficiencies
are not significant which implies that reported efficiencies do not depend on Q for values
ranging from 0.70 and 2.65 1/s.
Different factors can justify the high efficiency of sedimentation in gravel beds:
ii) Presence of other mechanisms such as screening and flocculation. The former may
be important at the end of the filter run when the pore sizes have decreased and the
latest, due to the fact that the packed bed of gravel provides ideal conditions for the
formation of compact settleable floes because of continuous recontacts provided by
the sinuous flow of water through the interticies formed by the gravel. Moreover high
total iron concentration in raw water can help to floe formations. According to
experimental conditions and based on the formula presented by Schultz and Okun
(1984) for gravel- bed flocculators the velocity gradient G can get a value of about
40 s’.
iii) Reduction of Q~during the filter run. This process of declining rate filtration
reduces S1 and increases the sedimentation efficiency.
On the basis of the high removal efficiencies and the fact that no significant differences
exist at a 95% confidence level it can be said that sedimentation is the main particle
removal mechanism in DyRF with average efficiencies ranging from 66.5% to 84.9%.
Based on the achieved results, Figure 7-1 illustrates the main particle removal mechanisms
in DyRF. Zone 1 corresponds with plain sedimentation on the exposed surface of the
gravel bed with an efficiency below 10% and possibly having a negative influence on the
hydraulic behaviour of the DyRF. Zone 2 is the area of major efficiency in particle
removal in DyRF where sedimentation takes place, but also other mechanisms such a
screening and flocculation may contribute to the removal efficiency.
60
1. Surface sedimentation
2. Sedimentation in
granular bed
+
other mechanisms
Remaining concentration
The effect of the overflow on the treated water quality has been analyzed according to the
design of the experiment indicated in Table 5-2. Although the concentrations of the
different parameters vary over time this has been solved by calculating removal
efficiencies [(Cm C~)/ Cmlx 100 with Cm = influent concentration (raw water) and C~,,,
-
As Q is the controlled variable and V varies over the ifiter run due to the process of
declining rate filtration, a relative simple analysis can be made of the established
information. The statistical data (average, STD, standard error, and confidence level) have
been calculated with the programme Slide.
i) Calculation of real efficiencies for each Q~and for each parameter under evaluation
based on the results of the experiments.
ii) Calculations of the statistical data on the basis of the data collected under i) and
help of the programmne Slide.
iii) Analysis of significant differences for each of the parameters for a 95%
confidence interval. This interval has been selected because of the small sample size
n < 30.
On the basis of the statistical data presented in Figures 6-6, 6-7 a) to 6-1 ~ and 6-8 it can
61
be established that:
i) The removal efficiencies for suspended solids in DyRF for V = 2.0 rn/h do not
depend on Q1 for values of Q1 = 0.70 Ifs and 2.65 Ifs. The average efficiency over
the total filter run is over 83.3% (STh = 9%) except for Q = 0.42 1/s which had
an efficiency of 59.38% (STh = 15.1%) due to poor hydraulic performance as
discussed earlier.
iii) At a 95% confidence level significant differences have been established for the
removal efficiency of total iron for different Q’s. This implies that the removal of
iron depends on the value of Q1. This can be explained as resulting from the
horizontal filtration process which may stimulate floe formation. Furthermore the
high level of iron in the raw water may be an important factor.
iv) Removal efficiencies for suspended solids for filtration rates V = 2.0, 3.0 and
4.0 rn/h do not present significant differences at a 95% confidence level. For V
5.0 m/h however significant lower efficiencies of 63.6% (STh 9.2%) were
obtained. This can be explained by the increase in the flow velocity which implies
an increase in the shear stress and thus the scouring capacity.
The removal efficiencies obtained in the evaluation of the impact of the overflow on the
quality of the treated water are justified primarily on the basis of sedimentation and the
effect of declining rate filtration. Other mechanisms such as screening and flocculation
also may contribute to an increase in efficiency, but its effect has not been quantified.
The analysis of the impact of the overflow on scouring of surface particles is based on the
following considerations:
i) The raw water of Cauca river has a high concentration of fine material: 70% of
the particles in Q and Q0 have diameters d, 3<and 5, 91% d respectively for
7 andhas13a ~m
particle densities of 2650, 2125 and 1365 kg/rn 3 < 27, 32 and 57
~zm respectively. Moreover the effective weight of the particle (p~ - p) acts
downwards and mechanical friction is a function of the size of the deposited particles,
the resistance to scouring is minimal.
62
iii) Minimum resistance against displacement appears to exist in the deposited
material on the exposed filter bed, which has an average density of 1564 kg/rn3 (STh
= 78.2). This behaviour is indicated in Figure 6.-il for cohesive material where
particles with diameter > 3.2 j~mcan be scoured with X = 10 cm/s and with d
1 >
0.5 ~m with X = 18 cm/s. The cohesive sediments tend to stick together forming
aggregates where the electro-chemical forces are acting between the particles and are
increasing their scouring velocities than those of the individual particles.
iv) The critical condition for scouring of surface particles per each Q, is taken place
ending each filter run when the real flow velocity (X) is taken the maximum value.
Therefore, the analises of scouring was based on the maximum velocities X for Q
= 0.7 1/s and 2.65 1/s which correspond with 0.10 and 0.18 cm/s respectively. For
these velocities the Reynolds number varies between 1074 and 3415 which generates
flow conditions between laminar and turbulent.
The experimental data show that with a minimum value X of 10 cm/s, 91% of the
particles in the sediment can be scoured and for X = 18 cm/s all particles can be
scoured. The drag force of the fluid on the sediments due to hydraulic shear acts in
the direction of the motion of the fluid and hydraulic shear is a function of the fluid
velocity therefore high flow velocities mean high drag forces. Scouring starts when
the hydraulic shear stress (r~)between the flowing water and the sludge deposits
exceeds the mechanical friction or attractive forces between layers of deposits or
between deposits and the grain surface. If r~is very much greater than mechanical
friction, scouring is not important and the particles will be in suspension in the water
and therefore scouring is not important aspect for DyRF.
v) The particlecorresponds
3, which with ansediments
size for cohesive were for
average value analysed for a density
clay. Specific = 2650
test toPidetermine
kg/rn
cohesive sediments have not been made but in view of the diversity of sediments
transported by Cauca river as supported by literature they are present. The literature
refers to cohesive material such as a mixture of sediments with clay (d, < 4 ~m) and
densities about 2650 kg/rn3. In this research this value was to determine particle
diameters and the calculation of U.
The high flow velocities and the low density of the deposited material (1564 kg/rn3),
contribute to the justification of the low efficiency-of plain sedimentation. The deposited
material can easily be removed because of the high drag force and only large non cohesive
particles can remain or small cohesive particles. In practice we see that quite some
material remains on the surface of the gravel bed even at the highest value for X and
particularly in the first half of the bed. This may be caused by either lower flow velocities
in the beginning of the bed due to a greater height of the water level causing a different
flow distribution over the heigh. This can be confirmed by a more detailed measurement
of the water level and flow velocities in the DyRF.
Another possibility is that the deposited material is strongly influenced by other forces
involved in absorption and therefore is not removed. Nevertheless even if all particles
would be removed scouring only affects the material which is deposited on the surface of
the grains and which is shown to be less than 10% of the total material removed by a
63
DyRF.
The sludge accumulation in the different layers of a DyRF does not depend on the value
of Q1 for values between 0.7 to 2.65 1/s. This is showing again that the impact of the
horizontal flow over the DyRF is very small. At a 95% confidence level it was established
that there was no significant difference in sludge retainment in the different layers for
different Q, except for Q, = 0.42 1/s which caused a poor hydraulic performance, with
dead zones at the surface and part of the bed at the start of the filter run.
The obtained averages for the retained sludge in each layer for V = 2.0 rn/h corresponds
withL1=80.5%(STD=9.5%),L2=11%(STD=4.2%)andL3=8.5%(STD=
6.2%). This clearly indicates the efficiency and importance of the first layer which
composes of the finest gravel.
The sludge accumulation for the higher filtration velocities V = 3.0 and 4.0 rn/h show a
similar pattern as for V = 2.0 rn/h but differs at a 95% confidence level for V = 5.0 rn/h
when more material was transported into the deeper layers. This can be explained by the
increase in flow velocity causing higher shear stress, taking deposited material deeper into
the bed.
For V 4.0 rn/h sludge accumulation distribution over L1, L2 and L3 does not depend
on V with average accumulation levels for L1 = 84%, L2 = 9.7% and L3 = 6%, whereas
for V = 5.0 rn/h the distribution was L1 = 50%, L2 = 25% and L3 = 25% showing
clearly a deeper penetration of the sludge into the bed.
64
8. Conclusions and Reconimendations
8.1. Conclusions
The Cauca river water has a typical characterization of untreated sewage which implies
a very high sanitary risk when it is used for drinking water. Also very fine particles were
found in raw water (about 70% particles < 5 ~&m).This water quality should better be
rejected as a source of drinking water and if used will need different treatment steps to
reduce the sanitary risk.
In DyRF Sedimentation is the main particle removal process which occurs in two
different locations:
i) Plain sedimentation onto the exposed surface of the gravel bed. Here the removal
efficiency is < 10%. -
ii) Sedimentation (as part of filtration process) into the gravel bed. The removal
efficiency ranged from 65% to 85% and is considered as the most important process
in DyRF.
The surface overflow did not have any impact on scouring of surface particles and on
treated water quality.
4. DyRF perfomance:
The following average removal efficiencies were obtained for DyRF units operating at
filtration rates between 2.0 and 4.0 rn/h with average surface flow velocities between 5
cmis and 18 cmis:
65
5. Desian auidelines:
The impact of the surface flow in DyRF is limited and may even result in uneven
distribution of water over the filter. Therefore it is better to design DyRF whithout surface
flow. The gravel sizes can be kept, although it may be explored to reduce to two sizes.
Preliminary design criteria for such a system are:
8.2. Recommendations
i) To eliminate the overflow as a normal hydraulic element in DyRF operation.
ii) To keep the weir in the out let structure as “tapon valve” in order to protect the water
treatment plant against suddenly raw water quality changes. This weir is also useful for
maintenance activities.
iii) To evaluate downflow roughing filter with declining rate filtration and constant rate
filtration in order to identify the best alternative and compare this with a DyRF with
overflow.
This evaluation preferably is carried out both in the Research Station and in another
location with different water quality.
66
REFERENCES
Ahsan, T., Alaerts, G.J. and Buiteman, J.P. (1991). “Direct Horizontal-Flow Roughing
Filtration: An improved pretreatment process for highly turbid water.” Proc. of the
Filtration Society Conference (Karlsruhi - Germany), October.
APHA, (1986). “Standard Methods for Examination of Water and Wastewater.” 16th
Edition. American Public Health Association. Washington.
APHA, (1989). “Standad Methods for Examination of Water and Wastewater.” 17th
Edition. American Public Health Association. Washington.
Arboleda, J. V. ,Giraldo, R., and Snel, H. (1985). “Hydraulic Behavior of Declining-Rate
Filtration.” Journal of the American Water Works Association. 67-74. December.
Chang, H. H. (1988). “Fluvial Processes in River Engineering.” John Wiley & Sons, New
York.
Cleasby, J. L. (1991). “Source Water Quality and Pretreatments options for Slow Sand
Filtation.” Chapter 3 in: Task Committe on Slow Sand Filtration. New York.
67
Deift Hydraulics. (1972). “Systematic Investigation of Two - Dimensional and Three -
Dimensional Scour.” Report M648/M863, Deift.
Frankel, R. J. (1974). “Series Filtration Using Local Filter Media.” JAWWA, 66, 2, 124
-127.
Galay, V. J., Yaremko, E. K., and Quazi, M. E. (1987). “River Bed Scour and
Construction of Stone Riprap Protection.” Sediment Transport in Gravel-Bed Rivers, Ed.
by Thorne, C. R., Bathurst, J. C., and Hey, R. D., John Wiley & Sons, New York.
Galvis, G., Fernandez, J., and Visscher, J. T. (1992). “Comparative Study of Different
Pretreatment Alternatives. Roughing Filters for Water Treatment.” Workshop in Zurich.
Galvis, G., Visscher, J. T., Fernandez, J., and Beron, F. (1993). “Pre-treatment
Alternatives for Drinking Water Supply Systems.” Occasional Paper 22, IRC. The Hague.
Ives, K.J. (1975). “Capture Mechanisms in Filtration, in the Scientific Basis of Filtration.”
Ed.Ives K.J., Noorrdhoff International, Leyden.
Lebcir, Rabia. (1992). “Factors Controlling the Perfomance of Horizontal Flow Roughing
Filters.” PhD thesis, The University of Newcastle upon Tyne, Department of Civil
Engineering. April. - -
Mantz, P. A. (1977). “Incipient Transport of Fine Grains and Flakes by Fluids - Extended
Shields Diagram.” Journal of the Hydraulics Division, ASCE, Vol. 103, HY6.
68
Miller, M. C., McCave, I. N., arid Komar, P. D. (1977). “Threshold of Sediment Motion
under Unidirectional Current Sedimentology.” Vol. 24, p. 507 527.
-
Perez, J., Pardon, M., Lavado, I. and Vargas, C. (1985). “Informe Preliminar de la
Investigacion sobre Prefiltros de Grava.” CEPIS/OPS/OMS. Lima.
Salazar, L. G. (1980). “Agua Potable para la Zona Rural Colombiana.” ACODAL No.
96, Bogota.
Smet, J.E. and Visscher, J.T. (1989). “Pre-treatment Methods for Community Water
Supply: an overview of techniques and present experiences. The Hague, Netherlands,
International Reference Centre for Community Water Supply and Sanitation.
Thanh, N. C. and Hettiaratchi, J.P.A., (1982). “Surface Water Filtration for Rural Areas:
Guidelines for Design, Construction, Operation and Maintenance.” ENSIC. Asian Institute
of Technology, Bangkok.
UNIVALLE, (1986). “Estudio Morfologico del rio Cauca - Fase 2.” Departamento
Mecanica de Fluidos y Ciencias Termicas, Cali.
Van Rijn, L. C. (1986). “Manual Sediment Transport Measurements.” Delft Hydraulics
Laboratory, March 1986, Deift.
Van Rijn, L. C. (1989). “Handbook Sediment Transport by Currents and Waves.” Delft
Hydraulics, Report H461, June, Delft.
69
Weber, W.M. (1972). “Physicochemical Processes forWater Quality Control.” John
Wiley & Sons, Inc.
Wolters, H., Smet, J. E. and Galvis, G. (1989). “Up Flow Roughing Filtration.”
-
Chapter 8. In: Treatment Methods for Community Water Suppiy. The Hague, IRC.
70
APPENDIX 1
Table 1-1 a)
Descriptive Statistics - Mean values for three filters runs
DyRF1: Q1 1.0 ]/s and DyRF2: Q2 = 0.42 us; runs 1,2 and 3
— ~1 imi
‘actom
4’ 4, 14 ii 14 14 —
KM III IN 03 III 47
7*1IIITV 511KV 14 44 17 44 16
74 17 34 14 34
4751 MI4li~
Nil.. 774 744 42 374 114
looks’ 77 34 17 27 37
IN 44 21 371 44
KM
17 27 ii~ 44 II
WEllES Ill ISV
ails mile IN I? 4.4’ 47 70
MIII SiltS 074 ~• 77 047 45
731k 55 27 24 27 27 57
53 7l 14 41
KM
a,TILO sis MV I’ 14 14 IS
ail5 MIlO 75 Ii 36 11
MIII slim, 14 71 54 42
701k 141101 07 77 27 77
45 47 41 12 44
.44
7705 Ct, 170 ISV 14 II 22
22 37 72 20
In’ 41410
NI IC Iii 57 44 11$ 52
707k 55 2 I $
44142 74744
KM 4~I~3 1417!
44757 21441
~4[C~ 410 ISV hIll ~
4744 I 4440
Ct!;,’ 41420 4744 s~4
107000 47000
IFEIIII.ll 44414.4 3~744~ ~
14 -- 4
747k54I~U IS - 14
1222 5.61
12.31
~M 1.15 4 ~ 3.10
707k 0 476 ISV 4.77 2.40
770 074 I
i~iII 4l4l4.4 7.34 444
74~ 2550
aiim, 2460 4
I 4
igiklSeti I
hIS! hIS?
ii I. I. Si Si 4.
KM IN 14 30 III 64
~lai 17$ ISV 21 14 II, 44 là
Will 414103 174 44 1% 44 21
sill 5110.1 74, 3t2 45 264 III
7016. 55 I, 1 II 14 24
KM 45 .1 44
WIlliS 00154! ii 10 II 11
Sill 114103 II 22 II
i,.eii Mum, 54 71 44 Ii
701k SeE IS IS IS I,
KM 43 ii 14 II
7c cm., 176 WV II ii 17
ilcvi ,1mm~ 22 27 77 ii 74
Nil’- 44 17 41 54 27
7406. iii 3 2 1 2
71
APPENDIX 1-1
Table 1-1 c) Descriptive Statistics run 2: 1.0 and 0.42 1/s
I
7WI 11W3
~nM —
Ii Si Si II Si Si
KM 131 147
44 45
IS 144
41 07
744111177 771 ISV 44 74 41 47 14
14751 hum, 7$ III
741 11 774
slim, 711 IS
I II II
707k 1107.0 hI
III N 21i II
KM 77$ 77 73 44 5$
WINOS 571400 44 7 4.4 • 144
WIll hills 114
IIqI1I SlitS 524
344 fl 340 145
701k $1042 I I I
74 II 37
KM ‘I I IS 10
WOTILI 171 ISV II 71
72 76
10.140 41410 35 52
47 44
1.4111 14110 44
I I
147k set
44 47
KM 44 7 14
7*1 Ct, 177KV 22 40 44 17
15(01 MIMIC 45 77
44 141
MIII) Ill
3 4
7076. 41011
47040 42174 Ilooc
KM 43120 77170 14702 11601
‘lOck 171 110 100204 4400
544 1414
01.25704 41kb 1400 17100
154000 742000
15(0/140411 4411401 74740
3 0
117k sal
447 1447 5.10
‘1*4 11 41 244
174 7.71
70o41 liii 477 MV 440 4’0
743 440
bill lisle 4.40 1.10
11$ 2144
14110 7554
3 7
701k ses
lyWl loW?
1004 -
Ii Ii Si Ii . Si Ii
KM 74 44 43 Si ‘2
714411174 $75 ISV 1.1 4.4 II II 44
1411.11 01410 41 51 34 56 14-
NIl... là 54 14 12 72
701k 51 IS II II IS II
00’s 44 17 II 04
t471L0 576KV I’ II II I?
SIlO 11410 74 4 24 15
1.4111 slimS 42 57 45 II
407k 14?Ii 4 I 4 4
4444 45 43 44 54
7*1 Ct00 III NV
11011 11410 41 54 74 51
slims 44 44 41 41
707k 57 2 7 3 2
72
APPENDIX 1
Table 1-1 e)
Descriptive Statistics Mean values for three filters runs
-
DyRF1: Q1 = 2.0 1/s and DyRF2: Q2 1.50 1/s; runs 1,2 and 3
,4WI ,457
144144107$
Is Si Si Ii Si Si
KM 44 44 54 53
W571L2 175KV 14 14 II I?
WIll 4I4ii41 37 77 13 21
154111 521105 51 74 ‘7 II
701k 57 Ii 17 IT IS
KM IN IN
1*1 CtN $75 ISO 07 22
(Poll 4IShO TI ii
Nile Ill Is?
710k sas 4
hISI IoN’?
4444107113 - —~~~--
Is Ii Si Ii Si 2.
KM 75 54 IS 34
SITItI 715100 71 II 55 IS
Sill 11114$ SI 31 75 40
154111 51111 47 74 47 71
70141 sao S
KM Ill Ill
7 CM00 171614 22 71
14021 41110 II? ill
a7I4.P ID? 177
757k44sV11 A I
73
APPENDIX 1-1
0*54477)7 __________
II Si Si Ii ,Si Si
KM 54 II 74 Ii
M,571L0 077 449 24 14 04 14
$40115 11115) 72 13 52 hI
MIII Solo 11 74 71 45
751k 4~( 5 4
KM $2 $2
741 CaY
14021
011KV
1101$)
II
V
112
50III) II?
107k 5~l
KM isiS. 7520
44005 ~7132 70374 774$
FStk $15 ISV
12471 3100 13400 7704
02.15354 41510
15(0/104*11 4411500 11754 15444 11114 12400
7 5 0 7
707k 40*11
Table 1-1 Ii) Descriptive Statistics run 3: 2.0 and 1.5 1/s
lyWl
14117
~Mlo4.1000 _________________
Is Si Si
II I. 0.
KM III II? II
711115177 174 Ni 47 s4 53 145 III 45
71 57 Dl
15001 11114.4 IV 74 5~
slim, 242 774 171 15 15 II
270 244 157
707k100sJ I I I I I I
KM 244 754 41
WDP70s7~3 172100 III 173 21 054 707 53
114 Ill 33
WIIO 41110 121 IS 1 L
110 II 15
114111 SIlO 447 471 47 444 3~ II?
117ksa0 S S S
4 S 5
KM so
51
*27(11 170 NO II Ii 13 12
$40115 41110 71 SI 71
MIII MIII) 71 43
44 — SI
717k sao
KM Ill
147
7OCt71 175400 57
2?
l~l 11110 II II
541401 III
Ill
74741 45*11 2
74
APPENDIX 1
Table 1-1 i)
Descriptive Statistics - Mean values for three filters runs
DyRF1: Q1 = 2.65 1/s and DyRF2: Q2 = 0.70 1/s; runs 1,2 and 3
I 1,S’2
444,5Th --
Ii Si I. Ii Si I.
KM II 53 57 70 33 54
730111171575KV 04 54 53 45 II 54
11701 41110 22 IS 57 30 Ii 52
SIlO 344 124 710 354 III 204
101k44510 34 34 14 35 23 33
KM 55 54 44 00
*47110 111Kv 14 Il 34 00
440254 11115) 17 II Ii 3
IS/Il 5170 154 73 124 II
Silk 40*11 72 33 32 24
55 14 SI 44 70
7#4a35477I$V 25 35 22 14
14741 17110 25 21 24 53
Silo l5~ Ill 154 124
747k 4041,5 I? 12 IS IS
lySI loll?
444,571,1
Si Si Ii Si Si -
KM 11 22 II SI 12 54
770114177 175 NV 4.0 7 11 4.Ds 4,52 4 40 5.13
11701 41110 II 35 so is 73 45
Sill) 30 40 54 71 44 44
747k sac 12 32 1? II I? IS
KM 104 41 II 77 SI
40915K) 471KV 15 II hO 24 SI
44.104 41410 4? IS Sb 57 04
IN/li SIlO) III 104 54 ‘ IS IS
747k sao I I I’ $
KM 04 44 577 57
54047110 171 ISV IS Il 00 SI
WIll 41410 52 12 30 2$
lsqiiI SilO 7? SI II 12
707k sal 7 1 I
KM 35 30 54 37
7001(41175 KS II II 13 II
14701 41110 34 SI 25 01
Slims 54 42 SI SO
707k oa 5 3 4
75
APPENDIX 14 -
1,51 lyLl
4*00Th
Is Si Si ii Si Si
KM 74 SI 74 74 74 55
710011117005101 S IS 74 42 II 47
11741 41410 22 IS 30 52 IS 52
SOlO) 240 74 241 110 a 75
717k1W10 IS 77 IS II 12 IS
55 04 44 54 25
440)1153 4111(1 44 24 44 22
$40117 41410 17 II IS
IS/il SOlO 774 74 155 57
757ksa0 4 4 I
55 47 52 57 42
7004041071KV 21 IS SI 74
14001 41110) 54 44 14 71
Silo III IS 105 IS
757k00ttE 3 S
hId IoN’?
4*00Th
Is Si Si ~Ii , I. I;—
554 07 15 55 7’
4 ~_~~77
730II1174 175 1(0 70 25 ,3 54
11701 4111415 7’I 37 Ii I’S 00 33
SIll) 277 131 373 250 17$
7l7kssrI,1 0 0 4 I 4 — I
KM SI 32 34 II
44,4711.0 111 500 IS S II
WIll 4111400 54 21 40
154/il Silo 5 Si S
741k10453 5 4 S s
KM S7 54 47 54
70040S$TIK/ 23 SI 37
14(01 17110 II SO SI I’
Silt. 15 IL? 110
317k$sNt( 5 5 S
76
APPENDIX 2
Table 24
Linear Regression Model between Suspended Solids and Raw Water Turbidity
Based on Column Settling Test Results.
Mean 55 Mean 55
STD 9 yU) Sli) 4
42-50 Minimum 32 1.1570X~ 0.9794 Minimum 49
Maximum 67 Maximum 58
sample size 39 sample size 39
Mean 88 Mean 87
STD 23 ~- STD 21
51-100 Minimum 49 l.2311X~ 0.9692 Mmimum 64
Maximum 158 Maximum 123
sample size 69 sample size 69
77
APPENDIX 3
Tabla 3-1
Real removal efficiency variations over the filter runs for different Q1’s and V 2.0 rn/h.
Descriptive Statistics for physical, chemical and bacteriological parameters.
TURSIDITY
SUSPENDE D SOLIDS
VOLATILE SOLIDS
TRUE C OLOR
M~ 00 00 . . . 0.0
Mean 13.3 24.2 . . . 23.8
~ 310 390 . . . 46.0
STD 139 18.4 . . III
Sta~.zderio 92 92 - . . 9.4
9S%~f~donoc 18.0 180 - . 18.0
99%o~&ncn 23.7 23.1 . . 343
SsndnaIoc 3 4 . . . 4
TOTAL IRON
78
APPENDIX A
Cumulative frequency distribution of settling velocities for different Q.
0~
I 0 10 20 30 40 50
—~ ----Os
Effkje,t Overflow
Figure A—i a) Nominal Q1 = 1.0 1/s
120
0.
100
80
60
40
20
I
0
0 10 20 30 40 50 60 70 80
—Qo
Influent Effluent Ov~flow
Figure A-i b) Nominal Q1 = 2.0 1/s.
I 0 10 20 30 40 5.0 ~0 70 80
79
APPENI)IX B
Typical variation in water flows Q, Q and Q0 over the filter runs.
l’s
fl1.IER RU . FILTER RUN 2 FiLTER RUN 3
1,4
1,3 -
~E~__
Figure B-i. a) Nominal Q, 0.42 1/s
DyRFI. INFLUENT : 01 1.0 pa
PILTRATION RATE 2.0 rn/h
—Qj~ ~Oo~Oe
1,2
— 0,1
0,4
0.2
0
2 3 4 6 S 2 3 4 I 6 1 2 3 4 6
I 2 3 1 2 5 4 1 2 3 4
DAY
S’s
LI
= 1,5
0.5
0
1 2 3 4 I 1 2 3 4 I 1 2 3
6,I
FiLTER RU FILTER RUN 2 FILTER RUN 3
= = =
0,5 —
1 2 2 4 1 2 3 4 1 1 2 5 4
DyRF1, INFLUENT 01 — 15 ps
FILTRATION RATE 4.0 rn/h
-Qi~Oo~3e
I’S
~l,5
0,1
1 2 3 4 1 2 3 4 1 2 S 4
~.N. ~
81
APPENDIX C
Typical variation of headloss for different Q1 and V = 2.0 rn/h.
FILTRATION RATE 2.0 rn/h
FILTER RUN 1
DEPTH
cm o
10
20
30
to
80
0 05 1 5 2 25 ff~ --
HEAD LOSS (cml - - - —-
DEPTH
~ 0
RUN DAY
—4— 2
* 3
—Q-- ‘
—s-- 5
-4—-- 8
Oh A - 10 2 _26 - --
HEAD LOSS 1cm) - -
DEP TI-I
cm o
to
20
30
40
50
80
05_ ~~J~i_ 115 2
HEADLQSS(cm( - - - — - - -
82
FILTRATION RATE 2.0 rn/h
FILTER RUN 2
DEPTH
cm
0
RUN DAY
10 ~ 2
20
—C--- 4
30
40
50
60
0 015 t L8 2 2h 3 as 4
HEAD LOSS (om)
DEPTH
cm 0
RUN DAY
10
—4— 3
20
—0t— 4
30
40
50
80
a as-4~
HEAD LOSS (cm)
Figure C-i e) Nominal Q1 = 2.0 1/s.
DEPTH
cm 0
140&ar ;‘ - .
- _--
-C--- ~
—a—
—+—— 5
0 05 1 ~I5 2 15
HEADLDSS (cM) - - -
83