On fluid dynamics of lamella separator modelling and process ...

On fluid dynamics of lamella

separator modelling and process

optimisation

Vom Fachbereich Produktionstechnik

der

Universitat Bremen

zur Erlangung des Grades

Doktor-Ingenieur

genehmigte

Dissertation

von

M.Sc. Ahmed Ibrahim Salem

Gutachter: Prof. Dr.-Ing. Jorg Thoming

Prof. Dr.-Ing. Udo Fritsching

Tag der mundlichen Prufung: 25. April 2012

1

Zusammenfassung

Eine Aufgabe bei der Behandlung von hauslichen Abwassern und Industrieabwassern

ist es, Suspensionen durch Schwerkraftabscheider in klare Flussigkeit und Feststoffe zu

trennen. Der Schragplatten-Schwereabscheider (IPS) kann dies mit eine relativ hohen

Abscheidegrad bei geringem Flachenbedarf leisten. Jedoch erreichen herkommliche

IPS hinreichende Trennleistung nur bei Volumenstromen, die unter den ausgelegten

Werten liegen.

Diese Art von Schwereabscheidern hangt vom Verhaltnis der Stromungsrichtun-

gen von Zulauf und Sediment-Supension ab, von der Geschwindigkeitsverteilung der

Suspensionen zwischen den Schwereabscheider-Kanalen und vielen anderen Faktoren.

Diese Faktoren werden sehr stark von der Zuleitungskonfiguration beeinflusst.

Zur Losung des Problems wurden drei verschiedene Zuleitungsstrukturen verwen-

det, um deren Effekt in einem Labormassstab IPS zu untersuchen. Versuche und

numerische Stromungssimulationen (CFD) wurden durchgefuhrt, um sowohl die hy-

draulischen Eigenschaften als auch die Trennleistung beurteilen zu konnen. Der Ver-

gleich der experimentellen Ergebnisse mit den vorhergesagten Ergebnissen der CFD-

Simulation zeigte eine gute Ubereinstimmung der Verweilzeitverteilungen (RTD-Kurven).

Es wurde deutlich, dass die Verwendung der Dusenverteiler die hydraulische Leistung

des IPS entscheidend verstarken und so zu einer Verbesserung der Trennleistung um

7% Leistung fuhren kann.

Basierend auf dem Ergebnis, dass die Trennleistung des IPS in der Regel von seinem

hydraulischen Verhalten eingeschrankt wird, welches wiederum vom Zulaufverteiler

bestimmt wird, wurde ein Optimierungsansatz entwickelt. Als Methode dazu wurde

die Response Surface-Methode(RSM) verwendet, basierend auf der Hypothese, dass die

Verringerung der Inhomogenitat der Stromungsverteilung zwischen Ausflussoffnungen

eines Dusenverteilers im IPS Zulauf das hydraulische Verhalten dieses IPS verbessert.

Die RSM wurde verwendet, um die Beziehung zwischen den Input-Parametern des

Verteilers und einer Zielfunktion zu beschreiben und mit einer kommerziellen Soft-

ware (Ansys) zur Losung des Optimierungsproblems zu fuhren. Durch die Opti-

mierung der Verteiler konnte die hydraulische Leistungsfahigkeit des IPS sowie Stro-

mungsgeschwindigkeitsverteilung innerhalb der einzelnen Kanale des Schwereabschei-

ders verbessert werden, so dass die Trennleistung um nahezu 10 % gesteigert wurde.

Zu Losung des Problems der Resuspension am Eingang der Schwereabscheider wur-

den Traufen-Kanale am unteren Rand der Platten mit einem veranderlichem Nei-

gungswinkel angebracht, um das Sediment zu sammeln und es uber Seitenabflusse

auszuschleusen. Durch diesen Ansatz konnte die Trennleistung des Schragkanalklarers

um weitere 7 % gesteigert werden.

3

Abstract

The suspensions in the treatment of water, domestic wastewater and industrial waste-

water can be separated into clarified fluid and solid matter by gravity settlers. Inclined

plate settler (IPS) is a type of gravity settler that can provide very high space time

yield. Its performance strongly depends on the direction of the feed flow relative to

sediment movement, the distribution quality of suspensions between settler channels,

the hydraulic characteristic of suspension within the settlers and many other factors.

These factors in turn are extremely influenced by the inlet configuration.

In this work, three different inlet structures were used to explore the effect of feeding

a lab scale IPS by a nozzle distributor in terms of separation efficiency. Experimental

work and Computational Fluid Dynamic (CFD) simulation analyses were carried out

to assess both the hydraulic characteristics and separation efficiency. Comparing the

experimental results with the predicted results by CFD simulation imply that the

CFD software can play a useful role in studying the hydraulic performance of the IPS

by employing residence time distribution (RTD) curves. Also, the results show that

using nozzle distributor can significantly enhance the hydraulic performance of the

IPS which has contributed to improve its separation efficiency.

This finding means that the flow pattern provided by the inlet structure can con-

strain the separation performance of IPS. An approach was used for optimising this

pattern by means of response surface methodology (RSM). The hypothesis could be

confirmed that reducing the discrepancy of flow distribution among outlet openings of

a nozzle distributor for feeding a bench scale IPS enhances the hydraulic behaviour of

this IPS. The RSM was used to establish an approximate model to describe the re-

lationship between the input parameters of the distributor and an objective function.

This model was solved by using Ansys CFD package. The results proved that optimi-

sation of the distributor has led to increase the hydraulic performance of the IPS in

4

terms of the flow pattern of IPS approaches plug flow condition, and also improved the

flow distribution within each settler, both of which improved the separation efficiency

to nearly 10%.

Attempting to decrease the resuspension problem at the entrance of the settlers, an

sediment gutters were placed on the lower edge of the plates with an inclination angle

to collect the sediment and dispose it via a lateral outlet. The results approved that

the proposed approach is an effective tool to improve the separation efficiency under

special conditions up to 7%.

Preface

This work has been carried out at Department of Chemical Engineering - Regeneration

and Recycling in Bremen. First of all, I wish to express my gratitude to my supervisor

Prof. J. Thoming for a wealth of suggestions and guidance throughout the thesis

project. I would like to thank Goerge Okoth who helped me when I came to deal with

the CFX software. Also, I would like to thank Waldemar Retkowski for conservations

and discussion on optimisation . I am grateful to Michael Birkner for manufacturing

the experimental set-up and to Dietmar Grotheer for assisting in some experiments.

Also I would like to thank Lydia Achelis, for her support in the particles size analysis.

I would like to thank The Bremer Institut fuer angewandte Strahltechnik (BIAS) for

providing access to the software ANSYS 12. Further, I gratefully acknowledge the

Egyptian government for providing the financial support for my scholarship through

the Missions Department in the Ministry of Higher Education. Finally I thank my

wife and parents for their encouraging me during the work in this thesis.

Contents

Zusammenfassung 3

Abstract 5

Preface 6

List of Figures 11

List of Tables 11

Symbols 14

1. Introduction 15

1.1. Inclined plate settler - state of the art . . . . . . . . . . . . . . . . . . . 181.2. Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2. Evolution design of inlet structure of countercurrent inclined plate settlers 25

2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.2. Material and method . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.2.1. Experimental set-up and procedures . . . . . . . . . . . . . . . 272.2.2. CFD Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.3. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 362.3.1. Flow distribution study . . . . . . . . . . . . . . . . . . . . . . . 362.3.2. Hydraulic behaviour study . . . . . . . . . . . . . . . . . . . . . 362.3.3. Separation efficiency study . . . . . . . . . . . . . . . . . . . . . 42

3. Effect of optimisation of inlet zone on the hydraulic behaviour 49

3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.2. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.2.1. Geometry Details of the IPS Model . . . . . . . . . . . . . . . . 503.2.2. Optimisation Methodology . . . . . . . . . . . . . . . . . . . . . 503.2.3. Numerical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.3. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 553.3.1. Optimisation results . . . . . . . . . . . . . . . . . . . . . . . . 553.3.2. Hydraulic performance of the IPS . . . . . . . . . . . . . . . . . 553.3.3. Separation efficiency . . . . . . . . . . . . . . . . . . . . . . . . 62

8 Contents

4. New concept for the disposal of sediment in countercurrent IPS 65

4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654.2. Material and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

4.2.1. Geometry details of the IPS test systems . . . . . . . . . . . . . 664.2.2. Experimental set-up and procedures . . . . . . . . . . . . . . . 67

4.3. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.3.1. Impact of the entrance zone length on the separation efficiency . 704.3.2. Assessment of the IPS with lateral sludge collector . . . . . . . 72

5. Concluding remarks and future work 83

Bibliography 87

List of Figures 9

List of Figures

1.1. Classification of IPS based on flow direction . . . . . . . . . . . . . . . 19

1.2. Typical commercial counter-current IPS consists of different functionalzones: Feeding zone contains both the inlet and feed box, effective sep-aration zone contains the incline plates, outlet zone contains both theeffluent collection channels and outlet pipe, and sludge hopper repre-sents the collected sludge zone. downloaded from website: www.nordic-water.de/docs/content.php, on 03.01.2012 . . . . . . . . . . . . . . . . 22

2.1. Sketches of the employed inlet structures. Where LS1 is fed by pipe, LS2is fed by a nozzle distributor which is surrounded by the IPS wall, andLS3 is fed by a nozzle distributor which is surrounded by an additionalcylindrical wall. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.2. Process flow diagram of experimental set-up with aerator (A), pump(B), rotameter (C), conductivity probe (D), data acquisition system (E),camera (F), and test section for measuring dye velocity through settlers(L). (1) and (2) denote the sample collection points in the inlet andoutlet streams, respectively, while (3) denotes the stream of withdrawalconcentration sediment. . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.3. The influence of both inlet configuration and flow rate on the effi-ciency of flow distribution quantified by standard deviation (SD) be-tween lamella plates: (a) experimental data, (b) κ-ε model, and (c) κ-ωmodel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2.4. The normalized experimental RTD curves as function of flow rate: (a)LS1, (b) LS2, and (c) LS3 . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.5. The impact of both inlet configuration and flow rate on the flow patternof the IPS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.6. Comparison of normalized RTD curves between simulations and exper-imental results for LS1 at flow rates: (a) Q= 200 l/h; (b) Q= 250 l/h;(c) Q= 300 l/h; (d) Q= 350 l/h. . . . . . . . . . . . . . . . . . . . . . . 43



10 List of Figures

2.9. Comparison of the predicted NTIS numerically with experimental data:(a) LS1, (b) LS2, and (c) LS3. . . . . . . . . . . . . . . . . . . . . . . . 46

2.10. Separation efficiency [η = (SSin - SSout) x100/SSin] as a function ofboth flow rate and inlet configuration. . . . . . . . . . . . . . . . . . . 47

3.1. Schematic of the IPS model . . . . . . . . . . . . . . . . . . . . . . . . 513.2. Geometry of distributor . . . . . . . . . . . . . . . . . . . . . . . . . . 543.3. sensitivity analysis for parameters D, H and h showing the standard

deviation (SD) as function of percent change of input parameters (dC):[(a),(b) and (c)] and [(d),(e) and (f)] for non-optimised and optimiseddistributor respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.4. Quality of flow distribution between lamella plates . . . . . . . . . . . . 573.5. Velocity distribution in the transverse direction in the middle of the

settlers: (a),(b) and (c) for non-optimised distributor;(d),(e) and (f) foroptimised distributor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

3.6. Influence of the optimisation on the RTD curves: (a)=250 l/hr; (b)=300l/hr;(c)=350 l/hr. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

3.7. Influence of the optimisation on the calculated NTIS . . . . . . . . . . 613.8. Influence of the optimisation on the separation efficiency of the IPS model 63

4.1. Isometric of IPS1 with variable entrance zone length (a) and IPS2 withlateral sludge collector (b) . . . . . . . . . . . . . . . . . . . . . . . . . 68

4.2. Experimental set-up with pump(A,D),valve (B),ground tank (T1), over-head tank (T2), rotameter (c).(1) and (2) denote the sample collectionpoints in the inlet and outlet streams, respectively, while (3) denotesthe stream of withdrawal concentrated sediment . . . . . . . . . . . . 69

4.3. Experimental separation efficiency as function of entrance zone lengthand inclination angle of the IPS1 . . . . . . . . . . . . . . . . . . . . . 70

4.4. Increment rate of separation efficiency as function of entrance zonelength and inclination angle of the IPS1 . . . . . . . . . . . . . . . . . 71

4.5. Turbulence eddy dissipation (TED) as function of entrance zone length 734.6. Details of plates used in the experiments . . . . . . . . . . . . . . . . . 744.7. Impact of the proposed plate on the separation efficiency . . . . . . . . 764.8. Velocity streamlines in two settler configurations with a bar of 6 mm

height and of two angles of inclination, 450 and 550. . . . . . . . . . . . 774.9. Effect of six plate structures with a sediment gutter inclined by 450 on

the normalised RTD curve predicted by numerical simulation. . . . . . 784.10. Effect of six plate structures with a sediment gutter inclined by 550 on

the normalised RTD curve predicted by numerical simulation. . . . . . 794.11. Effect of the different plate structures on the NTIS . . . . . . . . . . . 804.12. Effect of the different plate structures on the mean residence time of

IPS2 model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 814.13. Comparison of effect of the different plate structures on the NTIS and

the separation efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . 82

List of Tables 11

List of Tables

2.1. normalized variance values as function in flow rate and inlet configuration 42

3.1. Influence of optimisation on the NTIS . . . . . . . . . . . . . . . . . . . 61

4.1. Description of the experimental plan . . . . . . . . . . . . . . . . . . . 75

List of Tables 13

Symbols

B the sum of body forces,N

Cμ, Cε1, Cε2 constants in the κ− ε model dimensionless

Dφ kinematic diffusivity, m2.s−1

E(t) residence time distribution, s−1

E(θ) dimensionless residence time distribution, s−1

MRT mean residence time, s

NTIS number of tanks in series model

Pκ shear production, kg.m−1.s−3

SD standard deviation

tm mean residence time, s

tmean theoretical mean residence time, s

ti time interval, s

U the mean velocity vector, m.s−1

κ turbulent kinetic energy, m2.s−2

p′

modified pressure, Pa

Sφ volumetric source term

SCt turbulence Schmidt number dimensionless

ε turbulent dissipation rate,m2.s−3

ω turbulent frequency,s−1

ρ liquid density, kg.s−3

μeff effective viscosity,Pa.s

μt turbulent viscosity,Pa.s

14 List of Tables

σε constants in the κ− ε model dimensionless

σκ constants in the κ− ε and κ− ω models dimensionless

β′

, α, β, σω constants in the κ− ω model dimensionless

φ tracer concentration,kg.m−3

σ2 spread of residence time curve

σ2θ normalized variance of residence time distribution

θ dimensionless time

θpeak the time taken to reach maximum tracer concentration (dimensionless)

15

1. Introduction

In the recent days, the interest with environmental pollution control is raised as one of

the main world problems. With the increase of pollution rates due to the increase of

wastewater quantities and the growing demand for waste treatment and water recla-

mation and recycling, the need to preserve a high quality waste treatment effluent had

appeared to achieve the environmental limits.

The separation of particles from suspension is an important step in the water treat-

ment, the wastewater treatment and in many industrial processes because the per-

formance of these tanks have direct or indirect impact on other units of treatment

plant[1].

The main methods which are used in the separation of solid / liquid mixtures are

screening, settling (sedimentation), filtration and flotation.

A gravity sedimentation (settler) is the most economical method for separating

inorganic solids from suspensions under certain conditions, but it is not effective for

the separation of the organic flocs due to a negligible difference in densities between

particles and fluid or when the particles are very small . In spite of this drawback, the

settling by gravity is widely used in water treatment, wastewater treatment, and in

the chemical process industries because its cost is less than alternative systems [2, 3].

The separation efficiency depends on the characteristics of the solids and the hy-

draulic condition of flow field in the settling tank. These tanks represent a major

component of any water and wastewater treatment plant because the costs of their

construction and operation represent approximately 30% of the total investment of

any treatment plant [4].

16 1. Introduction

Settling tanks design based on the basis of theory developed by Hazen [5] and Camp

[6]. They concluded that the separation efficiency of ideal settling tank depends on

the overflow rate, which is function in the horizontal area of tank, rather than the

detention time. Consequently, the efficiency in this type is directly related to the

surface area available for settling.

The settling tanks are classified in two main categories: primary and secondary.

The concentration of suspended solids in primary tanks is considerably low comparing

with the secondary tanks which indicates that the concentration field has a minimal

effect on the flow field [7].

The behaviour of the settling of suspension is classified into four types: discrete,

flocculent, hindered, and compression. In the discrete type, the suspensions consist of

discrete particles which settle as individuals based on the Stokes’ low, and there is very

little interaction between them. In flocculent settling type, the suspensions contain

particles which tend to agglomerate together, resulting growth in their size and settle

faster. Hindered settling occurs when the solid concentration in the suspension is high,

causing a decrease in the fall velocity due to interaction with surrounding particles.

Compression settling occurs when the solid concentration become so high and the

particles support each other and to obtain further separation a compression tool is

used [8, 9] .

To decrease the footprint of settling tanks or if the available space for separation

process is limited, as it is in industrial wastewater treatment, the using of incline

plate settlers (IPS) is preferred. It provides high space-time yield due to short settling

distance and the available settling area is given by the total area of the plates projected

on a horizontal surface. Moreover,the IPS are attractive tools for the separation of

solid-liquid suspensions due to low energy demand because it does not need scrapers

which are used for collecting and removing the sediment from the settling tank floor

17

[10, 11].

IPS are a classical subject with a long history, and Boycott [12] was the first who

observed that the settling rate of suspension is better ”‘if the tube is inclined than

when it is vertical”’, this so-called Boycott effect. The settling behaviour in inclined

vessel was modelled firstly by Ponder [13] and latter by Nakamura and Kuroda [14]

and both so-called PNK theory [15]. This theory explains that the quality of the

clarified fluid is function in the vertical settling velocity of particles and the horizontal

projection area of the settler. Equation (1.1) represents the PNK theory.

S(t) = vs(bcosθ + Lsinθ) (1.1)

Where S(t) is the clarification rate of suspension per unit depth in the third dimension

of settler, vs is the vertical settling velocity, b is spacing between plates, L is the length

of settler, and θ is the inclination angle of settler with vertical.

Yao[10] analysed theoretically suspensions that flow through a duct and contain

discrete particles. He developed a general equation that describes the condition which

makes the trajectory of a particle stopping in this duct, and he discussed the influence

of the relative settler length (l/D) and inclination angle of the duct on the settling

performance, where l and D denote the length and height of settler (perpendicular

space between wall) respectively. He suggested that the relative settler length should

be below 40 and preferably about 20, while he found the performance of settler de-

creases as the inclination angle increases because the settling distance increases, and

the performance decreases rapidly when the angle becomes more than 400.

Several studies were performed based on the PNK theory. Acrivos and Herbolzheimer

[16] studied the behaviour of the three stratified layers: a sediment layer, a suspension

layer and a clarified layer, and they developed an equation to describe the thickness

of ”‘the clear fluid layer that formed underneath the downward facing plate”’ when

18 1. Introduction

the space between the plates is small compared with their length (low aspect ratio).

Also, they attributed the discrepancy from the PNK theory to the instability of the

interface between these layers which causes resuspension of particles and then reduces

the efficiency of separation.

Moreover, Leung and Probstein [17] analysed the behaviour of the above three layers,

and they developed equations which represent the velocity profiles for each layer.

Also, they observed that the increase of solids concentration and angle inclination

of settler decrease its efficiency. All previous analyses were conducted under laminar

flow condition, two-dimensional geometry and monodisperse suspension. Furthermore,

a number of studies were performed regarding the performance of inclined settler

containing bidisperse or polydisperse suspensions [18, 19, 20, 21, 22].

The flow instability problem in inclined settler was investigated by many authors

under different conditions [23, 24, 25, 26, 27, 28, 29]. These studies showed that the in-

stability of flow is a function of the settler angle, the feed flow rate and the suspension

concentration. Borhan [27]and Acrivos and Borhan [28]concluded that increasing con-

centration of suspension, angle of inclination with vertical and fluid viscosity improves

the stability of inclined settler. Also, they observed that the efficiency separation and

stability of high aspect-ratio (H/b) inclined settler is better than low aspect-ratio.

Here H and b denote the height of suspension in settler and the perpendicular space

between the plates respectively.

1.1. Inclined plate settler - state of the art

The IPS system can be constructed in one of three modes as shown in Figure 1.1:

With the counter-current, the direction of the suspension flow is opposite to the sedi-

ment flow. In a co-current flow, the direction of the suspension flow is the same as the

sediment flow direction. In a cross-flow, the direction of the suspension flow is per-

1.1. Inclined plate settler - state of the art 19

Figure 1.1.: Classification of IPS based on flow direction

pendicular to the sediment flow direction. Leung and Probstein [17] stated that the

feed of settler in the middle is complicated, , while Kowalski and Mieso [30] mentioned

that the counter-current system is widely used more than cross-flow mode although

its separation efficiency is more efficient. Rushton et al. stated that the co-current

mode is suitable for separation process that produce low bulk density sludge such as

metal hydroxides, and the best inclination angle for this type generally between 300

and 400. Also they concluded that the counter-current mode is preferred because this

system is simpler in design and operation.

Sarkar et al. [31] predicted the separation performance of conter-current IPS based

on the dimensional analysis approach, and they utilised seven non-dimensional vari-

ables to develop a model representing the efficiency under different geometrical and

dynamic conditions. Okoth et al. [32] summarized the factors which have effect on the

separation efficiency of IPS, and they modelled the suspension-sediment interaction

phenomenologically. He and Marsalek [11] investigated the effect of using vortex plates

20 1. Introduction

in clarifiers on the sedimentation efficiency experimentally, and numerically by using

computational fluid dynamics (CFD). They concluded that the efficiency is improved

by implementing the vortex plates up to 8% numerically and 26% experimentally.

All previous approaches are faced with a general problem, which is resuspension of

sediment due to flow instabilities as shown above and shear stress between the phases

as well as between the fluid and the lamella surfaces. This problem increases if the

particles are either cohesive or biological particles.

In more detail the problems of IPS are described as follows:

1. Flow instabilities and resuspension occur especially at the inlet zone, if the in-

let zone of IPS is not designed correctly, where the separation zone of IPS is

influenced greatly by the inlet structure [33].

2. Common designs of IPS have a problem by interference between the feed stream

and sediment path. When the sediment drops from one plate and encounter with

suspension which moves in the upflow direction, the sediment will be resuspended

easily [17].

3. Some designs use downflow direction for the feed stream; this maybe have more

problems in the exit zone. When the sediment drops from one plate and collides

with clarified supernatant which also moves in the downflow direction, likely

the sediment will go out with the temporarily clarified solution and become

suspension again.

The counter-current IPS mode is selected in this study. Figure 1.2 illustrates a

typical common industrial commercial countercurrent IPS. It consists of two main

parts: the upper tank inclines with the same angle of inclination of plates which are

in range between 500 and 600, and the second lower conical part that uses to collect

and thickening the sludge. To obtain clarified fluid that can be produced due the

1.2. Thesis outline 21

separation process in this unit, a suspension enters the IPS via a pipe into feed box

between the inclined plates. This suspension enters the plate cells from the sides

via the inflow openings, and then turns up between the plates and flows to effluent

channels via openings and then leaves the separator via the outlet pipe. Separation

occurs during the upward flow of the suspension to be clarified. The solids settle on

the plates and slide down into the sludge hopper. The sludge is thickened in the sludge

hopper and exits the settler via the sludge outlet [34].

This counter-current IPS system is faced with two general problems, which are the

flow distribution between the settlers due to feeding of the plate lamella (package)

from one side only and the resuspension of sediment occur at the inlet zone of settlers,

and both of which are influenced by the inlet zone configuration [32]. These problems

lead to decrease the separation efficiency. This study aims to focusing and understand

the causes of these problems in the counter-current mode in order to make technical

improvements for the IPS. These improvements will contribute in decreasing these

problems and obtain better separation efficiency.

1.2. Thesis outline

The thesis is organised in five chapters:

In Chapter 2, deals with the hypothesis that the separation zone of IPS is influ-

enced greatly by the inlet structure. Owing to this hypothesis, I assumed that IPS

performance will be improved due to enhancement of the quality of flow distribution

within each settler and its hydraulic characteristics. Three lab-scale IPS with different

inlet structure were investigated to explore the effect of feeding these IPS via a nozzle

distributor on their hydraulic performance and separation efficiency, and developing a

Computational Fluid Dynamic (CFD) model for simulating the hydraulic behaviour

of the IPS, and validate the CFD simulation with the lab-scale experimental results.

22 1. Introduction

Figure 1.2.: Typical commercial counter-current IPS consists of different functionalzones: Feeding zone contains both the inlet and feed box, effective sep-aration zone contains the incline plates, outlet zone contains both theeffluent collection channels and outlet pipe, and sludge hopper repre-sents the collected sludge zone. downloaded from website: www.nordic-water.de/docs/content.php, on 03.01.2012

1.2. Thesis outline 23

In Chapter 3, based on the results of Chapter 2, an objective function was suggested

to optimise the hydraulic performance of IPS. This was based on the hypothesis that

reducing the discrepancy of flow distribution among outlet openings of nozzle distrib-

utor, will enhance the hydraulic behaviour of IPS. A global optimisation-CFD tool for

the optimisation of a nozzle distributor was applied. Thereafter the optimised distrib-

utor is used in a lab-scale IPS to investigate the impact of optimised distributor on

the hydraulic behaviour and the separation efficiency of the IPS.

The interference between the feed stream and sediment path in the countercurrent

mode causes resuspension for the sediment at the lower edge of the settler. A new

structure of IPS is proposed in Chapter 4 to decrease this problem. Experimental

work and numerical simulations were performed to assess this approach.

Chapter 5 summaries the general conclusions of thesis and gives suggestions to

continue this research.

25

2. Evolution design of inlet structure

of countercurrent inclined plate

settlers

2.1. Introduction

Separation efficiency of IPS is usually well below the theoretical performance due

to many factors. Okoth et al. [32] summarized these factors by the modelling of

suspension-sediment interaction phenomenologically. In their experimental study, they

employed a nozzle distributor to avoid the problems which could be caused by feeding

these settlers by a lateral feed box as explained in Chapter one, consequently improve

the hydraulic performance of IPS which is one of the factors affecting the separation

efficiency. However, they did not investigate in detail the effect of using this nozzle

on both distribution of the total flow between the settler channels and flow patterns

of the entire IPS system, both of which have a significant impact on the separation

performance of the IPS. The equalized distribution of suspension within each settler is

important to obtain an equal overflow velocity on every plate, which contributes to the

improvement of the IPS efficiency, while the flow patterns are essential in determining

the flow characteristics in order to achieve a reliable design [35].

26 2. Evolution design of inlet structure of countercurrent inclined plate settlers

The residence time distribution (RTD) curve is an effective tool to understand the

reactor hydrodynamics, which characterise the operational shortcoming of a new sep-

aration equipment. Moreover, the residence time distribution is used to describe the

flow pattern in a reactor which is important in the design and modelling of reactors

[36, 37, 38].

The flow in tanks always deviates from the ideal plug flow or complete mixed flow

and is usually described as a non-ideal flow pattern. Levenspiel [39] described two

methods for the characterization of a non-ideal flow pattern. One is the longitudinal

dispersion (LD) model which represents a flow that deviates from plug flow, and

the other is the tanks in series (TIS) model which describes the mixing within a

single real reactor as a number of equally sized continuous stirred-tank reactors in

series (CSTRs). When the number of tanks in series is one, the model predicts the

performance of an ideal CSTR. As the number of tanks in the TIS model increases,

the flow within the reactor approaches that of a plug flow reactor (PFR) [40]. Both

the LD and TIS models are characterized by dispersion number and the number of

tanks in series (NTIS) respectively. The TIS model was used in this study because it

is mathematically much simpler than the LD model. Also, the NTIS does not depend

on the definition of the inlet and exit boundary conditions.

”‘Computational fluid dynamics (CFD) has become a robust tool in the design of

reactors and provides useful and detailed information prevailing in the reactors, such as

velocity field, concentration distribution, and phase hold-up distribution”’ [41]. CFD

is typically used in the simulation of RTD [42, 43, 44, 45]. The hydrodynamic of

flow field can be described and modeled by using Navier Stokes equations, which are

solved numerically using a CFD codes/models. Many software packages, such as CFX,

Fluent, and Comsol, offer several models, which should be experimentally verified [46].

Two turbulent models (κ − ε model and κ − ω model) were implemented in the

2.2. Material and method 27

present study to predict both velocity flow field and RTD curve. Thereafter, the

predicted results were compared with the experimental results to determine which of

the two models give the most realistic results.

Furthermore, the influence of hydraulic performance on the separation efficiency of

the lab-scale IPS fed by a nozzle distributor is demonstrated in this chapter.

Note: All figures in this chapter were published in Salem et al.[47].

2.2. Material and method

The separation zone of IPS is influenced greatly by the inlet structure [33]. Because

of this, three inlet configurations (LS1, LS2 and LS3) were suggested to investigate

the impact of feeding the IPS via a nozzle distributor on its hydraulic behaviour and

separation efficiency. Sketches of the employed inlet structures are shown in Fig. 2.1.

LS1 model is fed by pipe only, and it is investigated to identify the effect of using

distributor, used in LS2, on hydraulic performance of IPS, and this model is used by

Okoth et al. in their study [32]. LS3 model is suggested base on the hypothesis that

housing this distributor in cylinder shape will improve hydraulic performance of IPS.

The IPS was made of plexiglas of 15 mm in thickness with internal dimensions of

(100 mm x 80 mm x 480 mm) and was placed on a ramp with angle of inclination 450

in all tests. Three polyvinyl-chloride plates 300 mm long and 5 mm thick were used

in the IPS. The three plates could be fixed at any distance from the nozzle apex, and

the spacing between the plates was 16 mm.

2.2.1. Experimental set-up and procedures

Two techniques were used to identify the hydraulic behaviour of the IPS. The first

technique was measurement the velocity within every settler by using the colour ve-

locity measurement (CVM) method [48]. The second method was used to quantify


Figure 2.1.: Sketches of the employed inlet structures. Where LS1 is fed by pipe, LS2is fed by a nozzle distributor which is surrounded by the IPS wall, andLS3 is fed by a nozzle distributor which is surrounded by an additionalcylindrical wall.


the hydraulic behaviour of the IPS by using residence time distribution (RTD) ex-

periment. Furthermore, to explore the impact of using distribution nozzle on the

separation process, the removal of suspended solids (SS) efficiency for the IPS was

determined.

A small slug of concentration dye solution (potassium permanganate) was injected

impulsively into the IPS inlet, and a high resolution digital camera was used to provide

the data for the calculation of mean velocity within every settler by computing the

required mean time for the dye to travel a known length.

The RTD was measured by quickly injecting 3 mL of tracer (KCL, 3 g/l) into the

IPS inlet and the tracer concentration at the outlet was measured with a conductivity

probe every five seconds using a data acquisition system. The experimental procedures

were repeated five times for each flow rate. The fitting of curve was then performed

to minimize the deviation between the experimental data and the simulation data by

using exponentially modified Gaussian peak function which has given us adjusted R2

values between 0.94 and 0.97.

The separation efficiency was determined by specifying the concentration of SS in

the samples which were collected from the inlet stream and outlet stream. The samples

were filtered under pressure through a 0.45 μm pore size cellulose nitrate membrane

using a compressed air filter model 16249 from Sartorius AG. Thereafter, a dry mass

concentration analysis is performed using a moisture analyser model MA45 also from

Sartorius AG, Germany [32].

To carry out these tests, an experimental set-up was constructed. A process flow

diagram of the experimental set-up is shown in Fig. 2.2. It consists of a 35 L storage

tank with an aerator mounted at the bottom denoted as A. This tank was filled with

tap water in both velocity measurement and RTD experiments, while it was filled with

suspension - including crushed walnut shell particles - in the separation efficiency test.


The feeding of fluid was achieved by using a centrifugal pump denoted as B. The flow

rates were regulated by using variable direct-current voltage power supply in a range

of 0-24 volt. There is a flow meter between the pump and the inlet, denoted as C. Both

the conductivity meter and data acquisition system - denoted as D and E - were used

only in the RTD test. To determine the separation efficiency, three sets of samples

were collected from both the inlet stream (section 1) and the outlet stream (section

2).

The crushed walnut shell particles had a density of 1.35 g/cm3 and their size distri-

bution was analysed using the laser diffraction technique, with a Malvern (Mastersizer,

2000) analyser. Particle size distribution parameters like d(0.1), d(0.5), d(0.9) were

analysed, and the corresponding values were 35, 115, and 235 μm respectively. The

d(0.1), d(0.5) and d(0.9) values indicate that 10%,50% and 90% of the particles have

diameters which are smaller than or equal to the stated size.

The separation experiments were carried out as follows: At the beginning of each

experiment the suspension was mixed well in the feed tank for 5 minutes to achieve a

uniform distribution. The mixture was then pumped to the IPS. Three samples were

collected from the inlet stream after 5 minutes from the onset of pumping. On other

hand, no fixed time was set for the outlet stream sampling since this depended on the

hydraulic residence time. At the end of every experimental run, the IPS was drained

and the samples returned to the feed tank. The longest duration of each experiment

was 15 minutes.

2.2.2. CFD Simulations

CFD analysis using CFX-10 from ANSYS was performed by employing the standard

κ − ε model and κ − ω model, where κ, ε and ω denote turbulent kinetic energy,

turbulent dissipation rate, and turbulent frequency, respectively.


Figure 2.2.: Process flow diagram of experimental set-up with aerator (A), pump (B),rotameter (C), conductivity probe (D), data acquisition system (E), cam-era (F), and test section for measuring dye velocity through settlers (L).(1) and (2) denote the sample collection points in the inlet and outletstreams, respectively, while (3) denotes the stream of withdrawal concen-tration sediment.


For wall-bounded flows, as in IPS, the standard κ− ε model neglects the effects of

viscosity in the near-wall region, and it is valid for turbulent core flow [49]. A scalable

wall function is adopted in CFX to improve the near wall treatment by ”‘limiting the

value of the dimensionless distance from the wall (y+) to be = 11.06 (where 11.06 is the

intersection between the logarithmic and linear near wall profile), which prevents the

mesh points falling within the viscous sub-layer. Thus, all fine mesh inconsistencies

are avoided”’ [50]. On the other hand, an automatic near wall treatment is used in the

κ − ω model by Wilcox [51], which provides an analytical solution for ω in both the

logarithmic and the viscous regions. ”‘The idea behind the automatic wall treatment

is that the model shifts gradually between a viscous sub-layer formulation and wall

functions, based on the grid density near wall”’ [52].

The two models were selected in this chapter for purpose of comparison, and to

determine which model would be closed to experimental results.

The governing equations of mass and momentum were determined using the Reynolds

averaged Navier−Stokes Eqs. 2.1 and 2.2

∂ρ

∂t+∇ · (ρU) = 0 (2.1)

∂

∂tρU +∇ · (ρU ⊗ U) = ∇p

′

+∇ · (μeff (∇U + (∇U)T )) + B (2.2)

Where ρ is the liquid density, B is the sum of body forces, U is the mean velocity

vector, p′

is the modified pressure and μeff is the effective viscosity. The calculations

of p′

and μeff are:

p′

= p+2

3ρk (2.3)


μeff = μ+ μt (2.4)

Where μeff is the turbulent viscosity, which is given by:

μt = Cμρκ2

ε(2.5)

μt = ρk

ω(2.6)

Equations 2.5 and 2.6 were employed for the κ−ε and the κ−ω models respectively.

The values of κ and ε in the κ − ε model are estimated from equations 2.7 and 2.8

respectively.

∂(ρk)

∂t+∇ · [ρUK] = Pk − ρε+∇ · (

μeff

σκ

∇κ) (2.7)

∂(ρε)

∂t+∇ · [ρUε] =

ε

κ(Cε1Pκ − Cε2ρε) +∇ · (

μeff

σε

∇ε) (2.8)

Where Pκ is the shear production due to turbulence. The values of κ and ω in the

κ− ω model are estimated from equations 2.9 and 2.10 respectively.

∂(ρk)

∂t+∇ · [ρUK] = Pk − β

′

ρkω +∇ · [(μ+μt

σk

)∇k] (2.9)

∂(ρω)

∂t+∇ · [ρUω] = α

ω

kPk − βρω2 +∇ · [(μ+

μt

σω

)∇ω] (2.10)

Where Pκ is the production rate of turbulence.

The set of the κ− ε model constants is Cμ= 0.09, Cε1 = 0.1256, Cε2= 1.92, σκ= 0.9,

σε= 1.3.


While the values of the κ−ω model constants are β′

=0.09, α=5

9, β= 3

40, σk=2, σω=2.

To model the RTD in the ANSYS CFX 10.0, KCL was used as tracer. It was

introduced in the software as a volumetric additional variable with concentration 3

kg m−3 and it has molecular diffusivity of 1.703 x 10−9 m2s−1 [53]. The governing

equation for the tracer transport reads as follows:

∂φ

∂t+∇ · (Uφ) = ∇((ρDφ +

μt

Sct)∇ · (

φ

ρ)) + Sφ (2.11)

Where φ is the tracer concentration, φ

ρis the conserved quantity per unit mass,

Sφ is a volumetric source term, Dφ is the kinematic diffusivity for the scalar and

Sct is the turbulence Schmidt number which represents the ratio between the rate

of momentum transport and passive scalars. The simulation was performed in two

steps. In the first step, the simulation was performed in steady-state turbulent flow

condition to determine the hydraulic fluid characteristics such as velocity and kinetic

energy by using both the κ-εmodel and κ-ω model individually. In the second step, the

information obtained from the first stage was used to solve Eq. (2.11). By determining

the tracer concentration φ at the outlet from the transient simulations, the residence

time distribution could be computed as follows:

C(t) = φoutlet(t) (2.12)

The normalized concentration E(t) is used to compare RTD curves under different

flow rate conditions, and it is defined by the following equation:

E(t) =C(t)∫

∞

0C(t)dt

(2.13)


The mean residence time is given by:

tm =

∫∞

0tC(t)dt∫

∞

0C(t)dt

(2.14)

The spread of the residence time curve (variance) is measured by:

σ2 =

∫∞

0t2C(t)dt∫

∞

0C(t)dt

− (tm)2 (2.15)

The normalized variance is calculated from the following equation:

σ2

θ =σ2

t2m(2.16)

Finally the following equation estimates the NTIS [39]:

NTIS =1

σ2θ

(2.17)

The boundary conditions for the system were as follows: IPS wall, nozzle distri-

bution, and lamella plates were assumed as standard wall boundary conditions with

no-slip flow. The boundary condition at the inlet was set as the mass flow rate with

a value from 55 to 97 g/s, whereas the boundary condition at the outlet was set as

the average static pressure across the outlet area. The tetrahedral grid was used be-

cause it is the most common way of numerically solving problems in three-dimensional

domains of complex shapes [54, 55]. Furthermore, the inflated mesh was used in the

near-wall regions to capture the effects of the boundary layer. The mesh structures

for LS1, LS2 and LS3 have 1.09 x 106, 1.2 x 106 and 1.07 x 106 elements respectively.


2.3. Results and discussion

2.3.1. Flow distribution study

Owing to the hypothesis that the IPS efficiency depends on the quality of flow dis-

tribution within each settler, the velocity through every settler was determined. The

flow velocity in the CFD was derived from the calculation of velocity at a specific

section in the transverse direction, which was in the middle of the settler, while in

the experiment it resulted from the required mean time for the dye to travel a known

length in the longitudinal direction. Thereafter, the standard deviation (SD) of a set

of the values of flow velocities through every settler was utilized as a criterion for the

flow distribution efficiency. As shown in Figures 2.3a, 2.3b, and 2.3c, the SD increases

as the flow rate increases, indicating a decrease in the flow distribution efficiency. Fur-

thermore, the distribution efficiency clearly depends on the type of inlet structure. It

was optimal for LS3, where the average values of SD were between 0.09 and 0.22, while

the inlet structure of LS1 showed the worst performance, with average SD values from

1.3 to 2.1, and highest dependency of SD on flow rate. Obviously, the flow distribution

within the IPS can be significantly improved by using the nozzle distributor. Addi-

tionally, the simulation and experimental results clearly deviate significantly because

two different methods were used in the determination of the flow velocity through the

settlers, and the purpose here was only to perform a qualitative evaluation.

2.3.2. Hydraulic behaviour study

RTD Experiments

Runs were carried out at four different flow rates, with values of 200, 250, 300, and

350 l/hr and RTD curves were plotted between E(t) versus (t).

The influence of flow rate on the normalized concentration curves for different inlet

2.3. Results and discussion 37

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

�

�

��

��

��

��

�

��

��

��

��

��

Figure 2.3.: The influence of both inlet configuration and flow rate on the efficiency offlow distribution quantified by standard deviation (SD) between lamellaplates: (a) experimental data, (b) κ-ε model, and (c) κ-ω model.


structure is shown in Fig. 2.4. As expected, the peak of the curve appears earlier

shorter with increasing flow rate in the three cases (LS1, LS2 and LS3). Also a strong

tail of the RTD curve was observed in all cases. This leads to increase deviation

from the ideal case (i.e. plug flow) which is not desirable for the performance of the

separation process [56, 57]. Furthermore, the existence of a tail indicates presence

of dead space, which gives an indication of stagnant pockets or recirculation regions.

These regions reduce the effective volume and should be kept as small as possible.

To quantify the hydraulic performance of the IPS, the NTIS is calculated from the

RTD curve. Fig. 2.5 illustrates the impact of both inlet configuration and flow rate on

the NTIS. This figure reveals that the NTIS depends strongly on the inlet structure,

additionally, as expected, the hydraulic behaviour of LS1 is significantly different from

the plug flow due to non-use of an effective inlet device to distribute the suspension.

In contrast, the NTIS for LS2 and LS3 is about 7 to 10 which indicated tendencies to

plug flow.

Comparison between numerical simulation and experimental results

Figures 2.6,2.7 and 2.8 illustrate a comparison between the simulated data and ex-

perimental measurements of the normalized RTD. As can be seen from the figure,

the main characteristic for LS1 is an extreme initial peak with an average value of

θ ∼= 0.50 which indicates a short-circuiting stream where is dimensionless time [θ=

ti/Tmean; Tmean= theoretical mean retention time]. By contrast, the average values of

initial peak for both LS2 and LS3 reached maximum concentration at θ = 0.78, and

θ=0.90 respectively. This means the hydraulic behaviour of LS3 approaches plug flow

conditions [8].

Moreover, it can be observed from the figure that the simulated RTD curves of

both the κ-ω model and κ-ε model are deviated slightly for LS1. On the other hand,

for both LS2 and LS3, the predicated results by the κ-ε model for both types differ


� ��

��

��

��

��

� ��

��

��

��

��

� ��

��

��

��

��

��

�

��

��

��

��

��

�

��

��

��

��

��

��

��

��

�

�

��

�

��

Figure 2.4.: The normalized experimental RTD curves as function of flow rate: (a)LS1, (b) LS2, and (c) LS3


��

��

��

��

��

��

�

�

�

�

�

��

�

�

�

�

��

��

Figure 2.5.: The impact of both inlet configuration and flow rate on the flow patternof the IPS.


considerably from the predicated results by the κ-ω model. This means that a correct

modelling of the nozzle distributor plays an important role in the final results.

Also, it can be generally observed from the figures for both LS2 and LS3 that the

predicted results obtained by the κ-ω model better match the experimental observation

when compared with the κ-ε model. This is especially evident in the prediction of the

time at which the peak concentration of the tracer is observed, the end of the tail and

the normalized variance values. Whereas the discrepancy of the normalized variance

values between the predicted results by κ-ω model and experiments data was around

8% for LS2 and LS3, while the differences in the normalized variance values between

the predicted results by the κ-ε model and experiments data was within 20% for LS2

and 35% for LS3. This indicates that the κ-ω model provides a good overall description

of the IPS behavior.

As previously mentioned, having a correct model of the nozzle distributor has a

large impact on the final results, and despite the use of inflated mesh to resolve the

wall-layer accurately, there was little qualitative discrepancy between the predicted

RTD curve by the κ-ω model and the experimental results. This discrepancy could

be because of existing eddies of a wide range of length scale in the boundary layer,

which are totally modelled in the κ-ω model. Further work needs to be undertaken to

improve the prediction of the RTD curve by using other models, such as a large-eddy

simulation model. This model has the capability of resolving the three-dimensional

time-dependent details of the large and medium (i.e., resolved) scales, whereas the

effects of the small unresolved eddies are modelled with a sub-grid turbulence model

[58].

It is important to study the effect of flow rate on RTD curves to compare the

behaviour of simulated RTD curves with the experimental results. It is noticeable

that an increase in liquid flow rate leads to decreases in the average residence time


Table 2.1.: normalized variance values as function in flow rate and inlet configuration

IPS Derived data fromflow rate [l/h]

200 250 300 350σ2θ

LS1Experiment 0.34 0.54 0.52 0.57κ-ε model 0.37 0.42 0.53 0.60κ-ω model 0.43 0.50 0.56 0.61



as discussed in the RTD experiments section. Also,table (2.1) illustrates that, the

normalized variance values of LS1 increase as flow rate increases for both experiment

and simulation results. On the other hand, in both the LS2 and the LS3, flow rate did

not change the normalized variance values significantly, despite approximately 2-fold

increases in the flow rate. This indicates that a wide working range for both LS2 and

LS3. Figure 2.9 illustrates the hydraulic characteristics of the IPS as function in both

the flow rate and inlet structure. This figure shows that the calculated NTIS from CFD

simulations by using the κ-ω model also match well with the experimental findings.

Furthermore, it can be observed that the values of NTIS based on the normalized

variances are equivalent to 2 to 3 NTIS for LS1, 7 to 9 NTIS for LS2 and 9 to 10 NTIS

for LS3. These values again reveal that the LS3 provides significantly better hydraulic

behavior amongst the three.

2.3.3. Separation efficiency study

The impact of flow rate and inlet configuration on the removal of suspended solids is

illustrated in fig. 2.10. The results reveal that the separation efficiency decreases as

the flow rate increases for the three inlet structures. This could be expected based on


��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

�

��

��

��

��

�

��

��

��

��

�

��

Figure 2.6.: Comparison of normalized RTD curves between simulations and experi-mental results for LS1 at flow rates: (a) Q= 200 l/h; (b) Q= 250 l/h; (c)Q= 300 l/h; (d) Q= 350 l/h.


��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

�

��

��

��

��

�

��

��

��

��

�

��



��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

�

��

��

��

��

�

��

��

��

��

�

��



��

�

�

�

�

��

��

��

�

�

��

��

��

��

��

��

�

��

��

��

��

��

��

Figure 2.9.: Comparison of the predicted NTIS numerically with experimental data:(a) LS1, (b) LS2, and (c) LS3.


Figure 2.10.: Separation efficiency [η = (SSin - SSout) x100/SSin] as a function of bothflow rate and inlet configuration.

the previous results, which indicated that the hydraulic behavior of IPS decreases with

increases in the flow rate. It can also be observed that the separation efficiency for

both LS1 and LS2 are closed to each other in the flow range of 200l/h to 250l/h which

means that the separation efficiency at low flow rate does not depend significantly

on the configuration of the inlet zone. Moreover it can be seen that the separation

efficiency of LS1 drops rapidly after 250 l/h. On the contrary, the separation efficiency

of both LS2 and LS3 decreases gradually after 250l/h. In general, it can be said that

the inlet configuration employed in LS3 improved both the distribution of flow within

every settler and the hydraulic behavior of the IPS, which led to an improvement in

its separation efficiency.

These results proved that the inlet structure of LS3 helpful in improving both the


distribution of flow within every settler and hydraulic behaviour of the IPS, both of

which improve the separation efficiency of the IPS.

49

3. Effect of optimisation of inlet zone

on the hydraulic behaviour

3.1. Introduction

As it could be demonstrated in chapter two, the inlet configuration including a nozzle

distributor of a lab scale IPS has a great impact on the improving of its hydraulic per-

formance and the separation efficiency. To achieve convenient hydraulic behaviour, the

distributor design should have simultaneously two constraints: a sufficient total area

of inlet openings, and a equal flow distribution among the openings [59]. Inadequate

design of this distributor could lead to hydraulic flow problems which are often the

reason for poor separation. Thus, optimising of the distributor is necessary to achieve

higher separation efficiency. The CFD was used to optimise the performance of many

settling tanks because it is fast, reduces the experimental demand and provides more

confident scale-up [60, 61, 62, 63].

Among the available approaches to optimisation is the response surface methodology

(RSM). ”‘RSM is combination of statistical and mathematical techniques that are

useful for modelling, improving and optimising a process”’ [64]. The RSM is considered

a global optimisation method, and it has many advantage such as, ”‘the local sensitivity

analysis is not necessary. The information can be obtained from various sources and

50 3. Effect of optimisation of inlet zone on the hydraulic behaviour

by different tools. Multiple criteria as well as multiple design point optimisations can

be handled. Parallel computations can be easily performed. And, it smooths out the

high-frequency noise of the objective function and is thus expected to find a solution

near the global optimum”’[65]. ”‘Another advantage of the RSM is its capability of

escaping local minima”’ [66].

The objective of the study in this chapter is to optimise the nozzle distributor to

maximise the hydraulic performance of the IPS, which certainly will lead to improve

the separation efficiency of the IPS. To accomplish this goal, the RSM combined with

numerical simulation is used as tool for the distributor optimisation.

3.2. Methodology

3.2.1. Geometry Details of the IPS Model

A schematic representation of the IPS configuration is shown in Figure (3.1). It consists

of two zones. The first zone (A-1) comprises both the distributor nozzle and the

sediment collection chamber and it is a pipe with diameter 80 mm and 175 mm length.

The second zone (A-2) is used as separation chamber and it has internal dimensions

of (100 mm x 80 mm x 480 mm) and three plates 300 mm long with 5 mm thick.

The spacing between the plates is 16 mm. The distributor has totally twelve outlets

at three levels. Four outlets at every level have the same height and every level has

different diameter.

3.2.2. Optimisation Methodology

Outline Description of RSM

The RSM is created based on generating an explicit approximation to an objective

function, and thereafter this approximation function is used to carry out the optimi-

3.2. Methodology 51

Figure 3.1.: Schematic of the IPS model

sation. Moreover, the RSM is used widely to solve optimisation problems because it

has the ability to reduce the number of experiments, and the optimum solution can

be obtained quickly instead of perform additional expensive analysis [67, 66].

The RSM is carried out as follows:(1) the design of experiment (DOE) which is

necessary for selecting of design points;(2) development a mathematical model with

the best fit; (3) a mathematical form involving design variables is optimised to obtain

the optimum response value; (4) finding out the direct and interactive effects of the

input parameters on the objective function by constructing two and three-dimensional

plots [68].

Among several types of DOE techniques, a central composite design (CCD) approach

is selected in this study, which is a very efficient experimental design tool [64]. CCD

is originally proposed by Box and Wilson [69] and developed later by Box and Hunter

[70]. It requires number of experiments less than the full factorial design, and gives

more information about the design space because it is a five level fractional [71]. The


number of experiments depends on three parts: fractional factorial two-level, axial

points at distance α from its centre and a centre point and it is calculated according

to equation (3.1) [72].

N = 2N−f + 2N + one centre point (3.1)

where N and f denote the number of factors and fractional number respectively.

The usual approach to modelling the relationship between the variables and response

is the second order parametric model. But the disadvantage of these models is that

”‘they require strong assumptions on the functional form of possibly non-linear effects

of metrical covariates”’ [73]. This problem can be relieved by using non-parametric

regression to generate the response surface. Myers [67] suggested the use of non-

parametric regression in the following three cases: (i) the main purpose of the experi-

ment is the optimisation and not the parameter interpretation; (ii) the interpreting of

estimated regression coefficients is less important, while the shape of a response surface

is more important; or (iii) ”‘the mathematical model form of the relationship between

the variables and the response is highly non-linear and not well behaved”’. The non-

parametric regression is used here because the purpose of this study is optimisation of

distributor. Non-parametric models do not assume explicit function relationships be-

tween the input variables and the response, and ”‘they estimate the regression function

directly rather than to estimate the parameters in the function”’ [74].

There are several approaches in the non-parametric regression to fit the experimen-

tal data in the literatures such as local polynomial regression, kernel regression, and

support vector regression. The support vector method (SVM) has widely been used

for modelling non-linear systems due to ”‘its assurance of global solution, which is

achieved by transforming the regression problem into a convex optimisation problem

in dual space and its higher generalisation potential”’ [75]. For details on the basics

3.2. Methodology 53

of SVM, the reader is referred to Smola and Schoelkopf [76].

Objective Function and Design Variables

Owing to the hypothesis that a uniform flow distribution among the distributor outlet

openings will enhance the hydraulic behaviour of the IPS, minimising the standard

deviation (SD) of mass flow rate through these outlets was used as objective function

for the optimisation problem subject to SD≥0. The SD is given by the following

equation:

SD = s(q1, q2, q3, ...., qm) =

√√√√ 1

m− 1

m∑i=1

(qi − q)2 (3.2)

where qi,i=1,2,3,....,m denotes the mass flow rates at the outlet openings, q is the

average mass outflow rate and m= number of outlet openings=12. The Non-linear

Programming by Quadratic Lagrangian algorithm (NLPQL) is applied to optimise

the objective function. NLPQL is a widely used optimisation algorithm and can be

found in software packages like the IMSL Library, TOMLAB/Mathlab, OptiSLang,

ANSYS/Designexplorer and many others. The code NLPQL of Schittkowski is able to

minimise an objective function under non-linear equality and inequality constraints.

NLPQL is an implementation of a sequential quadratic programming (SQP) algorithm

[77]. SQP methods belong to the most powerful non-linear programming algorithms,

”‘which are based on the use of the gradient of the objective function and constraints,

for solving non-linear optimisation problems”’ [78]. NLPQL solves a quadratic pro-

gramming subproblem in each iteration. The quadratically approximation of the La-

grangian function and the linearising of the constraints is the basic idea of NLPQL.

Figure (3.2) shows nine design variables which were used in the distributor optimi-

sation: diameters of every level (D1, D2, D3), depth of every level (H1, H2, H3) and

height of outlet openings at every level (h1, h2, h3).


Figure 3.2.: Geometry of distributor

3.2.3. Numerical Analysis

In this chapter, the commercial software ANSYS CFX 12.0 was employed to optimise

the distributor and analyse the influence of the optimised distributor on the hydraulic

behaviour of the IPS. This software uses the Design Exploration approach for solving

the optimisation problems. The κ − ω model was chosen to predict the flow profile.

This model was selected in this work because the comparison between predicted data

by using the κ-ω model and experimental data in chapter two showed good agreement.

The conservation of mass and momentum were determined using the Reynolds aver-

aged Navier-Stokes as in chapter two. The residence time distribution (RTD) curve

for flow within the IPS was employed to characterise the flow pattern in the IPS, and

the tanks in series (TIS) model was used to characterise of a non-ideal flow pattern as

in chapter two.



3.3.1. Optimisation results

Numerical simulations for one hundred and forty seven design points were carried out

to optimise the shape of distributor at constant flow rate 350 l/hr. Thereafter, the

response surface was constructed from the predicted data numerically. As a result of

the optimisation, the objective function, i.e the SD, is successfully decreased for 71

% from 1.15e−4 to 3.29e−5 kgs−1. Sensitivity analysis is used to find out which input

parameter has influence on the objective function. Figure (3.3) illustrate the results

of sensitivity analysis of each input parameters for both the non-optimised and the

optimised distributor. Here, dC stands for the percent change of input parameters

dimension, and it is varied within±8% of the optimal value. The results reveal that

the objective function is more sensitive to the diameters (D1,D2) and the height (H1),

while the other input parameters have limited sensitivity on the objective function.

Moreover, it can be observed, for instance, that the non-optimised parameter (D3) has

a strong impact on the objective function, which is vanished via the optimisation as

shown in Figures (3.3a) and (3.3d) respectively.

3.3.2. Hydraulic performance of the IPS

Numerical simulations were carried out to investigate the effect of the optimised dis-

tributor on the hydraulic performance of the IPS at three flow rates with values of

250, 300 and 350 l/hr. The mesh structures for the IPS with non-optimised distributor

and optimised distributor have 274465 and 277199 elements respectively. Hydraulic

performance will be characterised by the quality of flow distribution and flow patterns

of the entire IPS.


��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

Figure 3.3.: sensitivity analysis for parameters D, H and h showing the standard devia-tion (SD) as function of percent change of input parameters (dC): [(a),(b)and (c)] and [(d),(e) and (f)] for non-optimised and optimised distributorrespectively.


��

�

�

�

�

�

��

��

��

��

��

Figure 3.4.: Quality of flow distribution between lamella plates

Quality of flow distribution

To asses the quality of flow distribution within each settler, the flow velocity through

every settler was determined, which was derived from the calculation of velocity in the

transverse direction in the middle of the settler. Then, the SD of a set of the values of

flow velocities through every settler was utilised as a criterion for the flow distribution

efficiency. Equation (3.3) was used to calculate the SD here.

SD =

√√√√ 1

m− 1

m∑i=1

(vi − v)2 (3.3)

where vi denotes the mean velocity within each settler , v is the average of these

velocities and m= number of settlers=4.

As shown in Fig. (3.4), the SD increases as the flow rate increases, indicating a

decrease in the flow distribution efficiency. Moreover it can be observed that the IPS


with optimised distributor improves the quality of flow distribution. The velocity

distributions on the plane at the middle of every settler are illustrated in Fig. (3.5). It

can be generally observed that in all cases, the velocities in the settlers are redistributed

due to optimisation of the distributor. Moreover, it is clear that the optimisation

causes a significant decrease in the velocity in the upper settler and increase in the

lower settler. On the other hand, this Figure shows existence of asymmetrical flow

pattern in settlers, which could be affected by the geometry of the distributor. This

pattern is created when turbulent flow passes through sudden expansion or contraction

based on the expansion or contraction ratio [79]. All these cases are realised in the

used distributor.

IPS flow pattern

RTD curves were plotted between E(θ) versus (θ) to investigate the influence of the

optimised distributor on the flow pattern of the IPS. Where θ is dimensionless time

[θ= ti/Tmean; Tmean= theoretical mean retention time]. As can be seen from Figure

(3.6), the optimised distributor increases slightly the initial peak, which represents the

time taken to reach maximum tracer concentration, (θpeak) approaching towards unity.

This means the hydraulic behaviour approaches plug flow conditions [8].

As mentioned earlier, the NTIS is used also to specify the flow pattern of the IPS.

Fig. (3.7) shows that the optimisation provides significantly better hydraulic perfor-

mance. Table (3.1) illustrates that the optimisation of distributor improves the NTIS

ranging between 55.7% to 68.3%. The main reason for this improvement as also shown

in Figure (3.6) could be due to the tail end of the normalised RTD curves by opti-

mised distributor is shorter which gives an indication of the recirculation regions are

decreased [47].


Figure 3.5.: Velocity distribution in the transverse direction in the middle of the set-tlers: (a),(b) and (c) for non-optimised distributor;(d),(e) and (f) for op-timised distributor


��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

�

��

��

��

��

��

�

��

��

��

��

Figure 3.6.: Influence of the optimisation on the RTD curves: (a)=250 l/hr; (b)=300l/hr;(c)=350 l/hr.


Table 3.1.: Influence of optimisation on the NTIS

Flow rate [l/hr] NTISNon-optimised shape Optimised shape Increment (%)

250 7.0 10.9 55.7300 6.0 10.1 68.3350 5.9 9.8 66.1

��

�

�

��

��

��

��

��

��

�

Figure 3.7.: Influence of the optimisation on the calculated NTIS


3.3.3. Separation efficiency

It is important to study the effect of distributor optimisation on the separation effi-

ciency of the IPS model. Either a Lagrangian or an Eulerian is commonly used to

model two-phase flow processes. Eulerian approach is used for all diffusion dominated

problems such as ultrafine particles. On the other hand, the Lagrangian approach

is applied for many two-phase flow applications due to their many capabilities. The

latter approach should not be applied when the particle volume fraction exceeds 10%.

In this approach, the fluid is treated as a continuum while the particles are tracked

individually in Lagrangian manner. The Lagrangian method was adopted in this

study. Because the modelling of seperation in the IPS has some limitations, where it

is difficult to simulate some physical process such as particle resuspension. The main

proposed of this simulation was to compare the effect of distributor optimisation on

the separation efficiency rather than determine its actual efficiency.

Numerical simulation with particle tracking was carried out by adding 200 particles

to the inflow and the slip velocity between the fluid and the particles was assumed equal

to zero. The used particles were the crushed walnut shell particles, which have diameter

50 μm and their density was 1350 kg/m3. The removal efficiency was determined at

three flow rates namely, 250 l/hr, 300 l/hr and 350 l/hr. Based on these flow rates

and solids concentration of 500 mg/l, the inlet flow rate of particles was estimated as

35 mg/s, 42 mg/s and 49 mg/s respectively. Because of the particle mass loading is

small, then the particles have not effect on the flow field (one-way coupling) [63]. The

inclination angle of the IPS was 45◦. The separation efficiency was calculated as shown

in equation(3.4). Figure (3.8) illustrates the effect of the distributor optimisation on

separation efficiency. It can be observed that the separation efficiency is improved


��

��

��

��

��

��

��

��

��

��

��

Figure 3.8.: Influence of the optimisation on the separation efficiency of the IPS model

significantly by the optimisation of the distributor ranging between 8.8% to 10.4%.

η =averaged volume fraction of particles @ outlet

averaged volume fraction of particles @ inlet(3.4)

65

4. New concept for the disposal of

sediment in countercurrent IPS

4.1. Introduction

As mentioned earlier,in Chapter two, separation efficiency of IPS is usually well below

the theoretical performance due to many factors. One of these factors is resuspension

of sediment due to flow instabilities and shear stress between the phases as well as

between the fluid and the lamella surfaces. ”‘The phenomenon of resuspension is

the process whereby, in the presence of the interference between the feed stream and

sediment path, an initially settled layer of negatively buoyant particles is entrained

into the bulk fluid and is convected away”’ [80]. Further, in case of countercurrent IPS

the resuspension occurs especially at the entrance zone of the settlers. In an attempt

to clarify this issue, when the sediment drops from one plate and encounter with

suspension which moves in the up flow direction, the sediment will be resuspended

easily. Moreover, the resuspension strongly depends on the local values of turbulence

dissipation rate at this region, where the turbulence has an adverse factor on the

separation process and causing sediment resuspension [32, 11].

Because the resuspension problem during the separation process is one of the reasons

for reduced IPS efficiency, new approach is suggested by construction one or two

66 4. New concept for the disposal of sediment in countercurrent IPS

sediment gutter on the plates to collect the sediment and dispose it via lateral outlets,

To the best of my knowledge there is no other study dealing with this approach.

This chapter aims to identify the best possible solution for the sediment path. This

was done by investigating a number of different plate designs experimentally and com-

paring their performance with the efficiency of the countercurrent in chapter three

with respect to a decrease of the resuspension problem. Further, as there is an in-

terdependence of resuspension and entrance length, the effect of distance between the

distributor tip and the stacked inclined plate on the separation efficiency was investi-

gated and the best distance was determined.

4.2. Material and methods

4.2.1. Geometry details of the IPS test systems

Two types of lab-scale IPSs were used in this study and both were made of plexiglass.

Figure (4.1) illustrates sketches of both models. The first model (IPS1) was used to

study the effect of distance between the nozzle tip and the staked inclined plate on

the separation efficiency. The details for this model were given in chapter three.

The another model (IPS2) was used to verify the effectiveness of the newly sug-

gestion to dispose the sediment via a lateral outlets.It consists of four compartments.

The first compartment (B-1) was used for housing the distributor and it was a pipe

with diameter 90 mm and 105 mm length. The second compartment (B-2) was used

as separation chamber and it has internal dimensions of (90 mm x 90 mm x 450 mm).

The third compartment (B-3) was used for receiving the sediment which collects via

the lateral outlets and it has internal dimensions of (50 mm x 90 mm x 350 mm).

The fourth compartment (B-4) was used to receive the sediment which comes from

the second compartment and the sediment which settled in the entrance zone. This

4.2. Material and methods 67

compartment has internal dimensions of (155 mm x 90 mm x 60 mm).

Three polyvinyl-chloride plates, 300 mm long and 5 mm thick, were used in both

models, and it could be fixed at any distance from the nozzle tip. The spacing between

the plates was 16 mm in the IPS1 and 19 mm in the IPS2. Both systems were placed

on a ramp with an angle of inclination between 00 − 550 from horizontal.

4.2.2. Experimental set-up and procedures

The process flow diagram of the experimental set-up is shown in Figure (4.2). The

setup consists of two tanks; ground tank (T1) and overhead tank (T2), both of which

have volume 25 litre. Crushed walnut shell particles with averaged diameter 160 μm

were used in experiments.Two stirrers were used to maintain turbulence in the two

tanks in order to keep the particles in suspension. Pump (A) was used to convey the

suspension from tank T1 to tank T2 and the excess water was drained back to tank

T1 via an overflow line.

The procedure for the start-up of the experiment is as follows. Initially the sus-

pension was mixed well in both T1 and T2 tanks to achieve a uniform distribution.

Thereafter, valve (B) was opened to allow the suspension to fill the IPS model; then

the valve (B) was adjusted to give a flow rate of 350 l/hr, and pump (A) was turned

off. Two samples were collected from the inlet stream (1) after 2 minutes from the

onset of opening the valve (B) . On the other hand, the outlet stream sampling was

set based on the hydraulic residence time for both IPS1 and IPS2 with 50s and 65s

respectively. At the end of every experimental run, the IPS was drained and the sus-

pension returned to the tank T1 via pump (D). The duration of each experiment was

not less than 45 minutes. The concentration of particles in suspension was 1050ppm ±

10% and 1080ppm ± 7% for IPS1 and IPS2 respectively. The method for determining

the separation efficiency was given in chapter two.


Figure 4.1.: Isometric of IPS1 with variable entrance zone length (a) and IPS2 withlateral sludge collector (b)

4.2. Material and methods 69

Figure 4.2.: Experimental set-up with pump(A,D),valve (B),ground tank (T1), over-head tank (T2), rotameter (c).(1) and (2) denote the sample collectionpoints in the inlet and outlet streams, respectively, while (3) denotes thestream of withdrawal concentrated sediment


� ��

��

��

��

��

��

��

��

�

�

��

��

��

��

��

��

Figure 4.3.: Experimental separation efficiency as function of entrance zone length andinclination angle of the IPS1


4.3.1. Impact of the entrance zone length on the separation

efficiency

The purpose of the present section is to study the dependence of the separation effi-

ciency on the distance between the distributor tip and the lower edge of the lamella

plates. The separation efficiency of the IPS1 model was investigated with four incli-

nation angles 300, 450, 500 and 550, and the tested entrance zone lengths were zero,

25mm, 50mm and 100mm.

Figure (4.3) illustrates the impact of the entrance zone length on the separation

efficiency. It can be observed that, the efficiency decreases as the inclination angle

increases as expected due to the settling distance increases as the inclination angle

increases [10]. Also, it can be seen that the efficiency increases as the length of entrance


Figure 4.4.: Increment rate of separation efficiency as function of entrance zone lengthand inclination angle of the IPS1

zone increases. Moreover, It is noticeable that the slope of curves at the distance

between zero and 25mm is larger than after this distance, but the behaviour of curve

slope for 550 is different. This implies that the increment rate of separation efficiency

decreases after the distance of 25 mm. This increment rate is calculated according to

equation (4.1).

Increment rate of separation efficiency =(η@ln − η@ln−1)x100

η@ln)(4.1)

Where η and l denote the separation efficiency and entrance length respectively, while

n = 1(0 mm), 2(25 mm), 3(50 mm) and 4(100mm).

Figure (4.4) shows the dependency of increment rate of efficiency upon the length

of the entrance zone. It can be seen that the increment rate increases sharply from

zero to a maximum value between the distance l = 0 mm and l = 25 mm and then

decreases as sharply from l = 25 mm to l = 50 mm. This Figure shows also that the


increment rate is directly proportional to the inclination angle of the IPS1 for 300, 450

and 500, where the increment rate increases as the inclination angle increases. One

interesting aspect of this figure is that the increment rate of efficiency increases slightly

from 50mm to 100mm for the previous angles which indicates the best entrance zone

distance is 50mm. On the other hand, this increment rate starts to decrease at 100mm

for 550.

In attempt to understand the reason behind these results, numerical simulations

were performed to investigate the effect of the entrance zone length on the average

of turbulence eddy dissipation (TED) at the lower edge of settlers. The TED is

selected based the discussion found in Krebs et al.[81], who mentioned that the energy

dissipation should be minimum at the inlet of clarifier, and this objective can be

accomplished by reducing the eddy scale.

The κ−ω model was chosen to predict the flow profile. The mesh structure for the

IPS1 was 620540 elements. Figure (4.5) illustrates the impact of the entrance zone

length on the TED. It can be observed that the TED value decreases dramatically be-

tween zero and 50 mm and thereafter it decreases slightly between 50mm and 100mm.

This graph explains the results discussed in Figure (4.4) for 300, 450 and 500. On

the other hand, the behaviour of the increment rate of efficiency with 550 is different

because the sediment velocity on plates increases as the inclination angle increases,

consequently the sediment flux rate at the lower edge of settlers could increase too.

To avoid this problem, a new approach is suggested by construction of a sediment

gutter on the plates to collect the sediment and disposes it via a lateral outlets.

4.3.2. Assessment of the IPS with lateral sludge collector

A new plate structure is proposed to eliminate the disadvantage described in the

previous section and to reduce the resuspension phenomena at the entrance of the


Figure 4.5.: Turbulence eddy dissipation (TED) as function of entrance zone length

settlers. Figure (4.6) illustrates the details of this plate. A bar with three different

heights was fixed on the plate. The tested heights (h) were 6, 8 and 10 mm representing

32%, 43% and 53%, respectively, of the total settler height. This proposed plate was

tested with one and two bars to find out the best structure. Also, the bar was placed

on the plate with two angles of inclination 450 and 550. The angle of inclination of

the IPS2 model was 450.

To examine the effectiveness of the proposed plate, thirteen experiments were carried

out to investigate the impact of this plate on the separation efficiency of the IPS2

model. Table (4.1) describes the plan of these experiments.

Figure (4.7) illustrates the influence of the proposed plate on the separation effi-

ciency at different bar heights. It can be observed generally that the efficiency de-

creases slightly as the height of bar increases. Furthermore, there is an enhancement

of efficiency by implementation of the proposed plate when the height of bar repre-

sents 32% and 43% of the total settler height for both the LS1-45-6 and LS2-45-6. On

the other hand, when the height of bar represents 53% of the total settler height, the


Figure 4.6.: Details of plates used in the experiments


Table 4.1.: Description of the experimental plan

Code No. of bars Inclination angle of bar Bar height [mm]LS1-45-6 1 45 6LS1-45-8 1 45 8LS1-45-10 1 45 10LS2-45-6 2 45 6LS2-45-8 2 45 8LS2-45-10 2 45 10LS1-55-6 1 55 6LS1-55-8 1 55 8LS1-55-10 1 55 10LS2-55-6 2 55 6LS2-55-8 2 55 8LS2-55-10 2 55 10LS0 — — —

efficiency of the IPS2 model becomes lower than the efficiency of LS0 in all cases.

It is also noticeable in this Figure that the efficiency decreases as the inclination

angle of the bar increases. To address this problem, two numerical simulations, as

example, were performed to investigate the behaviour of streamlines in one settler at

two cases. The configuration of this settler was similar as used in both LS1-45-6 and

LS1-55-6, and the feed flow rate represents one-fourth the total flow rate (350l/hr)

which is used in the experimental work. Figure (4.8) shows that the presence of

the bar generates waves in the settler, and the swing of these waves increases as the

inclination angle of bar increases. This increasing in the swing of these waves could

be the reason for this problem based on the conclusion which is made by Acrivos and

Herbolzheimer [16]. As is mentioned earlier in Chapter one, they observed that the

interface between the clarified layer that formed underneath the downward facing plate

and the suspension layer generate waves, which cause instability between these layers,

and then the clarified layer is resuspended again. Consequently, the efficiency of the

settler will decrease.

Owing to the impact of hydraulic settler characteristic on its separation efficiency,


Figure 4.7.: Impact of the proposed plate on the separation efficiency

it is important to study the influence of the proposed plate on the hydraulic behaviour

of the IPS2 model. Numerical simulations similar to simulations in Chapter two and

three were conducted to predict RTD curve and NTIS, both of which use to describe

the flow pattern of reactors. Figures (4.9) and (4.10) illustrate the impact of the

different plate structures on the normalised RTD curve of the IPS2 model. The two

figures show that the difference of normalised RTD curves between the conventional

plate and the different proposed plates is not distinctive.

The NTIS and the mean residence time (MRT) were estimated from the RTD curves

to capture accurately the effect of the different plate structures on hydraulic perfor-

mance of IPS2 model. Figure (4.11) shows that the NTIS values for the proposed plate

are higher than the conventional plate, which indicates that the hydraulic behaviour

is better. This improvement could be because of the existence of bars increase the

convection of the suspension within settlers. Moreover, it is noticeable generally in

Figure (4.13) the lowest NTIS gives the highest separation efficiency for up to h/H


Figure 4.8.: Velocity streamlines in two settler configurations with a bar of 6 mm heightand of two angles of inclination, 450 and 550.


��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

Figure 4.9.: Effect of six plate structures with a sediment gutter inclined by 450 on thenormalised RTD curve predicted by numerical simulation.


��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

Figure 4.10.: Effect of six plate structures with a sediment gutter inclined by 550 onthe normalised RTD curve predicted by numerical simulation.


Figure 4.11.: Effect of the different plate structures on the NTIS

= 0.43. Also it can be seen generally the efficiency decreases as the NTIS increases.

These results are contrary to the previous results which are shown in Chapters two

and three, where the separation efficiency was high, as the NTIS was higher. This

contradiction could be attributed to existence of waves which are generated by using

the bars, and they have negative effect on the settling process. While, in these two

chapters, the efficiency of separation has improved due to both the hydraulic perfor-

mance and the flow distribution in each settler has been improved, and both of which

have positive impact on the separation efficiency. These results also indicate that the

higher separation efficiency does not depend on the hydraulic performance alone, but

on many other factors as described by Kuoppamaeki and Okoth et al. [56, 32].

Figure (4.12) shows that the MRT decreases as the hight of bar increases as expected

and the longest MRT was for LS1-45-6 which has given the best efficiency.


Figure 4.12.: Effect of the different plate structures on the mean residence time of IPS2model


��

�

��

�

��

�

��

��

��

��

��

�

��

�

��

��

��

��

��

�

��

�

��

�

��

��

��

��

��

��

�

��

�

��

��

��

��

�

��

�

��

�

��

��

��

��

��

��

�

��

�

��

��

��

Figure 4.13.: Comparison of effect of the different plate structures on the NTIS andthe separation efficiency

83

5. Concluding remarks and future

work

The countercurrent inclined plate settlers (IPS) system is faced with two general prob-

lems, which are the flow distribution between the settlers and the resuspension of

sediment occur at the inlet zone of settlers, and both of which are influenced by the

inlet zone configuration. In this study, these problems were addressed to improve the

separation efficiency of IPS.

A nozzle distributor was used in feeding a lab-scale IPS to avoid the problems which

could be caused by feeding these settlers by a lateral feed box as explained in Chapter

one. A Computational Fluid Dynamic (CFD) model was used for simulating the

hydraulic behaviour of this IPS, and validate the CFD simulation with experimental

data. Two turbulent models (κ− ε model and κ−ω model) were implemented in this

simulation to predict both velocity flow field and RTD curve.

The results have shown that feeding the lab-scale IPS via this nozzle has a great

impact on the quality of flow distribution between the settlers, on flow patterns of the

entire IPS system and on separation efficiency could be clearly demonstrated via CFD

simulation and experiments at different flow rates. Also, the comparison between CFD

simulation and experimental data showed better agreement for the κ− ω model than

the κ− ε model.

Furthermore, the results show that the inlet structure of LS3 is helpful in improving

84 5. Concluding remarks and future work

both the distribution of flow within every settler and hydraulic behaviour of the IPS,

both of which could improve the separation efficiency of the IPS.

To the best of my knowledge there is no other study dealing with the effect of inlet

zone configuration on the hydraulic characteristic of countercurrent IPS. This study

proved that the RTD curve is an effective tool for modelling the hydraulic behaviour

of the IPS, and the results proved that the prediction of the RTD curve depends to a

great degree on the modelling accuracy of the distributor.

Based on the previous results which have shown that the distributor has a great

impact on the improving of hydraulic performance and the separation efficiency, the

response surface methodology (RSM) combined with numerical simulation was used to

optimise this distributor, and consequently achieve satisfactory separation efficiency.

The results have shown that the effectiveness of the objective function that is con-

structed based on the hypothesis that the uniformity of flow distribution provided by

the distributor outlet openings will enhance the hydraulic behaviour of the IPS.

Moreover, this objective function could be optimised successfully with the NLPQL

algorithms, and this objective function has appeared very well suited to improve the

hydraulic performance, flow distribution between the settlers and the separation effi-

ciency of the IPS by more than 68%, 65% and 10% respectively, but an asymmetrical

flow pattern is observed in settlers affecting by the geometry of this distributor.

The separation efficiency of the IPS also depends significantly on the entrance zone

length which can be determined based on the value of the turbulence eddy dissipation

at lower edge of the settlers.

The use of an sediment gutter on the plates with an inclination angle to collect

the sediment and disposes it via a lateral outlets is an effective tool to improve the

separation efficiency.

The results have shown that the inclined sediment gutter with heights up to 43%

85

from the total settler height and inclination angle of up to 450 give separation efficiency

better than the traditional plates. Furthermore, increase of the number of sediment

gutter caused decrease in the separation efficiency possibly due to formation of waves

which cause instability within settlers. Also, this inclined sediment gutter does not

have a great impact on the hydraulic behaviour of the entire IPS system.

Numerical simulations have shown that an increase of the inclination angle of sed-

iment gutter increase the swing of waves, which could be the reason for a decrease

in the efficiency of IPS. Moreover, these simulations demonstrated the existence of

asymmetry of flow in settlers due to the disposal the sediment from one side only,

which possibly also causes instability for the suspension in the settlers. The results

also indicate that separation efficiency does not depend on improving the hydraulic

performance of IPS alone, but on many other factors such as geometry of settlers,

inclination angle of IPS and properties of suspension.

The prediction of the RTD curve depends on the accuracy of the modelling of

the distributor, and there was little qualitative discrepancy between the predicted

RTD curve by the κ-ω model and the experimental results. This discrepancy could

be because of existing eddies of a wide range of length scale in the boundary layer,

which are totally modelled in the κ-ω model. Further work needs to be undertaken to

improve the prediction of the RTD curve by using other models, such as a large-eddy

simulation model. This model has the capability of resolving the three-dimensional

time-dependent details of the large and medium (i.e., resolved) scales, whereas the

effects of the small unresolved eddies are modelled with a sub-grid turbulence model.

To validate the use of a nozzle distributor in a larger size IPS instead of a small

laboratory model would provide inside whether the models describe the system reliably.

Due to observation of an asymmetrical flow pattern in settlers affecting by the

geometry of distributor, more constraints should be added to avoid this problem.

86 5. Concluding remarks and future work

Modelling the flow of suspension above an inclined sediment gutter could allow to

optimise both its height and inclination angle. Also to avoid asymmetry of flow in

settlers due to disposal the sediment from one side, more experiments are requested

on the disposal of sediment from two sides.

87

Bibliography

[1] A. Tamayol, B. Firoozabadi, and M. A. Ashjari. Hydrodynamics of secondary

settling tanks and increasing their performance using baffles. Journal of Environ-

mental Engineering, 136(1):32–39, 2010.

[2] Henry R. Bungay. Shallow-depth sedimentation of biological materials. Annals

of the New York Academy of Sciences, 369(1):335–340, 1981.

[3] E. Djoufac Woumfo, P. Djomgoue, P. Azinwi Tamfuh, D. Bitom, F. Figueras,

and D. Njopwouo. Clays from the bafang region (west cameroon): Properties and

potential application as decolorizing agent of river water. Applied Clay Science,

50(3):322 – 329, 2010.

[4] A. Tamayol, B. Firoozabadi, and G. Ahmadi. Effects of inlet position and baffle

configuration on hydraulic performance of primary settling tanks. Journal of

Hydraulic Engineering, 134(7):1004–1009, 2008.

[5] A. Hazen. On sedimentation. Transaction ASCE, 53:45–71, 1904.

[6] Thomas R. Camp. A study of the rational design of settling tanks. Sewage Works

Journal, 8(5):742–758, 1936.

[7] M. A. Md Said M. Shahrokhi, F. Rostami and S.Syafalni. Numerical modeling

of the effect of the baffle location on the flow field, sediment concentration and

88 Bibliography

efficiency of the rectangular primary sedimentation tanks. World Applied Sciences

Journal, 15(9):1296–1309, 2011.

[8] G. Tchobanoglous, F. Burton, and H.D. Stensel. Wastewater Engineering Treat-

ment and Reuse. Metcalf&Eddy, Inc., 2003.

[9] David Hendricks. Water Treatment Unit Processes Physical and Chemical. Taylor

& Francis Group., 2006.

[10] K. M. Yao. Theoretical study of high-rate sedimentation. Water Environment

Federation, 42(2):218–228, 1970.

[11] Cheng He and Jiri Marsalek. Vortex plate for enhancing particle settling.

135(8):627–635, 2009.

[12] A. E. Boycott. Sedimentation of blood corpuscles. Nature, 104:532, 1920.

[13] E. Ponder. On sedimentation and rouleaux formation. Quarterly Journal of

Experimental Physiology, 15:235–252, 1925.

[14] N. Nakamura and K. Kuroda. La cause de l’acceleration de la vitesse de sedi-

mentation de suspensions dans les recipients inclines. Keijo Journal of Medicine,

8:256–296, 1937.

[15] W.D. Hill, R.R. Rothfus, and Kun Li. Boundary-enhanced sedimentation due

to settling convection. International Journal of Multiphase Flow, 3(6):561 – 583,

1977.

[16] Andreas Acrivos and Eric Herbolzheimer. Enhanced sedimentation in settling

tanks with inclined walls. Journal of Fluid Mechanics, 92(03):435–457, 1979.

Bibliography 89

[17] Woon Fong Leung and Ronald F. Probstein. Lamella and tube settlers. 1. model

and operation. Industrial & Engineering Chemistry Process Design and Develop-

ment, 22(1):58–67, 1983.

[18] H. Nasr-El-Din, J.H. Masliyah, K. Nandakumar, and D.H.-S. Law. Continuous

gravity separation of a bidisperse suspension in a vertical column. Chemical

Engineering Science, 43(12):3225 – 3234, 1988.

[19] J.H. Masliyah, H. Nasr-El-Din, and K. Nandakumar. Continuous separation of

bidisperse suspensions in inclined channels. International Journal of Multiphase

Flow, 15(5):815 – 829, 1989.

[20] H.A. Nasr el Din, J.H. Masliyah, and K. Nandakumar. Continuous gravity sepa-

ration of concentrated bidisperse suspensions in an inclined plate settler. Inter-

national Journal of Multiphase Flow, 16(5):909 – 919, 1990.

[21] Robert H. Davis, Eric Herbolzhiemer, and Andreas Acrivos. The sedimentation

of polydisperse suspensions in vessels having inclined walls. International Journal

of Multiphase Flow, 8(6):571 – 585, 1982.

[22] R.H. Davis and H. Gecol. Classification of concentrated suspensions using inclined

settlers. International Journal of Multiphase Flow, 22(3):563 – 574, 1996.

[23] Woon Fong Leung. Lamella and tube settlers. 2. flow stability. Industrial &

Engineering Chemistry Process Design and Development, 22(1):68–73, 1983.

[24] Eric Herbolzheimer. Stability of the flow during sedimentation in inclined chan-

nels. Physics of Fluids, 26(8):2043–2054, 1983.

[25] E. S. G. Shaqfeh and A. Acrivos. The effects of inertia on the buoyancy-

driven convection flow in settling vessels having inclined walls. Physics of Fluids,

29(12):3935–3948, 1986.

90 Bibliography

[26] E. S. G. Shaqfeh and A. Acrivos. The effects of inertia on the buoyancy-

driven convection flow in settling vessels having inclined walls. Physics of Fluids,

29(12):3935–3948, 1986.

[27] A. Borhan and A. Acrivos. The sedimentation of nondilute suspensions in inclined

settlers. Physics of Fluids, 31(12):3488–3501, 1988.

[28] A. Borhan. An experimental study of the effect of suspension concentration on the

stability and efficiency of inclined settlers. Physics of Fluids A: Fluid Dynamics,

1(1):108–123, 1989.

[29] K. Zhang, A. Acrivos, and G. S. Triantafyllou. On the nature of the instability in

buoyancy-driven flows in inclined settlers. Physics of Fluids A: Fluid Dynamics,

4(6):1156–1164, 1992.

[30] W.P. Kowalski and R. Mieso. Computer simulations of cross-current sedimen-

tation process. In CAD Systems in Microelectronics, 2003. CADSM 2003. Pro-

ceedings of the 7th International Conference. The Experience of Designing and

Application of, pages 469 – 472, feb. 2003.

[31] Sudipto Sarkar, Dibyendu Kamilya, and B.C. Mal. Effect of geometric and process

variables on the performance of inclined plate settlers in treating aquacultural

waste. Water Research, 41(5):993 – 1000, 2007.

[32] G. Okoth, S. Centikaya, J. Brueggemann, and J. Thoeming. On hydrodynamic op-

timisation of multi-channel counter-flow lamella settlers and separation efficiency

of cohesive particles. Chemical Engineering and Processing: Process Intensifica-

tion, 47(1):90 – 100, 2008.

[33] Cheng He, Jim Wood, Jiri Marsalek, and Quintin Rochfort. Using cfd modeling to

Bibliography 91

improve the inlet hydraulics and performance of a storm-water clarifier. Journal

of Environmental Engineering, 134(9):722730, 2008.

[34] State of the art countercurrent inclined plate settlers from nordic water

gmbh, downloaded from website: www.nordic-water.de/docs/content.php, on

03.01.2012.

[35] Patricia Rodriguez Lopez, Antonio Gutierrez Lavin, Manuel M. Mahamud Lopez,

and Julio L. Bueno de las Heras. Flow models for rectangular sedimentation tanks.

Chemical Engineering and Processing: Process Intensification, 47(9-10):1705 –

1716, 2008.

[36] Maria Gavrilescu and Radu Z. Tudose. Residence time distribution of the liquid

phase in a concentric-tube airlift reactor. Chemical Engineering and Processing,

38(3):225 – 238, 1999.

[37] J. Behin and M. Aghajari. Influence of water level on oil-water separation by res-

idence time distribution curves investigations. Separation and Purification Tech-

nology, 64(1):48 – 55, 2008.

[38] A.D. and Martin. Interpretation of residence time distribution data. Chemical

Engineering Science, 55(23):5907 – 5917, 2000.

[39] O. Levenspiel. Chemical Reaction Engineering. John. Wiley and Sons Inc, 1999.

[40] L. Bircumshaw, K. Changunda, G. Hansford, and R. Rawatla. Development of a

mathematical model for continuous tank bioleaching. The Journal of the South

African Institute of Mining and Metallurgy, 106(11):277–282, 2006.

[41] Lifeng Zhang, Qinmin Pan, and Garry L. Rempel. Residence time distribution in a

multistage agitated contactor with newtonian fluids:cfd prediction and experimen-

92 Bibliography

tal validation. Industrial and Engineering Chemistry Research, 46(11):3538–3546,

2007.

[42] Ashwin W. Patwardhan. Prediction of residence time distribution of stirred re-

actors. Industrial & Engineering Chemistry Research, 40(24):5686–5695, 2001.

[43] Byung S. Choi, Bin Wan, Susan Philyaw, Kumar Dhanasekharan, and Terry A.

Ring. Residence time distributions in a stirred tank: Comparison of cfd

predictions with experiment. Industrial and Engineering Chemistry Research,

43(20):6548–6556, 2004.

[44] Yann Le Moullec, Olivier Potier, Caroline Gentric, and Jean Pierre Leclerc. Flow

field and residence time distribution simulation of a cross-flow gas-liquid wastew-

ater treatment reactor using cfd. Chemical Engineering Science, 63(9):2436 –

2449, 2008.

[45] J. Aubin, L. Prat, C. Xuereb, and C. Gourdon. Effect of microchannel aspect

ratio on residence time distributions and the axial dispersion coefficient. Chemical

Engineering and Processing: Process Intensification, 48(1):554 – 559, 2009.

[46] J. Thyn, M. Novy, P. Houdek, B. Portela, and R. ZitnY. Verification of cfd model

by stimulus response method. In In: 15th International Congress of Chemical

and Process Engineering CHISA. Praha., 2002.

[47] A.I. Salem, G. Okoth, and J. Thoeming. An approach to improve the separation of

solid-liquid suspensions in inclined plate settlers: Cfd simulation and experimental

validation. Water Research, 45(11):3541–3549, 2011.

[48] United States Department of the Interior Bureau of Reclamation. Water mea-

surement manual., 1997.

Bibliography 93

[49] C. M. Hrenya, E. J. Bolio, D. Chakrabarti, and J. L. Sinclair. Comparison of low

reynolds number k-epsilon turbulence models in predicting fully developed pipe

flow. Chemical Engineering Science, 50(12):1923 – 1941, 1995.

[50] H. Grotjans and F. Menter. Wall functions for general application CFD codes.

In Papailou, editor, ECCOMAS 98, pages 1112–1117, 1998.

[51] D. C. Wilcox. Multiscale model for turbulent flows. The American Institute of

Aeronautics and Astronautics Journal, 26(11):1211–1320, 1988.

[52] Xu Cheng and Nam il Tak. Investigation on turbulent heat transfer to lead-

bismuth eutectic flows in circular tubes for nuclear applications. Nuclear Engi-

neering and Design, 236(4):385 – 393, 2006.

[53] A. Griffiths, J. M. Dickson, and C. H. Griffiths. Determination of the coefficient

of diffusion of potassium chloride by an analytical method. Proceedings of the

Physical Society of London, 28:935 – 937, 1915.

[54] D. V. Kruglyakova, A. V .Neledova, V. F. Tishkin, and A. Filatov Yu. Unstruc-

tured adaptive grids for problems of mathematical physics (survey). Mathematical

Modeling (in Russian), 10(3):93–116, 1998.

[55] D. J. Mavriplis. Unstructured grid techniques. Annual Review of Fluid Mechanics,

29(1):473–514, 1997.

[56] Risto Kuoppamaeki. The applicability of tracer techniques for studies on sewage

treatment process dynamics. The International Journal of Applied Radiation and

Isotopes, 28(10-11):833 – 837, 1977.

[57] Derek Wilkinson, Brian Waldie, M. I. Mohamad Nor, and Hsio Yen Lee. Baf-

fle plate configurations to enhance separation in horizontal primary separators.

Chemical Engineering Journal, 77(3):221 – 226, 2000.

94 Bibliography

[58] C. Fureby, N. Alin, S. Menon, L. Persson, N. Svanstedt, and N. Wikstroem. Large

eddy simulation of high-reynolds-number wall bounded flows. The American In-

stitute of Aeronautics and Astronautics Journal, 42(3):457–468, 2004.

[59] Siping Zhou, Bevis W. L. Mak, Echo Leong, Kan Hon-Shing, Ip Shu-kuen, and Liu

Tze-Kwan. Optimized centre-feed clarifier design for the tai po sewage treatment

works, hong kong. Water and Environment Journal, 24(2):140–146, 2010.

[60] L.M. Oshinowo A. Bakker, A.H. Haidari. Realize greater benefits from cfd. Chem-

ical Engineering Progress, 97(3):45–53, 2001.

[61] N. Spogis and J.R. Nunhez. Design of a high-efficiency hydrofoil through the

use of computational fluid dynamics and multiobjective optimization. AIChE

Journal, 55(7):1723–1735, 2009.

[62] Savvas Xanthos, Minwei Gong, Krish Ramalingam, John Fillos, Alan Deur, Keith

Beckmann, and John McCorquodale. Performance assessment of secondary set-

tling tanks using cfd modeling. Water Resources Management, 25:1169–1182,

2011.

[63] Athanasia M. Goula, Margaritis Kostoglou, Thodoris D. Karapantsios, and Anas-

tasios I. Zouboulis. A cfd methodology for the design of sedimentation tanks in

potable water treatment: Case study: The influence of a feed flow control baffle.

Chemical Engineering Journal, 140(1-3):110–121, 2008.

[64] D. C. Montgomery. Design and analysis of experiments. New York: John Wiley

& Sons, 2001.

[65] Kwang-Yong Kim and Seoung-Jin Seo. Shape optimization of forward-curved-

blade centrifugal fan with navier-stokes analysis. Journal of Fluids Engineering,

126(5):735–742, 2004.

Bibliography 95

[66] Oktay Baysal, Mehti Koklu, and Nurhak Erbas. Design optimization of micro

synthetic jet actuator for flow separation control. Journal of Fluids Engineering,

128(5):1053–1062, 2006.

[67] R. H. Myers. Response surface methodology-current status and future direction.

Journal of Quality Technology, 31:30–44, 1999.

[68] Thuy Khanh Trinh and Lim-Seok Kang. Application of response surface method

as an experimental design to optimize coagulation tests. Environmental Engi-

neering Research, 15:63–70, 2010.

[69] G. E. P. Box and K. B. Wilson. On the experimental attainment of optimum

conditions. Journal of the Royal Statistical Society. Series B (Methodological),

13(1):1–45, 1951.

[70] G. E. P. Box and J. S. Hunter. Multi-factor experimental designs for exploring

response surfaces. The Annals of Mathematical Statistics, 28(1):195–241, 1957.

[71] Jianlong Wang and Wei Wan. Experimental design methods for fermentative

hydrogen production: A review. International Journal of Hydrogen Energy,

34(1):235 – 244, 2009.

[72] ANSYS-Fluent -Inc.- CFX -12.0, Germany GmbH, 2008.

[73] Stefan Lang, Samson B. Adebayo, Ludwig Fahrmeir, and Winfried J. Steiner.

Bayesian geoadditive seemingly unrelated regression. Computational Statistics,

18(1):263–292, 2003.

[74] Dongling Zhang, Yingjie Tian, and Peng Zhang. Kernel-based nonparametric

regression method. In Web Intelligence and Intelligent Agent Technology, 2008.

WI-IAT ’08. IEEE/WIC/ACM International Conference on, volume 3, pages

410–413, dec. 2008.

96 Bibliography

[75] S. Iplikci. Support vector machines-based generalized predictive control. Inter-

national Journal of Robust and Nonlinear Control, 16(17):843–862, 2006.

[76] Alex J. Smola and Bernhard Schoelkopf. A tutorial on support vector regression.

Statistics and Computing, 14:199–222, 2004.

[77] K. Schittkowski. Nlpql: A fortran subroutine solving constrained nonlin-

ear programming problems. Annals of Operations Research, 5:485–500, 1986.

10.1007/BF02022087.

[78] Y. Yan, G. Liu, and J. Chen. Integrated modeling and optimization of a parallel

hydraulic hybrid bus. International Journal of Automotive Technology, 11:97–104,

2010.

[79] M. P. Escudier, P. J. Oliveira, and R. J. Poole. Turbulent flow through a plane

sudden expansion of modest aspect ratio. Phyics of Fluids, 14(10):3641–3654,

2002.

[80] David Leighton and Andreas Acrivos. Viscous resuspension. Chemical Engineer-

ing Science, 41(6):1377–1384, 1986.

[81] Peter Krebs, Daniel Vischer, and Willi Gujer. Inlet-structure design for final

clarifiers. Journal of Environmental Engineering, 121(8):558–564, 1995.

ERKLARUNG

Hiermit versichere ich, Ahmed Ibrahim Salem Youssef geboren am 06. 02. 1971, dass

ich die vorliegende Dissertation mit dem Titel ”On fluid dynamics of lamella separator

modelling and process optimisation”

1. ohne unerlaubte Hilfe angefertigt habe.

2. keine anderen als die von mir angegebenen Quellen und Hilfsmittel benutzt habe

und

3. die den benutzten Werken wortlich oder inhaltlich entnommenen Stellen als

solche kenntlich gemacht habe.

Bremen, den 10. Januar 2012

Date post:	12-Jan-2017
Category:	Documents
Upload:	vuliem
View:	219 times
Download:	3 times

On fluid dynamics of lamella separator modelling and process ...

Documents