UNIVERSITY OF CALGARY

CFD Simulation of Multiphase Separators

by

Ali Pourahmadi Laleh

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF CHEMICAL AND PETROLEUM ENGINEERING

CALGARY, ALBERTA

SEPTEMBER, 2010

© Ali Pourahmadi Laleh 2010

 Library and Archives Bibliothèque et
Canada Archives Canada

Published Heritage Direction du

Branch Patrimoine de l’édition

395 Wellington Street 395, rue Wellington

Ottawa ON K1A 0N4 Ottawa ON K1A 0N4
Canada Canada

Your file Votre référence

ISBN: 978-0-494-69552-4
Our file Notre référence
ISBN: 978-0-494-69552-4

NOTICE: AVIS:

The author has granted a non- L’auteur a accordé une licence non exclusive
exclusive license allowing Library and permettant à la Bibliothèque et Archives
Archives Canada to reproduce, Canada de reproduire, publier, archiver,
publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public
communicate to the public by par télécommunication ou par l’Internet, prêter,
telecommunication or on the Internet, distribuer et vendre des thèses partout dans le
loan, distribute and sell theses monde, à des fins commerciales ou autres, sur
worldwide, for commercial or non- support microforme, papier, électronique et/ou
commercial purposes, in microform, autres formats.
paper, electronic and/or any other
formats.
.
The author retains copyright L’auteur conserve la propriété du droit d’auteur
ownership and moral rights in this et des droits moraux qui protège cette thèse. Ni
thesis. Neither the thesis nor la thèse ni des extraits substantiels de celle-ci
substantial extracts from it may be ne doivent être imprimés ou autrement
printed or otherwise reproduced reproduits sans son autorisation.
without the author’s permission.

In compliance with the Canadian Conformément à la loi canadienne sur la

Privacy Act some supporting forms protection de la vie privée, quelques
may have been removed from this formulaires secondaires ont été enlevés de
thesis. cette thèse.

While these forms may be included Bien que ces formulaires aient inclus dans
in the document page count, their la pagination, il n’y aura aucun contenu
removal does not represent any loss manquant.
of content from the thesis.
 UNIVERSITY OF CALGARY

FACULTY OF GRADUATE STUDIES

The undersigned certify that they have read, and recommend to the Faculty of Graduate

Studies for acceptance, a thesis entitled "CFD Simulation of Multiphase Separators"

submitted by Ali Pourahmadi Laleh in partial fulfilment of the requirements of the degree

of doctor of philosophy.

Supervisor, William Y. Svrcek, Chemical & Petroleum Engineering

Supervisory Committee, Wayne D. Monnery, Chemical & Petroleum Engineering

Supervisory Committee, Michael W. Foley, Chemical & Petroleum Engineering

Jalel Azaiez, Chemical & Petroleum Engineering

Om P. Malik, Electrical & Computer Engineering

Ronald J. Hugo, Mechanical & Manufacturing Engineering

External Examiner, Ted Frankiewicz, SPEC Services Inc.

Date
ii
 Abstract

The multiphase separators are generally the first and largest process equipment in an oil
production platform, furthermore this primary separation step is a key element in the oil
and gas production facilities in that downstream equipment, such as compressors, are
completely dependent on the efficient performance of these multiphase separators.
This research project applied Computational Fluid Dynamics (CFD) based simulation to
model multiphase separators. In order to capture both macroscopic and microscopic
aspects of multiphase separation phenomenon, an efficient combination of two
multiphase models of the established commercial CFD package, Fluent 6.3.26, was used.
The Volume of Fluid (VOF) model was used to simulate the phase behavior and fluid
flow patterns, and the Discrete Phase Model (DPM) was used to model the movement of
fluid droplets injected at the separator inlet. The “particle tracking” based simulation of
the multiphase separation process was the key aspect of this research project, and the
developed model did provide high-quality visualization of multiphase separation process.
The research project involved the CFD simulation of four pilot-plant-scale two-phase
separators and one industrial scale three-phase separator, including all the installed
internals. There was excellent agreement between simulated phase separation behavior
and the empirical observations and data gleaned from the pilot plant.
The research project also evaluated the classic separator design methodologies using
detailed CFD based simulations, and proposed improved design criteria. In order to
specify an effective optimum separator, a useful method was developed for estimation of
the droplet sizes used to calculate realistic separation velocities for various oilfield
conditions. The most important parameters affecting these efficient droplet sizes were the
vapor density and the oil viscosity. In contrast with classic design strategies, the CFD
simulation results showed that additional residence times are required for droplets to
penetrate through the interfaces. Moreover, the Abraham equation should be used instead
of Stokes’ law in the liquid-liquid separation calculations. The velocity constraints caused
by re-entrainment in horizontal separators were also studied via comprehensive CFD
simulations, and led to novel correlations for the re-entrainment phenomenon. Hence, this
research project does show the benefits that CFD analyses can provide in optimizing the
design of new separators and solving problems with existing designs.

iii
 Preface

The literature on the critical unit operation of multiphase separators abounds with macro
studies and design methodologies for two and three-phase vertical and horizontal
separators. There are very few studies/papers that provide the micro details of the actual
separation process. Hence, the purpose of this research project was to use Computational
Fluid Dynamics (CFD) as a method to simulate the detailed performance of multiphase
separators. The objectives, aside from developing the CFD model, were to use the
simulation results to validate and provide additional criteria for the separator design.
Chapter One provides the background and the objectives of the research project. Since
readers tend to skip over the introductory material and focus on the novel achievements
of a research study, only the most useful comments, which are necessary to set the stage
for the reminder of the thesis, have been presented in the first chapter.
Chapter Two presents a detailed literature review of the very few CFD based studies
performed on multiphase separators. Moreover, the classic guidelines for separator design
are also reviewed, providing a useful literature review within the research project scope.
Realizing the difficulties involved with CFD modeling of multiphase fluid flows and the
required high-quality simulation of multiphase separator performance, a proven
commercial CFD package, Fluent 6.3.26, was used. Chapter Three provides the details of
the developed procedure for the CFD models required in the Fluent software for
simulating the complex features of four pilot-plant-scale two-phase separators and one
industrial scale three-phase separator. In contrast with the published CFD based studies,
all details of the CFD modeling phase will be provided. Chapter Four presents the results
of using the developed CFD models of Chapter Three for the simulated separators.
In Chapter Five, the effective model assumptions and settings are used to establish
new/improved separator design criteria. These criteria will be combined with an industry
used algorithmic design method to specify a realistic optimum separator design. Finally,
Chapter Six presents the conclusions of this research project and recommendations for
future research projects.
Supplementary material of the thesis such as the fluid flow profiles, the separation
efficiency plots, and the developed computer codes (written in C++) has been included
on a CD attached to this thesis.

iv
 Acknowledgements

My sincere thanks go to Dr. William Y. Svrcek for his persuasive and powerful character
which made it possible to accomplish this research project. I am very proud of being his
last PhD student. I have learned a lot from his efficient and superb approaches to
directing my research. His precious moral/financial supports are gratefully appreciated.

I also express my gratitude to Dr. Wayne D. Monnery who contributed greatly to this
project. His excellent insights and recommendations were extremely conducive in
initiating and shaping this research.

I am in debt of Dr. Ramin B. Boozarjomehry, my supervisor in the MSc career, for his
everlasting encouragement and technical supports during my PhD career. His significant
role in my graduate career is most appreciated, and I wish all the best for him for good.

During the initial periods of my research, I took advantage of very helpful discussions
with Dr. Jalel Azaiez. I do appreciate him for providing me with good ideas and also
useful documents on the multiphase flow issues.

I would also thank Dr. Apostolos Kantzas as I have used his software license while
developing my CFD models. In addition, technical support of his research assistant,
Blake Chandrasekaran, is appreciated.

I am grateful to Dr. Thomas G. Harding for making the “Ursula & Herbert Zandmer
Graduate Recruitment” scholarship possible to me in consecutive years of 2006 to 2009.
This award played a major role in reaching my academic objectives.

As an expert on the computational resources, Dr. Doug S. Phillips helped me several

times to overcome the difficulties arisen while submitting/performing parallel simulation
cases on the WestGrid environment. His great responsibility and adorable diligence are
gratefully acknowledged.

v
I am indebted to my parents and sisters who set the stage for me to pursue my studies in
Canada. I extend my grateful appreciations to them for their encouragement, patience,
and support.

My wife, Roza Kazemi, joined the scenario in early 2008, and since then, she has been
taking all the proactive measures to my success in academic and personal life. Her
invaluable contribution is sincerely acknowledged.

Ehsan Aminfar and Ali Yazdani, with their lovely personality, have been more than
friends to me since I met them respectively in 2005. I would like to appreciate these
gentlemen for the moral assistance they have provided always.

Finally, I like to appreciate the administration staff of the Chemical & Petroleum
Engineering department for their technical support. Particularly, I am very thankful to
Andrea Cortes for her great personality and extraordinary responsibility.

vi
 Dedication

To My Mother, for Her Pure Love

&

To My Father, for His Profound Logic

vii
 Table of Contents
Approval Page..................................................................................................................... ii
Abstract .............................................................................................................................. iii
Preface................................................................................................................................ iv
Acknowledgements..............................................................................................................v
Dedication ......................................................................................................................... vii
List of Tables .................................................................................................................... xii
List of Figures and Illustrations ........................................................................................ xv
List of Abbreviations and Nomenclature......................................................................... xxi

CHAPTER ONE: INTRODUCTION..................................................................................1

1.1 Gas-Liquid Phase Separators .....................................................................................1
1.1.1 Multistage Separation ........................................................................................1
1.1.2 Multiphase Separator Terminology...................................................................2
1.1.3 Multiphase Separators .......................................................................................3
1.1.4 Operating Pressures and Capacities...................................................................5
1.1.5 Separator Internals.............................................................................................5
1.1.5.1 Inlet Diverters ..........................................................................................6
1.1.5.2 Controls....................................................................................................7
1.1.5.3 Mist Eliminators ......................................................................................7
1.1.6 Operational Difficulties ...................................................................................10
1.2 CFD as a Separator Modeling Method ....................................................................11
1.3 Motivation for the Research Project ........................................................................13

CHAPTER TWO: LITERATURE REVIEW....................................................................16

2.1 Classic Methods for Design of Multiphase Separators............................................16
2.1.1 Common Design Aspects ................................................................................17
2.1.1.1 Vessel Orientation..................................................................................17
2.1.1.2 The Aspect Ratio of a Separator ............................................................18
2.1.2 Two-Phase Separators .....................................................................................19
2.1.2.1 Droplet Settling Theory for Vapor-Liquid Separation ..........................19
2.1.2.2 Heuristics ...............................................................................................22
2.1.3 Three-Phase Separators ...................................................................................25
2.1.3.1 Droplet Settling Theory for Liquid-Liquid Separation..........................25
2.1.3.2 Heuristics ...............................................................................................26
2.2 CFD-Based Studies of Multiphase Separators.........................................................27
2.2.1 Contribution of NATCO Group ......................................................................27
2.2.2 Contribution of SINTEF Group.......................................................................32
2.2.3 Miscellaneous Studies .....................................................................................34
2.3 Two Relevant Theses...............................................................................................41
2.4 Summary..................................................................................................................44

CHAPTER THREE: DEVELOPED CFD MODELS .......................................................47

3.1 CFD Background .....................................................................................................47
3.1.1 Approaches to Solving Equations for Fluid Flow ...........................................47
3.1.1.1 SIMPLE Method....................................................................................49
3.1.1.2 PISO Method .........................................................................................50

viii
 3.1.2 Strategies for Simulation of Multiphase Fluid Flows......................................51
3.1.2.1 Discrete Phase Model (DPM) ................................................................51
3.1.2.2 VOF Model ............................................................................................52
3.1.2.3 Mixture Model .......................................................................................52
3.1.2.4 Eulerian Model ......................................................................................52
3.1.3 CFD Modeling of Flow-Distributing Baffles and Wire Mesh Demisters .......52
3.1.3.1 Momentum Equations for Porous Media...............................................53
3.1.3.2 Modeling Baffles via the Porous Media Model.....................................54
3.1.3.3 Modeling Wire Mesh Demisters via the Porous Media Model .............55
3.2 Simulation of Pilot-Plant-Scale Two-Phase Separators...........................................56
3.2.1 Experimental Equipment and Procedure .........................................................57
3.2.2 Material Definition ..........................................................................................58
3.2.3 Implementing DPM Approach ........................................................................59
3.2.3.1 Physical Models .....................................................................................59
3.2.3.2 Definition of Droplet Size Distribution .................................................63
3.2.3.3 Setting CFD Simulator Parameters........................................................64
3.2.4 Implementing VOF-DPM Approach ...............................................................65
3.2.4.1 Grid Systems..........................................................................................66
3.2.4.2 Definition of Droplet Size Distribution .................................................66
3.2.4.3 Setting CFD Simulator Parameters........................................................69
3.3 Simulation of a Large-Scale Three-Phase Separator ...............................................71
3.3.1 Material Definition ..........................................................................................73
3.3.2 Physical Model ................................................................................................75
3.3.3 Modeling the Baffles and the Mesh Pad Demister..........................................78
3.3.3.1 Modeling the Baffles .............................................................................78
3.3.3.2 Modeling the Wire Mesh Demister .......................................................80
3.3.3.3 Developed Model for Physical Verification ..........................................80
3.3.4 Definition of Droplet Size Distribution...........................................................81
3.3.5 Setting CFD Simulator Parameters .................................................................88
3.4 Summary..................................................................................................................89

CHAPTER FOUR: RESULTS AND DISCUSSION ........................................................91

4.1 Simulation of the Pilot-Plant-Scale Two-Phase Separators.....................................91
4.1.1 Results of DPM Only Model ...........................................................................91
4.1.1.1 Fluid Flow Profiles ................................................................................92
4.1.1.2 Droplet Coalescence and Breakup .........................................................92
4.1.1.3 Separation Efficiencies ..........................................................................99
4.1.2 Results of VOF-DPM Model.........................................................................101
4.1.2.1 Fluid Flow Profiles ..............................................................................101
4.1.2.2 Droplet Coalescence and Breakup .......................................................103
4.1.2.3 Separation Efficiencies ........................................................................103
4.1.2.4 Droplet Size Distribution in Gas Outlet...............................................112
4.2 Simulation of a Large-Scale Three-Phase Separator .............................................114
4.2.1 Physical Validation of the Developed Models for the Baffles and Demister114
4.2.2 Preliminary Considerations ...........................................................................116
4.2.3 Fluid Flow Profiles ........................................................................................118
4.2.4 Separation Efficiencies ..................................................................................127
ix
 4.2.5 The Effect of Minor Modifications on the Gullfaks-A Separator .................129
4.2.6 Redesign of the Gullfaks-A Separator...........................................................131
4.2.6.1 Classic Design Strategy .......................................................................131
4.2.6.2 Combining Improved Separation Velocities........................................132
4.2.6.3 On Substitution of Stokes’ Law by More Complicated Models..........141
4.2.7 Sensitivity Analyses ......................................................................................143
4.3 Summary................................................................................................................145

CHAPTER FIVE: IMPROVED DESIGN CRITERIA ...................................................147

5.1 Fluid Systems.........................................................................................................147
5.2 Phase Separation Components...............................................................................149
5.2.1 Vapor-Liquid Separation ...............................................................................150
5.2.2 Liquid-Liquid Separation ..............................................................................153
5.3 Vessel Orientation Considerations.........................................................................158
5.3.1 Horizontal Arrangement ................................................................................158
5.3.2 Vertical Arrangement ....................................................................................159
5.3.2.1 Vapor-Liquid Separation .....................................................................160
5.3.2.2 Liquid-Liquid Separation.....................................................................161
5.3.3 The Proposed Approach in a Vector Space ...................................................162
5.4 Designing the Realistic Optimum Separator for Gullfaks-A.................................162
5.4.1 Design Phase .................................................................................................163
5.4.2 Performance Validations via CFD Simulations ............................................165
5.5 Re-Entrainment Constraints...................................................................................174
5.5.1 Vapor-Liquid Re-entrainment .......................................................................174
5.5.2 Liquid-Liquid Re-entrainment.......................................................................176
5.6 Summary................................................................................................................180

CHAPTER SIX: CONCLUSIONS AND RECOMMENDATIONS ..............................183

6.1 CFD Simulation of Pilot-Plant-Scale Two-Phase Separators................................184
6.2 CFD Based Study of a Large-Scale Three-Phase Separator..................................186
6.3 Improved Design Criteria ......................................................................................188
6.4 Recommendations..................................................................................................190

REFERENCES ................................................................................................................191

APPENDIX A: DESIGN OF THREE-PHASE SEPARATORS……………….……....197

A.1 Vertical Configuration………………………………………….………………. 197 .

A.2 Horizontal Configuration……………………………………………….……….199

A.2.1 “Simple” Design…………………………………..…………………….... 200 .

A.2.2 “Boot” Design…………………………………………….……………….201

A.2.3 “Weir” Design……………………………………...……………………...203
A.2.4 “Bucket and Weir” Design…………………………………….…………..205

APPENDIX B: DROPLET BREAKUPS IN THE TWO-PHASE SEPARATORS…....208

B.1 Results for DPM Approach……………………………………………………...208
B.2 Results for VOF-DPM Approach……………………………...………………...208

x
APPENDIX C: SEPARATION PERFORMANCE OF WIRE MESH DEMISTERS…217

APPENDIX D: FLUID FLOW PROFILES FOR VARIOUS SEPARATOR DESIGNS 225

xi
 List of Tables

Table 2-1. The aspect ratios proposed by Walas (1990) for multiphase separators. ........ 19

Table 2-2. The K coefficients proposed by Walas (1990) for multiphase separators
equipped with stainless steel mesh pads. .................................................................. 21

Table 2-3. The K coefficients proposed by York Mist Eliminator Company and
regressed by Svrcek and Monnery (1993) for multiphase separators equipped
with stainless steel mesh pads................................................................................... 22

Table 3-1. The composition of the saturated gas at multiple pressures (Monnery and
Svrcek, 2000). ........................................................................................................... 58

Table 3-2. The values of material properties used in simulation case studies.................. 59

Table 3-3. Quality of the mesh produced in Gambit environment for two-phase
separators (DPM approach). ..................................................................................... 63

Table 3-4. Global quality of the original and modified mesh for two-phase separators
(DPM approach)........................................................................................................ 63

Table 3-5. The discrete phase parameters used for simulation of two-phase separators
through DPM approach............................................................................................. 65

Table 3-6. Quality of the mesh produced in Gambit environment for two-phase
separators (VOF-DPM approach). ............................................................................ 67

Table 3-7. Global quality of the original and modified mesh for two-phase separators
(VOF-DPM approach). ............................................................................................. 69

Table 3-8. The physical parameters of fluids in Gullfaks-A separator provided by

Hansen et al. (1993). ................................................................................................. 74

Table 3-9. Quality of the mesh produced for the Gullfaks-A separator in Gambit
environment. ............................................................................................................. 77

Table 3-10. The physical properties and correlation parameters given for wire mesh
pad of type E (Helsør and Svendsen, 2007).............................................................. 80

Table 3-11. The maximum stable droplet sizes calculated by various methods for
dispersions of Gullfaks-A separator.......................................................................... 87

Table 3-12. The discrete phase parameters used in CFD simulation of Gullfaks-A
separator. ................................................................................................................... 87

xii
Table 3-13. The calculated boundary condition inputs used in CFD simulation of
Gullfaks-A separator. ................................................................................................ 88

Table 4-1. The simulated and experimental values for incipient velocities (DPM
approach)................................................................................................................... 99

Table 4-2. The simulated and experimental values for incipient velocities (VOF-DPM
approach)................................................................................................................. 112

Table 4-3. The average values of Rosin-Rammler analysis on the droplets emerging
from the gas outlet (VOF-DPM approach). ............................................................ 113

Table 4-4. The droplet size distribution in important zones of the Gullfaks-A
separator. ................................................................................................................. 128

Table 4-5. The droplet size distribution in the important zones of the Gullfaks-A
separator without the internals (flow-distributing baffles and the mist
eliminator)............................................................................................................... 130

Table 4-6. The droplet size distribution in the important zones of the Gullfaks-A
separator with expanded flow-distributing baffles in the future conditions. .......... 130

Table 4-7. Quality of the mesh produced in Gambit environment for the redesigned
Gullfaks-A separator. .............................................................................................. 134

Table 4-8. The droplet size distribution in the important zones of the redesigned
Gullfaks-A separator. .............................................................................................. 140

Table 4-9. The results of sensitivity analyses on the developed CFD model for
Gullfaks-A separator. .............................................................................................. 144

Table 5-1. Physical properties for fluids of different oilfields at their corresponding
separator conditions. ............................................................................................... 148

Table 5-2. The discrete phase parameters used in CFD simulations of phase
separation. ............................................................................................................... 148

Table 5-3. The vapor-liquid separation characteristics for different oilfields. ............... 151

Table 5-4. The liquid-liquid separation characteristics for different oilfields. ............... 155

Table 5-5. The approximate separator weights calculated for various types of
separators processing Gullfaks-A. .......................................................................... 164

Table 5-6. Quality of the mesh produced in Gambit environment for various
separators processing Gullfaks-A. .......................................................................... 169

Table 5-7. Separation efficiencies for various separators processing Gullfaks-A.......... 170

xiii
Table 5-8. Quality of the mesh produced in Gambit environment for the “stabilized”
Gullfaks-A separator. .............................................................................................. 171

Table 5-9. The key geometric specifications and separation efficiencies for various
designs proposed for Gullfaks-A. ........................................................................... 173

Table 5-10. The maximum safe velocities of vapor phase determined for the various
oilfield case studies. ................................................................................................ 175

Table 5-11. The maximum safe velocities of water phase in “simple” and “weir”
designs predicted for various oilfield case studies.................................................. 178

Table A-1. The aspect ratio values suggested by Monnery and Svrcek (1994) for
multiphase separators.............................................................................................. 198

Table C-1. The discrete phase parameters used in CFD simulations of demister
performance. ........................................................................................................... 218

Table C-2. The droplet size distribution in inlet and outlet zones of the demister. ........ 220

Table C-3. The water droplet size distribution in inlet and outlet zones of the
demister................................................................................................................... 223

xiv
 List of Figures and Illustrations

Figure 1.1. Different Common Designs of Horizontal Three-Phase Separators; (a)

“Simple”, (b) “Boot”, (c) “Weir”, and (d) “Bucket and Weir”................................... 4

Figure 1.2. “Weir” Design Horizontal Three-Phase Separator with the Installed
Internals....................................................................................................................... 5

Figure 1.3. Vertical Three-Phase Separator with the Installed Internals. ........................... 6

Figure 1.4. Wire Mesh Pad Mist Eliminator....................................................................... 8

Figure 1.5. Vane-Type Mist Eliminator.............................................................................. 9

Figure 1.6. Schematic of a Typical Three-Phase Separator.............................................. 14

Figure 2-1. Design Heuristics Proposed by Watkins (1967) for Two-Phase Separators.. 24

Figure 2-2. Design Heuristics Proposed by Walas (1990) for Two-Phase Separators
Equipped with Wire Mesh Demisters. ...................................................................... 24

Figure 3-1. Process Flow Diagram for Separator Skid (Monnery and Svrcek, 2000)...... 57

Figure 3-2. The Geometrical Specifications of Models; (a) Vessel A, (b) Vessel B, (c)
Vessel C, (d) Vessel D. ............................................................................................. 60

Figure 3-3. The Grid System for Vessel B Produced in Gambit Environment (DPM
Approach). ................................................................................................................ 61

Figure 3-4. The Grid System for Vessel D Produced in Gambit Environment (DPM
Approach). ................................................................................................................ 62

Figure 3-5. The Grid System for Vessel C Produced in Gambit Environment (VOF-
DPM Approach)........................................................................................................ 67

Figure 3-6. The Grid System for Vessel D Produced in Gambit Environment (VOF-
DPM Approach)........................................................................................................ 68

Figure 3-7. Process Flow Diagram for Gullfaks-A (Hansen et al., 1993). ....................... 72

Figure 3-8. General Configuration of the First Stage Separator on the Gullfaks-A
Production Facility (Hansen et al., 1993). ................................................................ 72

Figure 3-9. Geometrical Specifications of the Gullfaks-A Separator............................... 75

Figure 3-10. The Grid System for the Gullfaks-A Separator Generated in Gambit
Environment.............................................................................................................. 76

xv
Figure 3-11. Geometry of the Square Hole Pattern in a Perforated Plate......................... 79

Figure 4-1. Fluid Flow Profiles in the Middle of Vessel C Operating at 70 kPa (DPM
Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s). ........ 93

Figure 4-2. Fluid Flow Profiles in the Middle of Vessel C Operating at 700 kPa
(DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity
(m/s). ......................................................................................................................... 94

Figure 4-3. Fluid Flow Profiles in the Middle of Vessel C Operating at 2760 kPa
(DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity
(m/s). ......................................................................................................................... 95

Figure 4-4. Fluid Flow Profiles in the Middle of Vessel D Operating at 70 kPa (DPM
Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s). ........ 96

Figure 4-5. Fluid Flow Profiles in the Middle of Vessel D Operating at 700 kPa
(DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity
(m/s). ......................................................................................................................... 97

Figure 4-6. Fluid Flow Profiles in the Middle of Vessel D Operating at 2760 kPa
(DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity
(m/s). ......................................................................................................................... 98

Figure 4-7. The Separation Efficiency versus Gas Velocity for Separator C at P = 70
kPa Assuming d = 150 µm (DPM Approach). .................................................... 100

Figure 4-8. Profiles of Volume Fraction in the Middle of Two-Phase Separators; (a)
Vessel A, (b) Vessel B, (c) Vessel C, and (d) Vessel D. ........................................ 102

Figure 4-9. Fluid Flow Profiles in the Middle of Vessel B Operating at 70 kPa (VOF-
DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity
(m/s). ....................................................................................................................... 104

Figure 4-10. Fluid Flow Profiles in the Middle of Vessel B Operating at 700 kPa
(VOF-DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of
Velocity (m/s).......................................................................................................... 105

Figure 4-11. Fluid Flow Profiles in the Middle of Vessel B Operating at 2760 kPa
(VOF-DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of
Velocity (m/s).......................................................................................................... 106

Figure 4-12. Fluid Flow Profiles in the Middle of Vessel D Operating at 70 kPa
(VOF-DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of
Velocity (m/s).......................................................................................................... 107

xvi
Figure 4-13. Fluid Flow Profiles in the Middle of Vessel D Operating at 700 kPa
(VOF-DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of
Velocity (m/s).......................................................................................................... 108

Figure 4-14. Fluid Flow Profiles in the Middle of Vessel D Operating at 2760 kPa
(VOF-DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of
Velocity (m/s).......................................................................................................... 109

Figure 4-15. The Separation Efficiency versus Gas Velocity (VOF-DPM Approach);
(a) Separator B at P = 700 kPa, and (b) Separator D at P = 70 kPa.................... 111

Figure 4-16. Profiles of Gas Flow in the Upper Half of the Gullfaks-A Separator for
1988; (a) Contours of Pressure (Pa), (b) Vectors of Velocity (m/s). ...................... 115

Figure 4-17. Vectors of Velocity (m/s) in the Inlet Zone of the Gullfaks-A Separator;
(a) 1988, and (b) the Future Condition. .................................................................. 119

Figure 4-18. Contours of Velocity in the x-direction (m/s) at the End of Splash Plate
of the Gullfaks-A Separator; (a) 1988, and (b) the Future Condition..................... 120

Figure 4-19. Vectors of Velocity (m/s) for the Gullfaks-A Separator at 1988
Conditions at (a) z = 0.999 m, and (b) z = 1.449 m. ............................................... 121

Figure 4-20. Vectors of Velocity (m/s) for the Gullfaks-A Separator at the Future
Conditions at (a) y = 1.4 m, and (b) y = 1.0 m. ....................................................... 122

Figure 4-21. Contours of Pressure (Pa) in the Middle of the Gullfaks-A Separator for
(a) 1988, and (b) the Future Condition. .................................................................. 124

Figure 4-22. Vectors of Velocity (m/s) in the Middle of the Gullfaks-A Separator for
(a) 1988, and (b) the Future Condition. .................................................................. 125

Figure 4-23. Contours of Density (kg/m3) in the Middle of the Gullfaks-A Separator
for (a) 1988, and (b) the Future Condition.............................................................. 126

Figure 4-24. Geometrical Specifications of the Redesigned Gullfaks-A Separator. ...... 134

Figure 4-25. The Computer Model for Redesigned Gullfaks-A Separator; (a)
Generated Mesh in Gambit Environment, and (b) with All the Internals in Fluent
Environment............................................................................................................ 135

Figure 4-26. Contours of Pressure (Pa) in the Middle of the Redesigned Separator for
(a) 1988, and (b) the Future Conditions.................................................................. 137

Figure 4-27. Vectors of Velocity (m/s) in the Middle of the Redesigned Separator for
(a) 1988, and (b) the Future Conditions.................................................................. 138

xvii
Figure 4-28. Contours of Density (kg/m3) in the Middle of the Redesigned Separator
for (a) 1988, and (b) the Future Conditions. ........................................................... 139

Figure 5-1. The Two-Phase Models Developed for CFD Simulation of Phase
Separation; (a) Vapor-Liquid Separation, and (b) Liquid-Liquid Separation......... 149

Figure 5-2. Efficient Oil Droplet Diameter for Estimation of Settling Velocity in
Vapor Phase versus Vapor Density......................................................................... 152

Figure 5-3. Efficient Water Droplet Diameter for Estimation of Settling Velocity in
Vapor Phase versus Vapor Density......................................................................... 153

Figure 5-4. Efficient Oil Droplet Diameter for Estimation of Rising Velocity in Water
Phase versus Water Viscosity. ................................................................................ 156

Figure 5-5. Efficient Water Droplet Diameter as a Function of Oil Phase Viscosity to
be Used in Abraham Equation. ............................................................................... 157

Figure 5-6. Efficient Water Droplet Diameter as a Function of Oil Phase Viscosity to
be Used in Stokes’ Law. ......................................................................................... 158

Figure 5-7. The Vertical Models Developed for CFD Simulation of Phase Separation;
(a) Vapor-Liquid Separation, (b) Separation of Water Droplets from Oil Phase,
and (c) Separation of Oil Droplets from Water Phase. ........................................... 160

Figure 5-8. Geometrical Specifications of Designed Separators for Gullfaks-A; (a)

Simple Type, (b) Weir Type, and (c) Bucket and Weir Type................................. 164

Figure 5-9. Geometrical Specifications of the Vertical Separator Designed for

Gullfaks-A (25% Capacity). ................................................................................... 165

Figure 5-10. The Grid System for the “Simple” Separator Designed for Processing
Gullfaks-A Generated in Gambit Environment. ..................................................... 166

Figure 5-11. The Grid System for the “Bucket and Weir” Separator Designed for
Processing Gullfaks-A Generated in Gambit Environment.................................... 167

Figure 5-12. The Grid System for the Vertical Separator Designed for Processing
Gullfaks-A (25% Capacity) Generated in Gambit Environment. ........................... 168

Figure 5-13. Geometrical Specifications of “Stabilized” Separator Designed for

Gullfaks-A............................................................................................................... 171

Figure 5-14. The Grid System for the Stabilized Separator Designed for Gullfaks-A
Generated in Gambit Environment. ........................................................................ 172

Figure 5-15. Maximum Safe Velocity for the Vapor Phase Estimated by Equation 5-
18 versus the Corresponding CFD Predictions. ...................................................... 177

xviii
Figure 5-16. Maximum Safe Velocity of Water Phase for “Simple” and “Weir”
Designs versus Oil Phase Viscosity. ....................................................................... 179

Figure A-1. The Vertical Three-Phase Separator. .......................................................... 198

Figure A-2. The “Simple” Design of Horizontal Three-Phase Separator....................... 201

Figure A-3. The “Boot” Design of Horizontal Three-Phase Separator. ......................... 202

Figure A-4. The “Weir” Design of Horizontal Three-Phase Separator. ......................... 204

Figure A-5. The “Bucket and Weir” Design of Horizontal Three-Phase Separator. ...... 206

Figure B-1. Droplet Breakup Diagrams for Separator A at Multiple Pressures (DPM
Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa. ......................................... 209

Figure B-2. Droplet Breakup Diagrams for Separator B at Multiple Pressures (DPM
Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa. ......................................... 210

Figure B-3. Droplet Breakup Diagrams for Separator C at Multiple Pressures (DPM
Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa. ......................................... 211

Figure B-4. Droplet Breakup Diagrams for Separator D at Multiple Pressures (DPM
Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa. ......................................... 212

Figure B-5. Droplet Breakup Diagrams for Separator A at Multiple Pressures (VOF-
DPM Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa................................. 213

Figure B-6. Droplet Breakup Diagrams for Separator B at Multiple Pressures (VOF-
DPM Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa................................. 214

Figure B-7. Droplet Breakup Diagrams for Separator C at Multiple Pressures (VOF-
DPM Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa................................. 215

Figure B-8. Droplet Breakup Diagrams for Separator D at Multiple Pressures (VOF-
DPM Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa................................. 216

Figure C-1. The Model Developed for CFD Simulation of Demister Separation
Efficiency. ............................................................................................................... 217

Figure C-2. Separation Efficiency of the Demister versus Gas Velocity for the
Original Droplet Size Distribution.......................................................................... 219

Figure C-3. Separation Efficiency of the Demister versus Gas Velocity for the
Extended Droplet Size Distribution. ....................................................................... 219

Figure C-4. The Demister Separation Performance Reported by Verlaan et al. (1989)
at; (a) Low Gas Velocity, and (b) High Gas Velocity. ........................................... 221

xix
Figure C-5. The Demister Separation Performance Reported by El-Dessouky et al.
(2000). ..................................................................................................................... 222

Figure C-6. Separation Efficiency of the Demister versus Gas Velocity for the Air-
Water Case Study.................................................................................................... 223

Figure D-1. Contours of Pressure (Pa) in the Middle of the “Simple” Separator
Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future Conditions... 225

Figure D-2. Vectors of Velocity (m/s) in the Middle of the “Simple” Separator
Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future Conditions... 226

Figure D-3. Contours of Density (kg/m3) in the Middle of the “Simple” Separator
Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future Conditions... 227

Figure D-4. Contours of Pressure (Pa) in the Middle of the “Bucket and Weir”
Separator Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future
Conditions. .............................................................................................................. 228

Figure D-5. Vectors of Velocity (m/s) in the Middle of the “Bucket and Weir”
Separator Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future
Conditions. .............................................................................................................. 229

Figure D-6. Contours of Density (kg/m3) in the Middle of the “Bucket and Weir”
Separator Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future
Conditions. .............................................................................................................. 230

Figure D-7. Contours of Pressure (Pa) in the Middle of the Vertical Separator
Designed for Processing Gullfaks-A (25% Capacity) at (a) 1988, and (b) the
Future Conditions.................................................................................................... 231

Figure D-8. Vectors of Velocity (m/s) in the Middle of the Vertical Separator
Designed for Processing Gullfaks-A (25% Capacity) at (a) 1988, and (b) the
Future Conditions.................................................................................................... 232

Figure D-9. Contours of Density (kg/m3) in Middle of the Vertical Separator

Designed for Processing Gullfaks-A (25% Capacity) at (a) 1988, and (b) the
Future Conditions.................................................................................................... 233

Figure D-10. Contours of Pressure (Pa) in the Middle of the “Stabilized” Separator
Designed for Gullfaks-A at (a) 1988, and (b) the Future Conditions. .................... 234

Figure D-11. Vectors of Velocity (m/s) in the Middle of the “Stabilized” Separator
Designed for Gullfaks-A at (a) 1988, and (b) the Future Conditions. .................... 235

Figure D-12. Contours of Density (kg/m3) in the Middle of the “Stabilized” Separator
Designed for Gullfaks-A at (a) 1988, and (b) the Future Conditions. .................... 236

xx
 List of Abbreviations and Nomenclature

Abbreviations
API American Petroleum Institute
CFD Computational Fluid Dynamics
DPM Discrete Phase Model
FLOSS Flow Simulator for Separators
FPSO Floating Production Storage and Offloading
FWKO Free Water Knock Out
GPSA Gas Processors Suppliers’ Association
LDA Laser Doppler Anemometry
NATCO National Tank Company
PC Personal Computer
PDA Phase Doppler Analysis
PISO Pressure-Implicit with Splitting of Operators
PR Peng Robinson
RAM Random Access Memory
SIMPLE Semi-Implicit Method for Pressure-Linked Equations
“Stiftelsen for industriell og teknisk forskning” which means:
SINTEF
Foundation for Scientific and Industrial Research
VDM Visual Dynamic Modeling
VOF Volume of Fluid

Nomenclature
a distance between holes of a perforated plate (Figure 3-11), m
surface area required for liquid-liquid separation or vessel cross-
A
sectional area, m2
Af open area of a perforated plate, m2
AH vessel cross-sectional area for heavy liquid phase, m2
AHLL surface area corresponding to high liquid level, m2
AL vessel cross-sectional area for light liquid phase, m2
ALLL surface area corresponding to low liquid level, m2
ALLV surface area of light liquid above vessel bottom, m2
Ap total area of a perforated plate, m2
AV surface area of vapor disengagement zone, m2
bi body force per unit mass in i direction, N / kg
C discharge coefficient for a perforated plate
C2 inertial resistance factor, m-1
CD drag coefficient
inertial resistance factor in j direction exerted on a plane
Cij
perpendicular to i axis, m-1
d diameter of holes in a perforated plate, m
d volume mean diameter in Rosin-Rammler equation, µm , m
d eff efficient diameter of droplet distribution, µm
dmin minimum droplet diameter, µm

xxi
dmax maximum droplet diameter, µm , m
dp particle diameter, µm , m
D diameter of separator or internal diameter of a pipe, m
DB boot diameter, m
viscous resistance factor in j direction exerted on a plane
Dij
perpendicular to i axis, m-1
f friction factor
g gravity acceleration, m / s 2
H liquid phase thickness, m
HH thickness of heavy liquid phase, m
HHLL height of high liquid level, m
HL thickness of light liquid phase, m
HLLB light liquid height in boot, m
HLLL height of low liquid level, m
HLLV light liquid height in separator, m
HolePitch diagonal distance between holes of a perforated plate, m
HR height of light liquid phase above the outlet, m
HS surge height, m
H sep separation height in horizontal separators, m
HT total height of a vertical separator, m
HV height of vapor disengagement zone, m
HW weir height, m
I turbulence intensity
k thermal conductivity, W/m°C
K settling velocity coefficient, m/s
L length of separator or residence length, m
Lmin minimum length required for vapor-liquid separation, m
m& mass flow rate through a perforated plate, kg/s
min mass of droplets entered to the demister, kg
mout mass of droplets exited from the demister, kg
n spread parameter in Rosin-Rammler equation
Nµ interfacial viscosity number as per Equation 5-18
P operating pressure, kPa
q& rate of volumetric heat addition per unit mass, W/kg
Q volumetric flow-rate of liquid phase, m3/s
Qc continuous phase flow-rate, m3/s
Qd dispersed phase flow-rate, m3/s
QHL heavy liquid phase flow-rate, m3/s
Q LL light liquid phase flow-rate, m3/s
QV vapor phase flow-rate, m3/s
Re Reynolds number

xxii
Re p particle Reynolds number
Si source term for the momentum equation in i direction, Pa/m
t HL separation time for heavy liquid droplets from light liquid phase, s
tinterface time required for droplet penetration through the interface, s
t LH separation time for light liquid droplets from heavy liquid phase, s
T temperature, K
Tr reduced temperature
u velocity component in x direction, m/s
U internal energy, J
Ud separation velocity of dispersed phase, m/s
v velocity component in y direction, m/s
V velocity, m/s
VB velocity of heavy liquid phase in the boot, m/s
Vc superficial velocity of continuous phase, m/s
VH holdup volume, m3
VS surge volume, m3
Vsep relative efficient separation velocity in a separator, m/s
Vsep ,vertical apparent efficient separation velocity in a vertical separator, m/s
VV velocity of vapor phase, m/s
VV ,max maximum safe velocity of vapor phase to avoid re-entrainment, m/s
VWater ,max maximum safe velocity of water phase to avoid re-entrainment, m/s
w velocity component in z direction, m/s
We Webber number
Wecrit critical Webber number as per Equation 3-25
′
Wecrit critical Webber number in Levich theory as per Equation 3-30
Yd is the mass fraction of droplets with diameter greater than d

Greek Letters
α permeability factor, m2
β volume fraction of dispersed phase
δ thickness of a perforated plate, m
∆H height difference between the light and heavy liquid weirs, m
∆P pressure gradient, Pa
∆x gradient along thickness of a perforated plate, m
ε porosity of a perforated plate
ϕ function of dp as per Equation 4-3
φ liquid dropout time from vapor phase, s
µc continuous phase viscosity, Pa.s
µm mixture viscosity as per Equation 4-2, Pa.s
µ Oil oil phase viscosity, Pa.s

xxiii
µV vapor phase viscosity, Pa.s
µW water phase viscosity, Pa.s
ν kinematic viscosity, m2/s
ρ density, kg/m3
ρc continuous phase density, kg/m3
ρd dispersed phase density, kg/m3
ρH heavy liquid phase density, kg/m3
ρL (light) liquid phase density, kg/m3
ρ Oil oil phase density, kg/m3
ρV vapor phase density, kg/m3
ρW water phase density, kg/m3
ψ energy dissipation rate per unit mass, W/kg
σ surface tension, N/m
τ dynamic pressure fluctuation, Pa
τ ij stress in j direction exerted on a plane perpendicular to i axis, Pa
θH residence time of the heavy liquid phase, s
θL residence time of the light liquid phase, s

Subscripts
B boot
c continuous
crit critical
d dispersed, a diameter
D drag
eff efficient
H heavy liquid, holdup
HLL high liquid level
in entered to
L liquid, light liquid
LLL low liquid level
m mixture
min minimum
max maximum
out exited from
p particle
r reduced
S surge
sep separation
T total
V vapor
W water

xxiv
CFD Simulation of Multiphase Separators 1

Chapter One: Introduction

1.1 Gas-Liquid Phase Separators

Once a crude oil has reached the surface, it must be processed so it can be sent either to
storage or to a refinery for further processing. In fact, the main purpose of the surface
facilities is to separate the produced multiphase stream into its vapor and liquid fractions.
On production platforms, a multiphase separator is usually the first equipment through
which the well fluid flows followed by other equipment such as heaters, exchangers,
distillation columns etc. Consequently, a properly sized primary multiphase separator can
increase the capacity of the entire facility.
The oil as produced at the platform varies significantly from oilfield to oilfield. In
some oilfields, water (brine) is not produced together with oil, and hence only the gas and
the oil need to be separated (two-phase separation). However, usually, three-phase
separation of oil, water, and gas is required in order to prepare the produced multiphase
fluid for downstream processing.

1.1.1 Multistage Separation

A low-pressure crude oil with a low gas to oil ratio can be passed directly into a “flow
tank” operating at atmospheric pressure in which the separation of the oil from other
phases can take place in one stage (Skelton, 1977), that is “single-stage separation”.
Single-stage separation was at one time also applied to high pressure crude oils with high
gas to oil ratio, but this process had several disadvantages such as a requirement for a
very large separation vessel, the loss of a valuable fraction of oil in the gas phase, and
extra mist and foam caused by abrupt pressure drop in the multiphase separator (Skelton,
1977). These drawbacks can be overcome by separating the phases from each other in a
number of succeeding stages, “multistage separation”. In this process, a number of
separators are used in series, and the multiphase flow passes from one separator to the
next while undergoing a controlled reduction of pressure. The intermediate operating
pressures are specified so as to obtain a maximum recovery of liquid hydrocarbons and,
at the same time, to provide maximum stabilization of the liquid and gas products. The
CFD Simulation of Multiphase Separators 2

optimum operating pressure of a separator is either calculated theoretically by flash

calculations and engineering judgment or determined by field tests so that the pressure
profile results in the highest economic yield from the sale of the liquid and gas
hydrocarbons (Smith, 1987; Arnold and Stewart, 2008). Although, in the past, as many as
seven stages of separation were used, the number has decreased to a maximum of five in
current surface production operations (Skelton, 1977). The optimum number of
separation stages can be determined by field testing or equilibrium calculations based on
laboratory tests of the well fluid, and the economical number of stages are typically three
or four, but five or six will pay out under special conditions (Smith, 1987).

1.1.2 Multiphase Separator Terminology

Multiphase separation can be carried out through various oil processing equipment with
the specific terminology corresponding to each system. Hence, it is worth defining the
most important multiphase separators in the oil industry before proceeding. The
conventional oil and gas separator, which is normally installed on a production facility or
platform, may be referred to as “oil and gas separator”, “separator”, “stage separator”, or
“trap”. A “knockout vessel” is used to remove either water or all liquid from the well
fluid flow. An “expansion vessel” is the first stage separator vessel usually operated at a
low temperature. A “flash chamber” or “flash vessel” normally refers to a conventional
oil and gas separator operated at low pressure as the second or third stage of the
multistage separation. A “gas scrubber” is an oil and gas separator with a high gas to
liquid ratio. In a “wet-type gas scrubber”, dust, rust, and other impurities of the gas phase
are washed using a bath of oil or other liquid, and the gas flows through a demister to
further remove liquid droplets from the gas stream. A “dry-type gas scrubber” or “gas
filter” is equipped with demisters and other coalescing media to aid in the removal of
most of the liquid from a gas stream.
CFD Simulation of Multiphase Separators 3

1.1.3 Multiphase Separators

The original phase separator was inclined, very long, and without any internal separating
aids. During the retention time (around one minute), the gas and oil underwent a very
limited separation. Sir Stephen Gibson designed this simple phase separator and also the
multi-stage separation process. He was the first to put this process into operation in 1930
at the Haft-Kel oilfield in Iran (Skelton, 1977).
From this very simple separator, a wide variety of separator designs and
configurations for multiphase separators in both vertical and horizontal orientations have
been developed. Various parameters such as space and operating restrictions, oilfield
variations, potential contaminants, and economic evaluations are considered in the design
of a multiphase separation system. For instance, some separators may be equipped with
special impingement internals to aid the separation process. Figure 1.1 shows some
different common designs composed of “simple”, “boot”, “weir”, and “bucket and weir”.
These designs offer a variety of methods to control the interface level in horizontal three-
phase separators. “Simple” design separator can easily be adjusted to handle unexpected
changes in oil or water density or flow rates (Arnold and Stewart, 2008). A boot,
typically, is used when the water fraction is not substantial (less than 15-20% of total
liquid by weight), and a weir is used if the water fraction is substantial (Monnery and
Svrcek, 1994). The bucket and weir design is usually used when interface level control is
difficult either in heavy oil applications or because of emulsions or paraffin problems
(Arnold and Stewart, 2008).
In spite of the variety of design configurations proposed for multiphase
separators, the phase separation process is accomplished in three zones: The first zone,
primary separation, uses an inlet diverter so that an abrupt change in flow direction and
velocity causes the largest liquid droplets to impinge on the diverter and then drop by
gravity. In this zone, the bulk of the liquid phase is separated from the gas phase. In the
next zone, secondary separation zone, gravity separation of fine droplets occurs as the
vapor and liquid phases flow through the main section of the separator at relatively low
velocities and little turbulence, and the liquid droplets settle out of the gas stream due to
gravity. The liquid collection section in the bottom half of separator provides the
CFD Simulation of Multiphase Separators 4

retention time required for entrained gas bubbles or other liquid droplets to join their
corresponding phases because of gravity and buoyancy. This section also provides the
holdup and surge volumes for safe and smooth operation of the separator. Gas flows
above the liquid phase while entrained small liquid droplets are again separated by
gravity. The final zone, coalescing media, is designed for mist elimination in which very
fine droplets that could not be separated in the gravity settling zone are separated by
passing the gas stream through a mist eliminator. In this zone, vanes, wire mesh pad, or
coalescing plates may be used to provide an impingement surface for very fine droplets to
coalesce and form larger droplets which can be separated out of gas stream by gravity.

(a) (b)

(c) (d)
Figure 1.1. Different Common Designs of Horizontal Three-Phase Separators; (a)
“Simple”, (b) “Boot”, (c) “Weir”, and (d) “Bucket and Weir”.
CFD Simulation of Multiphase Separators 5

1.1.4 Operating Pressures and Capacities

The operating pressure of separators may vary from a high vacuum to around 35 MPa
(Smith, 1987) and their capacities may range from a few hundred barrels per day to
100000 barrels per day or more (Skelton, 1977). As the operating pressure of a separator
increases, the density difference between the liquid and gas phases decreases. Therefore,
it is desirable to operate multiphase separators at as low a pressure as is consistent with
other process conditions and requirements. Most multiphase separators operate in a
pressure range of 138 kPa to 10340 kPa (Smith, 1987).

1.1.5 Separator Internals

The modern separator designs have much higher capacities than the original type and are
very short in length due to internals such as inlet diverters, controls, flow-distributing
baffles, and mist extractors which enhance the separator efficiency. A “weir” design
horizontal three-phase separator and a vertical three-phase separator, both with the
installed internals, are shown in Figure 1.2 and Figure 1.3, respectively. As represented,
both separators are equipped with inlet diverters, controls, wire mesh demisters and
pressure relief devices. In the vertical separator, a chimney is used to equalize gas
pressure between the lower and upper sections of the vessel. Important common internals
are explained in the following subsections.

Figure 1.2. “Weir” Design Horizontal Three-Phase Separator with the Installed Internals.
CFD Simulation of Multiphase Separators 6

Figure 1.3. Vertical Three-Phase Separator with the Installed Internals.

1.1.5.1 Inlet Diverters

Usually a deflector baffle or a cyclone is used as inlet diverter in the separator. Deflector
baffles come in various shapes and can be installed at different angles. However,
hemisphere or conical designs are preferred because they cause fewer disturbances than
plates or angle iron and reduce re-entrainment and emulsion problems (Arnold and
Stewart, 2008). Cyclone diverters are increasingly used in oil production facilities as they
promote foam breaking and mist elimination while performing the bulk gas-liquid
separation in the inlet zone (Chin et al., 2002). Hence, cyclone diverters can be used to
increase the operating capacity of multiphase separators.
CFD Simulation of Multiphase Separators 7

1.1.5.2 Controls

Separators are required to operate at a predetermined pressure which is specified by

economic and engineering studies. The fixed operating pressure in a separator is achieved
by using an automatic back pressure regulator on the gas outlet line. This device
maintains a steady operating pressure in the vessel. The liquid level controllers and liquid
outlet control valves are used to maintain constant oil and water levels in the separator.
Consequently, the operation of modern separator systems is completely automatic.

1.1.5.3 Mist Eliminators

Droplets with diameters of 100 µm and larger will generally settle out of the gas stream
in most average-sized separators. However, mist eliminators are usually required to
remove smaller droplets from the gas phase (Smith, 1987). As 95% of droplets entrained
in the gas stream can be separated in economically-sized separators without coalescing
media, the efficiency can be increased to around 100% by installing mist eliminators
(Walas, 1990; Sinnott, 1997; Arnold and Stewart, 2008). These coalescing media can
consist of a series of vanes, a knitted wire mesh pad, or cyclonic passages to remove the
very fine droplets from the gas phase by impingement on a large surface area where they
collect and collide with adjacent droplets. The mechanisms used in various mist
eliminators are gravity separation, impingement, change in flow direction and velocity,
centrifugal force, coalescence, and filtering (Smith, 1987). Mist eliminators can be of
many different designs exploiting one or more of these mechanisms.

1.1.5.3.1 Wire Mesh Demisters

The wire mesh pads, shown in Figure 1.4, are made of knitted wire mesh and are installed
by a lightweight support inside separators. Wire mesh pads have increasingly been used
for mist elimination in separators. Since the early 1950s, the wire mesh demisters have
been used in natural gas processing with the main use of removing fine droplets ranging
from 10 to 100 µm in diameter from the gas stream (Smith, 1987). Wire mesh demisters
are still frequently used in the natural gas industry and, with proper design, can separate
CFD Simulation of Multiphase Separators 8

even very fine droplets which are less than 10 µm in diameter from the gas stream
(Fewel and Kean, 1992; Sinnott, 1997). Generally, high separation efficiencies at a low
capital and maintenance cost are experienced from using standard wire mesh demisters.
The pressure drop is a function of the entrainment load, the mesh pad design, and gas
velocity but does not usually exceed 295 Pa (Lyons and Plisga, 2005). Mesh pad
manufacturers report excellent performance from 30% to 110% (with 75% as the
preferred operating value) of their recommended velocity (Evans, 1974; Walas, 1990).
Any common metal, such as carbon steel, stainless steel and aluminum, or plastic
material may be used in these devices. During 1960-1970, considerable development
effort was made to improve the performance of wire mesh demisters by using a
combination of filaments of different materials and diameters. These types of wire mesh
demisters do provide improved separation capacities at a cost of lower operating range as
they flood at velocities 50% below those of regular demisters (Smith, 1987).
Although pad thicknesses up to 0.9 m have been used, a pad thickness of 0.10 m
to 0.15 m is normally sufficient for most separator applications (Gerunda, 1981; Walas,
1990; Lyons and Plisga, 2005). However, the “fouling” tendency of wire mesh demisters
may restrict their applications to gas scrubbers (Smith, 1987). In fact, knitted wire mesh
may foul or plug from paraffin deposition and other impurities and thus reduce separation
efficiency dramatically after a short period of service. In such cases, vane-type or
centrifugal demisters are used.

Figure 1.4. Wire Mesh Pad Mist Eliminator.

CFD Simulation of Multiphase Separators 9

1.1.5.3.2 Vane-Type Demisters

Vane-type demisters, Figure 1.5, are widely used in oil and gas separators. The separation
mechanisms used in most of vane-type demisters are impingement, change in flow
direction and velocity, and coalescence. Vane-type demisters use the inertia of the liquid
droplets in the gas stream to collect a film of liquid on the vane surface. Vane-type
demisters are inexpensive and usually will not plug or foul with paraffin or other
contaminants, hence, providing a good separation performance under widely changing
field conditions (Smith, 1987). Pressure drop across the vane-type demisters are very
low, ranging from 250 Pa to 1 kPa (Smith, 1987). In order to establish a required
minimum pressure drop, vane-type demisters are sized by their manufacturers (Arnold
and Stewart, 2008). It is usually preferred to use vane-type demisters for most oil and gas
separators, particularly in high-capacity applications, because they can significantly
reduce the separator size (Fewel and Kean, 1992).

Figure 1.5. Vane-Type Mist Eliminator.

CFD Simulation of Multiphase Separators 10

1.1.6 Operational Difficulties

The most common factors which can reduce separator performance are very high or very
low liquid level, level control failure, improper design, damaged vessel internals, foam,
vortex formation in liquid outlet zones, plugged liquid outlets, and exceeding the design
capacity of the vessel (Arnold and Stewart, 2008).
The common approaches used for improving the separator performance in
difficult cases, as proposed by Blezard et al. (2000), are increasing droplet size of
dispersed phase (e.g., by promoting coalescence), inducing a high acceleration on
droplets (e.g., by using centrifugal force), increasing the difference between fluid
densities (e.g., by introducing diluents), and decreasing the viscosity of the liquid phases
(e.g., by heating). Some of these approaches may be combined to overcome a difficult
separation task. In the following, some different measures taken for operating foamy,
emulsified, or contaminated crude oils are outlined.
Foamy crude oils hinder liquid level control and also reduce the separation space
of the separator. To improve the separator performance, it is usually advantageous to
inject a silicon defoaming agent into the foamy oil stream (around 1 × 10 −6 m3 for 1 m3 of
oil) before it enters the separator (Skelton, 1977). This agent breaks up the foam and
keeps oil from being carried over by the gas phase, leading to an effective increase in the
capacity of the separator. The other approaches that assist in breaking the foam are
settling, baffling, heat, and centrifugal force (Smith, 1987). For separators suffering from
liquid carryover while processing foamy crude oils, or glycols, amines, and similar
materials (with high foaming tendency), a dual mist eliminator system composed of a
vane-type demister at a lower level and a wire mesh pad at higher level with a gap of 0.15
to 0.30 m between them is usually used (Lyons and Plisga, 2005).
The other separation difficulty is caused by thoroughly emulsified oil. The water
phase enters the bottom of the producing well and usually breaks up into fine droplets on
its way to the surface. These fine droplets form an emulsion with the oil phase which can
lead to a fully emulsified phase. Separation of the thoroughly emulsified phase is
extremely difficult, and it is often recommended that as much of the water as possible be
removed at the well head (Skelton, 1977). This treatment is done by processing the crude
CFD Simulation of Multiphase Separators 11

oil through a large vessel (long retention time) which would allow the larger water
droplets to settle out. If further treatment of the separated oil phase is necessary, the oil
phase may be heated to help break down the emulsion. Usually, a surface tension
reducing chemical is also added to enhance the treatment. Generally the combined
application of heat and chemicals is sufficient to reduce the water and, consequently, the
salt content of the oil phase to an acceptable level (Skelton, 1977). However, sometimes
the use of an electrical coalescing media may be necessary to achieve specification level
of oil product. In this treatment method, the oil-water stream is exposed to an electrical
field which agitates the water droplets and causes them to collide and coalesce into larger
droplets and then to settle out of the oil phase. The resultant water concentration in the
effluent oil stream is usually less than 0.5% (Blezard et al., 2000).
Another emulsion problem is experienced when some very fine particles cause a
stabilized rag layer at the oil-water interface. In order for the separator to operate
properly, the rag layer must be regularly broken or removed. For this purpose, some
techniques such as filtration, heating, and chemical injection are used (Hooper, 1997).
Contaminated crude oils are also difficult to process. The most common
contaminants are sand, silt, mud, and salt. Medium-sized sands in small quantities can be
removed by an oversized vertical settler. The residue should be removed periodically by
draining from the vessel bottom. Salt may be removed by washing the oil with water and
then separating the salty water from the oil phase.

1.2 CFD as a Separator Modeling Method

Computational Fluid Dynamics (CFD) is inherently connected with the “fluid” concept. It
is interesting that this “fluid” concept can still be defined as Isaac Newton proposed more
than 300 years ago in the following elegant way: “A fluid is any body whose parts yield
to any force impressed on it, and by yielding, are easily moved among themselves.” The
physical features of any fluid flow are governed by three fundamental physical principles:
mass is conserved, Newton’s second law applies, and energy is conserved. These
fundamental physical principles can be represented in terms of mathematical equations,
generally in the form of integral equations or partial differential equations. Computational
CFD Simulation of Multiphase Separators 12

fluid dynamics is the art of replacing the integrals or the partial derivatives in these
equations with their equivalent discretized algebraic forms. These discretized algebraic
equations are then solved to provide numbers for the flow field values at discrete points
in time and/or space. Therefore, in contrast with an analytical solution, the final product
of a CFD modeling is a collection of numbers. CFD solutions generally require the
iterative manipulation of many thousands, even millions, of numbers. This task is
obviously impossible without the aid of a high-speed digital computer which accelerated
the practical development of CFD. The historical development of CFD, as reviewed by
Anderson (1995), indicates that before 1970, there was no CFD in the way that we think
of it today, and although there was CFD in 1970, the storage and speed capacity of
computers limited all practical solutions essentially to two-dimensional flow problems.
However, by 1990, this story had changed dramatically. In today’s CFD modeling
applications, three-dimensional flow field solutions are abundant and such solutions are
becoming more and more prevalent within industry and government facilities. Indeed,
some computer programs for the calculation of three-dimensional flows have become
industry standards, resulting in their use as a tool in the design process. In short, CFD,
along with its role as a research tool, is playing an increasingly stronger role as a design
tool.
The high storage capacities and calculation speed of present computers and
advanced techniques devised in modern CFD solvers have culminated in today’s
common use of commercial software packages for CFD simulation of industrial
equipment. Currently, CFD software packages are routinely used to modify the design
and to improve the operation of most types of chemical process equipment, combustion
systems, flow measurement and control systems, material handling equipment, and
pollution control systems (Shelley, 2007). For implementation of a CFD simulation using
a commercial software package, the geometry of the object of interest is specified (with a
computer aided design drawing of the object) and the corresponding discretized grid
system is created using a mesh-generation tool. For mesh generation, present software
tools provide some predefined building units in a variety of forms such as tetrahedral,
pyramidal, hexahedral, and recently, polyhedral blocks. However, generating a high
CFD Simulation of Multiphase Separators 13

quality mesh for the system is still one of the most technical and time-consuming phases
in any CFD based analysis. After preparing the grid system, the initial and boundary
conditions of the problem are specified, and the CFD parameters are set. Finally, the CFD
software proceeds with the iterative process of solving the fundamental equations for
fluid flow. As noted, once a converged solution is achieved, CFD simulation output is a
collection of numbers which correspond to the defined points in space or time. In order to
visualize these CFD simulation results and obtain qualitative aspects of the system, the
post-processing tool of CFD software is used.
CFD complements the approaches of pure theory and pure experiment in the
analysis and solution of fluid dynamic problems. Although CFD will probably never
completely replace either of these approaches, it helps to interpret and understand the
results of theory and experiment. It should be noted that suitable problems for CFD often
involve predictions outside the scope of published data, where experimental studies are
too expensive or difficult or where the development of an insight is required (Sharratt,
1990). Therefore, CFD is primarily an insight tool which is useful for understanding the
important features of a system and for elucidating and solving some system uncertainties
and problems.

1.3 Motivation for the Research Project

Many industries have a variety of processes in which multiphase mixtures must be

separated. This is the situation in the hydrocarbon production, in which produced oil,
water and gas must be separated. Optimum and efficient design of the multiphase
separators, which are generally the first and largest elements of process equipment in an
oil production platform, will reduce capital cost and will improve operating performance
of the hydrocarbon production facility. In fact, the multiphase separation process is a key
element in the oil and gas production facilities in that downstream equipment, such as
compressors, are completely dependent on the efficient and proper performance of the
multiphase separators. Thus, multiphase separators play a critical role in onshore and
offshore oil production. Figure 1.6 shows a schematic of a typical three-phase separator,
in which a mixture of oil, gas and water enters the separator as a high momentum jet and
CFD Simulation of Multiphase Separators 14

proceeds through a cyclone designed to perform a bulk gas-liquid separation. The liquid
then falls into a pool in the lower part of the vessel and the gas together with the
entrained hydrocarbon liquid drops flow to the upper section of the separator. The
released gas rises to the outlet at the top of the vessel while oil and water flow to separate
outlets at the bottom of the vessel. In the upper part of the vessel, oil and water droplets
flow down to the liquid interface as the multiphase fluid flows to the outlet. Usually,
there is also a demister pad which removes the very fine droplets of liquid from the gas
by impingement on a surface where they coalesce. Several operational problems are often
experienced that include emulsion problems inside the separator, water level control
failure, a rising water content in oil exiting the separator because of increased production
of water, oversized vessels, low separation efficiency, and so on.
This research project applies CFD based simulation to model two and three-phase
separators. The CFD model is used to analyze velocity, pressure and concentration of
multiphase fluid flow within the phase separation process in the separators. CFD, which
involves the solution of the governing equations for fluid flow at thousands of discrete
points on a computational grid in the flow domain, provides a unique picture of the fluid
movement within the vessel. The provided fluid flow visualization helps to gain a better
understanding of multiphase separation process and the corresponding design issues.

Figure 1.6. Schematic of a Typical Three-Phase Separator.

CFD Simulation of Multiphase Separators 15

Although CFD simulation is not a substitute for experience, it does provide more
information than can be obtained from physical experiments and can bring far more
useful information to bear on the design and operation of the multiphase separation
process. CFD provides a flexible and economical means of testing the performance of
alternative designs or theoretical advances for cases that are very difficult or expensive to
physically test by experimental studies. For example, the geometry of a vessel and its
internals can be changed more easily and re-analyzed to determine the effect of the
change. In addition, scaling is not an issue with CFD because the simulation can easily be
adjusted to represent any equipment size. Therefore, CFD simulation is a useful tool that
once validated can reproduce conditions that would be impossible or impractical to
duplicate in physical testing. CFD studies can efficiently guide engineers to the sources
of performance inefficiencies. By developing a detailed CFD computer model, a possible
explanation for some of the problems is determined and possible solutions can then be
proposed. This research illustrates the benefits that CFD analysis can provide in
optimizing the design of new separators and solving problems with existing designs.
Thus, there is a direct tangible benefit to industry.
CFD Simulation of Multiphase Separators 16

Chapter Two: Literature Review

Multiphase separators and their performance are key issues to economical and stable
hydrocarbon fluid processing. Therefore, the public literature and corporate literature are
extensive, particularly regarding two-phase separators. Most of these documents propose
separator design guidelines and some address operating performance issues associated
with multiphase separators. In a very few studies, CFD based simulations have been
performed to provide a realistic picture of fluid phase separation phenomena and to
improve separator efficiency. These CFD based studies, however, have generally been
focused on performance issues in specific operating separators.
The present chapter will present a review of the CFD based studies performed on
multiphase separators. Furthermore, since this research project developed improved
criteria for designing multiphase separators, classic guidelines for design of separators are
also reviewed in this chapter to provide a comprehensive literature review within the
research project scope.

2.1 Classic Methods for Design of Multiphase Separators

In classic methods, multiphase separators are designed so as to provide sufficient

disengagement space for gas phase from which liquid droplets settle and to provide
adequate retention time for liquid phase(s) to establish satisfactory gas-in-liquid or liquid-
liquid separation (in three-phase operation). Some heuristics, rules-of-thumb, have also
been suggested which are usually conservative and lead to oversized separators
(Abernathy, 1993). In more systematic procedures, droplet settling theory is applied for
evaluating vapor-liquid or liquid-liquid separation requirements, and adequate retention
times may be assumed based on experience, scale model predictions, or field data. Using
systematic procedures, a software package was developed for design of two-phase and
three-phase separators (Grødal and Realff, 1999). The design procedure was based on
droplet settling theory, and the optimum solution was determined by applying Sequential
Quadratic Programming (SQP) techniques. As stated by Grødal and Realff (1999), the
most comprehensive approach among the classic methods was proposed by Svrcek and
CFD Simulation of Multiphase Separators 17

Monnery in two technical papers (Svrcek and Monnery, 1993; Monnery and Svrcek,
1994). In this approach, based on accepted industrial guidelines, an algorithmic method
was developed for designing the optimum (the most economical) multiphase separators
through iteration (for horizontal separators) or height adjustment (for vertical separators).
In order to provide more details of the approach, the main steps used in designing three-
phase separators proposed by Monnery and Svrcek (1994) are presented in Appendix A.
Two computer codes have also been developed in the present research project for
designing two-phase and three-phase separators based on their classical approach.

2.1.1 Common Design Aspects

2.1.1.1 Vessel Orientation

The first design issue connected with multiphase separators is their orientation.
Multiphase separators can be designed and installed in either horizontal or vertical
orientations. Vertical separators occupy little plot space, and they can be more easily
transported and installed than horizontal separators. However, other factors, such as high
operating capacity, and processing capabilities, do require horizontal separators.
Therefore, a choice is required for each design/processing situation. For instance, in the
Middle East, the separators are usually horizontal, but in the USA, where the quantities of
crude oil processed at each battery are much less than in the Middle East, vertical
separators are more common (Skelton, 1977).
To aid the separator design process, rules of thumb have been developed. As
concisely proposed by Evans (1974), vertical separators are used when there is small
liquid load, limited plot space, or where ease of level control is desired.
The GPSA Engineering Data Book (1998) states while horizontal separators are
most efficient for high capacity operations and where large amounts of solution gas are in
the liquid phase, vertical separators are usually used if the gas to liquid ratio is high or
total gas volumes are low.
Arnold and Stewart (2008) have shown that horizontal separators are more
economical than vertical types and provide better processing/operation when emulsions,
foam, or high gas to oil ratio fluids are present, but they have some limitations in sand
CFD Simulation of Multiphase Separators 18

and surge processing and require more plot space than equivalent vertical separators.
Vertical separators are often preferred for operating either low or very high gas to oil
ratio fluids.
A more comprehensive list of similar heuristics was proposed by Smith (1987).
However, although the presented guidelines are useful, as emphasized by Svrcek and
Monnery (1993), it is necessary for practical design of separators to compare both
horizontal and vertical orientation designs to determine which is more economical.

2.1.1.2 The Aspect Ratio of a Separator

Although there is not a unique set of diameter and lengths of a separator that satisfies a
given capacity requirement, the separator aspect ratio, defined as the ratio of length (or
height) to diameter in a separator ( L / D ), should be in a reasonable range which is
specified by economic analyses. Plot restrictions may also dictate the separator aspect
ratio. Economic studies have led to the following heuristics for separator aspect ratio.
Smith (1987) suggested applying a minimum of about 1.0 to 2.0 and a maximum
of about 8.0 to 9.0 to the aspect ratio. The preferred range of the aspect ratio for
horizontal separators is 2.0-6.0, and the value for vertical separators is either 2.0-3.0 if
gas flow-rate determines the size of the vessel or 2.0-6.0 if the liquid flow-rate
determines the separator size.
According to Walas (1990), separator aspect ratio should be in the range of 2.5 to
5.0 based on the operating pressure of separator, as presented in Table 2-1. This approach
was confirmed by Svrcek and Monnery (1993).1
Arnold and Stewart (2008) suggested applying a range of 1.0-4.0, supporting the
empirical observation that liquid droplets may be re-entrained by the gas stream if the
aspect ratio of the separator is higher than 4.0. However, as discussed by Grødal and
Realff (1999), it should be noted that the re-entrainment probability (and, hence, the
upper limit of the aspect ratio) is a strong function of the fluid physical properties which
vary significantly from oilfield to oilfield.
1
Please refer to Appendix A for more details.
CFD Simulation of Multiphase Separators 19

Table 2-1. The aspect ratios proposed by Walas (1990) for multiphase separators.
P (kPa ) 0-1700 1700-3400 >3400
Separator Aspect Ratio 3 4 5

2.1.2 Two-Phase Separators

In this subsection, the systematic guidelines which are based on droplet settling theory
are reviewed. Then, the most common heuristics used for designing two-phase separators
are presented.

2.1.2.1 Droplet Settling Theory for Vapor-Liquid Separation

In two-phase separators droplets of liquid are removed from a gas phase. Separation of
liquid droplets from gas phase is described by settling theory in that the terminal velocity
of a small droplet in a gas phase is governed by Equation 2-1 (Green and Perry, 2008):

4 × 10 −6 gd p (ρ d − ρ c )
Ud = (2-1)
3C D ρ c

where, U d is settling velocity in m/s, g is gravity acceleration in m / s 2 , d p is droplet

diameter in µm , ρ d and ρ c are dispersed phase (liquid) and continuous phase (vapor)

densities (respectively) in kg/m3, and C D is drag coefficient which can be calculated via
Equations 2-2 and 2-3 (Monnery and Svrcek, 2000) and is based on Gas Processors
Suppliers’ Association (GPSA) approach:

5.0074 40.927 44.07

CD = + + (2-2)
ln( x) x x

3.35 × 10 −9 ρ c (ρ d − ρ c )d p
3

x = (2-3)
µc 2
where µ c is viscosity of continuous phase (vapor) in Pa.s.
CFD Simulation of Multiphase Separators 20

Walas (1990) reported that sprays in process equipment usually are greater than 20 µm ,
mostly greater than 10 µm . His recommended droplet size for design purposes is 200
µm . Wu (1990) proposed that for steady state operations, separators can be designed
based on a droplet size of 250 µm .
Arnold and Stewart (2008), from field experience, showed that if droplets smaller
than around 140 µm are removed in the gravity separation zone of the separator, the
demister will not become flooded and will remove those droplets between 10 and 140
µm in diameter. So, their recommended droplet size for designing a two-phase separator
is 140 µm . They also suggested using a higher value of 500 µm for designing gas
scrubbers and a higher value in the range of 300-500 µm for designing flare or vent
scrubbers.
In the separator design literature, Newton's equation (Equation 2-1) is also
represented as the Sauders-Brown equation, Equation 2-4:

ρd − ρc
Ud = K (2-4)
ρc

where K is the settling velocity coefficient in m/s. Gerunda (1981) indicated that K
ranges from 0.03 to 0.107 m/s, and for satisfactory designs K can be assumed to be 0.07
m/s.
For separators equipped with wire mesh demisters, Branan (1983) developed an
empirical correlation for K coefficient, Equation 2-5:

0.0802
K= − 0.0022; 0.04 ≤ x ≤ 6.0 (2-5)
x + 0.573
1.294

where x is a function of the flow-rates and phase densities as per Equation 2-6:

Qd ρd
x= (2-6)
Qc ρc
CFD Simulation of Multiphase Separators 21

where Qd and Qc are dispersed phase (liquid) and continuous phase (vapor) flow-rates
in m3/s, respectively.
Smith (1987) indicated that K coefficient should be considered as a function of
many parameters such as separator orientation and aspect ratio (the most important
factor), the design of the internals, feed conditions, operating pressure, foaming tendency
of the fluids, and the degree of separation required. He suggested using a K coefficient
ranging from 0.03 to 0.051 m/s (typically, 0.051 m/s) for vertical separators and a K
coefficient ranging from 0.106 to 0.215 m/s (typically, 0.152 m/s) for horizontal
separators. He also recommended using conservative values for unknown operating
conditions or non-ideal cases such as slugging flow, high pressure, and excessive
platform vibration.
For vertical separators, Walas (1990) proposed a K value of 0.0427 m/s for “no
demister” cases and the K coefficients as shown in Table 2-2 for “with demister” cases.
He pointed out that if the wire mesh pad is installed in a vertical or inclined position, the
values of K should be 2/3 of the values reported in Table 2-2.
Evans (1974) and Walas (1990) recommended the use of higher values by a factor
of 1.25 for horizontal separators. The York Mist Eliminator Company recommended that
the K coefficients be based on the operating pressure. The values were regressed by
Svrcek and Monnery (1993) and are presented in Table 2-3.

Table 2-2. The K coefficients proposed by Walas (1990) for multiphase separators
equipped with stainless steel mesh pads.
Efficiency Density Special Surface K (m/s)
(kg/m3) (m2/m3) Under Pressure Vacuum
Low (99.0%) 80-112 213 0.122 0.061-0.082
Standard (99.5%) 144 279 0.107 0.061-0.082
High (99.9%) 192 377 0.107 0.061-0.082
Very High (>99.9%) 208-224 394 0.076 0.061-0.082
CFD Simulation of Multiphase Separators 22

Table 2-3. The K coefficients proposed by York Mist Eliminator Company and regressed
by Svrcek and Monnery (1993) for multiphase separators equipped with stainless steel
mesh pads.
Range of Absolute Pressure K (m/s)
(kPa)
6.7-101.3 0.02843 + 1.28 × 10 −4 P + 0.01402 ln (P )
101.3-276 0.1067
276-37911 0.1445 − 0.007 ln(P )

GPSA guidelines also recommended that K coefficients should be estimated based on the
operating pressure (Gas Processors Suppliers Association, 1998). The values proposed
for vertical separators equipped with demister have been fitted and the result is shown as
Equation 2-7:

K = 0.11[1.07 − 0.074 ln (1 + 0.01P )] (2-7)

where P is operating pressure in kPa. A factor of 1.25 and 0.5 was recommended in
Equation 2-7 to calculate K coefficient for horizontal separators and separators without a
demister pad, respectively. For vacuum services a value of K = 0.061 m/s for both vessel
orientations was recommended.

2.1.2.2 Heuristics

2.1.2.2.1 Liquid Retention Time

Evans (1974) recommended a liquid retention (or residence) time of 2-5 min to process
reasonable variations in fluid flow rates. For many cases, a liquid retention time of 5-10
min in a separator operating half-full of liquid is adequate (Walas, 1990). Gerunda (1981)
recommended the use of a vertical separator if a short liquid retention time is acceptable.
Svrcek and Monnery (1993) recommended using a systematic approach for
estimation of the required liquid retention in that the low liquid level is selected from the
CFD Simulation of Multiphase Separators 23

operating pressure and the vessel diameter, and then holdup2 and surge time3 are
estimated based on the process type, personnel and instrumentation quality. Typical
values for holdup and surge time are 2-10 min and 1-5 min, respectively (Svrcek and
Monnery, 1993).
For smooth operation and control, Sinnott (1997) recommended a typical value of
10 min as the liquid retention time. Smith (1987) and Arnold and Stewart (2008)
proposed that retention times of 30 s to 3 min are generally sufficient for most
applications and can also satisfactorily process “slugs” or “heads” of well fluids.
However, Smith (1987) pointed out that for satisfactory separation of foaming oil, the
retention time should be increased to 5-20 min based on the foam stability.

2.1.2.2.2 Overall Aspects of Separator Geometry

Watkins (1967) developed some heuristics that account for the influence of instrument
quality, labor quality, operating characteristics, and level control quality in specifying
liquid holdup. These heuristics are summarized and illustrated in Figure 2-1. This
guideline suggested that if the calculated aspect ratio is less than 3, the length is increased
making the ratio equal to 3, and if the aspect ratio in a vertical separator is more than 5, a
horizontal separator is instead specified.
Smith (1987) recommended that for a vertical separator, liquid depth above the oil
outlet should be from one to three times the vessel diameter. In a horizontal separator,
liquid phase should have a depth of 0.08 m to 60-70% of the separator cross-sectional
area, with a typical value of one-third of the vessel diameter.
Walas (1990) recommended some key dimensions for designing separators
equipped with mesh pad demisters which are illustrated in Figure 2-2. For vertical
separators, Sinnott (1997) suggested a vapor-liquid disengagement height equal to the
vessel diameter.

2
Holdup, considered for smooth and safe operation of downstream facilities, is defined as the time it takes
to reduce the liquid level from normal to low liquid level by closing inlet flow while maintaining a normal
outlet flow.
3
Surge time, considered for handling upstream or downstream variations, is defined as the time it takes to
increase the liquid level from normal to high liquid level by closing outlet flow while maintaining a normal
inlet flow.
CFD Simulation of Multiphase Separators 24

Figure 2-1. Design Heuristics Proposed by Watkins (1967) for Two-Phase Separators.

Figure 2-2. Design Heuristics Proposed by Walas (1990) for Two-Phase Separators
Equipped with Wire Mesh Demisters.
CFD Simulation of Multiphase Separators 25

2.1.3 Three-Phase Separators

In three-phase separators, both vapor-liquid and liquid-liquid separations must be

considered in the separator sizing calculations. The compartment for vapor-liquid
separation in a three-phase separator is designed based on the settling theory as presented
for two-phase separators. Therefore, vapor-liquid separation need not be discussed in this
section. However, the settling theory used for liquid-liquid separation and the most
common relevant heuristics are presented in this section. As noted previously, the
algorithmic procedure proposed by Monnery and Svrcek (1994) for designing three-phase
separators have also been presented in Appendix A, providing additional details for the
classic design methods.

2.1.3.1 Droplet Settling Theory for Liquid-Liquid Separation

Stokes’ law is used to determine the separation (settling or rising) velocity of the droplets
(Green and Perry, 2008):

d p2 g (ρ d − ρ c )
Ud = (2-8)
18 × 1012 µ c

where, U d is settling (or rising) velocity in m/s, d p is droplet diameter in µm , g is

gravity acceleration in m / s 2 , ρ d and ρ c are dispersed phase and continuous phase

densities (respectively) in kg/m3, µ c is continuous phase viscosity in Pa.s. Abernathy

(1993) and Monnery and Svrcek (1994) proposed that a maximum practical separation
velocity of 0.00423 m/s be used as an upper limit in Equation 2-8. The accepted droplet
size of 150 µm has been used as a standard in the API design method (Walas, 1990;
Hooper, 1997).
Recent papers suggest assuming a larger droplet size for design purposes.
Moreover, since the primary purpose of three-phase separation is to prepare the light
liquid (usually oil) for further treating, separator design methods have usually addressed
only water droplet removal from the oil phase. Field experience with separators sized in
this way indicates that the oil content in the separated water phase is very low ranging
CFD Simulation of Multiphase Separators 26

from a few hundred to 2000 mg/L (Arnold and Stewart, 2008). For the rare case of
operating very high-viscous water, Arnold and Stewart (2008) recommended that an oil
droplet size of 200 µm may be assumed for satisfactory separation of oil droplets from
water phase. Lyons and Plisga (2005) and Arnold and Stewart (2008) recommended
using a water droplet size of 500 µm for satisfactory separation of water droplets from the
oil phase. An effluent emulsion to be treated by downstream equipment is expected to
contain less than 5-10% water (Arnold and Stewart, 2008). For heavy crude oil cases,
Arnold and Stewart (2008) recommended a water droplet size of 1000 µm , with the
corresponding emulsion being expected to contain as much as 20-30% water.

2.1.3.2 Heuristics
2.1.3.2.1 Liquid Retention Time

Smith (1987) found that the operating liquid retention time in oil production separators
varies from 20 s to 1-2 h. However, a typical retention time in three-phase separators is
from 2 to 10 min, with 2-4 min being the norm (Smith, 1987).
For cases where emulsions are not likely to form, Sinnott (1997) recommended a
retention time of 5 to 10 min to be sufficient, while Perry and Green (1999) reported that
in the petroleum industry, oversized separators with liquid retention times of up to 1.0 h
have frequently been used, which in most cases are excessive and expensive.
Arnold and Stewart (2008) recommended using a retention time ranging from 3
min to 30 min depending on laboratory or field data. For cases such empirical data are not
available, they suggested using an oil retention time of 5 min to 10 min (proportional to
oil gravity or viscosity) and a water retention time of 10 min.

2.1.3.2.2 Overall Aspects of Separator Geometry

The liquid level in a horizontal separator for three-phase operation is normally assumed
to be at the horizontal centerline of the vessel (half full operation) to maximize the gas-
liquid interface area (Smith, 1987; Arnold and Stewart, 2008). However, the liquid level
can vary from 0.20-0.27 m up to 80-90% of the cross-sectional area of the vessel (Smith,
1987).
CFD Simulation of Multiphase Separators 27

2.2 CFD-Based Studies of Multiphase Separators

Most of the CFD-based studies of multiphase separators have been preformed by two
research groups: NATCO and SINTEF. Both groups have developed CFD models to
study the separation performance of large-scale separators. These studies are reviewed in
the following two subsections. The other CFD-based studies will be presented in a
separate subsection.

2.2.1 Contribution of NATCO Group

Frankiewicz et al. (2001) reviewed the effects of some design options, such as inlet
distributors and distributing baffles, on reducing the size and weight of separation trains
while maintaining or improving their performance. CFD results in terms of fluid
streamlines have been presented for oil and water separation inside a three-phase
separator with and without flow-distributing baffles. Moreover, contours of density have
been presented for oil and gas separation inside a vertical two-phase separator. The
sensitivity of operating quality of the installed vortex cluster to the inlet flow rate has
been demonstrated. This paper did not include any information on the developed CFD
models.

Frankiewicz and Lee (2002) used CFD simulations to study the flow pattern in two and
three-phase oilfield separators. The influence of inlet nozzle configuration, flow
distributors, perforated plates, and outlet nozzles were studied. It was realized that CFD
simulations can be used in designing separator internals such as perforated plates in order
to establish a reasonable flow distribution while minimizing liquid sloshing for offshore
applications. The well-known k − ε turbulence model and the Volume of Fluid (VOF)
multiphase model of Fluent were used, and the grid system was generated in the Gambit
environment (Frankiewicz and Lee, 2002). The CFD results in terms of streamlines and
velocity vectors for the fluid flows were considered as the criteria in modifying the
separator internals. The inlet nozzle, the inlet momentum breaker, perforated plates, weir
CFD Simulation of Multiphase Separators 28

or bucket plates, and outlet nozzles were taken as the key components affecting the fluid
streamlines throughout the CFD studies.
The single perforated plate installed in the separator was modeled as porous
media of finite thickness with directional permeability, and the CFD results showed that
the flow streamlines developed quickly downstream of the perforated plate were short-
circuiting (Frankiewicz and Lee, 2002). This prediction was also confirmed by some
experimental tests (Frankiewicz and Lee, 2002). It was recommended that fluids could
anticipate the outlet zone and chose the path of least resistance for traveling. This
behavior resulted in a significant loss in the effective liquid retention time. To prevent
this problem, CFD studies confirmed that a second perforated plate just upstream of the
outlet nozzle was required. CFD results also predicted a high velocity profile for liquids
flowing under perforated plates when an open area was retained to allow for sand
migration. Such flow profiles not only favored water short-circuiting, but also increased
the probability of oil and water mixing, particularly if the oil-water interface were too
close to the lower opening area. The position of perforated plates was determined by
experience and practical considerations such as location and configuration of the inlet
nozzle, position of the weir and water outlet nozzle, and the vessel dimensions. It was
emphasized that a fully symmetric setting of the internal baffles was not effective.
CFD simulations of Frankiewicz and Lee (2002) were validated by laboratory
tests and showed that the standard box distributor tended to bypass a significant fraction
of emulsion around the electrostatic grid, and the device should be replaced by a
shrouded pipe distributor. Revamping the operating separator with composite plate
electrodes and the shrouded pipe distributors led to an increased capacity of some 67%,
due mainly to improved flow distribution.

The primary goal of the CFD study performed by Lee et al. (2004) was to evaluate the
design of internals for a three-phase separator that would mitigate sloshing of liquid
phases due to offshore wave motions. The separator of interest was installed as part of a
Floating Production Storage and Offloading (FPSO) facility. The internals simulated
included perforated plate baffles, combo plate (comprised of both solid and perforated
CFD Simulation of Multiphase Separators 29

portions), and weir were modeled as either porous media or solid wall in the CFD
simulations.
Three case studies were performed. The first two were completed with a “no
flow” assumption since fluid flow practically had only a small effect on the sloshing
motion of interfaces. Fluid profiles obtained for the first two case studies demonstrated
that the separator would not operate properly because the water phase would spill over
the oil weir if the installed baffles were excluded from the separator. To prevent the water
phase from being pulled up toward the oil weir, a perforated baffle was designed and
placed near the oil weir, and the open areas of two preceding baffles were decreased.
Although the position of oil weir was fixed, its configuration was modified based on the
results of CFD simulations. These improvements led to a reasonable control of oil-water
interface, eliminating the water spillover problem of the second case study. Finally, as a
third case study, fluid flows of oil and water phases were taken into account while
evaluating the general performance of the installed baffles. In this case study, assuming
that there would be a minor impact from gas phase flow on the oil-water interface, gas
flow was ignored. The results of CFD simulations indicated that the designed baffle
arrangement, which was the same as the second case study, would mitigate the sloshing
problems and keep the water phase from spilling over the weir although both sources of
turbulence, wave motions and liquid flows through the vessel, were present. The field
data showed very low levels of impurities in the separated oil and water phases, thus
confirming both the validity of the CFD simulation studies and successful operation of
the separator. Unfortunately, the details of the CFD simulations and the obtained
solutions were not presented in the paper.

Lu et al. (2007) evaluated the effectiveness of perforated plate baffles for improving the
separation performance of a FWKO drum/separator. The FWKO separator of interest was
a horizontal vessel with diameter of 5.7912 m and length of 19.812 m. The normal liquid
level and oil-water interface level were 4.319 m and 2.858 m, respectively. Operating and
physical parameters including pressure, temperature, flow rates, densities, and viscosities
were also given.
CFD Simulation of Multiphase Separators 30

The separator design was based on providing a residence time of 18.8 minutes for the oil
phase and a residence time of 18.4 minutes for the water phase. Considering the
symmetric geometry and configurations of the separator, only half of the vessel was
modeled. This simplifying assumption was not reasonable because of the large deviation
from plug flow regime obtained for fluid flows inside the separator. The grid system was
generated in Gambit environment leading to some 245000 computational cells. The
quality of the produced mesh system was not verified in the study. Based on the vessel
dimensions, it seems reasonable that many more grid cells should be generated, providing
a fine mesh system. The k − ε turbulence model and the Mixture multiphase model of
Fluent were used. Multiphase modeling was not based on CFD simulation guidelines, but
on a balance between model capabilities and available computational resources.
The two perforated plate baffles, designed to improve the flow distribution in a
FWKO separator, with specific configurations were modeled as some porous jump
boundaries. Although details of modeling were not provided in the paper, it seems that
the porous jump model (Fluent 6.3 User’s Guide, 2006) was selected because of its ease
of implementation. However, the model recommended for flow distributing baffles is the
porous media model in spite of the more difficult solution convergence involved.
The CFD results in terms of velocity contours of fluid flows were presented to
visually confirm that by installing the perforated plate baffles, flow distribution was
improved as the large flow circulations were broken into small ones. Furthermore, the
Euler-Lagrange approach (probably through Discrete Phase Model (DPM) of Fluent) was
then used to track the trajectories of fluid particles. The mean residence time of fluid
particles were calculated and used as the criteria for evaluating separation performance of
the perforated plate baffles. As the result of the installation of distributing baffles, the
mean residence time increased from 630 s to 980 s for the water phase and from 520 s to
745 s for the oil phase. The volumetric utilization, defined as the ratio of actual residence
time to theoretical residence time, also increased from 46% to 66% for the oil phase and
from 57% to 89% for the water phase.
CFD Simulation of Multiphase Separators 31

Lee et al. (2009) presented some engineering suggestions and corresponding CFD based
verifications performed to revamp the phase separation inefficiencies experienced in a
major oil production facility. One of the separators of interest was operating at a high
pressure of 400 kPa, and the other was operating at a low pressure of 100 kPa. The high-
pressure separator of interest was a horizontal vessel with diameter of 4 m and length of
10.4 m, and the low-pressure separator was a horizontal vessel with diameter of 4.375 m
and length of 13.5 m. All the operating and physical parameters except for configuration
of distributing baffles were also provided for both separators. Separator design had been
based on providing a liquid (both oil and water) residence time of approximately 5.5
minutes in the high-pressure separator and a liquid residence time of approximately 4
minutes in the low-pressure separator. “Debottlenecking” studies led to some suggestions
for the weir height, liquid levels, and configuration and position of distribution baffles.
CFD simulations were performed to evaluate the overall improvements resulting
from modifications to the vessel internals and settings. Generation of grid system was
completed in Gambit environment (Gambit 2.4.6, 2006), and the CFD simulations were
performed using the Fluent software. Although details of modeling have not been
provided in the paper, it appears that almost the same strategies as proposed in the
previous study of Lee et al. (2007) had been used. Again, the volumetric utilization has
been considered as the criteria in evaluating separation performance of the modified
internals and settings. As the total result of all modifications, the volumetric utilization in
the high-pressure separator increased from 82% to 84% for the oil phase and from 55% to
75% for the water phase. In the case of the low-pressure separator, the volumetric
utilization increased from 87% to 94% for the oil phase and from 52% to 95% for the
water phase. Therefore, the CFD simulation results showed that the separator
modifications mainly influenced the water phase. Also, the CFD results for the fluid flow
streamlines were presented to visually confirm that by implementing the suggested
modifications, flow distribution was improved as the large flow circulations were broken
into small weak ones.
CFD Simulation of Multiphase Separators 32

2.2.2 Contribution of SINTEF Group

Hansen et al. (1991) introduced an under development computer code, Flow Simulator
for Separators (FLOSS), aimed for simulation of fluid flow behavior inside the phase
separators. A cubical transparent model with dimensions of 0.46 m × 0.46 m × 1.83 m
was also developed to validate the results of the separator simulator. The experimental
separator model was equipped with a spherical deflector baffle as the momentum breaker
and one flow distributing perforated baffle. The configuration and all geometrical
specifications of the momentum breaker and perforated baffle have been provided. The
experiments were performed using air, water, and a special oil mixture as the fluids, and
the fluid flow behavior in inlet zone and in liquid gravity separation zone, in which the
oil-water separation occurs, was of specific interest. Using a laser-Doppler flow-meter
and by tracking an injected tracer, velocity and residence time distribution in the zones of
interest were measured to produce the experimental data that was compared with the
results of the FLOSS simulations. The experiments showed that the fluid flows in the
gravity separation zone deviated from uniform plug flow pattern, and it was more than
likely that some recirculation regions existed.
Two multiphase models were implemented to simulate the various features of phase
separation:
1. Two-Fluid Model; based on the comments given in the paper, it appears that this
model was a reduced form of the more general multiphase model of “Volume of
Fluid”4 arranged for two immiscible phase separation simulation. This model was
used for simulation of vapor-liquid separation in the inlet zone of the separator.
2. Drift-Flux Model; similarly, it appears that this model was a specific version of
the more general “Mixture” multiphase model which was used for simulation of
separation of oil and water droplets dispersed in the liquid phases.
Finally, it has been shown that the simulation results in terms of velocity profiles and
actual liquid residence time were in good agreement with experimental data. For
example, average residence time of the tracer was less than the theoretical value in both

4
Please refer to section 3.1.2 for more details.
CFD Simulation of Multiphase Separators 33

the numerical simulation and the experimental study. Unfortunately, details of the CFD
modeling are missing from this paper, and the multiphase CFD model has been limited to
two-phase simulation.

Hansen et al. (1993) presented the simulation results of the developed CFD code, FLOSS,
for an industrial scale three-phase separator. The separator of interest, with diameter of
3.33 m and length of 16.30 m, was the first stage of the three-stage-dual-train production
process installed on Gullfaks-A offshore platform. The inlet three-phase fluid flow
entered the vessel as a high momentum jet striking a spherical deflector baffle as the
momentum breaker. The upper part of the vessel was equipped with internals, including
flow distribution baffles and demister, to enhance the separation of liquid droplets from
gas.
During the first four years of operation, some operating problems and separation
inefficiencies, such as emulsion problems, and water level control failure with increasing
produced water, were experienced. The CFD studies were performed by Hansen et al.
(1993) in order to develop a deep understanding of the complex three-phase separation
process taking place inside the separator. In their modeling effort, Hansen et al. (1993)
focused on two zones: the inlet and momentum breaker zone, and the bulk liquid flow
zone. The grid systems used for the numerical simulations of the inlet zone and the bulk
liquid zone were 11 × 8 × 15 and 23 × 4 × 5, respectively. The inlet section of the separator
in which all three phases are present was modeled as a two-phase gas-liquid flow, and the
results did provide the boundary conditions for the distributed velocity field in the liquid
pool. The approach was taken to simplify the complicated features of the problem. The
other important simplifying assumption, in both zones, was to assume a symmetrical fluid
flow profiles around the vertical plane in the middle of the separator, thus, only half of
each zone volume was modeled.
Hansen et al. (1993) proposed modifications to the separator based on the CFD
predictions of rotational flow regimes established between any two baffles. Testing of the
modified separator indicated improved data in terms of the water level control and the
produced oil quality.
CFD Simulation of Multiphase Separators 34

Because of the problem scale and importance, and also since almost all the operating and
physical parameters required for CFD simulations have been provided by Hansen et al.
(1993), this significant case study was selected for comprehensive CFD studies in the
present research project. Further comments on the performed CFD simulations and
results, as well as assessment of the original work approach and results are presented in
Chapter 4.

Hansen et al. (1995) addressed some practical aspects of the phase separation inside the
multiphase separators which led to using drift-flux and two-fluid models as the
multiphase models in their CFD simulator, FLOSS. Due to difficulties in trying to extend
the single-phase CFD model to multiphase version, the computer code development was
confined to two-phase fluid flow simulation.
Potential capabilities of the developed CFD code, FLOSS, in providing fluid flow
regimes and velocity vector profiles in the various zones of the phase separators were
demonstrated. The typical results were given for the inlet zone of a three-phase separator,
the first stage separator of Gullfaks-A platform, and the bulk liquid zone of a test
separator. Details of the CFD modeling have not been presented in their paper (Hansen et
al., 1995).

2.2.3 Miscellaneous Studies

Fewel and Kean (1992) provided a review of gas-liquid separation issues with a focus on
vane-type demisters. After addressing the advantages of vane-type demisters, such as
their higher capacity compared to mesh pads, the well-known relationships and
approaches for evaluating separation efficiency, pressure drop, and capacity of vane-type
demisters were presented. It was noted that evaluating the capacity of a separation device
is the most important stage in design process, and this significant issue is frequently
overestimated because of the errors involved in scale-up of experimental results.
Moreover, physical properties of fluids, such as their density, viscosity, and surface
tension, in the oilfield operation are typically significantly different from those of
laboratory test fluids. Therefore, both the flow pattern of gas phase and droplet size
CFD Simulation of Multiphase Separators 35

distribution of liquid phase are also different. To overcome this shortcoming, it has been
suggested that field experience be combined with CFD simulation analyses to obtain a
good estimation of the vane-type demister performance.
The authors then briefly described how porous cells can be defined in a CFD
model to simulate some separation devices such as mesh pads, vanes, and filters. Fewel
and Kean (1992) have obtained converged solution of the CFD model to check if
determined allowable velocity is exceeded in any areas of the vane. If so, the number of
vanes may be increased or some perforated baffles may be added to the design in order to
prevent liquid droplet re-entrainment and improve the separation efficiency of the vane-
type demister. CFD simulations design can also check the modified internals to make
sure that gas-liquid separation is effectively achieved.
Fewel and Kean (1992) have pointed out that the CFD analysis of separator
internals is very similar to a physical test because laboratory tests performed on various
arrangements usually match CFD results remarkably well. Therefore, it has been
concluded that CFD studies can provide accurate gas capacity predictions. For example,
in a case study (Fewel and Kean, 1992), velocity vectors of the CFD simulation matched
physical measurements within 5% in a redesigned vane-type demister, and successful
operation of the demister led to exceptional separation performance of the separator. Full
gas capacity of the separator was achieved and no trace of liquid hydrocarbon carryover
by gas phase was observed.

Wilkinson and Waldie (1994) used two pilot plant scale transparent horizontal separators;
one was two-dimensional and the other was three-dimensional. They also developed their
corresponding CFD models to study the flow pattern of oil-water separation
phenomenon. The two-dimensional separator was rectangular with dimensions of 0.23 m
× 0.25 m × 0.875 m, and the three-dimensional separator was cylindrical with diameter
of 1.0 m and length of 3.77 m. In both separators, geometrical specifications of inlet,
water outlet, weir plate, oil outlet, and liquid levels have been given. In order to measure
fluid velocity and particle size distribution in the experimental models, Laser Doppler
Anemometry (LDA) and Phase Doppler Analysis (PDA) techniques were used.
CFD Simulation of Multiphase Separators 36

Experiments were conducted using dispersed oil in a oil-water mixture (with volumetric
fractions of 0.01 to 0.10%) at superficial velocities from 0.005 m/s to 0.012 m/s for the
two-dimensional model and oil-water mixtures at superficial velocities from 0.005 m/s to
0.010 m/s for the three-dimensional model. Although, from a practical point of view,
using oil-water mixtures with higher oil concentrations is more realistic for field
separator operations, this led to some technical problems in PDA measurements.
Therefore, the measureable fluid flow systems were studied.
The k − ε turbulence model and the Discrete Phase Model (DPM) of the
commercial CFD software, Fluent 2.99, were used for the CFD simulations. The grid
systems used for numerical simulations of the two and three-dimensional models were
25 × 75 and 11 × 11 × 41, respectively. In the case of three-dimensional model, flow was
considered to be symmetrical around the vertical plane in the middle of the separator,
thus, only half of the volume was modeled.
The CFD results in terms of velocity profiles and particle size distribution in the
inlet vicinity were in reasonable agreement with experimental data in the two-
dimensional model. However, in contrast with the CFD results, the PDA measurements
detected a significant presence of oil droplets in the inlet zone of the separator. This
discrepancy was justified by the fact that in CFD simulations, an ideal steady state
operation of separator was assumed even though the practical operating conditions were
fully dynamic.
Furthermore, there were substantial differences between measured flow patterns
and CFD simulations in the three-dimensional model. This shortcoming was mainly
associated with some limitations in computational resources which restricted the number
of defined grid cells. Actually, the minimum grid cell size which could be used was
restricted by the maximum number of cells in the grid system of 20001. Therefore, a
relatively coarse grid system with the average cell size of 0.036 m × 0.045 m × 0.092 m
had to be used for CFD simulation of three-dimensional model.
The authors concluded that although the results were not necessarily
representative for practical oil-water separators, they did demonstrate some potential
CFD Simulation of Multiphase Separators 37

problems while applying CFD simulations or measurement techniques to multiphase

separation phenomenon (Wilkinson and Waldie, 1994).

Hallanger et al. (1996) developed a CFD model for a three-phase separator by extension
of Two-Fluid model. The phases of interest were free gas, oil containing dispersed water,
and free water. The water droplets dispersed in the oil phase were assumed to be
spherical and obey the drag law for solid particles. The Mixture model was used for
modeling the oil phase. The water droplets were distributed among different classes
based on their diameter size, and a momentum equation for the mixture phase together
with continuity equations for each class were solved. Interactions between dispersed
droplets such as coalescence and breakup were neglected. The pressure correction
approach, as proposed through the Semi-Implicit Method for Pressure Linked Equations
(SIMPLE), with some adjustments for the mixture phase was used to obtain the
numerical solution of the system.
The model was used to simulate the first-stage separator, with diameter of 3.15 m
and length of 13.1 m. The other geometrical aspects of the separator are missing from the
paper. The separator was equipped with a deflector baffle, two perforated baffles, a
demister, and a weir plate. Assuming symmetrical flow profiles, only half of the vessel
was modeled. The grid system used for numerical simulation was 10 × 20 × 49,
corresponding to the average cell size of 0.1575 m × 0.1575 m × 0.27 m. For definition
of dispersed water droplets, an average diameter of 0.25 mm with spread parameter of 3.0
was used, and droplets were distributed in seven particle classes.
The velocity profiles for gas, oil, and water phase flows have been presented and
discussed. CFD results also indicated that most of the smallest water droplets would
remain in the oil phase while almost all of the largest droplets would join the free water
phase. The CFD results in terms of concentrations of water droplets in the oil outlet
versus oil residence time were in reasonable agreement with experimental measurements.
The largest deviation from experimental data occurred for a long oil residence time of
around 130 s.
CFD Simulation of Multiphase Separators 38

Wilkinson et al. (2000) performed small scale experiments and some CFD simulations to
investigate the flow of a single liquid phase in a two-dimensional model separator prior to
switching to larger three-dimensional two-phase (oil and water) models. The best
separation was achieved when the flow of liquids was close to plug flow in the separator
without regions of re-circulating flow. The aim was to design simple internal fittings to
achieve a plug flow velocity distribution in the separator without inducing extra
dispersion of the phases.
These results were compared with experimental data obtained from a larger three-
dimensional cylindrical model with baffles using only water flows and a 20% by volume
oil in water mixture. Flow distribution quality of a pair of baffle plates with various
configurations were also measured in the three-dimensional cylindrical model.
Two pilot-plant-scale transparent horizontal separators and their corresponding
CFD models were used to study the effect of distributing perforated baffles on fluid flow
pattern in oil-water separation phenomenon. The two-dimensional separator was the same
as in their previous study (Wilkinson and Waldie, 1994), but a smaller three-dimensional
cylindrical separator with diameter of 0.60 m and length of 2.26 m was also used in this
follow-up study. In both separators, geometrical specifications of inlet, water outlet, weir
plate, oil outlet, and liquid levels have been provided in the paper.
Experiments were conducted using water at superficial velocity of 0.005 m/s for
the two-dimensional model and either water or dispersed oil in water mixture (20% by
volume) at superficial velocity of 0.011 m/s for the three-dimensional model. The
velocities were chosen to provide a residence time of 100–200 s which is similar to
oilfield production separators.
In order to measure fluid velocity in the experimental models, Laser Doppler
Anemometry (LDA) and Phase Doppler Analysis (PDA) techniques were used.
Significant velocity fluctuations, with typical standard deviation of 40% around the
average velocity, were reported based on the PDA measurements in the two-dimensional
model. In the three-dimensional model, flow distributing effects of perforated baffles
were studied by installing a single baffle just downstream of the momentum breaker.
Various perforated baffles with open area fractions of 5% to 20% were tested. The
CFD Simulation of Multiphase Separators 39

standard deviations of time averaged axial velocity measurements were used as criteria
for evaluating the effectiveness of the flow distributing baffle. Since measurements in
water only flows were consistent with those in dispersed oil in water flow, only two runs
were performed using the oil-water mixture.
A two-dimensional CFD model was developed for each of the experimental
separators using the k − ε turbulence model of the CFD software, PHOENICS v.1.5.
Without providing detailed experimental data, Wilkinson et al. (2000) concluded that the
developed CFD model can provide a reasonably good simulation of the flow regime in
the two-dimensional experimental separator. However, as indicated by the comparison
between the results of CFD simulations and experimental data, two-dimensional CFD
modeling of the three-dimensional separator was not so successful. Although the velocity
profiles of the CFD simulation were partially the same as the experimental
measurements, the optimum suggested by CFD was at a different value of baffle open
area. There was also a significant difference in the magnitudes of standard deviation of
velocities between the two-dimensional CFD model and the three-dimensional
experimental model. However, the CFD model did predict that there should be an
optimum value for baffle open area to maximize flow uniformity.

Swartzendruber et al. (2005) modeled a vertical two-phase separator by using the

commercial CFD software, Fluent. The separator was equipped with a deflector baffle
and a vane-type demister. The focus of the study was on the quality of gas flow
distribution through the demister. Again, unfortunately, details of the developed CFD
model are missing from the paper. CFD results such as the velocity profiles and fluid
flow streamlines were presented. In order to mitigate the uneven flow distribution in the
vane demister, as shown by the CFD simulations, the following changes were proposed:
1. The deflector baffle should be moved away from the inlet and installed parallel to
vane demister.
2. A 90° elbow with turning vanes should be installed between the inlet and the
deflector baffle.
CFD Simulation of Multiphase Separators 40

It was also concluded that CFD simulations can be used as an effective design tool in
identifying potential problems and modifying low-efficiency separator designs.

Newton et al. (2007) presented an introduction to the two modeling tools for multiphase
separators, CFD and Visual Dynamic Modeling (VDM), and discussed some issues and
limitations involved with implementing each approach. The VDM approach was
described as a qualitative modeling tool which uses a scaled model of a separator to
simulate the flow pattern inside the actual separator. No measurements are taken, and the
model is used to visualize the behavior of multiphase flow through the separator.
CFD, as a modeling tool, was then introduced and its well-known multiphase
modeling approaches, Euler-Lagrange approach and Euler-Euler approach, and the
restrictions caused by available computational resources were addressed. To overcome
the computational limitations, it has been suggested that the separator volume be divided
into compartments; each including only one continuous phase and dispersed phases. Each
compartment is simulated independently, while gas-liquid and liquid-liquid interfaces are
defined as frictionless walls that will trap the dispersed droplets coming into contact with
them. Therefore, the droplets are supposed to be eliminated from computational space
after reaching these interfaces.
When specifying a dispersed phase in a CFD model, basic concepts such as using
Rosin-Rammler equation for defining particle size distribution, and estimating the
maximum stable particle size were presented. Then, some of the necessary steps for CFD
simulation of multiphase separators using Euler-Euler approach, i.e. initializing the
solution, “patching” volume fractions of the phases, setting under-relaxation factors
based on convergence and stability trend of iterative solution, and periodic
checking/adjusting the level of interfaces while iterations are proceeding, were addressed.
The authors placed emphasis on careful implementation of the Euler-Euler approach due
to its extremely complex calculations at each iteration.
In the last section of the paper, some important shortcomings of VDM approach
for modeling multiphase separators were discussed. It has been pointed out that, because
of some economic restrictions, VDM systems typically utilize two-phase flow of air and
CFD Simulation of Multiphase Separators 41

water at moderate pressures for visualizing multiphase flow of other materials inside the
separators operating at higher pressures. It was also noted that in order to achieve an
acceptable representation of the multiphase flow through the actual separator, the
involved internals should be scaled accurately, and the Reynolds number of VDM flows
should be equal or very close to the Reynolds number of the actual separator flows.
This paper presents some useful comments and remarks on basic concepts of CFD
simulation of multiphase separators. However, it should be noted that independent
simulation of separator compartments, as proposed by Newton et al. (2007) for
simplification of the separator model, may lead to some abnormal results as will be
discussed in more detail in Chapter 4 of the thesis.

2.3 Two Relevant Theses

Two academic theses have been developed to empirically study the performance of
multiphase separators in oil production facilities:

1. “Gravity Separator Revamping” by Arntzen (2001)

2. “Experimental Characterization of High-Pressure Natural Gas Scrubbers” by
Austrheim (2006)

In both of these research projects, pilot plant scale separators were used for the study of
phase separation phenomenon. The Arntzen thesis focused on the study of liquid-liquid
separation, while the Austrheim thesis focused on the study of vapor-liquid separation.
These two significant research projects are of special interest to the present research
project and are reviewed in detail in this section.

Arntzen (2001) focused on the mechanism of formation of oil-water interface (referred as

dispersion layer in the thesis) and its mathematical modeling. A pilot-plant scale
horizontal separator with diameter of 0.630 m and length of 2.80 m operating at room
temperature and atmospheric pressure and equipped with an inlet cyclone and a flow
distributer (diffuser) was used in the experiments. Different multiphase fluid systems,
CFD Simulation of Multiphase Separators 42

composed of mixtures of Exxsol oils with a mixture viscosity of 1.6 mPa.s and a crude
oil with viscosity of 1.38 mPa.s, saline water, and air, were tested. The experimental data
showed that the dispersion layer thickness increases almost linearly with dispersed phase
concentration within the limited range investigated.
Coalescence and penetration of droplets through the interface were also analyzed
by studying the water outlet quality. With water as the dispersed phase, the oil
concentration at the water outlet increased by increasing pressure drop in the inlet valve,
water flow-rate, and coalescence rate. With water as the continuous phase, the oil
concentration at the water outlet increased by increasing pressure drop in the inlet valve,
and water flow-rate. Furthermore, the data showed that the magnitude of pressure drop in
the inlet valve would affect both dispersed and continuous phases, but the concentration
of the dispersed phase and position of the inlet valve would affect only the continuous
phase. Therefore, it was concluded that the dispersed phase quality might be affected
only by the initial turbulence, while the continuous phase quality was affected by
downstream coalescence.
To improve the pilot-plant-scale separator performance, three different
mechanical alternatives, i.e. adjusting the position of inlet nozzle, removing a water-rich
fraction of inlet flow before the flow diffuser, and increasing the coalescence area by
installing some parallel horizontal baffles, were studied. The water outlet quality and oil-
water interface thickness were used as the separation efficiency criteria. The performed
experiments showed that the entrance position of the liquid was important, and the
dispersion layer increased by injecting the liquid phase in a non-continuous liquid phase.
These experimental results were also confirmed by tests on the large scale separators
operating in the Gullfaks and Statfjord oilfields. The second mechanical alternative,
removing a water-rich fraction of inlet flow upstream of the flow diffuser, was promising
only for water dominant feeds. Finally, the use of coalescing baffles did not lead to a
significant improvement in the separator performance.
CFD Simulation of Multiphase Separators 43

The Austrheim (2006) study focused on the performance of gas scrubbers operating at
low and high pressures. Three different vertical pilot-plant scale separators operating at
various pressures were used:

1. A small scrubber with diameter of 0.389 m operating at ambient temperature and

low pressures ranging from 200 kPa to 650 kPa was used. The multiphase fluid
system was composed of the Exxsol D60 oil with viscosity of around 1.4 mPa.s,
water and air.
2. A small scrubber with diameter of 0.150 m operating at ambient temperature and
high pressures ranging from 2000 kPa to 9200 kPa was used. The involved two-
phase fluid system was composed of the Exxsol D60 oil with viscosity of around
1.4 mPa.s and nitrogen or, for some cases, synthetic natural gas containing
methane, ethane, and pentane.
3. A large scrubber with diameter of 0.840 m operating at ambient temperature and
high pressures ranging from 2800 kPa to 11300 kPa was used. A two-phase fluid
system formed by processing a real natural gas was used.

All the pilot-plant separators were equipped with a vane-type inlet and a mesh pad
demister. The experimental data showed that the separation performance of the scrubbers
was excellent in non-flooded operating region of the demister. The flooding point
occurred for the large scrubber at high pressure and processing a real natural gas. For this
case, a significant liquid carryover was measured due to droplet re-entrainment (mainly)
and some very small droplets. Therefore, it was recommended that more attention should
be paid to understanding and evaluating re-entrainment mechanisms particularly while
dealing with live hydrocarbon fluids at high pressures. In fact, separation efficiency of
the scrubber generally decreased at high pressure operation with the live hydrocarbon
fluid system.
Some significant discrepancies between the results of large scale scrubber and the
small scale scrubber were reported. These discrepancies were associated with different
distributions of the multiphase fluids in the cross-sectional area of the scrubbers. Thus, it
CFD Simulation of Multiphase Separators 44

was emphasized that in order to predict the performance of real natural gas scrubbers, the
experimental tests should be carried out on large scale separators with oilfield fluids at
relevant pressures.
Furthermore, droplets size distribution in a high-pressure scrubber was determined
for the first time through experimental measurements. The results indicated that in most
cases the maximum droplet size was greater than 400 µm and the minimum droplet size
was in the range of 1-10 µm .

2.4 Summary

The important relevant literature for the research project scope has been reviewed in this
chapter. The heart of this chapter is review of the few CFD-based studies of multiphase
separators. Two research groups, NATCO and SINTEF, and other researchers have
provided a few papers on CFD-based study of multiphase separators. Although these
studies have led to some beneficial modifications and useful scientific achievements,
from an academic point of view, this part of “multiphase separator literature” is subjected
to some significant shortcomings:

1. In the published documents, only the overall steps of CFD modeling have been
provided and the details of developed CFD models are missing.
2. Generally, symmetrical fluid flow profiles have been assumed and only half of the
separator volume has been modeled. Considering the fact that the plug flow
regime assumption is not valid based on even these simplified CFD models,
assuming symmetrical flow profiles is not realistic.
3. The quality of produced computational grid system generally has not been
validated. In some studies, the computational restrictions have led to employing
some coarse grid systems and, hence, dubious results.
4. In some studies, in order to reduce the required computational memory and effort,
two-phase simulation of three-phase fluid flow has been performed. This
approach evidently reduces the validity of produced CFD profiles.
CFD Simulation of Multiphase Separators 45

5. Generally, some indirect factors such as the liquid retention time, the volumetric
utilization, and the standard deviation of time averaged velocity have been used as
the criteria for evaluating separation efficiencies. Although improving these
factors can lead to more plug flow regimes inside separators, the actual separation
efficiency is not necessarily improved. Therefore, the criteria are not real
measures of separator efficiency.
6. CFD based modifications have generally been confined to the separator internals
such as flow-distributing baffles. However, as emphasized by Lyons and Plisga
(2005), optimizing the separator internals has only a minor effect on the separator
performance, and an inefficient/poorly designed separator cannot be significantly
improved by optimizing its internals.

As will be explained through the following chapters of the thesis, the current research
project will modify the above noted CFD simulation shortcomings with new
comprehensive approaches and efficient strategies:

1. In this thesis, all details of developed CFD models will be provided and
discussed.
2. Total volume of multiphase separators will be modeled.
3. The high quality of produced computational grid system will be validated.
4. All fluid phases present in a multiphase separator will be considered in CFD
simulations, and more high quality details of the phase separation features will be
provided.
5. Separation efficiency will be evaluated directly based on mass distribution of fine
droplets as they are tracked by a suitable multiphase CFD model. Thus, the model
will provide a realistic performance of the simulated separators.
6. Multiphase separators with all their internals will be simulated, and realistic
modifications on an operating large-scale separator will be proposed. Moreover,
the current research project will establish improved design criteria for existing
design methods.
CFD Simulation of Multiphase Separators 46

In this chapter, classic guidelines for design of multiphase separators have also been
reviewed. In the classic methods, vapor-liquid and liquid-liquid separation compartments
are designed based on droplet settling theory. Moreover, the retention time of liquid
phase is selected based on empirical data or heuristics for establishing a safe and smooth
operation of the separator and downstream equipment.
Finally, two academic research projects, developed to empirically study the vapor-liquid
and liquid-liquid separations in oil production systems, have been reviewed.
CFD Simulation of Multiphase Separators 47

Chapter Three: Developed CFD Models

When one considers the scope of the research project, the size of the separators of
interest, and the challenge involved with CFD simulation of multiphase fluid flows, a
proven commercial CFD package should necessarily be selected for the development of
the CFD models of the multiphase separators. In this research project, the commercial
CFD package, Fluent 6.3.26, was selected for the model development. This choice was
recommended for simulation of multiphase fluid flows by the CFD experts, i.e. Ramin B.
Boozarjomehry, Jalel Azaiez, and Apostolos Kantzas, with whom the author had a
chance to consult. Note, this software package had been used successfully for simulation
of large-scale multiphase separators by the NATCO group in numerous case studies.
This chapter will outline the steps required for the successful development of
efficient CFD models using the Fluent software for simulating complex features of
multiphase separation. In contrast with the published CFD based studies, which are
mostly industrial assessments, all the details of the CFD modeling will be presented.

3.1 CFD Background

Before proceeding to the simulation case studies, the pertinent CFD concepts and some
modeling strategies, particularly those involved in simulating multiphase separators, are
introduced in this section.

3.1.1 Approaches to Solving Equations for Fluid Flow

In the modern CFD literature, mathematical representation of the governing equations for
viscous fluid flow is addressed as the Navier-Stokes equations. In order to demonstrate
their complex nature, these equations for a dynamic, three-dimensional, compressible,
viscous flow are presented in the following:

• Continuity Equation (mass is conserved)

∂ρ
+ ∇.( ρV ) = 0 (3-1)
∂t
CFD Simulation of Multiphase Separators 48

• Momentum Equations (Newton’s second law)

∂ ( ρu ) ∂P ∂τ xx ∂τ yx ∂τ zx
+ ∇.( ρuV ) = − + + + + ρb x (3-2)
∂t ∂x ∂x ∂y ∂z

∂ ( ρv ) ∂P ∂τ xy ∂τ yy ∂τ zy
+ ∇.( ρvV ) = − + + + + ρb y (3-3)
∂t ∂y ∂x ∂y ∂z

∂ ( ρw ) ∂P ∂τ xz ∂τ yz ∂τ zz
+ ∇.( ρwV ) = − + + + + ρb z (3-4)
∂t ∂z ∂x ∂y ∂z

• Energy Equation (energy is conserved)

∂   V 2    V2  ∂  ∂T  ∂  ∂T  ∂  ∂T 
ρ
  U +   + ∇. ρ
  U + V  = ρq& +  k  + k  + k −
∂t   
2    2   ∂x  ∂x  ∂y  ∂y  ∂z  ∂z 

∂ (uP ) ∂ (vP ) ∂ (wP ) ∂ (uτ xx ) ∂ (uτ yx ) ∂ (uτ zx ) ∂ (vτ xy ) ∂ (vτ yy ) ∂ (vτ zy )

− − − + + + + + + +
∂x ∂y ∂z ∂x ∂y ∂z ∂x ∂y ∂z

∂ (wτ xz ) ∂ (wτ yz ) ∂(wτ zz )

+ + + + ρa.V (3-5)
∂x ∂y ∂z

Various fluid properties are required in these equations: ρ is density, V is velocity with
its components in x, y, and z directions represented by u, v, and w (respectively), P is
pressure, τ ij is stress in the j direction exerted on a plane perpendicular to the i axis, bi is

i component of the body force per unit mass, U is internal energy, q& is rate of volumetric
heat addition per unit mass, k is thermal conductivity, and T is temperature.
The incompressible Navier-Stokes equations can be obtained from the
compressible form simply by assuming that the density and viscosity are constant
throughout the flow regime.
As explained by Churchill (1988), the momentum equations for a viscous flow
were named the Navier-Stokes equations after Navier5, who first derived these equations

5
Claude-Louis Marie Henri Navier (1785-1836), a French civil engineer, was one of the first to develop a
theory of elasticity.
CFD Simulation of Multiphase Separators 49

in 1822 on the basis of intermolecular arguments (Navier, 1822), and Stokes6, who first
derived these equations in 1845 for compressible fluid flow and without the molecular
hypotheses of Navier (Stokes, 1845). The terminology of Navier-Stokes equations was
then expanded to include the entire system of viscous flow equations (not only
momentum equations).
The Navier-Stokes equations are a coupled system of nonlinear partial differential
equations, hence, are very difficult to solve. In fact, there is no general analytical solution
to these equations. Finding effective approximation methods for solution of Navier-
Stokes equations is at the heart of a broad range of engineering applications, from
airplane design to nuclear reactor safety evaluation or to weather prediction (Elman et al.,
2005).
As explained by Anderson (1995), because of some stability issues, the
incompressible Navier-Stokes equations can not be solved explicitly, and solution
techniques for the incompressible equations are usually different from those used for
solution of the Navier-Stokes equations for compressible flow. To overcome this
difficulty, the pressure correction approach has been proposed. This accepted and widely
used approach has been applied to both compressible and incompressible flows with good
success (Anderson, 1995).

3.1.1.1 SIMPLE Method

The pressure correction approach is embodied in an algorithm called SIMPLE (Semi-

Implicit Method for Pressure-Linked Equations). The SIMPLE algorithm was originally
published in the often-cited paper by Patankar and Spalding (1972). The method was
developed for incompressible fluid flow but has been extended successfully to
compressible flows as well. The method has found widespread application over the past
30 years for both compressible and incompressible flows. The main features of the
SIMPLE method are as follows:

6
George Gabriel Stokes (1819-1903), an Irish mathematician and physicist, made extensive contributions
to fluid mechanics
CFD Simulation of Multiphase Separators 50

1. The finite difference equations are obtained from the governing equations to form
a coupled nonlinear system.
2. The pressure profile is estimated.
3. Using the values of pressure field, the momentum equations are solved for
velocity profile.
4. The values of velocity profile will not necessarily satisfy the continuity equation.
Hence, using the continuity equation the pressure profile is updated.
5. After obtaining the corrected pressure profile, return to step 3, and repeat the
process until a velocity profile is found that does satisfy the continuity equation.
When this is achieved, the Navier-Stokes equations have been solved.

The key point in this iterative method is obtaining a rational formula for correcting the
pressure field. To obtain this, using pressure-velocity relationships from the linearized
momentum equations, a pressure correction equation is derived from the continuity
equation.

3.1.1.2 PISO Method

The Pressure-Implicit with Splitting of Operators (PISO) method is an improved version

of SIMPLE algorithm. In the SIMPLE algorithm, updated velocities through the pressure
correction equation do not necessarily satisfy the momentum equations. To improve the
efficiency of the required iterative method, two complementary corrections, neighbor
correction and skewness correction, are applied by the PISO method.
Issa (1986) noted that the main idea of the PISO algorithm is to bring the repeated
calculations required by SIMPLE method inside the solution stage of the pressure
correction equation. This iterative process is called “momentum correction” or “neighbor
correction” and the updated velocities better satisfy the continuity and momentum
equations, hence, rapidly move to convergence. Although the neighbor correction
technique requires more CPU time at each iteration, it does reduce the number of
iterations required for convergence, especially for dynamic simulation problems (Fluent
6.3 User’s Guide, 2006). An internal iterative technique, similar to neighbor correction, is
CFD Simulation of Multiphase Separators 51

applied by PISO method for highly skewed meshes to reduce the impact of cell skewness
on the quality of the updated velocities (Ferzieger and Peric, 1996). The technique,
referred to as “skewness correction”, significantly reduces convergence difficulties
caused by highly skewed meshes.
The PISO method with neighbor correction has been recommended for all
dynamic flow simulations, and this method with skewness correction has been
recommended for highly skewed mesh systems (Fluent 6.3 User’s Guide, 2006).

3.1.2 Strategies for Simulation of Multiphase Fluid Flows

There are two approaches to modeling multiphase flows: the Euler-Lagrange approach
and the Euler-Euler approach. In the Euler-Lagrange approach, a continuous fluid phase
is modeled by solving the time-averaged Navier-Stokes equations, and the dispersed
phase is simulated by tracking a large number of particles through the flow field based on
Newton’s second law. The Euler-Euler approach, however, deals with the multiple phases
as continuous phases that interact with each other. Since the volume of a phase can not be
occupied by the other phases, phase volume fractions are assumed to be continuous
functions of space and time, and their sum is equal to one. Three different Euler-Euler
multiphase models available in Fluent 6.3.26 are: the Volume of Fluid (VOF) model, the
Mixture model, and the Eulerian model (Fluent 6.3 User’s Guide, 2006).

3.1.2.1 Discrete Phase Model (DPM)

Following the Euler-Lagrange approach leads to the Discrete Phase Model (DPM) in
Fluent. This model works well for flow regimes in which the discrete phase is a fairly
low volume fraction, usually less than 12% (Fluent 6.3 User’s Guide, 2006).
Various forces are taken into account while Fluent tracks the particles through the
flow field that include the gravity force, the drag force (with the option of involving
dynamic drag coefficient to account for particle deformation), the virtual mass force
(accelerating the fluid surrounding the particle), the thermophoretic force (exerted on
small particles suspended in a gas phase with a temperature gradient), the Brownian
force, and the lift force.
CFD Simulation of Multiphase Separators 52

Coalescence of particles and their breakups can also be modeled by DPM. For this
purpose, based on the particle Webber number a proper model within the spray model
theory is used.

3.1.2.2 VOF Model

The VOF model is a surface tracking model which is designed for simulation of
immiscible multiphase flows where the position of the interface between any two
adjacent different phases is of interest. In the VOF model, a single set of momentum
equations is shared by the fluids, and the volume fraction of each of the fluids in each
computational cell is tracked throughout the domain. Applications of the VOF model
include free-surface flows, sloshing, the motion of large bubbles in a liquid, the motion of
liquid after a dam break, the simulation of jet breakup, and the steady or dynamic
tracking of any liquid-gas interface.

3.1.2.3 Mixture Model

In the Mixture model, the phases are assumed to be completely interpenetrating. The
Mixture model solves for the mixture momentum equation and the dispersed phases are
modeled via calculation of their relative velocities. Applications of the mixture model
include bubbly flows, sedimentation, and cyclone separators.

3.1.2.4 Eulerian Model

The Eulerian model solves a set of momentum and continuity equations for each phase.
The pressure and inter-phase exchange coefficients are defined among phases based on
the type of phases involved. Applications of the Eulerian multiphase model include
bubble columns, risers, and fluidized beds.

3.1.3 CFD Modeling of Flow-Distributing Baffles and Wire Mesh Demisters

The porous media model in Fluent 6.3.26 with appropriate modifications can be used to
model the flow through baffles and demisters. This available model can also be used for a
CFD Simulation of Multiphase Separators 53

wide variety of equipment, such as modeling flow through packed beds, filter papers,
perforated plates, flow distributors, and tube banks (Fluent 6.3 User’s Guide, 2006). In
order to use the model, a cell zone in which the porous media model is to be applied is
defined and the pressure loss is modeled by setting appropriate parameters.

3.1.3.1 Momentum Equations for Porous Media

By using the porous media model, a momentum source equation is added to the
governing momentum equations. The source term, Equation 3-6, is composed of two
parts: a viscous loss term (Darcy), and an inertial loss term:

 3 3
1 
S i = −  ∑ Dij µV j + ∑ C ij ρ V V j 
 (3-6)
 j =1 j =1 2 

where S i is the source term for the ith (x, y, or z direction) momentum equation, V is the

magnitude of the velocity and D and C are prescribed matrices. On the right hand side of
Equation 3-6, the first term is the viscous loss term and the second term is the inertial loss
term. The porous media momentum equation contributes to the pressure gradient in the
porous cell, creating a pressure drop that is proportional to the fluid velocity or velocity
squared in the cell.
In the case of a simple homogeneous porous media model, Equation 3-6 will
reduce to Equation 3-7:

µ 1 
S i = −  Vi + C 2 ρ V Vi  (3-7)
α 2 

where α is the permeability factor and C2 is the inertial resistance factor.

In laminar flow through porous media, the pressure drop is typically proportional to the
velocity and the constant C2 can then be set to zero. Ignoring convective acceleration and
diffusion, the porous media model then reduces to Darcy's Law, Equation 3-8:
CFD Simulation of Multiphase Separators 54

µ r
∇P = − V (3-8)
α

At high flow velocities, the constant Cij in Equation 3-6 provides a correction for inertial
losses in the porous media. This constant can be viewed as a loss coefficient per unit
length along the flow direction, thereby allowing the pressure drop to be specified as a
function of dynamic head.

3.1.3.2 Modeling Baffles via the Porous Media Model

In modeling perforated plates, the permeability term in a porous media formulation can
usually be eliminated and only the inertial loss term needs to be considered. This
simplification leads to the following form of the porous media equation, Equation 3-9:

3
1
∇P = − ∑ C 2ij ρV Vj (3-9)
j =1 2

Kolodzie and Van Winkle (1957), have shown that the following single-orifice based
equation can be used as the flow equation through a perforated plate:

2 ρ∆P
m& = CA f 2
(3-10)
 Af 
1−  
 Ap 
 

where, m& is mass flow rate through the plate in kg/s, Af is the total area of the holes in
m2, Ap is the total area of the plate in m2, and C is the discharge coefficient.
Based on experiments, the discharge coefficient diagram has been provided as a
function of Reynolds number, plate thickness, hole diameter, and hole pitch (Kolodzie
CFD Simulation of Multiphase Separators 55

and Van Winkle, 1957). Setting m& = ρVA p , Equation 3-10 can be rearranged to provide

the pressure drop equation for a perforated plate, Equation 3-11:

1  A 
2

∆P = 2  p  − 1 1 ρV 2  (3-11)
C  A f 
  2 
 

Dividing both sides of Equation 3-11 by the plate thickness, ∆x = δ , results in Equation
3-12:

∆P 1  A 
2

= 2  p  − 1 1 ρV 2  (3-12)
∆x C δ  A f 
  2 
 

where V is the superficial velocity (not the velocity in the holes). Combining Equation 3-
12 with Equation 3-9, results in Equation 3-13 for the constant C2 in the direction normal
to the plate:

1  A 
2

C2 = 2  p  − 1 (3-13)
C δ  A f 
 
 

Therefore, if the configuration and dimensions of flow-distributing baffles are provided,

Equation 3-13 can be used to calculate C2, and the porosity of baffles, which is necessary
for the porous media model, can also be calculated.

3.1.3.3 Modeling Wire Mesh Demisters via the Porous Media Model

Knitted wire mesh demisters can be modeled as a porous media. For this purpose, the
porous media parameters, which are used for pressure drop calculations in the media,
need to be set. Since the mesh pad demisters generally result in very low pressure drops,
their pressure drop was not modeled and assumed to be negligible. Fortunately, just
CFD Simulation of Multiphase Separators 56

recently, a comprehensive and practical research study has appeared in the public
literature that deals with the characterization of pressure drop in knitted wire mesh
demisters. Helsør and Svendsen (2007) have reviewed the two other relevant studies in
this field and presented their model for pressure drop calculation in mesh pads. In their
experimental studies, the data have been collected and analyzed for seven different wire
mesh demisters, at four different system pressures (ranging from atmospheric pressure to
9.2 MPa), and with three different fluids (air, nitrogen, and natural gas). The data have
been fit to a Hazen-Dupuit-Darcy type equation, Equation 3-14:

∆P µ
= V + CρV 2 (3-14)
∆x α

and the correlating parameters ( α and C) have been provided for the different mesh pad
types. Comparing Equation 3-14 with the momentum source equation for porous media
and its simplified form (Equations 3-6 and 3-7), results in Equation 3-15 for the constants
D and C2 in a flow direction normal to the mesh pad:

1
D= ; C 2 = 2C (3-15)
α

Therefore, provided that the type and characteristics of mesh pad are available, Equation
3-15 can be used through the correlations developed by Helsør and Svendsen (2007) to
calculate the parameters required for the porous media model.

3.2 Simulation of Pilot-Plant-Scale Two-Phase Separators

In this section, the steps and input data required for CFD simulation of pilot-plant-scale
two-phase separators are presented. The paper titled “Analytical Study of Liquid/Vapor
Separation Efficiency”, by Monnery and Svrcek (2000), and a field pilot plant skid at the
Prime West East Crossfield gas plant have been used as the basis for the CFD model.
CFD Simulation of Multiphase Separators 57

3.2.1 Experimental Equipment and Procedure

The process schematic of the experimental equipment is shown in Figure 3-1. The system
consists of two-phase separators, gas inlet piping, liquid pumping and injection, and a
high efficiency filter to collect entrained separator liquids. The aim of experiments was
studying phase separation performance in the three horizontal separators and one vertical
separator operating at three different pressures 70 kPa, 700 kPa, and 2760 kPa.
The experimental procedure consisted of setting the gas pressure using the outlet
manual globe valve and the flow was adjusted using the inlet globe valve to have
established the desired flow-rate at the desired pressure. Once the gas flow stabilized, the
liquid phase was injected in an amount that over-saturated the gas. The system was left
until the flows and temperatures showed a steady-state operating condition.

Figure 3-1. Process Flow Diagram for Separator Skid (Monnery and Svrcek, 2000).
CFD Simulation of Multiphase Separators 58

The experimental data consisted of separator liquid level, gas flow-rate, and liquid
carryover. These data were then used to determine the gas velocity and corresponding
separation efficiency. The incipient velocity of gas phase which results in liquid droplet
carryover was measured and the value was used to estimate the entrained droplet
diameter.

3.2.2 Material Definition

The composition of saturated gas at three different pressures, as provided by Monnery

and Svrcek (2000), is given in Table 3-1. To determine the saturated liquid composition
and other physical properties, the Peng-Robinson (PR) equation of state was used in the
commercial process simulator, HYSYS 3.2. The predicted values used in the CFD
simulations are presented in Table 3-2.

Table 3-1. The composition of the saturated gas at multiple pressures (Monnery and
Svrcek, 2000).
Mole Fraction
Component
70 kPa 700 kPa 2760 kPa
Nitrogen 0.0377 0.0446 0.0454
Methane 0.7573 0.8958 0.9113
Ethane 0.0183 0.0217 0.0221
Propane 0.0055 0.0065 0.0066
i-Butane 0.0231 0.0041 0.0020
n-Butane 0.0766 0.0132 0.0060
i-Pentane 0.0298 0.0051 0.0022
n-Pentane 0.0271 0.0046 0.0021
n-Hexane 0.0157 0.0028 0.0014
n-Heptane 0.0029 0.0005 0.0003
n-Octane 0.0005 0.0001 0.0001
n-Nonane 0.0001 0.0000 0.0000
Methylcyclopentane 0.0018 0.0003 0.0001
Cyclohexane 0.0014 0.0003 0.0001
Methylcyclohexane 0.0008 0.0001 0.0001
Benzene 0.0009 0.0002 0.0001
Toluene 0.0005 0.0001 0.0001
p-Xylene 0.0001 0.0000 0.0000
CFD Simulation of Multiphase Separators 59

Table 3-2. The values of material properties used in simulation case studies.
Case Study Case Study Case Study
(I) (II) (III)
Operating Pressure ( kPa ) 70 700 2760
Continuous Density ( kg / m 3 ) 1.8992 6.6164 23.846
Phase Viscosity ( Pa.s ) 1.062 × 10-5 1.0812 × 10-5 1.1473 × 10-5
Density ( kg / m 3 ) 691.52 690.04 670.05
Discrete
Phase Viscosity ( Pa.s ) 4.0934 × 10-4 4.0882 × 10-4 3.3683 × 10-4
Surface Tension ( N / m ) 0.019905 0.019676 0.01707

3.2.3 Implementing DPM Approach

The first approach used to simulate vapor-liquid separation in the two-phase separators
was the DPM multiphase model available in Fluent 6.3.26. In order for this model to
provide useful results, the dispersed phase should be a very low volume fraction, less
than 12% (Fluent User’s Guide, 2006). This was the situation that existed in the vapor
phase of the two-phase separators in question. The overall strategy adopted was that of
Newton et al. (2007), in which only separation of liquid dispersion from vapor phase was
simulated. It was also assumed that the gas-liquid interface in the separators was defined
as a frictionless wall which traps the droplets coming into contact with it.

3.2.3.1 Physical Models

The physical models consist of four separators labeled vessel A, B, C (all horizontal) and
D (vertical). Figure 3-2 provides the individual vessel specifications.
Mesh generation step was completed using Gambit 2.4.6, the Fluent preprocessor
tool. In order to have a good grid quality, the external edges of nozzles were discretized
at first. Then the faces and the whole volume were swept by the Gambit mesh generation
tool. The supplementary CD includes screen photos from the Gambit environment during
the development of the grid systems for all the separators. However, for convenience,
Figure 3-3 and Figure 3-4 provide example photos for separators B and D, respectively.
CFD Simulation of Multiphase Separators 60

(c)

(b) (d)

(a)
Figure 3-2. The Geometrical Specifications of Models; (a) Vessel A, (b) Vessel B, (c)
Vessel C, (d) Vessel D.
CFD Simulation of Multiphase Separators 61

Figure 3-3. The Grid System for Vessel B Produced in Gambit Environment (DPM
Approach).

Quality of the produced mesh was examined based on the cell skewness factor. The
results, outlined in Table 3-3, indicate that a negligible fraction of cells, less than 0.01%,
is of poor quality. However, the grids with a cell skewness factor above 0.8 were
converted to polyhedral grids by employing a new capability of the latest version of
Fluent. The values of maximum cell squish, maximum skewness, and maximum aspect
ratio for original and modified mesh are reported in Table 3-4.
CFD Simulation of Multiphase Separators 62

Figure 3-4. The Grid System for Vessel D Produced in Gambit Environment (DPM
Approach).
CFD Simulation of Multiphase Separators 63

Table 3-3. Quality of the mesh produced in Gambit environment for two-phase
separators (DPM approach).
Skewness
Model 0-0.20 0.20-0.40 0.40-0.60 0.60-0.80 0.80-1.0
A 29.56% 47.22% 19.49% 3.72% 0.00%
B 29.56% 46.47% 19.96% 4.01% 0.00%
C 30.05% 45.64% 19.79% 4.51% 0.01%
D 39.30% 34.42% 20.64% 5.64% 0.01%

Table 3-4. Global quality of the original and modified mesh for two-phase separators
(DPM approach).
Number of Maximum Maximum Maximum
Cells Squish Skewness Aspect Ratio
A Original 320602 0.787114 0.829467 22.1483
Converted 320597 0.779131 0.829057 22.1483
B Original 204008 0.787114 0.829467 22.1484
Converted 204003 0.779131 0.829057 22.1484
C Original 143064 0.770505 0.821138 20.1374
Converted 143050 0.770505 0.820576 19.8758
D Original 62903 0.770509 0.821137 20.1372
Modified 62900 0. 770509 0.805442 19.0316

3.2.3.2 Definition of Droplet Size Distribution

As noted, the experimental data contained the incipient carryover velocity of gas phase
for multiple separator operating pressures. These values were then used to estimate the
entrained liquid droplet size. In this CFD simulation study, however, the probable droplet
size distribution was input and the incipient gas velocity that resulted in the liquid
carryover was determined. The common particle size distribution function of Rosin-
Rammler (1933), Equation 3-16, was used:

  d n 
Yd = exp −    (3-16)
  d  

where Yd is the mass (or volume) fraction of droplets with diameter greater than d. The

Rosin-Rammler equation contains two parameters: volume mean diameter, d , and spread
parameter, n, which specifies how narrow the distribution around the mean diameter is.
CFD Simulation of Multiphase Separators 64

Monnery and Svrcek (2000) presented maximum stable and average entrained liquid
droplet diameter at two different pressures 70 and 700 kPa. Based on their results and
assuming a minimum droplet size of 100 µm (which is small enough for simulation
purposes), the discrete phase parameters were set to the values shown in Table 3-5. As
Table 3-5 indicates, three different values were assumed as the mean droplet size to study
the effect of this parameter on the predicted incipient velocity. These values were the
well-known design droplet size of 150 µm , the average of experimental mean droplet
size of 550 µm , and 1000 µm .

3.2.3.3 Setting CFD Simulator Parameters

After importing the mesh file and making the modifications, the necessary material
properties for continuous and discrete phase were input. The discrete phase model could
then be specified by setting the droplet size distribution, the surface from which droplets
were injected to the vessel (nozzle surface), and the velocity of injection which was
assumed to be equal to the continuous phase velocity.
In all case studies the Reynolds number was much more than the transient value
(Re = 2300). Therefore, a suitable turbulence model was selected as the viscous model. In
this study, the standard k − ε (Launder and Spalding, 1972) model was selected because
of its robustness, economy and accuracy for a wide range of turbulent flows in industrial
flow simulations (Fluent 6.3 User’s Guide, 2006). This semi-empirical model has been
accepted as the most cost-effective and widely applicable turbulence model and has been
selected as the default in most commercial packages (Sharratt, 1990; Gosman, 1998).
Proper boundary condition setting was very important in this study. For the
continuous phase, inlet velocity and outlet pressure were set. The values of inlet velocity
were varied in order to find the incipient velocity. When setting the flow regime in the
inlet and outlet nozzles, the turbulence intensity and hydraulic diameter of the nozzles
were determined. For calculation of the turbulence intensity, the empirical correlation for
pipe flows, as recommended by Fluent 6.3 guidelines, was used:

I = 0.16 Re −0.125 (3-17)

CFD Simulation of Multiphase Separators 65

Table 3-5. The discrete phase parameters used for simulation of two-phase separators
through DPM approach.
Number of dmin dmax d n Total Flow-rate
Tracked Particles ( µm ) ( µm ) ( µm ) ( kg / s )
150
1000 100 3000 550 1.0 6 × 10-5
1000

In the developed CFD model, the droplets reaching the gas-liquid interface were assumed
to become part of the liquid phase. In fact, this assumption tends to simplify the
separation task for liquid droplets and might lead to higher separation efficiencies. In
order to compensate partially, the interface levels were specified somewhat lower than in
normal design practice. Therefore, the interface level in the vessels was assumed to be
around 5 cm from the vessel bottom. When a droplet contacted with walls other than
defined interface, it was assumed that 95% of its normal momentum and 90% of its
tangential momentum were lost. Thus, normal and tangent reflection coefficients were set
to be 0.05 and 0.10, respectively.
Careful choice of the solution method and under-relaxation factors is a major
contribution toward both the rate of convergence and the solution existence (Sharratt,
1990; Anderson, 1995). Following the Fluent 6.3 guidelines and the trend of solution
convergence, the solver parameters were set as follows:

Discretization Method for Momentum: Second Order Upwind

Solution Method: SIMPLE
Under-Relaxation Factor for Pressure ≈ 0.9
Under-Relaxation Factor for Momentum ≈ 0.1

3.2.4 Implementing VOF-DPM Approach

As the second approach to simulating vapor-liquid separation in the two-phase separators,

an effective combination of DPM and VOF multiphase models within Fluent 6.3.26 was
used. This approach was selected considering the nature of the empirically studied phase
separation process and the features of multiphase models available in Fluent.
CFD Simulation of Multiphase Separators 66

Consequently, the VOF model was used to obtain the overall picture of the fluid flow
behavior in the two-phase separators, and then, in order to move the simulation toward a
realistic situation, liquid droplets were injected from the inlet to be tracked by DPM
model. The equations governing the injected discrete phase (liquid droplets) and
coexisting continuous phases (gas and liquid phases) were solved simultaneously. This
approach required significant computational time because after having the continuous
fluid phases converged to their preliminary solutions, interactions among multiple
discrete and continuous phases were included.

3.2.4.1 Grid Systems

In Gambit environment, the external edges of nozzles were first discretized, and the faces
and all the volume were then swept. The screen photos from the Gambit environment
while developing the grid systems for all the separators are provided on the
supplementary CD, and for convenience, Figure 3-5 and Figure 3-6 include such photos
for separators C and D, respectively.
The quality of the produced mesh was examined based on the cell skewness
factor. The results, Table 3-6, show that a negligible fraction of cells (less than 0.01%) is
of poor quality. However, in order to reach an even better grid quality, the grids with a
cell skewness factor above 0.8 were converted to polyhedral grids in the Fluent
environment. The values of maximum cell squish, maximum skewness, and maximum
aspect ratio for original and modified mesh are reported in Table 3-7.

3.2.4.2 Definition of Droplet Size Distribution

Droplet size distribution was defined similar to the DPM approach, presented in the
previous section (Table 3-5), with the difference for this phase being the mean droplet
size of 150 µm . This size is an industrial accepted separator design droplet size, and its
suitability in prediction of incipient velocities was also confirmed using the DPM
approach7.

7
Please refer to Chapter Four for details.
CFD Simulation of Multiphase Separators 67

Table 3-6. Quality of the mesh produced in Gambit environment for two-phase
separators (VOF-DPM approach).
Skewness
Model 0-0.20 0.20-0.40 0.40-0.60 0.60-0.80 0.80-1.0
A 29.58% 46.72% 19.81% 3.89% 0.004%
B 29.97% 46.34% 19.75% 3.93% 0.004%
C 29.66% 46.12% 20.10% 4.11% 0.01%
D 38.30% 36.27% 20.75% 4.60% 0.08%

Figure 3-5. The Grid System for Vessel C Produced in Gambit Environment (VOF-DPM
Approach).
CFD Simulation of Multiphase Separators 68

Figure 3-6. The Grid System for Vessel D Produced in Gambit Environment (VOF-DPM
Approach).
CFD Simulation of Multiphase Separators 69

Table 3-7. Global quality of the original and modified mesh for two-phase separators
(VOF-DPM approach).
Number of Maximum Maximum Maximum Aspect
Cells Squish Skewness Ratio
A Original 326550 0.790256 0.829117 22.9059
Converted 326530 0.776824 0.820576 19.8753
B Original 219586 0.787082 0.829389 22.1426
Converted 219579 0.778983 0.828974 22.1426
C Original 165921 0.777096 0.821001 20.1331
Converted 165914 0.776824 0.820576 19.8753
D Original 68174 0.770605 0.888568 30.0719
Modified 68114 0.770605 0.884907 19.0413

3.2.4.3 Setting CFD Simulator Parameters

The CFD simulator parameters were set after importing the mesh file into Fluent 6.3.26
environment. Similar to DPM approach, the material properties were input and discrete
phase was specified by setting its droplet size distribution and injection characteristics.
Again, the standard k − ε model was selected for turbulence modeling.
The liquid depth in the separators was set using the normal practice as described
by Smith (1987)8. Therefore, the depth of liquid in vertical separator was set to be 29.1
cm , and in horizontal separators was set to be 6.6 cm . In order to initiate the iterative
solution procedure, the position of interface between the gas and liquid phases was
specified. For this purpose, the volume fractions of phases above and below the assumed
interface plane were set using the “Patching” tool of Fluent. Since the solver might distort
the shape and position of the interface surface, the iterations were stopped at frequent
intervals and the interface level was checked and corrected (if necessary) by patching the
volume fractions of phases. The liquid level must be maintained at the correct height
during the iterative solution process.
At the boundary setting stage, the inlet velocity and volume fractions were
specified for inlet continuous phases. Then, the outlet pressure and volume fractions (as

8
Please refer to “Overall Aspects of Separator Geometry” in “Two-Phase Separators” subsection of
Chapter Two for more details.
CFD Simulation of Multiphase Separators 70

pure gas) were set for the gas-outlet boundary, and the outlet velocity and volume
fractions (as pure liquid) were set for the liquid-outlet boundary.
In the experimental research project, the amount of liquid flow-rate was constant
while the gas flow-rate was changed to provide different superficial velocities in the
separators (Monnery and Svrcek, 2000). The flow-rate of liquid was assumed to be
3.59 × 10-4 m 3 / s which was corresponding to occupying 1% of inlet nozzle surface area
by liquid phase to give an overall 0.10 m / s superficial velocity in horizontal separators.
Hence, the liquid outlet velocity was set to 0.7811 m / s . Note, the values of inlet
velocities and volume fractions were always set so that the liquid flow-rate was
maintained constant. For example, in order to have a superficial gas velocity of 0.5 m / s
in the horizontal separators, the mixture inlet velocity and the liquid volume fraction were
set to 9.11 m / s and 0.0198. For the same superficial velocity in the vertical separator,
the mixture inlet velocity and the liquid volume fraction were set to 2.02 m / s and 0.089.
Similar to the DPM approach, the turbulence intensity in the inlet and outlet zones was
estimated using Equation 3-17.
For discrete phase modeling, the destiny of droplets after contacting with internal
walls of vessels must also be specified. For this purpose, it was assumed that the droplets
reaching the bottom of vessels (containing liquid) were drained, that is not returning to
the vapor phase. Furthermore, it was assumed that 95% of the normal momentum and
90% of tangential momentum of a droplet contacting with vessel walls (other than bottom
wall) was lost. Thus, normal and tangent reflection coefficients were set to 0.05 and 0.10,
respectively.
The process of solving a multiphase simulation problem is inherently difficult,
and stability and/or convergence problems would be encountered (Fluent User’s Guide,
2006). The Fluent 6.3 guidelines and the solution convergence trend were used to set the
solver parameters so as to overcome stability and convergence problems:
CFD Simulation of Multiphase Separators 71

Discretization Method for Pressure: Body Force Weighted

Discretization Method for Momentum: First Order Upwind
Solution Method: PISO
Under-Relaxation Factor for Pressure = 0.1
Under-Relaxation Factor for Momentum = 0.001
Under-Relaxation Factor for Volume Fraction = 0.005
Under-Relaxation Factor for Turbulent Groups (Kinetic Energy,
Dissipation Rate, and Viscosity) = 0.7

3.3 Simulation of a Large-Scale Three-Phase Separator

A three-phase separator that is part of a large North Sea production platform was next
simulated. The separator is located in the Gullfaks oil field in the Norwegian sector of
the North Sea. The production on the Gullfaks-A platform started in 1986-87. Figure
3-7, taken from Hansen et al. (1993), shows the flow diagram for the production process.
Represented in Figure 3-7, the separator of interest is the first stage of the three-stage
dual-train production process. Figure 3-8 presents the details, as provided by Hanson et
al. (1993), of the separator as it was designed and installed on the platform.
The separator operates at a fixed pressure of 6870 kPa. The inlet multiphase fluid
flow enters the vessel as a high intensity momentum jet that strikes the inlet momentum
breaker (deflector baffle). The liquid drops into the liquid pool in lower part of the vessel
and the gas together with liquid droplets flows in the upper part of the vessel. In the vapor
phase, liquid droplets will settle to the liquid interface, flowing to the corresponding
liquid phase outlets. The upper part of the vessel is equipped with internals, including
distribution baffles and demister, to enhance the separation of liquid droplets from vapor.
During the first four years of operation, separation inefficiencies have been
experienced due to the following operating problems:
• water level control failure with increasing production of water,
• emulsion problems inside separator,
• sand accumulation, and
• increased water content in the oil outlet.
CFD Simulation of Multiphase Separators 72

Figure 3-7. Process Flow Diagram for Gullfaks-A (Hansen et al., 1993).

Figure 3-8. General Configuration of the First Stage Separator on the Gullfaks-A
Production Facility (Hansen et al., 1993).

In order to develop a deep understanding of the complex three-phase separation process

Hansen et al. (1993) decided to model the overall fluid flow regimes inside the separator.
To simplify this complicated simulation task, they focused on two zones: the inlet and
momentum breaker zone and the bulk liquid flow zone. Based on operational experience
and the simulation results, Hansen et al. (1993) proposed modifications to improve the
CFD Simulation of Multiphase Separators 73

separation efficiency that included redesigning the sand removal system, modifying the
liquid level measurement device, and reducing the height of the demister and baffles to
be completely out of the liquid phase. The modifications to the separator did improve
both the water level control and the produced oil quality.
In the current CFD simulation research project, all the separation zones of the
separator will be simulated. Therefore, the results should provide an overall picture of
separation quality not only in the inlet and bulk liquid zones but also in the gas and
interface zones. Exploiting the various multiphase models available in Fluent, the
combined VOF-DPM model is the best choice for modeling both the macroscopic and
microscopic features of the three-phase separator and was used to model the phase
separation behavior in this field separator. The necessary model settings will be described
in detail in the following sections.

3.3.1 Material Definition

The physical parameters for the fluids in the Gullfaks-A separator are taken from Hansen
et al. (1993) and presented in Table 3-8. In the CFD simulation, the vessel was assumed
to operate at a constant pressure, hence the pressure drops of the fluid phases flowing
through the baffles and the demister pad were assumed to be negligible. Note, in Chapter
Four, this criterion will be utilized to physically verify the modeling of the baffles and the
mist eliminator.
The other important physical properties necessary for CFD simulation of the
three-phase separation are interfacial surface tensions. Since these data were not given in
the original paper, estimated values were used in the current study. For this purpose, a
four component mixture of n-C27H56, n-C28H58, n-C29H60, and n-C30H62, with mole
fractions of 0.85, 0.05, 0.05, and 0.05, respectively, were used in HYSYS 3.2 to simulate
the oil phase. The criterion for setting the composition of the mixture was the accuracy of
the mixture density and viscosity at operating temperature and pressure compared to the
values given in the original study. Using the PR equation of state and the TRAPP model,
density and viscosity of the mixture were estimated to be 783.59 kg/m3, and 0.005296
Pa.s, respectively. The estimation errors for the oil density and viscosity were 5.76% and
CFD Simulation of Multiphase Separators 74

0.88%, respectively. Thus, it was assumed that the oil-gas surface tension estimated by
HYSYS should be reasonable, and a surface tension of 0.0238 N/m was assumed for the
oil-gas interface. The assumed value compared well with the oil surface tension range of
0.023 to 0.038 N/m at 20°C proposed by Streeter and Wylie (1985). HYSYS was also
used for estimation of water-gas surface tension, and the estimated value was 0.0668
N/m. Using the chart provided by Heidemann et al. (1987), surface tension of pure water
is 0.067 N/m at 55.4°C which is in agreement with the HYSYS estimate. Finally, an
empirical study was used to estimate oil-water surface tension: Kim and Burgess (2001)
have reported a value of 0.052 N/m as the surface tension for a mineral oil and water
interface at 25°C. They noted that although oil is a mixture of various hydrocarbons, each
constituting hydrocarbon in contact with water, has almost the same interfacial surface
tension. Thus, it was assumed that the surface tension for oil-water interface at 25°C
would be almost the same as reported by Kim and Burgess (2001). Furthermore, in order
to account for surface tension temperature functionality, the reported value was modified
using the method proposed by Poling et al. (2001), Equation 3-18, which provides a
correlation between the surface tension and the reduced temperature, Tr:

σ (Tr 2 )  1 − Tr 2 
1.2

=  (3-18)
σ (Tr1 )  1 − Tr1 

Note, while using Equation 3-18 for alcohols, the right hand side exponent should be
changed to 0.8.

Table 3-8. The physical parameters of fluids in Gullfaks-A separator provided by Hansen
et al. (1993).
1988 Production Future Production Density Viscosity
3
Rate (m /h) Rate (m3/h) (kg/m3) (Pa.s)
Gas 1640 1640 49.7 1.30 × 10-5
Oil 1840 1381 831.5 5.25 × 10-3
Water 287 1244 1030 4.30 × 10-4
Operating Conditions Temperature = 55.4°C and Pressure = 6870 kPa
CFD Simulation of Multiphase Separators 75

Using Equation 3-18 and assuming a pseudo-critical temperature of 576°C for the oil as
estimated by HYSYS, a surface tension of 0.0486 N/m was estimated for the oil-water
interface at 55.4°C. The estimated value is in agreement with the Antonoff’s rule in that
the oil-water surface tension should approximately be equal to the absolute difference
between oil and water surface tensions (Antonoff, 1907), which are 0.0238 N/m and
0.0668 N/m, respectively.

3.3.2 Physical Model

Figure 3-9 provides the specifications of Gullfaks-A separator and does show that not all
the required dimensions were provided by Hansen et al. (1993) in their paper. The
missing information, however, may be estimated using the given dimensions for the
realistic schematic of the separator shown in Figure 3-8. To complete the information
required for a CFD simulation, the inlet nozzle and the outlet nozzles were assumed to
have a diameter of 0.48 m and 0.38 m, respectively, while the diameter of hemispherical
momentum breaker and the height of weir were assumed to be 0.80 m and 1.25 m,
respectively.
Generation of the corresponding mesh system was performed in the Gambit 2.4.6
environment. Screen photos of the generated grid system in the Gambit environment have
been included in Figure 3-10.

Figure 3-9. Geometrical Specifications of the Gullfaks-A Separator.

CFD Simulation of Multiphase Separators 76

Figure 3-10. The Grid System for the Gullfaks-A Separator Generated in Gambit
Environment.
CFD Simulation of Multiphase Separators 77

In order to have a discretized model with “good” grid quality, the mesh generation
process should be completed in a step by step sequence. The vessel was split into areas
and the inlet nozzle, momentum breaker, splash plate, weir, and outlet nozzles were first
discretized. In doing so, the edges of nozzles and other internals were discretized before
the mesh generation for the separator surfaces and volumes. Then, the cylindrical part of
the vessel was discretized such that some cells in this part were separated and referred to
as the porous media and did include mesh for distribution baffles and the demister pad.
The horizontal surrounding surfaces of each baffle (with thickness of 0.02 m) and those
of demister pad (with thickness of 0.15 m) were assumed to be flat surfaces. Therefore, in
the cylindrical part of vessel, the grids must be fine enough and arranged horizontally in
regular and constant intervals. After generating the mesh for the cylindrical part of the
vessel, the remaining parts of the vessel were “swept” by the Gambit mesh generation
tool.
The global quality of the produced mesh in terms of number of cells, maximum
cell squish, maximum skewness, and maximum aspect ratio are presented in Table 3-9.
Furthermore, to ascertain the quality of the generated mesh, the cell skewness was
evaluated. As shown by the mesh results of Table 3-9 only a negligible fraction of cells
(0.0025%) was of poor quality. However, by employing a new feature of Fluent 6.3.26,
the grids with cell skewness factor above 0.8 were converted to polyhedral grids.
Although this minor modification did not reduce the maximum values reported in Table
3-9, the number of cells was reduced from 884847 to 884805.

Table 3-9. Quality of the mesh produced for the Gullfaks-A separator in Gambit
environment.
Number of Cells Maximum Squish Maximum Skewness Maximum Aspect Ratio
884847 0.748182 0.895873 44.1708
Skewness of the produced mesh
Skewness Range 0-0.20 0.20-0.40 0.40-0.60 0.60-0.80 0.80-1.0
Density of Cells 79.0416% 15.4785% 3.8489% 1.6285% 0.0025%
CFD Simulation of Multiphase Separators 78

3.3.3 Modeling the Baffles and the Mesh Pad Demister

As described in section 3.1.3, the porous media model can be used to model the flow
through baffles and the demister. This section will explain how the required adjustments
were made in the Fluent 6.3.26.

3.3.3.1 Modeling the Baffles

As noted in section 3.1.3.2, the configuration and dimensions of flow-distributing baffles

are taken into account for calculation of C2 and ε parameters which are necessary for the
porous media model. However, these crucial specifications were not provided in the
original paper of Hansen et al. (1993). Fortunately, Hansen et al. (1991) have presented
useful experimental data previously. Hansen et al. (1991) performed several experiments
to obtain data in order to validate the results of their developed computer code, FLOSS,
which was used for the simulation of Gullfaks-A separator. Inside their experimental
model, with dimensions of 0.46 m × 0.46 m × 1.83 m, the flow-distributing baffle was
specified as a perforated plate with 173 holes, each with diameter of 6.4 mm and distance
between centers of 25 mm. Since the model has been used for validation of the computer
code, and the computer code was used to simulate the Gullfaks-A separator, it can be
expected that the model baffle had the same overall configuration (hole pattern) as the
baffles of Gullfaks-A separator. Therefore, assuming a thickness of 20 mm for
distribution baffles, the porosity ( ε ), and constant C2 could be calculated using Equations
3-19 to 3-23:

Calculation of ε
πd 2 π (6.4) 2
Af 4 4
From Figure 3-11, we have ε = = = ≈ 0.05 (3-19)
Ap a2 25 2

Calculation of C2
First, HolePitch was calculated based on Figure 3-11:

a = 25 mm ⇒ HolePitch = 25 2 = 35.355 mm (3-20)

CFD Simulation of Multiphase Separators 79

Figure 3-11. Geometry of the Square Hole Pattern in a Perforated Plate.

Then Re number was calculated for gas flow through the baffle holes:
Vd 2.095 × 0.0064
Re = = = 51270 (> 4000) (3-21)
ν 2.6157 × 10 − 7
δ 20
Considering that = = 3.125, and using the correlation developed by Kolodzie and
d 6 .4
Van Winkle (1957), C could be calculated as:
0.1 0.1
 d   6 .4 
C = 0.98  = 0.98  = 0.826 (3-22)
 HolePitch   35.355 
Finally, combining Equations 3-13, 3-19 and 3-22, C 2 was calculated:

1 100  2  -1
C2 =   − 1 ≈ 29240 m (3-23)
(0.826) × 0.02   
2
5

By adjusting C2 in the x-direction, the resistance of the baffles to flow in x-direction was
taken into account. As recommended in the Fluent 6.3 user manual, for convergence
purposes, a factor of 1000 is used for the constant C2 in the y and z directions if the
resistance to flow in y and z directions through the baffles is much higher than that in the
x-direction. Therefore, C 2 was set to 2.924 × 107 m-1 for flow in y and z directions
through baffles.
CFD Simulation of Multiphase Separators 80

3.3.3.2 Modeling the Wire Mesh Demister

As noted in section 3.1.3.3, constants D and C2 in the x-direction should be calculated for
simulation of the flow through the wire mesh demister using the porous media model.
However, the specifications of the demister were not provided in the original paper of
Hansen et al. (1993). Therefore, the most commonly used wire mesh properties were
assumed for calculation of these constants. A wire mesh pad with a thickness of 0.15 m is
commonly used in separators (Walas, 1990; Lyons and Plisga, 2005; Coker, 2007), hence
a type E as specified by Helsør and Svendsen (2007) was selected for simulation
purposes. Specifications of the selected mesh pad as well as correlating parameters were
taken from Helsør and Svendsen (2007) and are presented in Table 3-10.
Based on the parameters reported in Table 3-10 and using Equation 3-15, the
constants D and C2 in the x-direction were calculated as follows: D = 3.85 × 106 m-2; C 2 =
126 m-1. Again, as was the case when adjusting parameters in y and z directions for the
baffles, a factor of 1000 was used for constants D and C2 in the y and z directions: D =
3.85 × 109 m-2; C 2 = 1.26 × 105 m-1.

3.3.3.3 Developed Model for Physical Verification

Although the best parameter estimates, based on experience, were made while modeling
the baffles and demister via the porous media model in Fluent, the upper section of
separator, in which the internals were installed, was simulated to make certain that the
pressure and velocity profiles were of an expected shape9. The steps for mesh generation
are the same as those previously described for the whole separator.

Table 3-10. The physical properties and correlation parameters given for wire mesh pad
of type E (Helsør and Svendsen, 2007).
Thickness Wire Mesh Mesh Specific Porosity α C
(m) Diameter Density Surface Area (%) (m2) (m-1)
3 2 3
(mm) (kg/m ) (m /m )
0.15 0.27 186.9 345 97.7 2.6 × 10-7 63

9
The results will be presented in Chapter Four.
CFD Simulation of Multiphase Separators 81

Inlet gas velocity and outlet pressure of 6870 kPa were input as the boundary conditions,
and the SIMPLE method with the following settings was used as the solver:

Discretization Method for Pressure: Standard

Discretization Method for Momentum: Second Order Upwind
Under-Relaxation Factor for Pressure = 0.7
Under-Relaxation Factor for Density = 0.7
Under-Relaxation Factor for Body Force = 0.7
Under-Relaxation Factor for Momentum = 0.0001
Under-Relaxation Factor for Turbulent Groups (Kinetic Energy,
Dissipation Rate, and Viscosity) = 0.7

3.3.4 Definition of Droplet Size Distribution

In order to model the dispersion of oil and water droplets in the fluid flow domain, the
specification of particle size distribution is a key step. Any relevant experimental data
would have been very useful. However, this crucial information was not given in the
paper of Hansen et al. (1993). Thus, a reliable method was required for prediction of
particle size distribution for oil and water droplets entering the separator. There are
numerous research studies that predict the size distribution of fluid dispersions. However,
most of them have focused on prediction of maximum stable droplet size. Because, the
other necessary size distribution parameters such as spread parameter, minimum and
mean droplet size can be estimated based on the predicted (or measured) maximum stable
droplet size and the nature of the fluid phases. In the present research project, the Rosin-
Rammler (1933) particle size distribution function (Equation 3-16) has been used. In
Equation 3-16, the volume mean diameter, d , can be estimated from maximum droplet
diameter, dmax, via Equation 3-24 (Green and Perry, 2008):

d = 0.4 d max (3-24)

CFD Simulation of Multiphase Separators 82

Furthermore, to specify the spread parameter, n, for the Gullfaks-A dispersions, two
experimental studies, performed by Karabelas (1978) and Angeli and Hewitt (2000),
were used. The experiments of Karabelas were carried out with kerosene ( ρ = 798
kg/m3, µ = 0.00182 Pa.s) and a more viscous transformer oil ( ρ = 892 kg/m3, µ ≅
0.0156 Pa.s) as continuous phases and water as dispersed phase. The experiments of
Angeli and Hewitt were performed with both water and the oil ( ρ = 801 kg/m3, µ =
0.0016 Pa.s) as dispersed and/or continuous phases. The experimental distributions of
Angeli and Hewitt produced a value between 2.1 and 2.8 for the Rosin-Rammler spread
parameter. This result agrees with the values of 2.13 to 3.30 reported by Karabelas (1978)
for water dispersed in two different oils. As one of the most interesting experimental
results, Karabelas (1978) emphasized that the spread parameter can be assumed to be
constant and close to their measured average value while dealing with oil in water or
water in oil dispersions. Therefore, the arithmetic average value of 2.6, as reported by
Karabelas (1978), was used while setting the particle size distribution in this study.
The next step involved finding a reliable method for prediction of maximum
stable oil and water droplet sizes. As noted, many methods have been presented. The
fundamental theoretical work in the field of droplet dispersions in the turbulent flow was
conducted independently by Kolmogorov (1949) and Hinze (1955). They assumed that
the maximum stable droplet or bubble size, dmax, could be determined by the balance
between the turbulent pressure fluctuations, tending to deform or break the droplet or
bubble, and the surface tension force resisting any deformation. In their developed
theory, the ratio of these forces was included through the critical Webber number,
Equation 3-25:

τ
Wecrit = (3-25)
σ d max

where τ is dynamic pressure fluctuations in Pa, and σ is surface tension in N/m.

Equation 3-25 can be combined with the equations for turbulent fluid flow in a pipeline
CFD Simulation of Multiphase Separators 83

to calculate the maximum stable droplet or bubble size. In this regard, Equation 3-26
calculates the dynamic pressure fluctuations:

τ = 2 ρ c (ψd max )2 / 3 (3-26)

where ψ is energy dissipation rate per unit mass in W/kg which can be calculated by
Equation 3-27:

3
2 fVc
ψ = (3-27)
D

where f is friction factor, Vc is superficial velocity of continuous phase in m/s, and D is

inside diameter of pipe in m. The friction factor can be calculated using the Blasius
equation, Equation 3-28:

f = 0.079 Re −0.25 (3-28)

By substitution of Equations 3-26 to 3-28 into Equation 3-25, d max can then be estimated
using Equation 3-29:

0.6  σ 0.6  D 0.5 

d max = 1.38(We crit )  0.5 0.1  1.1  (3-29)
 ρ µ  V 
 c c  c 

where ρ c and µ c are the continuous phase density in kg/m3 and viscosity in Pa.s,
respectively.
CFD Simulation of Multiphase Separators 84

The other important theory for maximum stable bubble size was developed by Levich
(1962). He assumed that the maximum stable droplet or bubble size d max , could be
determined by the balance between the internal pressure of the droplet or bubble and the
capillary pressure of the deformed droplet or bubble. In his developed theory, the ratio of
these forces was taken into account by the critical Webber number in a new form,
Equation 3-30:

1/ 3
τ  ρd 
′ =
Wecrit   (3-30)
σ d max  ρc 

Similar to the Kolmogorov-Hinze theory (Kolmogorov, 1949; Hinze, 1955), Equation 3-

30 can be combined with Equations 3-26 through 3-28 to calculate the maximum stable
droplet or bubble size d max , Equation 3-31:

 σ 0.6  D 0.5 
d max = 1.38(Wecrit
′ )0.6  0.3 0.2 0.1  1.1  (3-31)
 ρ ρ µ  V 
 c d c  c 

where ρ d is dispersed phase density in kg/m3, and the other parameters were defined
while deriving Equation 3-29.
In an excellent study, Hesketh et al. (1987) modified the Levich theory to develop
an equation that includes all the salient physical fluid properties required to describe
droplet or bubble size in turbulent flow. Hesketh et al. (1987) considered both the
Kolmogorov-Hinze and Levich theories and recognized that in predicting maximum
particle size for liquid-liquid and gas-liquid dispersions, only the latter gives consistent
results. The critical Webber number should depend only on the mechanism of breakup,
not on the fluid physical properties. For the liquid-liquid dispersions, the values of Wecrit
calculated from the Kolmogorov-Hinze theory are some 10% of those calculated for the
gas-liquid dispersions. This discrepancy implies that the breakup mechanism for liquid
droplets differs from that for gas bubbles. However, the empirical findings contradict this
CFD Simulation of Multiphase Separators 85

implication, and values of Wecrit for liquid-liquid and gas-liquid dispersions should be
equal from an empirical perspective. This is what is proposed by Levich theory in that
′ for both liquid-liquid and gas-liquid dispersions are close to 1.0. Based on the
Wecrit
Levich theory, Equation 3-31 includes all the variables necessary to describe turbulent
dispersion except for the dispersed phase viscosity, µ d . The viscous forces within a
droplet or bubble increase its stability, and become important when they are of the same
order of magnitude as the surface tension forces (Hesketh et al., 1987). By combining a
viscosity grouping term originally proposed by Hinze (1955) with Equation 3-31, and
evaluating the coefficients using experimental data, Hesketh et al. (1987) have developed
the following generalized equation, Equation 3-32:

 σ 0.6  D 0.5 
d max = 1.38 0.3 0.2 0.1  1.1  ×
 ρ ρ µ  V 
 c d c  c 

( )
0.6
  µ µ 0.25V 2.75 ρ −0.25 D −1.25 d 1/ 3
 ρ 

× 1 + 0.5975 d c c c max
 c 
(3-32)
  σ  ρ d 


Note that d max should be calculated from Equation 3-32 in an iterative manner, since

d max is also present in the right hand side of Equation 3-32.

Although estimation of d max using Equation 3-32 seems a bit tedious (yet not
difficult if common mathematical software is used), its strong theoretical background as
well as its very satisfactory representation of empirical data provided sufficient
confidence in using this method for prediction of the maximum droplet size in this study.
Note, Hesketh et al. (1987) showed that this approach provided excellent results when
dealing with experimental data that included a broad range of physical properties: surface
tension of 0.005 to 0.072 N/m, the continuous phase viscosity of 0.001 to 0.016 Pa.s, and
the dispersed phase density of 1 to 1000 kg/m3.
CFD Simulation of Multiphase Separators 86

As mentioned, a number of attempts has been made to correlate the maximum droplet
size in fluid dispersions. In order to compare the results of the selected method with those
of some other available methods, in addition to the Hinze (1955) method that has been
described in detail, four other correlations are as follows:

1- Sleicher (1962) correlation:

 σ 1.5   µ Vc  
0.7

d max = 38  1 + 0.7  d

  (3-33)
 ρ V 2.5 µ 0.5    σ  
 c c c 

2- Tatterson et al. (1977) correlation:

 σ 0.5 D 0.6 
d max = 0.265 0.4 0.9 0.1  (3-34)
ρ V µ 
 c c c 

Note, the original equation, with the coefficient of 0.106 on the right hand side, was
proposed for estimation of volume median droplet size. The presented form is based
on Equation 3-24.

3- Karabelas (1978) correlation:

 σ 0.6 D 0.4 
d max = 4.0 0.6 1.2  (3-35)
ρ V 
 c c 

4- Angeli and Hewitt (2000) correlation:

4.2 × 10 −8
d max = 1.8
(3-36)
Vc f 3.13

Note, all the variables present in these correlations have been defined while developing
Equation 3-32.
CFD Simulation of Multiphase Separators 87

The maximum stable droplet sizes calculated by various methods or correlations for
dispersions of Gullfaks-A separator are reported in Table 3-11. Assuming that Equation
3-32 predicts d max for the Gullfaks-A dispersions with the best reliability, the other
approaches can be sorted (in a descending order) based on their average accuracies:
1- Tatterson et al. (1977), 2- Karabelas (1978), 3- Hinze (1955), 4- Sleicher (1962), 5-
Angeli and Hewitt (2000).
Based on the discussed procedures, all the discrete phase parameters necessary for
CFD simulation of Gullfaks-A separator have been calculated and presented in Table
3-12.

Table 3-11. The maximum stable droplet sizes calculated by various methods for
dispersions of Gullfaks-A separator.
Calculated Values Calculated Values for
Method for 1988 Condition (m) the Future Conditions (m)
Oil Drops Water Drops Oil Drops Water Drops
Hinze (1955) 0.004345 0.008071 0.003744 0.006954
Sleicher (1962) 0.01799 0.04924 0.01374 0.03632
Tatterson et al. (1977) 0.003434 0.005753 0.003071 0.005145
Karabelas (1978) 0.003615 0.006715 0.003115 0.005785
Hesketh et al. (1987) 0.002267 0.004006 0.001955 0.003452
Angeli and Hewitt (2000) 0.03857 0.03857 0.03088 0.03088

Table 3-12. The discrete phase parameters used in CFD simulation of Gullfaks-A
separator.
1988 Condition Future Condition
Discrete Phase Parameters Oil Drops Water Drops Oil Drops Water Drops
Maximum Diameter ( µm ) 2267 4000 1955 3450
Mean Diameter ( µm ) 907 1600 780 1380
-4 -3 -4
Total Mass Flow-rate ( kg / s ) 6.5 × 10 4.4 × 10 4.2 × 10 2.8 × 10-3
Number of Tracked Particles 1000
Minimum Diameter ( µm ) 100
Spread Parameter 2.6
CFD Simulation of Multiphase Separators 88

3.3.5 Setting CFD Simulator Parameters

After importing the mesh file and making the modification to it, the necessary material
properties for various phases were input. Mesh modification consisted of converting the
highly skewed grids to polyhedral grids which resulted in a minor reduction in the
number of cells. Then, since the Reynolds number was much more than the transient
value (Re = 2300) for all fluid phases, the common turbulent flow model of k − ε was
selected.
In order to set the boundary conditions for inlet, the velocity and volume fractions
of phases were set. For the gas-outlet boundary, outlet pressure and volume fractions (as
pure gas) were set while for the liquid-outlet boundaries, outlet velocities and volume
fractions (as pure liquid) were set. For setting the flow regimes in inlet and outlet nozzles,
the turbulence intensity and hydraulic diameter of flow through the nozzles were
determined. Similar to the CFD modeling of the two-phase separators, discussed in
section 3.2, the turbulence intensity in the inlet and outlet zones was estimated using
Equation 3-17. Therefore, the boundary condition inputs could be calculated based on the
separator dimensions and various phase properties. Table 3-13 provides the calculated
input values. Note, that in setting the gas outlet boundary condition, the pressure outlet
was set to a constant value of 6870 kPa.
For the dispersed liquid droplets the momentum reduction of droplets after hitting
the inside walls of vessel, deflector baffle, splash plate, and weir was set. In the current
study, it was assumed that droplets lose 90% of their momentum after coming into
contact with the solid surfaces. Thus, normal and tangent reflection coefficients were set
to be 0.10 for all the solid surfaces.

Table 3-13. The calculated boundary condition inputs used in CFD simulation of
Gullfaks-A separator.
1988 Velocity Future Velocity I (1988) I (Future)
(m/s) (m/s)
Inlet Mixture 5.8788 6.6560 2.90% 2.73%
Gas Outlet —— —— 2.28% 2.28%
Oil Outlet 4.4359 3.3293 3.35% 3.48%
Water Outlet 0.6919 2.9991 3.01% 2.51%
CFD Simulation of Multiphase Separators 89

During normal operation, the separator was half-filled with liquid (Hansen et al., 1993).
Furthermore, based on the “horizontal separator with weir” design procedure (Monnery
and Svrcek, 1994), the normal liquid level for the heavier phase is typically set to half of
the weir height. Therefore, the liquid levels for oil and water phases were set to be 1.664
m and 0.625 m, respectively. To set the position of interface between phases, the volume
fractions of phases above and below the assumed interface planes were set to the
reasonable values by the “Patching” tool of Fluent. Again, the iterations need to be
stopped regularly, and the position of interfaces should be checked and corrected, if
necessary, by “patching” the volume fractions of phases.
Based on the provided Fluent guidelines and the trend of solution convergence,
the solver parameters were set as follows which did overcome stability and convergence
problems:

Discretization Method for Pressure: Body Force Weighted

Discretization Method for Momentum and Volume Fraction: First Order Upwind
Solution Method: PISO
Under-Relaxation Factor for Pressure = 0.1
Under-Relaxation Factor for Density = 0.9
Under-Relaxation Factor for Body Force = 0.9
Under-Relaxation Factor for Momentum = 0.0005
Under-Relaxation Factor for Volume Fraction = 0.005
Under-Relaxation Factor for Turbulent Groups (Kinetic Energy,
Dissipation Rate, and Viscosity) = 0.7

3.4 Summary

The steps required for CFD simulation of multiphase separators have been presented.
First, the pertinent concepts of CFD modeling were introduced, and then the requirements
for the use of Fluent 6.3.26 for the simulation of multiphase separators were described.
Two cases, i.e. pilot-plant-scale two-phase separators and the large-scale three-phase
separator, were CFD simulated. Two approaches, Discrete Phase Model (DPM) and a
CFD Simulation of Multiphase Separators 90

combination of Volume of Fluid (VOF) and DPM, each with appropriate model
assumptions and settings have been used. The CFD guidelines recommend the DPM
approach for flow regimes in which the discrete phase is a small volume fraction (less
than 12%), while the VOF model is recommended for the simulation of immiscible
multiphase flows with specific interface surfaces. In the first case study, both the DPM
and the VOF-DPM approaches were tested. However, based on the results obtained from
this case study, only the VOF-DPM approach was used for simulation of the complex
features of three-phase separation in the large-scale separator. To implement the
combined VOF-DPM approach, having developed a CFD simulation of the overall phase
behavior of the fluid flows using the VOF model, droplets of oil and water were injected
through the inlet nozzle to be tracked by DPM model. Consequently, the VOF model was
used to provide the overall picture of fluid flow behavior in the separators, and the DPM
model was used to track the liquid droplets injected, moving the simulation toward a
realistic situation.
The distribution baffles and mist eliminator were modeled using the Porous
Media Model of Fluent. The use of the Porous Media Model required utilization of the
available specifications and design information for three-phase separators.
By exploiting the available theoretical approaches and experimental correlations,
a useful methodology for estimation of particle size distribution, which is necessary for
implementing DPM approach, has been developed.
CFD Simulation of Multiphase Separators 91

Chapter Four: Results and Discussion

The CFD simulation results for the pilot-plant scale two-phase separators and the large-
scale three-phase separator will be presented. For the two-phase separators, two
simulation approaches, DPM and VOF-DPM, were used. Based on the results for the
two-phase separators, only the VOF-DPM modeling approach was implemented for
simulation of the three-phase separator.

4.1 Simulation of the Pilot-Plant-Scale Two-Phase Separators

Two approaches were implemented for the simulation of the pilot-plant-scale two-phase
separators. The first simulation used only the DPM multiphase modeling of Newton et al.
(2007), and only the vapor-liquid compartment of the separators was simulated. In this
approach, the gas-liquid interface was assumed as a frictionless wall which trapped the
droplets coming into contact with it. In the second modeling method a combination of
DPM and VOF multiphase models within Fluent 6.3.26 was used. This CFD model was
based on the type of involved phase separation process and the characteristics of
multiphase models of Fluent.

4.1.1 Results of DPM Only Model

In order to obtain the incipient velocity and separation efficiency diagram for each case,
the inlet velocity was gradually increased by 0.1 m / s increments, and the flow regimes
in inlet and outlet nozzles were set. The minimum velocity at which a minor fraction of
the droplets (about 1wt%) were carried over in the gas phase was reported as the incipient
velocity. After determining the incipient velocity, the velocity was gradually increased to
5 m/s, generating the separation efficiency versus velocity data. This procedure was used
for the twelve cases that included four separators operating at three different pressures.
Three different mean droplet sizes were also tested for each case. After setting all the
CFD parameters, the number of iterations required for the continuous-phase solution
convergence in vessels A, B, C, and D was 600, 500, 250, and 300, respectively. The
CPU time for each iteration on a Pentium D (3.20 GHz) and 2.00 GB of RAM PC was 4,
CFD Simulation of Multiphase Separators 92

2, 1.5, and 1 s for vessels A, B, C, and D, respectively. Therefore, a relatively short PC

run-time of from 5 to 40 min was found for the solution of continuous-phase flow
regimes. However, a further PC run-time of 1 to 2 h was required for simulation of
discrete phase (liquid droplets) interaction with continuous gas phase for each case study.

4.1.1.1 Fluid Flow Profiles

The pressure and velocity profiles for the continuous phase at the incipient velocity for
one of the horizontal separators, vessel C, and the vertical separator are presented in
Figures 4-1 to 4-6. The attached CD contains the profiles for all the case studies. The
simulated pressure profiles do indicate that the separators can be considered to operate at
a constant pressure with some negligible low pressure zones in the vicinity of the gas
outlet.

4.1.1.2 Droplet Coalescence and Breakup

Coalescence and breakup of droplets was also modeled and required the development of
software, written in C++, that took data from the screen report of Fluent and calculated
the number of droplet collisions and breakups. The results of this calculation showed that
droplet coalescence at a low rate of about 0.2% was taking place. The results did show
that high operating pressures and low vessel aspect ratios increased droplet coalescence.
CFD Simulation of Multiphase Separators 93

(a)

(b)
Figure 4-1. Fluid Flow Profiles in the Middle of Vessel C Operating at 70 kPa (DPM
Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 94

(a)

(b)
Figure 4-2. Fluid Flow Profiles in the Middle of Vessel C Operating at 700 kPa (DPM
Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 95

(a)

(b)
Figure 4-3. Fluid Flow Profiles in the Middle of Vessel C Operating at 2760 kPa (DPM
Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 96

(a) (b)
Figure 4-4. Fluid Flow Profiles in the Middle of Vessel D Operating at 70 kPa (DPM
Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 97

(a) (b)
Figure 4-5. Fluid Flow Profiles in the Middle of Vessel D Operating at 700 kPa (DPM
Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 98

(a) (b)
Figure 4-6. Fluid Flow Profiles in the Middle of Vessel D Operating at 2760 kPa (DPM
Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 99

Unlike droplet coalescence, the droplet breakup was noticeable and did show a
significant variation at different operating conditions. The droplet breakup variations
versus continuous phase velocity for all case studies have been presented in Appendix B.
These CFD data show, as expected, that higher velocities and pressures usually increase
the number of droplet breakups.

4.1.1.3 Separation Efficiencies

Table 4-1 compares the predicted incipient velocities to the experimental values. In Table
4-1, only few of the simulated values shown in italic letters are not in good agreement
with the measured experimental values. In other words, for most of the case studies, there
is excellent agreement between simulated incipient velocities and the experimental data.

For the droplet mean diameter of d = 150 µm , only one case study (separator B at 70
kPa) does not provide a good match. Therefore, one can conclude that the industry

assumed mean diameter of d = 150 µm is an acceptable value for the estimation of

separator incipient velocity.

Table 4-1. The simulated and experimental values for incipient velocities (DPM
approach).
Incipient Velocity (m/s)
Separator Pressure Simulated
(kPa) Experimental
d = 150 µm d = 550 µm d = 1000 µm
70 1.7 1.7 1.7 ---
A 700 1.0 0.8 0.8 1.0 -1.7
2760 0.8 0.6 0.4 0.4-0.8
70 1.5 1.4 1.4 2.2-2.6
B 700 0.8 0.6 0.6 0.4-1.4
2760 0.4 0.4 0.4 0.1-0.5
70 2.2 1.5 1.6 2.2-2.6
C 700 0.8 0.6 0.5 0.4-1.4
2760 0.4 0.4 0.4 0.1-0.5
70 1.5 1.7 1.7 1.3-2.0
D 700 0.8 0.9 0.9 0.7-1.7
2760 0.6 0.6 0.6 0.4-0.8
CFD Simulation of Multiphase Separators 100

Although this simple modeling approach was successful in predicting the separator
incipient velocities, the resultant plots for separation efficiency versus gas velocity were
not correct. A complete set of separation efficiency plots for all case studies at three
different mean droplet sizes are included on the supplementary CD. As an example,
Figure 4-7 shows such a plot for vessel C operating at 70 kPa and indicates that by
gradually increasing the gas velocity, two zones with separation efficiencies less than
100% appear: one at about 2.2 m/s and the other at 2.6 m/s. In fact, based on the
simulation results for some cases, there could be more than one of these zones, which
may be referred as “carryover rush” zones. This behavior, however, is opposite to
practical experience which would suggest that after reaching the incipient velocity, by
increasing the velocity of continuous phase, the separation efficiency drops sharply to
lower values. The other anomaly shown in this case study was a perfect separation
efficiency at very high velocities (> 3.0 m / s ).

Figure 4-7. The Separation Efficiency versus Gas Velocity for Separator C at P = 70
kPa Assuming d = 150 µm (DPM Approach).
CFD Simulation of Multiphase Separators 101

These anomalies occurred in several case studies in which the DPM only model proposed
by Newton et al. (2007) was used in the separator simulations. Thus, it can be concluded
that modeling only the gas phase compartment of a separator does not capture the phase
separation phenomenon even though the model did predict the empirical incipient
velocities. The shortcoming of this very simple modeling approach is that the interactions
between the liquid and gas phases and, more significantly, the dynamic interaction
between the liquid droplets and continuous liquid phase are totally neglected.

4.1.2 Results of VOF-DPM Model

The realistic VOF-DPM model did include the continuous liquid phase in the
calculations. The equations governing the injected discrete phase (liquid droplets) and
coexisting continuous phases (gas and liquid phases) were solved simultaneously. This
model required significantly more computational time for the calculation of the
interactions among multiple discrete and continuous phases.
The number of iterations required for continuous-phase solution convergence for
all the case studies was about 400. The CPU time for each iteration on a Pentium D (3.20
GHz) and 2.00 GB of RAM PC was 8, 5, 3.5, and 1 s for vessels A, B, C, and D,
respectively. Therefore, a relatively short PC run-time of 7 to 55 min was sufficient for
the solution of continuous-phase flow regime, but a further PC run-time of about 1 to 3 h
was required for the simulation of interactions among multiple discrete and continuous
phases.

4.1.2.1 Fluid Flow Profiles

In order to provide a macroscopic scale picture of phase separation process, the fluid flow
profiles were used as the simulation output. The volume fraction contours for the case
study separators operating at various pressures were essentially the same. Thus, selecting
the intermediate pressure of 700 kPa for illustration purposes, Figure 4-8 shows volume
fraction contours for all the separators. As represented, the coexisting liquid and gas
phases have been separated from each other by a relatively clear interface.
CFD Simulation of Multiphase Separators 102

(c)

(b)
(d)

(a)
Figure 4-8. Profiles of Volume Fraction in the Middle of Two-Phase Separators; (a)
Vessel A, (b) Vessel B, (c) Vessel C, and (d) Vessel D.
CFD Simulation of Multiphase Separators 103

The pressure and velocity profiles corresponding to the incipient velocity for one of
horizontal separators, vessel B, and the vertical separator are presented in Figures 4-9 to
4-14. The supplementary CD includes a complete set of profile plots for all the case
studies. Based on the separator pressure profiles, all the separators can be assumed to
operate at a constant pressure.

4.1.2.2 Droplet Coalescence and Breakup

The droplet coalescence and breakup were again modeled, and the simulation results
confirmed that droplet coalescence at a rate of less than 1% was not a common
phenomenon. A general conclusion for droplet coalescence could not be reached based
on the simulation case study results. However, droplet breakup was a common
phenomenon and did show significant variations to operating conditions. Based on the
simulation results, higher velocities usually intensified the number of droplet breakups in
horizontal separators, and in vertical separators, higher pressures stabilized the number of
breakups to a constant rate of about 20%. Additional details on the droplet breakup
variations versus operating conditions have been presented in Appendix B.

4.1.2.3 Separation Efficiencies

In order to obtain the incipient velocities and separation efficiency plots, the inlet velocity
was gradually increased by 0.2 m / s increments to a maximum of 4 m/s. Data output
surfaces were defined in the Fluent environment to enable collection of particle
characteristic data as the injected droplets moved through the separator. A software
program in C++ was developed to calculate the separation efficiencies based on the mass
distribution of liquid droplets between the gas and liquid outlets.
CFD Simulation of Multiphase Separators 104

(a)

(b)
Figure 4-9. Fluid Flow Profiles in the Middle of Vessel B Operating at 70 kPa (VOF-
DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 105

(a)

(b)
Figure 4-10. Fluid Flow Profiles in the Middle of Vessel B Operating at 700 kPa (VOF-
DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 106

(a)

(b)
Figure 4-11. Fluid Flow Profiles in the Middle of Vessel B Operating at 2760 kPa (VOF-
DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 107

(a) (b)
Figure 4-12. Fluid Flow Profiles in the Middle of Vessel D Operating at 70 kPa (VOF-
DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 108

(a) (b)
Figure 4-13. Fluid Flow Profiles in the Middle of Vessel D Operating at 700 kPa (VOF-
DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 109

(a) (b)
Figure 4-14. Fluid Flow Profiles in the Middle of Vessel D Operating at 2760 kPa (VOF-
DPM Approach); (a) Contours of Pressure (Pa), and (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 110

The minimum velocity, at which a minor mass fraction of droplets (1wt%) were carried
over in the gas phase, was assumed as the incipient velocity. The other characteristic of
the incipient velocity, as also observed by Monnery and Svrcek (2000) in the pilot plant
operation, was that after the incipient velocity, the carryover rate increased sharply and
the separation efficiency dropped rapidly. After determining the incipient velocity, the
velocity was gradually increased up to 4 m/s to produce the data required for separation
efficiency versus velocity plots. The predicted incipient velocities and corresponding
experimental values have been presented in Table 4-2. The data of Table 4-2 does show
that only two cases, the values shown with italic letters, are not a good match. It is
interesting to note that the predicted incipient velocity for one of these cases, separator C
at 70 kPa, using the simple DPM model does match the experimental data quite well.
Therefore, by adjusting the current model assumptions toward the DPM model
assumptions, i.e. reducing the liquid level, the simulation results can be made to match
the experimental data for this case.
The complete set of separation efficiency plots for all the case studies (twelve
case studies) have been presented on the supplementary CD. However, for illustration
purposes, the separation efficiency diagrams for separator B at 700 kPa and separator D
at 70 kPa are shown in Figure 4-15. From a practical point of view, the obtained data
curves are reasonable in that the separation efficiencies have dropped sharply after the
predicted incipient velocities are exceeded. Moreover, the unacceptable results produced
by the DPM simple model, i.e. perfect separation at velocities higher than the incipient
velocity, have been completely eliminated. In summary, there is excellent agreement
between CFD simulated phase separation behavior and the experimental
data/observations for almost all of the case studies.
CFD Simulation of Multiphase Separators 111

(a)

(b)
Figure 4-15. The Separation Efficiency versus Gas Velocity (VOF-DPM Approach); (a)
Separator B at P = 700 kPa, and (b) Separator D at P = 70 kPa.
CFD Simulation of Multiphase Separators 112

Table 4-2. The simulated and experimental values for incipient velocities (VOF-DPM
approach).
Incipient Velocity (m/s)
Separator Pressure (kPa) Simulated Experimental
70 1.6 ---
A 700 1.0 1.0 -1.7
2760 0.6 0.4-0.8
70 1.2 2.2-2.6
B 700 0.8 0.4-1.4
2760 0.5 0.1-0.5
70 1.2 2.2-2.6
C 700 0.6 0.4-1.4
2760 0.5 0.1-0.5
70 1.5 1.3-2.0
D 700 1.0 0.7-1.7
2760 1.0 0.4-0.8

4.1.2.4 Droplet Size Distribution in Gas Outlet

Obtaining a measure of particle size distribution for the droplets in the gas outlet can also
be used as a key factor in the separator performance analysis. In an efficiently designed
separator, the particle size distribution in the gas-outlet zone should be very narrow and
consist only of very small droplets which can be further separated by a demister. With a
properly designed demister, all liquid droplets in the range of 10 to 100 µm can be
removed (Smith, 1987).
A software program written in C++ was developed to use the Rosin-Rammler size
distribution model to calculate droplet size distribution in the gas-outlet. The results did
show a variation with the gas velocity when the incipient velocity was gradually
increased up to a velocity resulting in very low separation efficiency. However, only the
average values of the Rosin-Rammler analysis, reported in Table 4-3, were used in
assessing the separator performance.
CFD Simulation of Multiphase Separators 113

Table 4-3. The average values of Rosin-Rammler analysis on the droplets emerging from
the gas outlet (VOF-DPM approach).
Average Values
Separator Pressure dmin dmax d n
(kPa) ( µm ) ( µm ) ( µm )
70 1.27 26.11 12.11 8.34
A 700 1.13 29.12 12.00 7.84
2760 1.79 33.03 13.84 7.13
70 3.21 52.44 27.84 6.28
B 700 5.28 49.95 30.76 7.77
2760 3.76 34.89 16.98 6.29
70 1.82 37.31 18.65 7.72
C 700 1.45 35.64 15.66 7.27
2760 1.78 26.04 11.97 5.65
70 7.94 349.64 82.86 4.21
D 700 14.88 421.79 143.39 4.60
2760 13.79 273.67 99.15 5.06

The data of Table 4-3 for horizontal separators do show that the particle size distribution
was always very narrow with corresponding spread parameter of about 7.1. Therefore,
the emerging droplets were totally dominated by droplets of the mean diameter size.
Since the average values for the mean diameter, shown in Table 4-3, are always greater
than 10 µm , a properly designed demister at reasonable operating condition would
separate these fine droplets. Provided that the load/amount of carried over fine droplets is
within the capacity of demister, high separation efficiencies may be established at
velocities higher than incipient velocity. As reported in Table 4-3, the average values of
the spread parameter are generally lower for the vertical separator than those for
horizontal separators, and the particle size distribution is also wider for the vertical
separator. The data does show that both very small (<15 µm ) and very large droplets
(>>100 µm ) are carried over in the vertical separator. These droplets cannot be
efficiently removed by a demister and would flood it. Therefore, from a CFD point of
view, separation performance of a demister for a vertical separator would be lower than
that for a horizontal separator.
CFD Simulation of Multiphase Separators 114

4.2 Simulation of a Large-Scale Three-Phase Separator

Based on the CFD simulation results obtained for the tested two-phase models and their
comparison to the pilot-plant scale separator data, the combined VOF-DPM model of
Fluent was used.

4.2.1 Physical Validation of the Developed Models for the Baffles and Demister

To physically verify the models (the Porous Media model in Fluent) used to simulate the
baffles and demister, the pressure and velocity profiles in upper section of separator, in
which these internals were installed, were studied. After preparing the grid system and
setting all the required CFD parameters (Chapter Three), the solution converged in some
3000 iterations with a PC run-time of 10 s/iteration. The resultant pressure and velocity
profiles are shown in Figure 4-16. Figure 4-16a does show that a low pressure drop for
the gas flow through the distribution baffles and a very low pressure drop for the gas flow
through the mesh demister were predicted. Therefore, the assumed geometrical
specifications for baffles were validated, and the separator was shown to be operating at a
constant pressure of 6870 kPa. Furthermore, since knitted wire mesh demisters are often
causing a very low pressure drop, in the order of 250 Pa (Coker, 2007), the resultant
negligible pressure drop is realistic. The velocity vectors shown in Figure 4-16b are
reasonable and further demonstrate that the assumptions made while adjusting the porous
media parameters for the distribution baffles and the wire mesh demister are realistic.
Separation performance of the CFD model developed for wire mesh demister is detailed
in Appendix C.
CFD Simulation of Multiphase Separators 115

(a)

(b)
Figure 4-16. Profiles of Gas Flow in the Upper Half of the Gullfaks-A Separator for
1988; (a) Contours of Pressure (Pa), (b) Vectors of Velocity (m/s).
CFD Simulation of Multiphase Separators 116

4.2.2 Preliminary Considerations

Prior to presenting the results of this case study, it is important to highlight the most
important modifications of this study when compared with the original research project as
presented by Hansen et al. (1993) at the 6th International Conference on Multiphase
Production. In their CFD simulation of this multiphase separator, Hansen et al. (1993)
made a number of simplifying assumptions as follows:

1- Fluid flow analysis was confined to the inlet zone and the bulk liquid flow
zone. Therefore, the interaction between multiple zones was ignored.
2- In both zones, the flow was considered to be symmetrical around the vertical
plane in the middle of the separator (xz-plane), thus, only half of each zone
volume was modeled. Apparently, this is a questionable assumption,
particularly, when no plug flow regime was established as shown by their
results.
3- The inlet section of the separator in which all three phases are present was
modeled as a two-phase gas-liquid flow, and the results did provide the
boundary conditions for the distributed velocity field in the liquid pool. A
two-phase simulation of a three-phase zone will reduce the accuracy of the
results for the inlet zone and the incorrect boundary condition will also
decrease the accuracy of downstream bulk liquid solution flow.
4- The grid systems used for numerical simulations of the inlet zone and the bulk
liquid zone were 11 × 8 × 15 and 23 × 4 × 5, respectively. Given the vessel
dimensions, the generated grid systems are rather coarse. Note, if the
assumed grid system was developed such as to cover the whole vessel, the
generated grid would include some 4480 cells which is 0.51% of the
generated mesh cells of this research project CFD model. So, the grid system
of the current study is almost 200 times finer than that used in the original
work of Hansen et al. (1993).
CFD Simulation of Multiphase Separators 117

The other major improvement of this research project, when compared to not only
Hansen et al. (1993) but also all the previous projects on the CFD-based study of
separator performance, is the direct and quantitative evaluation of the separator
efficiency. For this purpose, some data recording planes were defined to record the
characteristics of the droplets passing through them. The recording surfaces of interest
were two vertical yz-planes at the start and at the end of the gravity separation section,
the gas outlet, the oil outlet, and the water outlet. Again, computer codes were developed
to analyze the databases provided by the recording planes or the screen output of Fluent:

1- “Separation Efficiency Analyzer” dealt with databases recorded by the capturing

surfaces at gas, oil and water outlets to calculate the separation efficiencies. The
calculations were based on mass distribution of droplets among gas, oil and water
outlets.
2- “Rosin-Rammler Analyzer” dealt with databases recorded by all the five
capturing surfaces to provide the corresponding particle size distributions based
on Rosin-Rammler equation.
3- “Coalescence-Breakup Analyzer” dealt with screen output of Fluent (a large
transcript file) to calculate the number of droplet coalescence and breakup.

Having prepared the grid system and set all the CFD parameters, some 4000 iterations
were required for the continuous-phase solution convergence. Note, the iterative process
was stopped regularly to check the interface levels and correct them if necessary (Newton
et al., 2007). Each iteration took about 22 s on a Pentium D (3.20 GHz) and 2.00 GB of
RAM PC. Therefore, a PC run-time of about 24 h was required for solution of
continuous-phase fluid flows with a further PC run-time of about 3 h required for the
simulation of interactions among the dispersed droplets and the continuous phases.
CFD Simulation of Multiphase Separators 118

4.2.3 Fluid Flow Profiles

The results of the CFD simulation of the three-phase separator in terms of fluid flow
profiles are presented. The profiles of the original work of Hansen et al. (1993) have also
been included (in monochrome) to make the comparison more convenient. Figure 4-17
provides the velocity vectors in the inlet zone for the 1988 production level and the
predicted upcoming production conditions (referred as “future condition”). Note that the
velocity vectors have not been presented for the inlet nozzle in the original work. The
velocity vectors indicate the turn of inlet flow after striking the deflector baffle and the
role of splash plate in stabilizing the reflected flow. Under the splash plate, an almost
stagnant fluid zone can be seen. The results are comparable with those obtained in the
original study of Hansen et al. (1993).
Figure 4-18 shows the velocity contours in the x-direction over the splash plate
for 1988 and for the future production condition. Although the range of velocities differs
from that in the original work, the overall features are similar, in that no plug flow regime
is established for the inlet zone. On the splash plate, there are back flows even though the
velocity distribution is more homogeneous (less back flows) for the future production
condition. In the original work, back flows are predicted to occur only for the 1988
production condition.
Velocity vectors on four parallel horizontal planes (z = 0.252, 0.627, 0.999, and
1.449 m) and four parallel vertical planes (y = 1.4, 1.0, 0.6, and 0.2 m) for both 1988 and
the future production condition were obtained and compared with the corresponding
profiles from the original study. While Figures 4-19 and 4-20 provide some of these
profiles for illustration purposes, a complete set of the profiles is on the supplementary
CD. Using the original work profiles, Hansen et al. (1993) addressed the rotational flow
regimes established between any two internals. In the present study, however, the large-
scale fluid flow circulations were not present even though some minor flow circulations
or back flows have been predicted. As was noted during the solution convergence trend,
if the over-relaxation parameters are not adjusted correctly or the correct solver is not
selected, large rotational flow patterns can be produced and the solution fluctuates
without approaching a realistic converged solution.
CFD Simulation of Multiphase Separators 119

(a)

(b)
Figure 4-17. Vectors of Velocity (m/s) in the Inlet Zone of the Gullfaks-A Separator; (a)
1988, and (b) the Future Condition.
CFD Simulation of Multiphase Separators 120

(a)

(b)
Figure 4-18. Contours of Velocity in the x-direction (m/s) at the End of Splash Plate of
the Gullfaks-A Separator; (a) 1988, and (b) the Future Condition.
CFD Simulation of Multiphase Separators 121

(a)

(b)
Figure 4-19. Vectors of Velocity (m/s) for the Gullfaks-A Separator at 1988 Conditions
at (a) z = 0.999 m, and (b) z = 1.449 m.
CFD Simulation of Multiphase Separators 122

(a)

(b)
Figure 4-20. Vectors of Velocity (m/s) for the Gullfaks-A Separator at the Future
Conditions at (a) y = 1.4 m, and (b) y = 1.0 m.
CFD Simulation of Multiphase Separators 123

In addition to this issue, the major simplifying assumptions used in the original work are
another probable source of the inaccuracy. So, it would seem that the large flow
circulations predicted by Hansen et al. (1993) are a result of poor adjustments or
assumptions, e.g. the poor setting of the over-relaxation parameters. Furthermore, the
recent CFD-based study by Lu et al. (2007) does show that the distribution baffles
generally improve the quality of liquid flow distribution in the vessel, break the large-
scale circulations into smaller ones, and reduce the short-circuiting flow streams.
The pressure, velocity and density profiles for the fluid flows on the central xz-
plane of the separator are shown in Figures 4-21 to 4-23. As was expected from
simulation of upper section of the separator, the pressure drops assigned to the baffles
and demister are small (reasonable) and the velocity vectors are also realistic. However,
based on the simulated density contours, Figure 4-23, it would seem that the separator at
both of operating conditions (particularly, in the future production condition) may suffer
from foam and emulsion problems. The distortion of interfaces in the inlet and outlet
zones does indicate a potential for foam and emulsion problems. The other detectable
problem is the flow behavior near the water outlet predicted for the future production
condition (Figure 4-23b). With the large increase in the produced water flow-rate, the
water phase should be pumped from the vessel at much higher rates. Therefore, as
indicated by the present CFD simulations, there is an increasing tendency for the oil
phase to be pushed towards water outlet. This, at least, will increase the risk of turbulence
in the water outlet zone and may lead to mixing of the phases. In order to overcome this
problem, one should minimize the risk of mixing liquid phases by improving the vessel
design. For instance, redesign of the vessel as a “bucket and weir” design may be a
plausible solution.
CFD Simulation of Multiphase Separators 124

(a)

(b)
Figure 4-21. Contours of Pressure (Pa) in the Middle of the Gullfaks-A Separator for (a)
1988, and (b) the Future Condition.
CFD Simulation of Multiphase Separators 125

(a)

(b)
Figure 4-22. Vectors of Velocity (m/s) in the Middle of the Gullfaks-A Separator for (a)
1988, and (b) the Future Condition.
CFD Simulation of Multiphase Separators 126

(a)

(b)
3
Figure 4-23. Contours of Density (kg/m ) in the Middle of the Gullfaks-A Separator for
(a) 1988, and (b) the Future Condition.
CFD Simulation of Multiphase Separators 127

4.2.4 Separation Efficiencies

The analysis of the oil and water droplets exiting at the separator outlets resulted in
predicting a total separation efficiency of 98.0% at the 1988 production conditions. The
result is based on mass distribution of injected oil and water droplets among the separator
outlets. This mass distribution analysis indicated that 100% of oil droplets and 96.9% of
water droplets were separated and came out through their corresponding outlets. Note,
there were no droplets present in the gas phase outlet, hence all the injected droplets came
out in either the oil outlet or the water outlet.
As expected from practical field experience and also shown in the density
contours of Figure 4-23b, with an increase in the produced water flow-rate in the
upcoming years, the separation efficiency for oil droplets was predicted to decrease to
1.3%. Again, there was no predicted carry over in the gas outlet, thus all the injected
droplets came out with either the oil outlet or the water outlet. So, gas-liquid separation
efficiency was still 100%. The very low separation efficiency for oil droplets indicates
that water phase does not provide enough residence time for oil droplets to rise up and
join oil phase, and almost all of oil droplets are carried by water phase to the water outlet.
Although the separation efficiency was calculated to be 100% for water droplets, because
of the difficulty in separating the oil droplets, the total separation efficiency has been
reduced to 70.4%.
The result of droplet size distribution analysis on the selected surfaces of the
separator is shown in Table 4-4. As Table 4-4 shows, compared with the initially defined
droplet size distribution (Table 3-12), droplets have become smaller. The reason is that
droplet breakup occurs when the injected droplets strike the deflector baffle. Therefore,
the volume median diameter has decreased to 70% of its initial value for oil droplets, and
to 67% of its initial value for water droplets for the 1988 production condition. With
upcoming increase in the inlet water flow-rate, these values change to 54% for oil
droplets and to 46% for water droplets. Table 4-4 also shows that droplet size distribution
before and after gravity separation zone is almost the same. This implies that there should
be no further breakup while droplets are traveling through the main part of the separator;
hence, droplet size distribution remains essentially constant.
CFD Simulation of Multiphase Separators 128

The CFD simulation results show that for the 2000 injected droplets, the number of
breakups was predicted to be 1590 for 1988 condition and 1543 for the future condition.
Similar to the case of two-phase separators, free coalescence of droplets was not a
common phenomenon in the three-phase separator. Droplet coalescence may happen at a
very low rate of about 0.1%, without any noticeable trend.

Table 4-4. The droplet size distribution in important zones of the Gullfaks-A separator.
Before After
Discrete Oil Water
Gravity Gravity
Phase Outlet Outlet
Separation Separation
Parameters
Zone Zone
dmin ( µm ) 26 26 26 –––––––
Oil dmax ( µm ) 1848 1830 1209 –––––––
Droplets d ( µm ) 640 627 419 –––––––
1988
Production n 3.18 3.20 3.80 –––––––
Condition dmin ( µm ) 85 85 93 90
Water dmax ( µm ) 2424 2338 438 2315
Droplets d ( µm ) 1077 1009 257 1008
n 2.81 2.79 6.79 3.05
dmin ( µm ) 37 37 47 25
Oil dmax ( µm ) 1009 1009 219 855
Droplets d ( µm ) 422 423 154 300
Future
Production n 4.42 4.43 6.55 3.60
Condition dmin ( µm ) 34 34 76 34
Water dmax ( µm ) 1596 1596 148 1593
Droplets d ( µm ) 643 633 132 545
n 4.11 3.55 6.60 3.67
CFD Simulation of Multiphase Separators 129

4.2.5 The Effect of Minor Modifications on the Gullfaks-A Separator

The first modification that was tested was optimizing the position of the distribution
baffles. To test this effect, all the baffles were removed from the model to see if any
tangible effect on the separator performance might be observed. With this change, the
separation efficiency was calculated as 99.5% for oil droplets and 96.8% for water
droplets at the 1988 conditions, and 0.8% for oil droplets and 100% for water droplets at
the future production conditions. Compared with the reported values for the separator
equipped with the baffles, a negligible difference is seen. On the other hand, as concluded
by Lu et al. (2007) in their recent CFD-based study, it can be expected that the baffles
will improve the quality of liquid flow distribution in the vessel and increase the
separation efficiency. Noting that the baffles were originally installed in the upper (vapor
disengagement) section of the separator, it seemed that by expanding the baffles so that
they cover the lower areas as well, some improvements might be achieved. In order to
investigate this issue, baffles in the inefficient operating separator were expanded to
cover the whole cross-sectional area. The resultant fluid flow profiles are presented on
the supplementary CD, and the droplet size distributions for all the case studies discussed
above are presented in Tables 4-5 and 4-6. The CFD simulation of droplet dispersions
within the separator resulted in separation efficiencies of 0.5% and 100% for oil and
water droplets, respectively. Therefore, it can be concluded that adjusting the
arrangement of distributing baffles cannot really improve the separation efficiency in the
Gullfaks-A separator. In fact, as emphasized by Lyons and Plisga (2005), although
properly designed internals are generally helpful in reducing liquid carryover at design
conditions, they cannot overcome separation problems in a basically inefficient design or
in an undersized separator.
CFD Simulation of Multiphase Separators 130

Table 4-5. The droplet size distribution in the important zones of the Gullfaks-A
separator without the internals (flow-distributing baffles and the mist eliminator).
Before After
Discrete Oil Water
Gravity Gravity
Phase Outlet Outlet
Separation Separation
Parameters
Zone Zone
dmin ( µm ) 58 58 19 62
Oil dmax ( µm ) 1852 1852 1843 102
Droplets d ( µm ) 702 702 435 89
1988
Production n 4.59 4.59 2.11 7.26
Condition dmin ( µm ) 56 56 149 56
Water dmax ( µm ) 2106 2106 495 2106
Droplets d ( µm ) 1073 969 260 968
n 2.95 2.98 4.34 3.01
dmin ( µm ) 34 24 81 14
Oil dmax ( µm ) 916 916 212 842
Droplets d ( µm ) 430 434 179 274
Future
Production n 4.00 4.02 9.79 2.68
Condition dmin ( µm ) 40 40 ––––––– 27
Water dmax ( µm ) 1532 1532 ––––––– 1371
Droplets d ( µm ) 653 653 ––––––– 475
n 4.27 4.28 ––––––– 3.79

Table 4-6. The droplet size distribution in the important zones of the Gullfaks-A
separator with expanded flow-distributing baffles in the future conditions.
Before After
Discrete Oil Water
Gravity Gravity
Phase Outlet Outlet
Separation Separation
Parameters
Zone Zone
dmin ( µm ) 39 39 55 20
Oil dmax ( µm ) 1398 1398 191 1398
Droplets d ( µm ) 492 493 172 346
n 4.15 4.16 4.60 2.77
dmin ( µm ) 34 34 ––––––– 34
Water dmax ( µm ) 1557 1557 ––––––– 1475
Droplets d ( µm ) 662 662 ––––––– 498
n 3.79 3.79 ––––––– 3.31
CFD Simulation of Multiphase Separators 131

The other modification analyses were performed by changing the liquid levels. Although
in all of the following liquid level analyses, the separation efficiency remained about
100% for water droplets, the separation efficiency did generally decrease for the oil
droplets to be:
- 0.08% with an increase of 20 cm in the thickness of water layer,
- 0.03% with a decrease of 20 cm in the thickness of water layer, and
- 0.02% with an increase of 20 cm in thickness of both water and oil layers.

Thus, it may be concluded from the CFD simulations that minor modifications cannot
change the separation inefficiencies in the Gullfaks-A separator.

4.2.6 Redesign of the Gullfaks-A Separator

The CFD simulation results showed that a major separation inefficiency would be
encountered by the existing separator for the upcoming increase in the produced water
flow-rate. The CFD simulations also showed that minor adjustments could not mitigate
this separation inefficiency. Therefore, redesign of the separator is the only reasonable
solution. To accomplish this redesign task, two approaches would be used. In the first
approach, the separator was redesigned by one of the well-known classic methods. In the
second strategy, the same classic design method was improved by considerations of
appropriate retention times of separator at 1988 production conditions.

4.2.6.1 Classic Design Strategy

The algorithmic design method proposed by Monnery and Svrcek (1994) was used. The
method uses industry standard settling or rising velocities to design the most economical
separator. Unfortunately, the exact values of settling or rising velocities of droplets may
be totally different from the industry accepted values. As noted in Chapter Two, the usual
value for liquid droplet size is 150 µm and is used as a standard in the API design
method (Walas, 1990; Hooper, 1997). However, assuming this droplet size, no realistic
results could be obtained for the Gullfaks-A separator. That is, a separator with diameter
of 14.4 m and length of 87 m would result for separating the mixture at 1988 production
CFD Simulation of Multiphase Separators 132

conditions, and a separator with dimensions of 12.4 m × 75 m would be proposed for the
future production conditions. The calculated separator dimensions imply that, through
classic design strategy, there is no feasible design for separating the mixture with the
Gullfaks-A production conditions. On the other hand, both CFD simulation results and
the oilfield experience showed that the original separator has operated in an acceptable
way for the 1988 conditions, and the separation inefficiencies have been associated with
the increase in the produced water flow-rate at the future production conditions. This
would indicate that a realistic design could be made by combining the classic method
with new separation velocities developed from CFD simulations.

4.2.6.2 Combining Improved Separation Velocities

As noted by Svrcek and Monnery (1993), although the basic equations used for separator
sizing are widely known; subjectivity exists during the selection of the parameters used in
these equations. The most controversial parameters are settling (or rising) velocities of
droplets in liquid-liquid separation. These parameters are to be tuned via CFD simulation
using the 1988 production conditions at which the separation efficiencies have been
acceptable.

- Estimation of settling velocity of oil droplets in the gas phase:

Two industrial methods, i.e. Gas Processors Suppliers’ Association (GPSA)
method and York demister method, and one theoretical method were evaluated.
Using these methods, the terminal settling velocity of oil droplets was calculated
to be 0.460 m/s, 0.328 m/s, and 0.216 m/s, respectively. Thus, the corresponding
minimum length required for the vapor-liquid separation in the original separator
was calculated as 0.51 m, 0.71 m, and 1.08 m, respectively. The CFD simulations
indicated that around 10 cm after the momentum breaker, almost all the liquid
droplets have settled down, which implies that the calculated values are somewhat
conservative. However, while sizing the Gullfaks-A separator, the classic
procedure did indicate that the liquid-liquid separation is controlling. Therefore,
CFD Simulation of Multiphase Separators 133

the focus of the current phase of this study was placed on the proper estimation of
oil-water separation velocities.

- Estimation of the rising velocity of oil droplets out of the water phase:
In classic methods, the flow of rising oil droplets in the water phase is assumed to
be laminar and the rising velocity can be estimated using Stokes’ law. By using
Stokes’ law, a velocity of 0.00566 m/s was calculated for the industry accepted
droplet size of 150 µm . This value is higher than the value implied by CFD
simulation results as follows. The separation efficiency of 100% for oil droplets
indicates an appropriate separator design/size. The provided residence time of this
original separator was 165.6 s for water phase with a water layer thickness of
0.625 m, hence the rising velocity of oil droplets should be 0.00377 m/s as was
assumed while redesigning the separator.

- Estimation of the settling velocity of water droplets out of the oil phase:
The flow of settling water droplets in the oil phase is assumed to be laminar,
hence Stokes’ law is used in the classic methods. A velocity of 0.000464 m/s was
calculated assuming a liquid droplet size of 150 µm . Compared with the CFD
simulation results, this value is too conservative and leads to an oversized
separator. The installed separator with dimensions of 3.33 m × 16.30 m has been
working efficiently for the designed 1988 production conditions. From a CFD
point of view, the predicted separation efficiency of 96.9% for water droplets
indicates that the design has been successful in having the water droplets
separated from the oil phase. Therefore, the provided residence time of 73.4 s for
oil phase was reasonable. Accounting for the oil layer thickness of 1.039 m, the
settling velocity of water droplets should be 0.01416 m/s. In order to further
improve the water separation efficiency, a value of 0.0134 m/s (95% of 0.01416
m/s) was assumed while redesigning the separator.
CFD Simulation of Multiphase Separators 134

Instead of the estimated values calculated from Stokes’ law, new “separation velocities”
were used in the classic design method proposed by Monnery and Svrcek (1994). In order
to specify a useable separator, the separator was redesigned for the maximum flow-rates
of produced gas, oil, and water. Figure 4-24 provides the sizing specifications of the
redesigned Gullfaks-A separator. As expected, a larger separator is required. To keep the
design in line with the original separator, four extra baffles with gaps of 4.0 m between
them were also added to the new separator design. Figure 4-25a shows the vessel in the
Gambit environment after assigning the required mesh, and Figure 4-25b represents the
separator in the Fluent environment with all the installed internals. The global quality of
the produced mesh in terms of number of cells, maximum cell squish, cell skewness, and
maximum aspect ratio are presented in Table 4-7. Table 4-7 indicates that only a
negligible fraction of cells (around 0.0044%) was of poor quality, hence the mesh is
correct.

Figure 4-24. Geometrical Specifications of the Redesigned Gullfaks-A Separator.

Table 4-7. Quality of the mesh produced in Gambit environment for the redesigned
Gullfaks-A separator.
Number of Cells Maximum Squish Maximum Skewness Maximum Aspect Ratio
1155813 0.961646 0.999632 128.621
Skewness of the Produced Mesh
Skewness Range 0-0.20 0.20-0.40 0.40-0.60 0.60-0.80 0.80-1.0
Density of Cells 76.4772% 17.8354% 4.4223% 1.2607% 0.0044%
CFD Simulation of Multiphase Separators 135

(a)

(b)
Figure 4-25. The Computer Model for Redesigned Gullfaks-A Separator; (a) Generated
Mesh in Gambit Environment, and (b) with All the Internals in Fluent Environment.
CFD Simulation of Multiphase Separators 136

Having set all the CFD parameters for the redesigned separator, some 5100 iterations are
required for continuous-phase solution convergence. Each iteration takes about 31 s and a
PC run-time of around 44 h is hence required per solution of the continuous phases. A
further PC run-time of around 4 h is also required for simulation of interactions between
the dispersed droplets and continuous phases. The fluid flow profiles inside the
redesigned separator are presented in Figures 4-26 to 4-28. These figures provide the
pressure, velocity and density profiles for the fluid flows on the central xz-plane of the
separator. As was expected, the macroscopic features of the redesigned separator are very
similar to those of the original separator. Again the pressure drops predicted for the
baffles and demister are low and hence reasonable, and the velocity vectors and density
contours are realistic.
The modifications made as a result of the CFD simulation of this research project
did enhance the separator performance. The redesigned separator dealt satisfactorily with
1988 production conditions, in that the total separation efficiency was as high as 99.1% (a
bit higher than the original separator efficiency) with its components of 100% and 98.7%
as separation efficiencies for oil and water droplets, respectively. For the projected future
production conditions, the redesigned separator had a total separation efficiency of 99.7%
with its components of 93.4% and 100% as separation efficiencies for oil and water
droplets, respectively. Note, although some 6.6wt% of the oil droplets were predicted to
exit in the water outlet for future production case, this would not decrease the separated
water purity dramatically, in that the water outlet composition would be 99.7% water and
only 0.3% oil. As was the case with the original separator, there would be no droplet
carry-over in the gas phase outlet, i.e. all the injected droplets exited in either the oil
outlet or the water outlet. Droplet size distribution is given for the selected surfaces of the
separator in Table 4-8. Table 4-8 indicates that compared with the initially defined
droplet size distribution (Table 3-12) the volume median diameter has decreased to
become 87% of its initial value for oil droplets and to 75% of its initial value for water
droplets. At the new project operating conditions the values changed to 63% for oil
droplets and to 73% for water droplets.
CFD Simulation of Multiphase Separators 137

(a)

(b)
Figure 4-26. Contours of Pressure (Pa) in the Middle of the Redesigned Separator for (a)
1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 138

(a)

(b)
Figure 4-27. Vectors of Velocity (m/s) in the Middle of the Redesigned Separator for (a)
1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 139

(a)

(b)
3
Figure 4-28. Contours of Density (kg/m ) in the Middle of the Redesigned Separator for
(a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 140

Table 4-8. The droplet size distribution in the important zones of the redesigned
Gullfaks-A separator.
Before After
Discrete Oil Water
Gravity Gravity
Phase Outlet Outlet
Separation Separation
Parameters
Zone Zone
dmin ( µm ) 17 17 17 –––––––
Oil dmax ( µm ) 2058 2058 1755 –––––––
Droplets d ( µm ) 791 799 683 –––––––
1988
Production n 2.52 2.18 2.19 –––––––
Condition dmin ( µm ) 29 16 29 33
Water dmax ( µm ) 2509 1615 476 496
Droplets d ( µm ) 1218 390 403 223
n 2.95 2.27 1.98 2.04
dmin ( µm ) 17 25 39 25
Oil dmax ( µm ) 1835 1846 1504 352
Droplets d ( µm ) 490 683 664 270
Future
Production n 3.94 2.47 5.62 2.28
Condition dmin ( µm ) 40 35 ––––––– 35
Water dmax ( µm ) 3007 3007 ––––––– 1240
Droplets d ( µm ) 1004 1074 ––––––– 480
n 2.54 3.28 ––––––– 4.07

While tracking the 2000 injected droplets, the number of breakups predicted in the CFD
simulation was some 1363 for 1988 production conditions and 1568 for the future
production conditions.
Similar to the case of the original separator, the effect of installed distribution
baffles was simulated. Again, all the baffles were removed from the model and this was
assumed to be a base case. The resultant separation efficiency was calculated to be 100%
for oil droplets and 99.3% for water droplets at the 1988 production conditions. For the
future production conditions, a separation efficiency of 97.7% for oil droplets and of
100% for water droplets was calculated. When compared with the corresponding values
reported for the separator equipped with the baffles, some small differences in favor of
eliminating the baffles results.
CFD Simulation of Multiphase Separators 141

The second phase of the internals simulation again consisted of expanding the baffles so
that they cover all the cross-sectional area of the vessel. The results of simulation in terms
of the pressure, velocity and density profiles are presented on the supplementary CD. The
results of the CFD modeling predicted efficiencies of 100% and 99.7% for the oil and
water droplets at the 1988 production conditions, respectively. Thus, the separation
efficiency of water droplets has shown a minor improvement with the baffles expanded in
the separator. However, a noticeable decrease in the oil separation efficiency (from
93.4% to 75.2%) was predicted for the separator operating at the future production
conditions. For this case, the predicted separation efficiency was still 100% for the water
droplets.
The results generated for the original separator and its modified versions would
point to the fact that distribution baffles have a minor effect on the separation efficiency.
They are generally helpful in improving the quality of flow distribution, however,
installing distribution baffles cannot overcome poor designs. Moreover, there may be
some rare cases with decreased separation efficiency after installing the baffles.10

4.2.6.3 On Substitution of Stokes’ Law by More Complicated Models

Using Stokes’ law in the classic design method led to an oversized separator for Gullfaks-
A. In fact, Stokes’ law assumes a single solid particle falling or rising in a stationary fluid
and calculates the particle terminal velocity. This, of course, is not the case in an actual
multiphase separator, in which a population of fluid droplets (of different size, position
and velocities) is to be separated from a moving fluid flow. So, the complicated aspects
of the described process make it impossible to come up with a mathematical model to
precisely describe the process. However, some modified relations have been proposed to
mitigate the Stokes’ law restrictions. The relations provided by Ishii and Zuber (1979) are
useful in that the authors have provided empirical relations for estimation of drag
coefficients and relative terminal velocities for dispersed droplets. The relations are rather

10
Please also refer to Hansen et al. (1993) for a practical example.
CFD Simulation of Multiphase Separators 142

complicated, as expected, and only the pertinent equations are presented here. Equation
4-1 has been proposed by Ishii and Zuber (1979) for calculation of the relative terminal
velocity in the “undistorted droplet regime”, which satisfies the Gullfaks-A separation
conditions:

21.6 µ c µ c ϕ 4 / 3 (1 + ϕ )
Ud = (1 − β ) (4-1)
ρcd p µm µ 
6/7

1+ϕ c 1− β 
 µm 

where U d is relative velocity of droplets in m/s, dp is diameter of droplets in m, µ m is

mixture viscosity in Pa.s as defined by Equation 4-2, β is volume fraction of dispersed

phase, and ϕ is a function of dp as given by Equation 4-3.

µc
= (1 − β )
2.5 ( µ + 0.4 µ c ) ( µ d + µ c )
d
(4-2)
µm

0.75
 0.01ρ ρ − ρ d 3 g  4 / 7 
ϕ = 0.551 +  − 1
c d c p
(4-3)
 µc 2  
  

Using Equations 4-1 to 4-3 and assuming a volume fraction of 5% for the dispersed phase
and the usual droplet size of 150 µm , the separation velocities are calculated to be
0.00452 m/s and 0.000416 m/s for oil and water droplets, respectively. It is interesting
that the calculated separation velocities partially agree with the Stokes’ law results, with a
difference of 25.2% for oil droplet velocity and 11.5% for water droplet velocity. In order
for the model proposed by Ishii and Zuber (1979) to predict the more realistic separation
velocities, the oil and water droplet diameters should be assumed to be 135 µm and 876
µm , respectively. The corresponding oil and water droplet diameters for proper velocity
estimation by Stokes’ law are 122 µm and 806 µm , respectively. Again, there are small
CFD Simulation of Multiphase Separators 143

differences of about 9% between the corresponding results for the Stokes’ law and Ishii
and Zuber equations. Therefore, it would seem that having assumed proper/efficient
droplet sizes, the separation velocities can be estimated by simpler models such as
Stokes’ law. This issue will be discussed in more detail in Chapter Five.

4.2.7 Sensitivity Analyses

In this subsection, sensitivity of the results with respect to the most important parameters
used in the CFD simulations is discussed. The first issue of interest is the repeatability of
the CFD predicted separation efficiencies. For this purpose, the original Gullfaks-A
separator CFD simulation at the two production conditions was used and the injected
droplets were tracked for four times. A statistical analysis of the produced results
indicated that the standard deviation of results was generally less than 1% for the average
values in all cases. Therefore, the number of performing DPM model of Fluent was
reduced to one in the modification case studies.
The other important CFD parameters for sensitivity analyses are as follows:

1- Assigned Grid Size

Quality of the produced mesh for Gullfaks-A separator was verified in Chapter
Three. As discussed, the grid system was composed of 884805 cells,
corresponding to an average interval size of 0.037 m. However, in order to further
verify the quality of applied grid system, a relatively coarse grid system,
composed of 500130 cells, with an average interval size of 0.065 m was used to
see if the CFD results would change significantly. The tendency with denser grids
is to reveal more details of the fluid flow regimes, and if the obtained solution
remains unchanged in its important features, it can be concluded that the assumed
grid system is sufficiently fine (Sharratt, 1990).

2- Droplet Size Distribution

To analyze the sensitivity of the CFD results to the defined droplet size
distribution, the maximum droplet size was alternatively estimated by the
CFD Simulation of Multiphase Separators 144

Karabelas (1978) correlation instead of Hesketh et al. (1987) approach, and the
mean droplet size was changed using Equation 3-24.

3- Surface Tensions
The surface tension values were estimated using the methods discussed in Chapter
Three. However, for sensitivity analysis purposes, the oil and oil-water surface
tensions were set to 0.038 N/m, the maximum value given by Streeter and Wylie
(1985), and 0.03 N/m, based on Antonoff’s rule (1907), respectively.

After implementing these changes in the CFD model or its parameters, the fluid flow
profiles were obtained for each of the case studies and were compared with those of the
original case study. The CFD results were also compared from a separation efficiency
point of view. These comparisons indicated that both the macroscopic and microscopic
features of the CFD solution for Gullfaks-A separator were not so sensitive to the applied
changes. For example, Table 4-9 provides the separation efficiencies of the studied cases
as well as the original Gullfaks-A values. The data of Table 4-9 would indicate that there
are some negligible differences between the “sensitivity analysis” case studies and the
original separator values. Therefore, it can be concluded that the original grid system is
sufficiently fine and that some uncertainties in the defining droplet size distribution or
estimated surface tensions have little effect on the CFD simulation results of multiphase
separators.

Table 4-9. The results of sensitivity analyses on the developed CFD model for Gullfaks-
A separator.
Separation Efficiencies
1988 Condition Future Condition
Oil Water Oil Water
Sensitivity Grid Size 98.0% 99.8% 0.0% 100%
Analysis Maximum Particle Size 100% 99.6% 2.1% 99.8%
Parameter Surface Tension 99.9% 96.3% 0.1% 100%
Original Values 100% 96.9% 1.3% 99.9%
CFD Simulation of Multiphase Separators 145

4.3 Summary

In Chapter Four, the CFD simulation results of selected pilot-plant scale two-phase
separators and one large scale three-phase separator have been presented. Two simulation
approaches, DPM and VOF-DPM, were used in the simulation of the two-phase
separators. The DPM approach was useful in predicting the incipient velocities, but the
resultant separation efficiency plots were not correct. In fact, some unacceptable results
were obtained for several case studies when using the DPM only model as proposed by
Newton et al. (2007). These poor results were caused by the model assumption in that the
liquid phase is ignored. Thus, the gas-liquid interactions and, particularly, the dynamic
interaction between liquid droplets and continuous liquid phase cannot be taken into
account. The more complex VOF-DPM model does include these liquid phase
interactions. From a practical perspective, the obtained separation efficiency data and
diagrams were reasonable and the poor results of the DPM model were completely
eliminated. Moreover, there was excellent agreement between simulated phase separation
results, such as separator efficiencies etc., and the experimental data and observations in
most of the case studies. Droplet coalescence and breakup were also modeled through the
CFD simulations. The results indicated that while droplet coalescence occurred rarely,
droplet breakup was a common phenomenon particularly at higher velocities. The study
of droplet size distribution in the gas outlet for the selected separators showed that mist
eliminators may operate more efficiently in the horizontal separators than in the vertical
separators.
Based on the obtained results for the CFD simulation of two-phase separators,
only VOF-DPM approach was used in the CFD simulation of the three-phase separator.
Compared to the original study of Hansen et al. (1993), the developed model did provide
high-quality details of fluid flow profiles, leading to a very realistic overall picture of
phase separation in all zones of the separator. This realistic CFD simulation of the three-
phase separator performance did provide an understanding of both the microscopic and
macroscopic features of the three-phase separation phenomenon. The CFD simulations
did show that droplet breakage was common with an average rate of 76%, when
dispersed droplets came into contact with the deflector baffle. Because of droplet
CFD Simulation of Multiphase Separators 146

breakup, the volume median diameter of droplets decreased to about 67% of the initial
value. However, the droplet size distribution remained almost the same while the droplets
were traveling through the gravity separation zone of the separator.
In line with the oilfield experience, the CFD simulations showed that serious
separation inefficiencies may be encountered with the projected increase in the flow-rate
of produced water in the upcoming years. To overcome these inefficiencies, first, minor
modifications such as adjusting baffle positions and liquid levels were evaluated.
However, these minor modifications could not resolve the essential separation
inefficiencies in the Gullfaks-A separator, and the separator was redesigned using a
modified version of the classical design method of Monnery and Svrcek (1994) that
involved improved separation velocities. It was also realized that the popular classic
methods would be too conservative and specify an extremely oversized separator for
Gullfaks-A. The redesigned separator operated satisfactorily at both the 1988 and the
projected production conditions and did confirm that a realistic optimum separator can be
specified using the classical method of Monnery and Svrcek (1994) when the new
developed separation velocities are incorporated in the design procedure. Furthermore, it
was demonstrated that simple models such as Stokes’ law may still be used for estimation
of realistic separation velocities, but suitable/efficient droplet sizes should be used for the
oil and water phases.
Finally, sensitivity analysis was performed on the developed CFD model for
Gullfaks-A separator. The results reconfirmed the quality of the grid system and
robustness of the DPM outputs with respect to solution repetitions. The sensitivity
analysis also showed that small (<10%) uncertainties in droplet size distribution or in
estimating surface tensions will have negligible effect on the CFD simulation results of
multiphase separators.
CFD Simulation of Multiphase Separators 147

Chapter Five: Improved Design Criteria

CFD simulations of four pilot-plant scale two-phase separators and an industrial scale
three-phase horizontal separator were developed. In order to successfully simulate these
separators, an efficient combination of two multiphase simulation models of Fluent, VOF
and DPM, with appropriate model assumptions and settings was used. The obtained
results were in excellent agreement with the empirical observation and pilot plant data. In
the case of three-phase separator, the developed model did provide useful fluid flow
profiles and a more realistic picture of phase separation quality when compared with the
original study of Hansen et al. (1993). Hence, the goal of the present chapter is to keep
the implemented model assumptions/settings and provide updated phase separator design
criteria. These criteria can then be combined with the algorithmic design method of
Monnery and Svrcek (1994) to specify the realistic optimum separator design.

5.1 Fluid Systems

The following oilfield separator data ranging from light oil conditions to heavy oil
conditions were used to obtain the necessary physical properties required in the CFD
simulations:

• Oilfield-1: Black Oil Reservoir Case Study (Grødal and Realff, 1999)
• Oilfield-2: Gullfaks A (Hansen et al., 1993)
• Oilfield-3: A Diluted Bitumen Treatment (Lu et al., 2007)
• Oilfield-4: Kuito Field Offshore Cabinda (Lee et al., 2004)
• Oilfield-5: Beta Offshore California (Visser, 1989)

The physical properties used in CFD simulations are presented in Table 5-1. Except for
Oilfield-5, all the physical properties necessary for CFD simulations have been provided
by original papers. In the case of Oilfield-5, the data for water and vapor phases are
missing from original paper (Visser, 1989), and these data had to be estimated. Since the
water phase density does not differ significantly from one oilfield to another, the value
from Oilfield-4 was used in the simulations. For estimation of the water phase viscosity,
CFD Simulation of Multiphase Separators 148

the approach recommended by McCain (1973) was used. This approach estimates the
water phase viscosity using the oilfield temperature and pressure. For estimation of vapor
phase properties, the natural gas composition of Vetter et al. (1987) for a rather similar
heavy oilfield (22.20°API) was used in the HYSYS 3.2 simulator. The recommended
Peng-Robinson (PR) equation of state was used as the thermodynamic model for this gas
phase hydrocarbon system.
As noted in Chapter Four, separation efficiencies are not that sensitive to the
assumed maximum droplet size. Therefore, a maximum droplet size of 2267 µm and
4000 µm , as estimated for Gullfaks-A oilfield at the 1988 production condition, were
assumed for the oil and water droplets, respectively. Then, the required particle size
distribution parameters were set using the strategy outlined in Chapter Three. The results
are shown in Table 5-2.

Table 5-1. Physical properties for fluids of different oilfields at their corresponding
separator conditions.
Oilfield-1 Oilfield-2 Oilfield-3 Oilfield-4 Oilfield-5
Pressure (kPa) 2000 6870 1280 690 345
Temperature (°C) 80 55.4 135 50 60
Density (kg/m3) 17.46 49.7 9.50 5.70 2.44
Gas Viscosity (Pa.s) 1.07 × 10-5 1.30 × 10-5 1.38 × 10-5 1.20 × 10-5 1.26 × 10-5
Density (kg/m3) 767.7 831.5 874 907 960
Oil Viscosity (Pa.s) 0.73 × 10-3 5.25 × 10-3 6.90 × 10-3 42.0 × 10-3 100 × 10-3
Density (kg/m3) 974.6 1030 931 1026 1026
Water Viscosity (Pa.s) 3.70 × 10-4 4.30 × 10-4 2.0 × 10-4 9.0 × 10-4 5.2 × 10-4

Table 5-2. The discrete phase parameters used in CFD simulations of phase separation.
Discrete Phase Parameters Oil Droplets Water Droplets
Minimum Diameter ( µm ) 100 100
Maximum Diameter ( µm ) 2267 4000
Mean Diameter ( µm ) 907 1600
Total Mass Flow-rate ( kg / s ) 6.5 × 10-4 4.4 × 10-3
Number of Tracked Particles 1000 1000
Spread Parameter 2.6 2.6
CFD Simulation of Multiphase Separators 149

5.2 Phase Separation Components

In order to efficiently model the phase separation process, two independent sets of CFD
simulations were performed; one for vapor-liquid separation and the other for liquid-
liquid separation. For this purpose, two two-phase models shown in Figure 5-1 were
used. These useful models were developed by verification of the results for various two-
phase systems when compared to those of industrial scale separators. These
investigations confirmed that although the mesh generation stages in the Gambit
environment and the setting of CFD parameters in the Fluent environment would be more
straightforward for these models, the results produced for phase separations are the same
as those for large-scale separators.
The other reason for using these models is connected with their horizontal layout.
By using the horizontal layout, the results can be directly used in designing horizontal
separators, and the subjectivity of applying the relationships obtained from the vertical
orientation force balance to the horizontal orientation is not encountered. Before
reviewing the results, two additional comments are worth noting:

(a) (b)
Figure 5-1. The Two-Phase Models Developed for CFD Simulation of Phase Separation;
(a) Vapor-Liquid Separation, and (b) Liquid-Liquid Separation.
CFD Simulation of Multiphase Separators 150

Comment 1: In the developed models, for the first time, a distribution of droplets is
being tracked. Therefore, in order to evaluate the separation velocities, a mass averaged
time taken by droplets to separate out of continuous phases was used. So, the reported
values are the “efficient” separation velocities predicted for various oilfield systems.
Having predicted the efficient separation velocities for a specific system, the
corresponding efficient droplet size can then be calculated using available relationships.
Comment 2: In classic separator design methods, it is assumed that when oil or water
droplets come in contact with vapor-liquid or liquid-liquid interfaces, they immediately
penetrate the interface and join their own phases. Therefore, assuming an extra residence
time for droplet penetration through interfaces has been ignored in the classic design
methods. However, the CFD simulation results indicate that depending on the involved
fluid properties, a residence time of up to 100 s may be necessary for droplets to pass
through the interfaces.

5.2.1 Vapor-Liquid Separation

The CFD simulation results of vapor-liquid separation for various oilfields have been
presented in Table 5-3. For calculation of the efficient droplet diameter, the well-known
equation, Equation 5-1, was used (Green and Perry, 2008):

4 × 10 −6 gd eff (ρ L − ρ V )
Vsep = (5-1)
3C D ρ V

where Vsep is separation velocity in m/s, g is gravity acceleration in m/s2, d eff is efficient

diameter of droplet distribution in µm , ρ L and ρ V are liquid and vapor densities

(respectively) in kg/m3, and C D is drag coefficient which can be calculated based on the
Gas Processors Suppliers’ Association (GPSA) approach (Monnery and Svrcek, 2000)
via Equations 5-2 and 5-3:
CFD Simulation of Multiphase Separators 151

5.0074 40.927 44.07

CD = + + (5-2)
ln( x) x x

3.35 × 10 −9 ρ V (ρ L − ρV )d eff
3

x = (5-3)
µV 2

where µV is viscosity of vapor phase in Pa.s.

Table 5-3 presents the residence times required for oil droplets to pass through the
vapor-liquid interface. The values vary from 4 s to 58 s almost linearly with respect to
vapor phase density which varies from 2.44 kg/m3 to 49.7 kg/m3. For water droplets, the
interface residence times decrease to about 4.5 s with a reported maximum value of 11.4
s. Also, the settling velocities of water droplets are higher than those of oil droplets in all
cases. Therefore, in the vapor-liquid compartment of a three-phase separator, the efficient
separation of oil droplets guaranties satisfactory separation of the water droplets.
Figure 5-2 shows the variations of calculated efficient oil droplet size versus gas
phase density. The CFD simulation results shown in Figure 5-2 can be fit with an
exponential function, Equation 5-4:

d eff ,Oil −Gas = 206.06 − 189.42 exp(− 0.08635 ρ V ) (5-4)

Table 5-3. The vapor-liquid separation characteristics for different oilfields.

Oilfield-1 Oilfield-2 Oilfield-3 Oilfield-4 Oilfield-5
Vsep (m/s) 0.7215 0.4967 0.8623 0.9944 1.0869
Oil tinterface (s) 12.4 58.0 30.0 14.5 4.0
Droplets
deff ( µm ) 164.7 203.6 119.3 94.6 51.0
Vsep (m/s) 0.9691 0.7055 1.0620 1.1905 1.2644
Water tinterface (s) 4.0 5.9 5.5 11.4 2.0
Droplets
deff ( µm ) 223.6 311.6 163.5 116.3 62.7
CFD Simulation of Multiphase Separators 152

Figure 5-2. Efficient Oil Droplet Diameter for Estimation of Settling Velocity in Vapor
Phase versus Vapor Density.

Similarly, Figure 5-3 shows the variations of calculated efficient water droplet size versus
gas phase density, and Equation 5-5 presents the result of an exponential function fit for
d eff - ρ V data:

d eff ,Water −Gas = 322.14 − 302.66 exp(− 0.06619 ρV ) (5-5)

As noted, only oil-gas separation needs to be used for separator design. Therefore,
Equations 5-1 to 5-4 can be used for estimation of vapor-liquid separation velocity.
Furthermore, an additional vapor residence time proportional to the vapor density (or
about 60 s for a conservative design), should also be used in the separator design.
CFD Simulation of Multiphase Separators 153

Figure 5-3. Efficient Water Droplet Diameter for Estimation of Settling Velocity in
Vapor Phase versus Vapor Density.

5.2.2 Liquid-Liquid Separation

A force balance on the oil droplets rising in the water phase or water droplets settling in
the oil phase leads to Equation 5-6 (Green and Perry, 2008):

4 × 10 −6 gd eff (ρ W − ρ Oil )
Vsep = (5-6)
3C D ρ c

where ρ W , ρ Oil , and ρ c are densities of water, oil, and continuous liquid phase (oil or
water), respectively, in kg/m3.
In the classic design methods, the drag coefficient for the flow of rising oil
droplets in the water phase or settling water droplets in the oil phase is described using
Stokes’ law, Equation 5-7 (Green and Perry, 2008):
CFD Simulation of Multiphase Separators 154

24 24 × 10 −6 ρ c d eff Vsep
CD = = (5-7)
Re p µc

where Re p is particle Reynolds number, and µ c is viscosity of continuous liquid phase

in Pa.s.
Substitution of drag coefficient from Equation 5-7 into Equation 5-6 leads to a
straightforward relation for calculation of separation velocity in oil-water systems:

1 × 10 −12 gd eff
2
(ρW − ρ Oil )
Vsep = (5-8)
18µ c

Note, the use of Equation 5-8 is limited to the laminar settling or rising because Stokes’
law cannot be used if Re p exceeds the upper limit of 0.10 (Green and Perry, 2008). King

(2002), who has provided a pertinent and useful survey of the fluid-particle interactions,
suggested an even lower Reynolds number of 0.01 as the upper limit for Stokes’ law.
An analysis of the CFD simulation results showed that the calculated Re p

exceeded the upper limit in several of the case studies (see Table 5-4). Alternatively, as
recommended by King (2002), for a Re p < 2 × 10 3 , which would cover all the oilfield

case studies, the drag coefficient can be estimated accurately by the Abraham equation,
Equation 5-9 (Abraham, 1970):

2 2
 9.06   µc 
C D = 0.281 +  = 0.281 + 9.06 × 10 3  (5-9)
 Re p   ρ c d eff Vsep 
   

Table 5-4 presents the CFD simulation results for the liquid-liquid separation for the
oilfield cases. Similar to vapor-liquid separation, a residence time is required for the oil
and water droplets to penetrate through the liquid-liquid interface. Table 5-4 shows that
as viscosity of oil phase varies from 0.73 × 10-3 Pa.s to 0.10 Pa.s, the interface residence
CFD Simulation of Multiphase Separators 155

times vary almost linearly from 3.8 s to 97.2 s for the oil droplets and from 0.9 s to 8.4 s
for the water droplets.
In Table 5-4, the efficient droplet sizes have been calculated from both Stokes’
law and the Abraham equation. A comparison between the results indicates that there is a
significant difference that occurs while estimating the efficient oil droplet sizes in all five
cases and while estimating the efficient water droplet sizes in three cases. Hence, it can
be concluded that the use of Stokes’ law for estimation of efficient droplet sizes generally
results in totally different values from those estimated when using the Abraham equation.
Figure 5-4 shows the variations in the calculated efficient oil droplet size versus viscosity
of the water phase. The data of Figure 5-4 would indicate that a constant-value line,
d eff ,Oil − Liquid = 597 µm , can represent the CFD simulation results when the Abraham

equation is used. Therefore, an efficient oil droplet size of 600 µm can be assumed for
the Abraham equation to estimate the oil rising velocity in separator design procedures.
Figure 5-4 also shows that the incorrect use of Stokes’ law for interpretation of CFD
results does lead to a totally different and weak correlation between efficient oil droplet
size and the viscosity of water phase.

Table 5-4. The liquid-liquid separation characteristics for different oilfields.

Oilfield-1 Oilfield-2 Oilfield-3 Oilfield-4 Oilfield-5
Vsep (m/s) 0.03175 0.03002 0.01800 0.01494 0.01344
tinterface (s) 3.8 27.7 24.3 74.5 97.2
Stokes’ deff ( µm ) 322.7 345.4 340.4 455.4 440.8
Oil Law Rep 26.99 24.84 28.52 7.76 11.69
Drops Abraham deff ( µm ) 566.9 591.6 608.4 595.2 622.9
Equation Rep 47.41 42.55 50.98 10.14 16.52
Vsep (m/s) 0.05266 0.01813 0.006056 0.003249 0.001341
tinterface (s) 0.9 2.7 3.1 4.0 8.4
Stokes’ deff ( µm ) 583.9 938.1 1159.8 1450.4 1930.6
Water Law Rep 32.33 2.69 0.89 0.10 0.02
Drops Abraham deff ( µm ) 1086.8 1072.8 1235.5 1450.4 1930.6
Equation Rep 60.18 3.08 0.95 0.10 0.02
CFD Simulation of Multiphase Separators 156

Figure 5-4. Efficient Oil Droplet Diameter for Estimation of Rising Velocity in Water
Phase versus Water Viscosity.

It would seem that such a constant value assumed for the Abraham equation may
similarly be used in Stokes’ law. However, while the maximum variation from the
assumed constant value (597 µm ) is 5.0% for the Abraham equation; for Stokes’ law, the
maximum variation from the average value of 381 µm will be 19.5% which would result
in errors up to 43% for the estimation of the separation velocities.
Figure 5-5 presents the variations of efficient water droplet size calculated using
the Abraham equation versus oil phase viscosity. A linear equation, Equation 5-10, can
represent the correlation between efficient water droplet size and oil phase viscosity with
corresponding R-squared value of 0.9774.

d eff ,Water − Liquid = 1095 + 8382 µ Oil (5-10)

CFD Simulation of Multiphase Separators 157

The variation of efficient water droplet size calculated by Stokes’ law versus oil viscosity
is shown in Figure 5-6. Compared with the Abraham equation based modeling, a much
broader range of efficient droplet size as well as a non-linear regression of provided data,
Equation 5-11, are required if Stokes’ law is used for interpretation of CFD results.

d eff ,Water − Liquid = 3170µ Oil

0.2253
(5-11)

In summary, Equations 5-6 and 5-9 with the oil droplet size of 600 µm and the water
droplet size as per Equation 5-10 can be used iteratively for estimation of liquid-liquid
separation velocities. Furthermore, an additional liquid residence time proportional to the
oil viscosity of some 100 s for water phase and 10 s for oil phase should be used for
separator designs.

Figure 5-5. Efficient Water Droplet Diameter as a Function of Oil Phase Viscosity to be
Used in Abraham Equation.
CFD Simulation of Multiphase Separators 158

Figure 5-6. Efficient Water Droplet Diameter as a Function of Oil Phase Viscosity to be
Used in Stokes’ Law.

5.3 Vessel Orientation Considerations

The influence of continuous phase on the movement of discrete phase in horizontal and
vertical configurations will be explained from a CFD simulation perspective. The
emphasis will be placed on estimating the allowable velocities for the continuous phases.

5.3.1 Horizontal Arrangement

Using the two-phase separation models (Figure 5-1) and performing CFD simulations of
phase separation at various continuous phase velocities did confirm that oil and water
droplets follow the continuous phase flow in their horizontal movements. In other words,
the mass averaged horizontal velocities of oil and water droplets as simulated were
almost identical to the continuous phase velocities. Therefore, in the iterative design
procedure proposed by Svrcek and Monnery (1993), the residence length (L) can be set
CFD Simulation of Multiphase Separators 159

using the separation height ( H sep ), the separation velocity (Vsep), the continuous phase

velocity ( Vc ) and the required interface residence time (tinterface) as per Equation 5-12:

 H sep 
L ≥  + t int erface Vc (5-12)
V 
 sep 

Note, although the continuous phase can flow with a velocity that carries droplets up to
outlet vicinity, these high velocities will lead to re-entrainment of droplets by the
continuous phase. The other limit is due to economic considerations in which separators
with aspect ratios higher than 9 are rarely economical (Smith, 1987). Thus, there is an
upper limit for the continuous phase velocity and the separator aspect ratio.

5.3.2 Vertical Arrangement

In order to investigate how changing the separator orientation from horizontal to vertical
may influence the phase separation process, CFD simulations were performed using the
models represented by Figure 5-7. Similar to the horizontal arrangement, CFD
simulations confirmed that the movement of oil and water droplets is completely
influenced by the flow of continuous phase. In the vertical configuration, the continuous
phase is flowing in the opposite direction to the settling or rising droplets, hence the
apparent separation velocities are lower than those in the horizontal case, where forces do
not directly oppose one another. However, the relative separation velocities were almost
the same as the separation velocities predicted for horizontal arrangement by the CFD
simulations. From a practical point of view, the drag force exerted on droplets is a
function of the relative velocities (not the apparent velocities). For instance, the drag
force on a particle fixed in space with a lighter fluid moving upward is almost the same
as the drag force on the particle freely settling in the stationary fluid at the same relative
velocity (Green and Perry, 2008). Therefore, the separation velocities and relationships
for the horizontal arrangement can also be used for designing vertical separators.
However, the apparent separation velocity for a vertical separator will be a function of
continuous phase velocity, Equation 5-13:
CFD Simulation of Multiphase Separators 160

Vsep ,vertical = Vsep − Vc (5-13)

In designing vertical separators, Equation 5-13 should be combined with the vapor-liquid
and liquid-liquid separation relationships provided for the horizontal separator models.
The most important differences are presented in the following two subsections.

5.3.2.1 Vapor-Liquid Separation

Separation velocity of oil droplets, Vsep , is calculated using Equations 5-1 to 5-4. The

calculated value is an upper limit for the vapor velocity at which oil droplets would be
suspended in the vapor phase. Hence, according to Monnery and Svrcek (1994), a vapor
velocity of around 75% of Vsep may be assumed for a realistic design. Note, with this

assumption, the apparent settling velocity of oil droplets would be 25% of Vsep . Having

set the vapor velocity, the internal diameter of separator can be initially determined.

(a) (b) (c)

Figure 5-7. The Vertical Models Developed for CFD Simulation of Phase Separation; (a)
Vapor-Liquid Separation, (b) Separation of Water Droplets from Oil Phase, and (c)
Separation of Oil Droplets from Water Phase.
CFD Simulation of Multiphase Separators 161

5.3.2.2 Liquid-Liquid Separation

To start the sizing calculation, the separation velocities of oil and water droplets are
calculated using Equations 5-6, 5-9 and with the oil droplet size of 600 µm and the water
droplet size as per Equation 5-10. These are the maximum apparent separation velocities
that may occur in a vertical separator with a diameter of infinity. Since the initial
diameter of separator has been set based on vapor-liquid separation requirements,
superficial velocities of oil and water phases can be calculated. These values are used in
Equation 5-13 to calculate the apparent separation velocities for liquid phases.
Furthermore, similar to the horizontal separator, an extra appropriate liquid residence
time should also be added. The assumed separator diameter may need to increase if the
provided liquid residence time is not sufficient for the required separation. Thus, some
iterative calculations are required to specify the vessel diameter.
Alternatively, the following equivalent approach is proposed for a vertical
separator configuration. Assume that liquid phase (oil or water) height (H) and
volumetric flow-rate (Q), and interface residence time (tinterface) have been given or set,
and the surface area (A) required for liquid-liquid separation is to be calculated. Using the
widely accepted design procedures of Monnery and Svrcek (1994), Equation 5-14 should
be satisfied:

HA H
= + t int erface (5-14)
Q Vsep

Solving Equation 5-14 for “A” results in Equation 5-15:

 1 t int erface 
A = + Q (5-15)
V H 
 sep 

Equation 5-15 provides the minimum surface area required for efficient liquid-liquid
separation and should be satisfied for both the oil phase and water phase. For the water
phase, calculation of required area leads to a straightforward calculation of the required
CFD Simulation of Multiphase Separators 162

diameter. For the oil phase, however, if there is a baffle plate, the cross-sectional area of
down-comer should also be added to the required area. After calculation of the two
required diameters for water and oil phases, the values will be compared with the
diameter assumed for efficient vapor-liquid separation. The largest value will be selected
as the required internal diameter for the vertical separator.

5.3.3 The Proposed Approach in a Vector Space

This subsection summarizes the approach proposed in section 5.3 using a vector space.
CFD simulation results show that the apparent velocity vector of droplets is the sum of
separation velocity vector and continuous phase velocity vector independent of the
separator orientation:

r r r
Vapparent = Vsep + Vc (5-16)

Note, Ishii and Zuber (1979) have also employed the same vector space approach when
using the relative settling or rising velocities to present their relationships for various
dispersed phase-continuous phase fluid regimes.
Therefore, as concisely demonstrated in subsection 4.2.6.3, the simple models can
be used for estimation of realistic separation velocities for both horizontal and vertical
arrangements, but efficient droplet sizes, proposed in section 5.2, should be estimated and
used with these models.

5.4 Designing the Realistic Optimum Separator for Gullfaks-A

The aim of this section of Chapter Five is to provide the optimum separator design for
processing Gullfaks-A (Hansen et al., 1993). For realistic design purposes, the results of
the CFD simulations presented in previous sections have been used, and both horizontal
and vertical arrangements have been considered. The performance of designed separators
has been verified based on the CFD simulations. Appendix D presents CFD results in
terms of pressure, velocity, and density profiles for the resultant fluid flows.
CFD Simulation of Multiphase Separators 163

5.4.1 Design Phase

The algorithmic design method proposed by Monnery and Svrcek (1994) was used. Their
design procedure, however, used the “separation velocities” and interface residence times
obtained from CFD simulations to design the most economical separator. The physical
parameters for the fluids in the Gullfaks-A separator, presented in Table 3-8, have been
taken from Hansen et al. (1993). The maximum flow-rates of the produced gas, oil, and
water phases were used in the design. Consequently, a “simple” separator with a diameter
of 2.75 m and length of 19.40 m and an approximate weight of 135270 kg was calculated
as the most economical design. A vertical separator with a diameter of 10.8 m and a
height of 16.2 m and an approximate weight of 3014660 kg turned out to be the most
expensive design.
The estimated weights corresponding to all design types are presented in Table
5-5. Based on the volumetric flow-rates and calculated separator dimensions, the “boot”
design was excluded from the feasible design set. Due to the large volumetric flow-rate
of water, the boot diameter (7.67 m) was calculated to be very large compared with the
vessel diameter (3.06 m). Note, the “boot” design is typically used when the volumetric
flow-rate of heavy liquid is not high (Monnery and Svrcek, 1994). Figure 5-8 presents the
dimensions of the feasible horizontal separators designed for processing Gullfaks-A.
Because of its large diameter (10.8 m) and very large weight (12.8 times heavier
than the heaviest horizontal separator, Table 5-5), the designed vertical arrangement is
not economical and was excluded from further CFD performance validations. However,
in order to study the separation efficiency of the vertical arrangement, the production
capacity of Gullfaks-A was assumed to be 25% of the original values, and an additional
smaller vertical separator was designed. Figure 5-9 provides the dimensions of the
smaller vertical separator designed for processing Gullfaks-A at 25% of total capacity.
CFD Simulation of Multiphase Separators 164

(a)

(b)

(c)
Figure 5-8. Geometrical Specifications of Designed Separators for Gullfaks-A; (a)
Simple Type, (b) Weir Type, and (c) Bucket and Weir Type.

Table 5-5. The approximate separator weights calculated for various types of separators
processing Gullfaks-A.
Various Designs Simple Boot Weir Bucket & Weir Vertical
Approximate Weight (kg) 135270 181740 228720 234500 3014660
CFD Simulation of Multiphase Separators 165

Figure 5-9. Geometrical Specifications of the Vertical Separator Designed for Gullfaks-
A (25% Capacity).

5.4.2 Performance Validations via CFD Simulations

The separation performance of all the feasible horizontal separators as well as the vertical
separator was validated using CFD simulations. Using the developed procedure outlined
in Chapter Three for simulation of the original separator, the generation of the grid
system was completed for all separators in the Gambit and Fluent environments (Gambit
2.4.6, 2006; Fluent 6.3.26, 2006). Figure 5-10 to Figure 5-12 provide additional detail for
the generated mesh systems. The global quality of the produced mesh systems in terms of
number of cells, maximum cell squish, cell skewness, and maximum aspect ratio are
presented in Table 5-6. Table 5-6 also indicates that a negligible fraction of cells,
generally less than 0.08%, are of poor quality. Having set all the CFD parameters for the
1988 or future production conditions, the number of iterations required for convergence
varied from 6000 (for “simple” and “weir” designs) to 12000 (for “bucket and weir”
design and vertical arrangement). Each iteration takes about 27 s on a Pentium D (3.20
CFD Simulation of Multiphase Separators 166

GHz) and 2.00 GB of RAM PC. Therefore, a PC run-time of around 45 h (for “simple”
and “weir” designs) or 90 h (for “bucket and weir” design and vertical arrangement) was
required for solution of continuous-phase flow regimes. Note, the iterative process was
stopped regularly to check the interface levels and correct them (if necessary). Also, an
additional PC run-time of around 3 h per each case study was required for simulation of
interactions between the discrete phase (liquid droplets) and continuous phases.

Figure 5-10. The Grid System for the “Simple” Separator Designed for Processing
Gullfaks-A Generated in Gambit Environment.
CFD Simulation of Multiphase Separators 167

Figure 5-11. The Grid System for the “Bucket and Weir” Separator Designed for
Processing Gullfaks-A Generated in Gambit Environment.
CFD Simulation of Multiphase Separators 168

Figure 5-12. The Grid System for the Vertical Separator Designed for Processing
Gullfaks-A (25% Capacity) Generated in Gambit Environment.
CFD Simulation of Multiphase Separators 169

Table 5-6. Quality of the mesh produced in Gambit environment for various separators
processing Gullfaks-A.
Number of Maximum Maximum Maximum
Cells Squish Skewness Aspect Ratio
Weir 985834 0.735457 0.895869 44.4761
Simple 920205 0. 745314 0. 907619 42.7887
Bucket and Weir 1019479 0. 829993 0. 895890 44.0854
Vertical 938765 0. 997151 0. 999998 1108.15
Skewness of the produced mesh
Skewness Range 0-0.20 0.20-0.40 0.40-0.60 0.60-0.80 0.80-1.0
Weir 80.1525% 14.9260% 3.4968% 1.4233% 0.0014%
Density
Simple 72.6225% 20.5087% 5.6634% 1.2040% 0.0014%
of
Bucket & Weir 72.5747% 19.0717% 5.8814% 2.4626% 0.0096%
Cells
Vertical 36.7367% 43.2999% 17.0732% 2.8129% 0.0773%

The macroscopic features of CFD simulations consisting of the pressure, velocity and
density profiles for the fluid flows on the central xz-plane of the separators are presented
in Appendix D. As was expected, in all separators, the pressure drops assigned to the
baffles and demister are low and hence as expected. The velocity vectors and density
contours are also as expected.
Table 5-7 presents the separation efficiencies for all the designed separators. Based
on the data of Table 5-7, the following comments can be made:

1- From a separation efficiency point of view, “bucket and weir” design is superior
to other horizontal designs.
2- Although water separation efficiency is high enough for almost all designs, some
horizontal designs suffer from low oil separation efficiencies with the projected
increase in the water flow-rate.
3- Separation efficiencies are very high (about 100%) for both the oil and water
droplets in the vertical separator.

As noted, the separation efficiencies for horizontal case studies (particularly, for the weir
design) are not high enough to ensure complete phase separation. To further study this
issue, additional CFD simulations were performed using the available model for “weir”
design for the future production condition.
CFD Simulation of Multiphase Separators 170

Table 5-7. Separation efficiencies for various separators processing Gullfaks-A.

Oil Droplets Water Droplets
Weir 1988 100% 100%
Future 68.0% 100%
Simple 1988 99.8% 81.3%
Future 84.0% 100%
Bucket and Weir 1988 98.4% 96.6%
Future 99.9% 96.5%
Vertical 1988 100% 100%
Future 100% 100%

In all of the case studies, several parallel xy-planes were defined inside the vessel to
record the characteristics of the droplets passing through them. Software was developed
to do an analysis on the database provided by the recording surfaces. This analysis did
confirm the values shown in Table 5-4 for the separation velocities. However, the
captured data also showed that some oil droplets after reaching the oil-water interface
were not able to pass through. Further studies indicated that these droplets, while
bouncing near the interface along the separator, were carried out by the water phase.
Therefore, the existence of “re-entrainment” phenomenon, in which a fraction of oil
droplets are re-entrained by the water phase at high water velocities, was observed as part
of the CFD simulations.
To further study this phenomenon, additional case studies were performed at
various oil and water flow-rates using the same CFD model. Based on the production
conditions at 1988 and the future, the oil and water flow-rates were varied from their
minimum values to their maximum values in 10% increments. For each case study, the
oil and water droplets were injected and tracked to determine if they were carried out by
the other liquid phase. The results of CFD simulations indicated that independent of the
oil flow-rate value, re-entrainment by water phase occurred at a water flow-rate of greater
than 0.1595 m3/s. Hence, the previous lower water flow-rate of 0.1329 m3/s was assumed
as the maximum value that would avoid re-entrainment of the oil droplets. Using the
separator dimensions, the maximum water velocity was calculated to be 0.063 m/s. Based
on the maximum water production rate, a minimum cross-sectional area of 5.485 m2 was
calculated for water phase in order to avoid re-entrainment of the oil droplets.
CFD Simulation of Multiphase Separators 171

Using the data provided by the CFD simulations, the “stabilized” optimum separator was
then designed using the procedure of Monnery and Svrcek (1994), and the performance
of this separator was checked using CFD simulations. Figure 5-13 presents the
dimensions of the “stabilized” separator. The approximate vessel weight was calculated
to be 231530 kg which is some 1.23% higher than the non-stabilized separator design
shown in Figure 5-8b.
Table 5-8 presents the quality of the grid system produced in the Gambit
environment and does indicate that an absolutely negligible fraction of cells, around
0.0015%, is of poor quality. Figure 5-14 provides additional features of the generated
mesh system, and Appendix D includes the detailed fluid flow profiles. As expected, a
very high separation efficiency of 100% was calculated for the oil and water droplets for
the stabilized separator.

Figure 5-13. Geometrical Specifications of “Stabilized” Separator Designed for

Gullfaks-A.

Table 5-8. Quality of the mesh produced in Gambit environment for the “stabilized”
Gullfaks-A separator.
Number of Cells Maximum Squish Maximum Skewness Maximum Aspect Ratio
966574 0.746627 0.895882 44.5911
Skewness of the Produced Mesh
Skewness Range 0-0.20 0.20-0.40 0.40-0.60 0.60-0.80 0.80-1.0
Density of Cells 79.8406% 15.0993% 3.5702% 1.4884% 0.0015%
CFD Simulation of Multiphase Separators 172

Figure 5-14. The Grid System for the Stabilized Separator Designed for Gullfaks-A
Generated in Gambit Environment.

In summary, three different separators have been redesigned and proposed for Gullfaks-
A: The first separator design, labeled “redesigned” separator in Chapter Four, was based
on the CFD simulation results and the oilfield experiences. The originally installed
separator had a reasonable separation efficiency of 98.0% for the 1988 production
conditions. Hence, it was assumed that the provided residence times of continuous phases
in 1988 might be necessary for a new workable design. Using the maximum fluid flow-
rates, the separator was redesigned. The separation performance of this separator, as
CFD Simulation of Multiphase Separators 173

demonstrated by CFD simulations, was acceptable, but the design suffered from being
oversized. An interesting fact emerged from the CFD simulation of the “redesigned”
separator was connected with the water phase re-entrainment phenomenon. The
“redesigned” separator, interestingly, has the same cross-sectional area of 5.485 m2 for
water phase (suggested at above in the present subsection).
The second separator, labeled “weir” design, was designed based on the data
provided by CFD simulations and the two-phase models. Although this separator was
smaller than the “redesigned” separator, its separation efficiency was also lower (see
Table 5-9). Further CFD studies indicated that this minor inefficiency was caused by
water phase re-entrainment of oil.
Using both the data of two-phase models and the findings of the present studies,
the third separator, labeled “stabilized” separator, was designed for Gullfaks-A. This
design took advantage of all phase separation data provided by CFD simulations as well
as the logical optimization methodology presented by Monnery and Svrcek (1994). The
separation performance of the “stabilized” separator and its smaller dimensions compared
with those of the “redesigned” separator confirmed that “redesigned” separator was
oversized and a more economical separator that can accomplish the separation task of
Gullfaks-A could be designed. For comparison purposes, Table 5-9 shows the key
dimensions and separation efficiencies for the original, “redesigned”, “weir” and
“stabilized” separators proposed for Gullfaks-A.

Table 5-9. The key geometric specifications and separation efficiencies for various
designs proposed for Gullfaks-A.
Separation
Geometric Specifications
Efficiency
Design Layer Thickness (m) Gravity
D (m) L (m) Separation 1988 Future
Oil Phase Water Phase
Length (m)
Original 3.328 16.301 1.0390 0.6250 10.216 98.0% 70.4%
Redesigned 4.845 32.081 1.6300 1.6300 25.830 99.1% 99.7%
Weir 3.48 18.37 0.9525 0.9525 12.290 100% 90.4%
Stabilized 3.84 18.40 0.6000 1.9300 12.316 100% 100%
CFD Simulation of Multiphase Separators 174

5.5 Re-Entrainment Constraints

As noted in subsection 5.3.1 and demonstrated in subsection 5.4.2, continuous-phase high

velocities can lead to re-entrainment of droplets in horizontal separators. Note, based on
the developed design procedure for the vertical arrangement, continuous phase velocities
are always lower than droplet separation velocities, and re-entrainment is not an issue
with vertical separators. Thus, focusing on horizontal separators, a comprehensive CFD-
based study was carried out on the velocity constraints caused by re-entrainment. For this
purpose, the developed large-scale CFD models for different horizontal designs
(subsection 5.4.1) were used. The simulations were performed for all the oilfield
conditions (Table 5-1), and the results were analyzed to see if some general rules could
be proposed for the minimization of re-entrainment.

5.5.1 Vapor-Liquid Re-entrainment

Since the vapor-liquid separation compartment is essentially the same for the various
horizontal designs, only the “weir” designed separator developed for Gullfaks-A was
used as the CFD model. The oil and water flow-rates were set as per Gullfaks-A oilfield
at the 1988 production conditions. The vapor flow-rate was gradually changed from the
normal flow-rate of 1988 in increments of 0.2278 m3/s (corresponding to vapor velocity
increments of 0.054 m/s) to determine the vapor velocity at which the liquid droplets can
be re-entrained by the vapor phase from the vapor-liquid interface vicinity. To investigate
the issue, several parallel xy-planes were defined inside the vessel to record the
characteristics of the droplets passing through them, and the developed software was used
to analyze the database provided by the sampling planes. The maximum vapor phase
velocities predicted for various oilfields are reported in Table 5-10. Although the aim of
current study was not to study the re-entrainment phenomenon in detail, the data of Table
5-10 does indicate that high vapor densities and high oil viscosities reduce the maximum
allowable velocity of vapor phase. This trend is also in line with practical experience that
indicates high pressures and high oil viscosities limit the allowable vapor velocity, and
hence, reduce the gas capacity of the separator (Viles, 1993).
CFD Simulation of Multiphase Separators 175

Viles (1993) proposed a method for estimation of the maximum vapor phase velocity.
The method was, in fact, the explicit representation of the experimental correlations of
Ishii and Grolmes (1975) for predicting the onset of droplet re-entrainment. In the
experimental study, concurrent two-phase fluid flow systems composed of water-nitrogen
or water-helium inside a small-scale rectangular transparent apparatus with dimensions of
0.00317 m × 0.0254 m × 0.762 m were used. To specify which equation should be used,
an interfacial viscosity number ( N µ ) and the Reynolds number for surface liquid (oil

phase) are evaluated. The interfacial viscosity number is evaluated by Equation 5-17:

µ Oil
Nµ = (5-17)
σ Oil −Gas
ρ Oil σ Oil −Gas
g (ρ Oil − ρV )

where, µ Oil is oil viscosity in Pa.s, ρ Oil and ρ V are densities of oil and vapor phases,

respectively, in kg/m3, σ Oil −Gas is oil-gas surface tension in N/m, and g is gravity
acceleration in m/s2. For evaluation of the surface liquid Reynolds number, physical
properties of oil phase and hydraulic diameter of the total liquid phase (oil and water) are
used.

Table 5-10. The maximum safe velocities of vapor phase determined for the various
oilfield case studies.
Oilfield-1 Oilfield-2 Oilfield-3 Oilfield-4 Oilfield-5
3
Vapor Density (kg/m ) 17.46 49.7 9.50 5.70 2.44
Oil Viscosity (Pa.s) 0.73 × 10-3 5.25 × 10-3 6.90 × 10-3 42.0 × 10-3 100 × 10-3
CFD
U V , max Predictions 1.63 0.54 1.85 1.14 1.09
(m/s) Estimated by
2.62 1.06 2.36 0.82 0.54
Equation 5-18
CFD Simulation of Multiphase Separators 176

For the region of rough turbulent flow regimes ( Re Oil >1635), which is the case for all the

various oilfield case studies, the maximum vapor phase velocity ( VV ,max ) is estimated by

Equation 5-18 (Viles, 1993):

 0.8 σ Oil −Gas ρ Oil

N µ ; N µ ≤ 0.0667
 µ Oil ρ V
VV ,max = (5-18)
0.1146 σ Oil −Gas ρ Oil ; N > 0.0667
µ
 µ Oil ρV


Based on Equation 5-18 and the available oilfield data, Viles (1993) has presented some
general guidelines for the common case of rough turbulent flow regimes. The guideline
for the high-pressure operating region (P > 6900 kPa) is to use a vessel aspect ratio of
less than 5 to avoid droplet re-entrainment. Note, higher aspect ratios may be used at
lower pressures. Care should also be taken for heavy oil (<30°API) separator sizing as re-
entrainment becomes more likely as oil viscosity increases. For such cases, increasing the
operating temperature of the separator in order to reduce the oil viscosity has been
suggested (Viles, 1993).
Table 5-10 presents maximum safe velocities for several oilfield case studies as
predicted by Equation 5-18. These results are compared with the corresponding CFD
predictions in Figure 5-15. Figure 5-15 would indicate that some of the CFD simulation
data are underestimated and some are overestimated by Equation 5-18. Therefore, it is
concluded that the methodology proposed by Viles (1993) and the resultant guidelines are
only partially confirmed by the CFD simulations.

5.5.2 Liquid-Liquid Re-entrainment

The geometry of the liquid-liquid separation compartment is often different for horizontal
separator design configurations. Thus, all the horizontal vessel designs except for “boot”
design were part of the CFD simulation case studies. The reason why the “boot” design
has been excluded from the study will be explained later in this subsection.
CFD Simulation of Multiphase Separators 177

Figure 5-15. Maximum Safe Velocity for the Vapor Phase Estimated by Equation 5-18
versus the Corresponding CFD Predictions.

The CFD models are those developed for Gullfaks-A oil production platform (Subsection
5.4.1). The vapor flow-rate was set to the Gullfaks-A oilfield rate, and the oil and water
flow-rates were gradually varied from their minimum values in increments of 0.02658
m3/s which corresponded to velocity increments of 0.0083 m/s and 0.0126 m/s for oil and
water phases, respectively. For each case study, having set all the necessary CFD
parameters and obtaining the converged solution for continuous phases, oil and water
droplets were injected and tracked to see if they were re-entrained by the other liquid
phase. Again, parallel xy-planes were defined in the liquid-liquid interface vicinity
(sampling planes) providing data on the passing droplets. These data were analyzed using
the developed software.
The results of CFD simulations using “simple”, “weir”, and “bucket and weir”
models at selected oilfield conditions indicated that the oil phase does not re-entrain the
CFD Simulation of Multiphase Separators 178

water droplets even at a very high velocity of 1.11 m/s. Therefore, the focus was placed
on studying re-entrainment by water phase. Hence, the “boot” design was excluded from
the study since the flow-rate of water phase is typically very low in this design and
cannot re-entrain the oil droplets.
The CFD studies of the “simple” and “weir” separators showed that re-entrainment
by water phase might strongly influence the separation efficiency in these designs. Table
5-11 presents the maximum safe cross-sectional velocities for the “simple” and “weir”
designs predicted by the CFD simulations. The data of Table 5-11 do show that high oil
viscosities reduce the maximum safe velocity of water phase. Figure 5-16 shows the
variations in the maximum safe velocities of the water phase versus viscosity of the oil
phase for both “simple” and “weir” designs. As shown in Figure 5-16, a linear equation
can reasonably represent the CFD-based correlations for each design. Equation 5-19
presents the result of linear regression fit for VWater ,max - µ Oil data for the “simple” design:

VWater ,max_ simple = 0.08734 − 0.5696 µ Oil (5-19)

The R-squared value for this linear regression is 0.9522. Similarly, a linear regression fit
for VWater ,max - µ Oil data for the “weir” design is presented as Equation 5-20 with

corresponding R-squared value of 0.9595:

VWater ,max_ weir = 0.065767 − 0.4982 µ Oil (5-20)

Table 5-11. The maximum safe velocities of water phase in “simple” and “weir” designs
predicted for various oilfield case studies.
Oilfield-1 Oilfield-2 Oilfield-3 Oilfield-4 Oilfield-5
Oil Viscosity (Pa.s) 0.73 × 10-3 5.25 × 10-3 6.90 × 10-3 42.0 × 10-3 100 × 10-3
VWater ,max Simple Design 0.085 0.093 0.077 0.062 0.031
(m/s) Weir Design 0.069 0.063 0.063 0.038 0.019
CFD Simulation of Multiphase Separators 179

Figure 5-16. Maximum Safe Velocity of Water Phase for “Simple” and “Weir” Designs
versus Oil Phase Viscosity.

As Figure 5-16 shows, the regressed lines for the “simple” and “weir” designs are
approximately parallel to each other. In fact, the VWater ,max values for “weir” design are

around 0.018 m/s lower than the corresponding VWater , max values for “simple” design.

The results of CFD simulation for the “bucket and weir” design indicated that, in
contrast with the “simple” and “weir” designs, the oil droplets were not re-entrained by
the water phase even at a high velocity of 0.82 m/s. The obvious reason is the different
geometry of water phase compartment in the “bucket and weir” design. To clarify the
issue from a CFD point of view, various separation sections of “bucket and weir” design
can be compared with those of “simple” and “weir” designs. As shown in Figure 5-8, the
gravity separation zones for all three designs are the same. However, after the gravity
separation zone, the water phase flows through a complicated and totally different path in
the “bucket and weir” design resulting in no virtual path for the oil droplets to move with
CFD Simulation of Multiphase Separators 180

the water phase. Therefore, oil droplets separate from water phase near the oil weir and
become part of the continuous oil phase.
In summary, the CFD simulations did verify that oil droplets may be “re-
entrained” by the water phase at a high velocity in horizontal separators. Although
significant separation inefficiencies caused by “liquid-liquid” re-entrainment are not
likely to be experienced in “boot” and “bucket and weir” designs, a safe upper limit
should be empirically determined or estimated by Equations 5-19 and 5-20 for the water
phase velocity to avoid “liquid-liquid” re-entrainment in the “simple” and “weir” designs.
Therefore, the geometry of water phase compartment in horizontal separators should be
considered as a key factor in the liquid-liquid re-entrainment phenomenon.

5.6 Summary

CFD simulations were developed for selected aspects of phase separation. The focus was
placed on hydrocarbon-water systems, and the oilfield separator data ranging from light
oil conditions to heavy oil conditions were used in these multiphase separator
simulations. An efficient combination of two multiphase simulation models available in
Fluent, VOF and DPM, with appropriate model assumptions and settings was used. Two
independent sets of CFD simulations, one for vapor-liquid separation and the other for
liquid-liquid separation, were performed using simple and efficient grid systems.
When compared to classic design strategies, CFD simulations indicated that
additional residence times are necessary for droplets to pass through the interfaces. The
interface residence time may be as high as around 100 s depending on the fluid
properties.
In the vapor-liquid separation compartment, the efficient droplet size and the
appropriate extra vapor residence time (for droplet penetration through the interface)
were estimated as a function of the vapor density. It was shown that for the three-phase
separator case study, the efficient separation of oil droplets from gas phase results in total
separation of water droplets from the gas phase.
For the liquid-liquid separation process, the use of Abraham equation instead of
Stokes’ law was recommended since the upper limit of Stokes’ law was exceeded in
CFD Simulation of Multiphase Separators 181

several case studies. Using the Abraham equation in liquid-liquid separation calculations,
the efficient droplet size was estimated based on continuous phase viscosity. Hence, for
water droplets, a linear regression fit based on the oil phase viscosity was developed. An
efficient oil droplet size of 597 µm (or simply 600 µm ) resulted when the Abraham
equation was used for estimation of oil rising velocity in separator design procedures.
Furthermore, it was shown that the use of Stokes’ law for interpretation of CFD results
does lead to a weak correlation between efficient droplet sizes and continuous phase
viscosities.
The additional liquid-liquid interface residence times were estimated to be
proportional to the oil viscosity, however, 100 s for water phase and 10 s for oil phase are
recommended for conservative separator designs.
CFD simulations confirmed that the movement of oil and water droplets is
governed by the flow of continuous phase. So, in horizontal arrangements, oil and water
droplets completely follow the continuous phase flow in the horizontal direction, while in
vertical arrangements, apparent separation velocities of oil and water droplets are directly
affected by the continuous phase flow.
The relative separation velocities in vertical separators were almost identical to
the separation velocities predicted for horizontal separators. Thus, the separation
velocities and relationships for the horizontal separators can also be used for designing
vertical separators. However, it should be noted that the apparent separation velocity for a
vertical separator is equal to the separation velocity in horizontal separator minus the
continuous phase velocity in vertical separator. Therefore, apparent separation velocities
in horizontal separators are always higher than those in vertical separators.
The algorithmic design method of Monnery and Svrcek (1994) was modified to
use CFD simulation results to specify a realistic optimum separator design/size. The
performance of the designed separator was also verified using the CFD simulations.
Finally, a comprehensive CFD-based study on the velocity constraints caused by
re-entrainment in horizontal separators was carried out. Practical experience and CFD
simulations show that high vapor densities and high oil viscosities reduce the maximum
safe velocity of the vapor phase. An empirically-based method proposed by Viles (1993)
CFD Simulation of Multiphase Separators 182

for predicting vapor-liquid re-entrainment was also tested and the results were compared
with the CFD results of the current study. This method was based on the empirical study
of Ishii and Grolmes (1975) in which only water-nitrogen and water-helium fluid systems
in a small-scale rectangular apparatus were used for data generation. Therefore, the
equations developed by Viles (1993) could only approximate the predicted CFD results.
The results of CFD simulations, using all the feasible horizontal designs and selected
oilfield conditions, indicated that the oil phase does not re-entrain the water droplets, but
the oil droplets may be “re-entrained” by the water phase at a high velocity. In the
“simple” and “weir” designs, it was observed that re-entrainment by the water phase may
strongly influence the separation efficiencies. For these designs, the maximum safe
velocity of water phase was regressed as a linear function of the oil viscosity. The
regressed lines were almost parallel, and the maximum safe water velocities for the
“weir” design were 0.018 m/s lower than the corresponding values for the “simple”
design.
The “boot” design, because of its very low water phase flow-rate, was excluded
from the study. Moreover, CFD simulation performed on “bucket and weir” design
indicated that oil droplets are not re-entrained by water phase. Therefore, it was
concluded that significant separation inefficiencies caused by “liquid-liquid” re-
entrainment are not likely to be experienced in “boot” and “bucket and weir” designs, but
a correct upper limit should be assumed for water phase velocity to avoid “liquid-liquid”
re-entrainment in the “simple” and “weir” designs. Consequently, the geometry of water
phase compartment in horizontal separators is a key factor affecting the liquid-liquid re-
entrainment phenomenon.
CFD Simulation of Multiphase Separators 183

Chapter Six: Conclusions and Recommendations

This research project has provided a detailed approach to the use of CFD as a tool for the
modeling and simulation of the performance of the multiphase separators. The developed
CFD models provided both macroscopic and microscopic understanding of the phase
separation phenomenon. Compared with the previous CFD based studies of multiphase
separators, the current study does provide realistic strategies for CFD simulation of
multiphase separators, and the thesis does provide all details of developed CFD models.
The other significant accomplishments are as follows:

1- Symmetrical fluid flow profiles have generally been assumed in previous studies,
and only half of the separator volume has been modeled. However, even the
results of these simplified CFD models prove that the plug flow regime cannot be
assumed, hence, the assumption of symmetrical fluid flow profiles is not realistic.
In this study the total volume of multiphase separators was modeled.
2- The quality of produced computational grid system has not been verified in the
previous studies, and coarse grid systems that have been used because of
computational restrictions in some studies, would produce poor/doubtful results.
The high quality of produced computational grid systems was verified in the
present study.
3- In order to reduce the required computational memory and time in some of the
previous studies, three-phase fluid flows have been simulated by two-phase CFD
models. This approach would reduce the validity of produced CFD profiles and
was not used for simulation of three-phase separators in the current research.
Instead, all fluid phases present in a multiphase separator with their corresponding
physical properties were considered in the CFD simulations, thus providing very
high quality details of the phase separation features.
4- In the previous studies, only indirect criteria such as the liquid retention time, the
volumetric utilization, and the standard deviation of time averaged velocity have
in general been used for the evaluation of separation efficiencies. Improving these
factors can lead to better plug flow regimes inside separators, but it cannot
CFD Simulation of Multiphase Separators 184

necessarily lead to increased separation efficiencies. In the current CFD study, to

define an effective criterion, separation efficiencies were evaluated directly based
on the mass distribution of fine fluid droplets as they were tracked by a suitable
multiphase CFD model, DPM. Thus, the realistic performance of the separators
was simulated.
5- In the previous studies, CFD based modifications have generally been
concentrated on the separator internals such as flow-distributing baffles.
However, as demonstrated in this study and previously emphasized by Lyons and
Plisga (2005), optimizing the separator internals has only a minor effect on the
separator performance, and an essentially inefficient separator cannot become
workable simply by optimizing its internals. To overcome this shortcoming, the
current research applied realistic modifications to an operating large-scale
separator and also proposed useful and improved design criteria for existing
design methods.

Three major CFD based studies were carried out during this research project. The most
significant features and results of these studies will be presented in the following
sections.

6.1 CFD Simulation of Pilot-Plant-Scale Two-Phase Separators

The paper titled “Analytical Study of Liquid/Vapor Separation Efficiency”, by Monnery

and Svrcek (2000), and a field pilot plant skid at the Prime West East Crossfield gas plant
were used as the basis for the CFD model and to provide experimental data. Two
approaches were implemented in the CFD simulations. The overall strategy in the first
approach, DPM multiphase modeling, was adopted from Newton et al. (2007) and only
the vapor-liquid compartment of the separators was simulated. In this approach, the gas-
liquid interface was assumed to be a frictionless wall which trapped the droplets coming
into contact with it. In the second approach, an effective combination of DPM and VOF
multiphase models within Fluent 6.3.26 was used. This CFD model was based on the
physics of involved phase separation process and the characteristics of the Fluent
CFD Simulation of Multiphase Separators 185

multiphase models. To implement the combined VOF-DPM model, having developed a

CFD simulation of the overall phase behavior of the fluid flows using the VOF model,
droplets of oil and water were injected at the inlet nozzle and tracked by DPM model.
The important conclusions that can be drawn from these two approaches are:

• Although the DPM approach was quite successful in predicting the incipient
velocities, the produced diagrams for separation efficiency versus gas velocity were
not realistic. In fact, modeling only the gas phase compartment was not sufficient to
capture all details of the phase separation phenomenon. The approach shortcoming is a
result of ignoring the existence of the continuous liquid phase. Thus, gas-liquid
interactions and, more significantly, the dynamic interactions between liquid droplets
and continuous liquid phase are neglected in the DPM approach.
• Three different mean droplet sizes were tested for prediction of incipient velocities in
the DPM approach. The CFD results showed a minor difference in favor of the case of

d = 150 µm . Thus, the use of this well-known design value for prediction of
incipient velocity was validated by the study.
• The VOF-DPM approach was a substantial modification to the DPM approach as the
model did include the continuous liquid phase within the calculations. From a practical
point of view, the obtained separation efficiency data and diagrams were correct and
the poor behavior of the DPM-only approach, such as perfect separation at velocities
higher than the incipient velocity, was completely eliminated. There was excellent
agreement between simulated phase separation behavior and the experimental data and
observations in most of the case studies.
• Based on the obtained fluid flow profiles, all the two-phase separators were shown to
essentially operate at a constant pressure.
• In the developed CFD simulations, for the first time, both droplet coalescence and
breakup were modeled. The results showed that while droplet coalescence occurred
rarely, droplet breakup was a common phenomenon particularly at high velocities. The
simulation results do show that higher velocities intensified the number of droplet
CFD Simulation of Multiphase Separators 186

breakups in horizontal separators while in vertical separators, higher pressures

stabilized the number of breakups to a constant rate of about 20%.
• Using the developed C++ software, the size distribution of the droplets exiting through
the gas-outlet in different separators was calculated. The results showed that mist
eliminators may operate more efficiently in horizontal separators than in vertical
separators. In horizontal separators, the particle size distribution was always very
narrow and the emerging droplets were totally dominated by droplets of mean
diameter larger than 10 µm , which can be separated by a properly designed demister
within reasonable operating conditions and separator capacity. However, the average
values of the spread parameter were generally lower for the vertical separator than
those for the horizontal separators, and both very fine and very large droplets were still
present in the vertical separator gas outlet.

6.2 CFD Based Study of a Large-Scale Three-Phase Separator

A three-phase separator located in the Gullfaks oilfield in the Norwegian sector of the
North Sea was simulated. Based on the characteristics of the Fluent multiphase models
and the obtained results for the pilot-plant scale separators, the combined VOF-DPM
model of Fluent 6.3.26 was implemented. In this study, the installed distribution baffles
and mist eliminator were modeled using the Porous Media Model of Fluent which
required the available specifications and design information for the three-phase separator.
Using the available theoretical approaches and experimental correlations, a useful
methodology for estimation of droplet size distribution, which is necessary for
implementing DPM approach, was developed. In order to overcome serious separation
problems experienced by the projected increase in the water flow-rate and to enhance the
separation efficiency, design approaches were also tested using CFD simulations. The
significant results and conclusions are presented in the following:

• Compared to the original study of Hansen et al. (1993), the developed model did
provide more rational and high-quality details of fluid flow profiles leading to a
realistic overall picture of the phase separation in all zones of the separator. Thus, the
CFD Simulation of Multiphase Separators 187

realistic performance of the separator was simulated and the microscopic features of
the three-phase separator were studied in detail.
• The CFD simulations indicated that the droplet breakup with an average rate of 76%
was a common phenomenon when the dispersed droplets came into contact with the
deflector baffle. Because of these droplet breakups, the volume median diameter of
droplets was predicted to decrease to about 67% of its initial value.
• In line with the oilfield experience, the CFD simulations showed that serious
separation inefficiencies would be encountered with the existing separator at the
projected increase in the flow-rate of produced water.
• To overcome the inefficiencies, minor modifications such as adjusting baffle positions
and liquid levels were tested via CFD simulations. However, the results did show that
the minor modifications can not resolve the essential separation inefficiencies in the
Gullfaks-A separator. Therefore, the separator was redesigned using the classical
design method of Monnery and Svrcek (1994) and implementing appropriate liquid
retention times as were implied by the satisfactory performance of the separator at the
1988 production conditions.
• The excellent separation performance, shown in the CFD simulation, of the redesigned
separator did confirm that, a realistic optimum separator can be specified if the
algorithmic method of Monnery and Svrcek (1994) is modified to use realistic
separation velocities.
• It was shown that the popular classic methods, mostly due to a lack of a useable
mathematical model for estimation of droplet “separation velocities”, do result in a
conservative design and would specify extremely oversized separators for Gullfaks-A.
• Sensitivity analyses performed on the developed CFD model for Gullfaks-A separator
reconfirmed the high quality of the grid system and robustness of the DPM results
with respect to solution repetitions. These analyses also indicated that some
uncertainties while defining droplet size distribution or estimating surface tensions had
only little effect on the CFD simulation of multiphase separators.
• Although the overall performance of wire mesh demisters, such as pressure and
velocity profiles and some separation features, could be properly simulated, the
CFD Simulation of Multiphase Separators 188

separation efficiency of wire mesh demisters and their operability ranges could not be
simulated using the existing CFD models. The CFD models appear to under-predict
the demister efficiency. This shortcoming is possibly a result of the complicated
separation mechanisms involved with wire mesh demisters.

6.3 Improved Design Criteria

The aim of this phase of the research project was to exploit the implemented VOF-DPM
model assumptions/settings and provide some generally useful updated phase separator
design criteria. These criteria could then be combined with the algorithmic design method
proposed by Monnery and Svrcek (1994) to specify a realistic optimum separator. In
order to simulate the various aspects of the phase separation phenomenon, oilfield
separator data ranging from light oil conditions to heavy oil conditions were used. As a
result, a systematic method for estimation of realistic separation velocities was
developed. The velocity constraints caused by re-entrainment in horizontal separators
were also studied via the comprehensive CFD simulations. The important CFD-based
findings are highlighted in the following:

• In contrast with classic design strategies, additional residence times should be

assumed for droplet penetration through the interfaces. The appropriate extra vapor
and liquid residence times for droplet penetration through the interfaces were shown to
be function of the vapor density and the oil viscosity, respectively. For conservative
separator designs, some 60 s for vapor phase, 100 s for water phase and 10 s for oil
phase can be assumed as the required interface residence times.
• In the liquid-liquid separation compartment, the use of the Abraham equation instead
of Stokes’ law was recommended because the upper limit of Stokes’ law was
exceeded in several case studies. Moreover, it was shown that incorrect use of Stokes’
law for interpretation of CFD results does lead to a weak correlation between design
droplet sizes and continuous phase viscosities.
• The efficient/design droplet size for estimation of settling velocities in vapor and oil
phases was shown to be function of the vapor density and oil viscosity, respectively.
CFD Simulation of Multiphase Separators 189

An oil droplet size of around 600 µm can be assumed if the Abraham equation is used
for estimation of the rising velocity of oil droplets out of the water phase.
• CFD simulations do show that the movement of oil and water droplets is completely
influenced by the continuous phase flow. Thus, in horizontal arrangements, oil and
water droplets are part of the continuous phase flow in their horizontal movements,
while for vertical arrangements, apparent separation velocities of the oil and water
droplets are directly affected by the continuous phase flow.
• The relative separation velocities in vertical arrangements are almost identical to the
separation velocities predicted for horizontal arrangement. Therefore, the separation
velocities and relationships obtained for the horizontal arrangement can be used for
designing vertical separators. However, note that the apparent separation velocity for a
vertical separator is equal to the separation velocity in horizontal separator minus the
continuous phase velocity in vertical separator. Thus, apparent separation velocities in
horizontal separators are always higher than those in vertical separators.
• As verified by CFD simulations, the algorithmic method proposed by Monnery and
Svrcek (1994) can be modified to use realistic separation velocities to specify an
effective optimum separator design/size.
• In line with practical experience, CFD simulations indicated that high vapor densities
and high oil viscosities reduce the maximum safe velocity of vapor phase.
• The results of CFD simulations, using all the feasible horizontal designs and various
oilfield conditions, indicated that the oil phase does not re-entrain the water droplets.
However, it was observed that re-entrainment by water phase may strongly influence
the separation efficiencies in the “simple” and “weir” designs. For these designs, the
maximum safe cross-sectional velocity of water phase was regressed as a linear
function of the oil viscosity.
• The geometry of water phase compartment in horizontal separators was shown to be a
key affecting factor in the liquid-liquid re-entrainment phenomenon.
CFD Simulation of Multiphase Separators 190

6.4 Recommendations

This research project did clearly show the benefits CFD analyses can provide in
optimizing the design of new separators and solving problems with existing designs.
Consequently, a direct tangible benefit to industry is expected through applying the
developed CFD simulation strategies and the established new/improved design criteria.
Recommendations for future research would include the following:

• With expected development in CFD modeling tools, the complicated separation

mechanisms involved with wire mesh demisters can be simulated, and the realistic
separation performance of wire mesh demisters would be studied.
• As the rate of droplet coalescence has always been predicted to be very low even for
the CFD model developed for the wire mesh demister, the validity of the droplet
collision model of the Fluent need be investigated using experimental data.
• In the previous CFD studies, indirect criteria such as the fluid flow profiles, the liquid
retention time, and the volumetric utilization have been used for improving the
position or the configuration of separator internals. As noted, these criteria do not
necessarily lead to increased separation efficiencies. Thus, it is recommended that the
direct separation efficiency criterion developed in this research project be used for the
CFD optimization of separator internals.
• Although the developed new/improved design criteria resulted in a realistic optimum
separator for the Gullfaks-A case study, it is recommended that these criteria be
empirically validated using different oilfield fluids in the large-scale multiphase
separators.
CFD Simulation of Multiphase Separators 191

References

Abernathy, M.W.N., “Gravity Settlers, Design”, in “Unit Operation Handbook”, J.J.

McKetta (Ed.), Vol. 2, Marcel Dekker, 1993.
Abraham, F.F., “Functional Dependence of Drag Coefficient of a Sphere on Reynolds
Number”, Physics of Fluids, 13, 1970, 2194-2195.
Anderson, J.D., “Computational Fluid Dynamics, The Basics with Applications”,
McGraw-Hill, 1995.
Angeli, P., Hewitt, G.F., “Drop Size Distribution in Horizontal Oil-Water Dispersed
Flows”, Chem. Eng. Sci., 55, 2000, 3133-3143.
Antonoff, G.N., J. Chim. Phys., 5, 1907, 372.
Arnold, K., Stewart, M., “Surface Production Operations”, 3rd Edition, Elsevier, 2008.
Arntzen, R., “Gravity Separator Revamping”, Dr.-Ing. Dissertation, Norwegian
University of Science and Technology, Trondheim, Norway, 2001.
Austrheim, T., “Experimental Characterization of High-Pressure Natural Gas Scrubbers”,
PhD Dissertation, University of Bergen, Bergen, Norway, 2006.
Blezard, R.G., Bradburn, J., Clark, J.G., Cohen, D.H., Costaschuk, D., Downie, A.A.,
Fowler, P., Hassoun, L., Hunt, A.P., Kirton, D., Knight, F.I., Lach, J.R., Law, E.J.,
McDonald, P.A., Morrison, A.K., Cairney, J.M., Naik, H., Sutton, W.J.E.,
Thompson, P., “Production Engineering”, in “Modern Petroleum Technology”, R.A.
Dawe (Ed.), 6th Edition, Vol. 1, Institution of Petroleum, John Wiley & Sons, 2000.
Branan, C., “The Process Engineers Pocket Handbook”, Vol. 2, Gulf, 1983.
Chin, R.W., Stanbridge, D.I., Schook, R., “Increasing Separation Capacity with New and
Proven Technologies”, Society of Petroleum Engineers, SPE-77495, 2002, 1-6.
Churchill, S.W., “Viscous Flows, The Practical Use of Theory”, Butterworths, 1988.
Coker, A.K., “Ludwig's Applied Process Design for Chemical and Petrochemical Plants”,
Vol. 1, 4th Edition, Elsevier, 2007.
El-Dessouky, H.T., Alatiqi, I.M., Ettouney, H.M., Al-Deffeeri, N.S., “Performance of
Wire Mesh Mist Eliminator”, Chemical Engineering and Processing, 39, 2000, 129-
139.
CFD Simulation of Multiphase Separators 192

Evans, F.L., “Equipment Design Handbook for Refineries and Chemical Plants”, Vol. 2,
Gulf, 1974.
Ferziger, J.H., Peric, M., “Computational Methods for Fluid Dynamics”, Springer-
Verlag, Heidelberg, 1996.
Fewel, K.J., Kean, J.A., “Computer Modeling Aids Separator Retrofit”, Oil & Gas
Journal, 90(27), 1992, 76-80.
Fluent 6.3.26, "Fluent 6.3 User’s Guide", Fluent Inc., Centerra Resource Park, 10
Cavendish Court, Lebanon, USA, 2006.
Fluent 6.3.26, Commercial Computational Fluid Dynamics Simulator, Fluent Inc., 2006.
Frankiewicz, T., Browne, M.M., Lee, C-M., “Reducing Separation Train Sizes and
Increasing Capacity by Application of Emerging Technologies”, Offshore
Technology Conference, OTC-13215, 2001, 1-10.
Frankiewicz, T., Lee, C-M., “Using Computational Fluid Dynamics (CFD) Simulation to
Model Fluid Motion in Process Vessels on Fixed and Floating Platforms”, Society of
Petroleum Engineers, SPE-77494, 2002, 1-9.
Gambit 2.4.6, Preprocessor Tool of Fluent, Fluent Inc., 2006.
Gas Processors Suppliers Association, GPSA Engineering Data Book, 11th Edition, Vol.
1, Gas Processors Association, 1998.
Gerunda, A., “How to Size Liquid-Vapor Separators”, Chemical Engineering, May 4,
1981, 81-84.
Gosman, A.D., “Developments in Industrial Computational Fluid Dynamics”, Trans
IChemE, 76(A), 1998, 153-161.
Green, D.W., Perry, R.H., “Perry’s Chemical Engineers’ Handbook”, 8th Edition,
McGraw-Hill, 2008.
Grødal, E.O., Realff, M.J., “Optimal Design of Two- and Three-Phase Separators: A
Mathematical Programming Formulation”, Society of Petroleum Engineers, SPE
56645, 1999, 1-16.
Hallanger, A., Soenstaboe, F., Knutsen, T., “A Simulation Model for Three-Phase
Gravity Separators”, Society of Petroleum Engineers, SPE-36644, 1996, 695-706.
CFD Simulation of Multiphase Separators 193

Hansen, E.W.M., Celius, H.K., Hafskjold, B., “Fluid Flow and Separation Mechanisms in
Offshore Separation Equipment”, 1st International Symposium on Two-Phase Flow
Modeling and Experimentation, 1995, 117-129.
Hansen, E.W.M., Heitmann, H., Lakså, B., Ellingsen, A., Østby, O., Morrow, T.B.,
Dodge, F.T., “Fluid Flow Modeling of Gravity Separators”, 5th International
Conference on Multiphase Production, 1991, 364-380.
Hansen, E.W.M., Heitmann, H., Lakså, B., Løes, M., “Numerical Simulation of Fluid
Flow Behavior Inside, and Redesign of a Field Separator”, 6th International
Conference on Multiphase Production, 1993, 117-129.
Heidemann, R.A., Jeje, A.A., Mohtadi, F., “An Introduction to the Properties of Fluids
and Solids”, The University of Calgary Press, 1987.
Helsør, T., Svendsen, H., “Experimental Characterization of Pressure Drop in Dry
Demisters at Low and Elevated Pressures”, Trans IChemE, 85(A3), 2007, 377-385.
Hesketh, R.P., Fraser Russel, T.W., Etchells, A.W., “Bubble Size in Horizontal
Pipelines”, AIChE J., 33(4), 1987, 663-667.
Hinze, J.O., “Fundamentals of the Hydrodynamic Mechanism of Splitting in Dispersion
Processes”, AIChE J., 1(3), 1955, 289-295.
Hooper, W.B., “Decantation”, Section 1.11 in “Handbook of Separation Techniques for
Chemical Engineers”, Ph.A. Schweitzer (Ed.), 3rd Edition, McGraw-Hill, 1997.
HYSYS 3.2, Commercial Chemical Process Simulator, Hyprotech Ltd., 2003.
Ishii, M., Grolmes, M.A., “Inception Criteria for Droplet Entrainment in Two-Phase
Concurrent Film Flow”, AIChE J., 21(2), 1975, 308-317.
Ishii, M., Zuber, N., “Drag Coefficient and Relative Velocity in Bubbly, Droplet or
Paticulate Flows”, AIChE J., 25(5), 1979, 843-855.
Issa, R.I., “Solution of the Implicitly Discretized Fluid Flow Equations by Operator-
Splitting”, Journal of Computational Physics, 62, 1986, 40-65.
Karabelas, A.J., “Droplet Size Spectra Generated in Turbulent Pipe Flow of Dilute
Liquid/Liquid Dispersions”, AIChE J., 24(2), 1978, 170-180.
Kim, H., Burgess, D.J., “Prediction of Interfacial Tension between Oil Mixtures and
Water”, J. Collide Interface Sci., 241, 2001, 509-513.
CFD Simulation of Multiphase Separators 194

King, R.P., “Introduction to Practical Fluid Flow”, Butterworth-Heinemann, 2002.

Kolmogorov, A.N., “On the Breaking of Drops in Turbulent Flow”, Doklady Akad.
Nauk. SSSR, 66, 1949, 825-828.
Kolodzie, P.A., Van Winkle, M., “Discharge Coefficients through Perforated Plates”,
AIChE J., 3(3), 1957, 305-312.
Launder, B.E., Spalding, D.B., “Mathematical Models of Turbulence”, Academic Press,
1972.
Lee, C-M., Dijk, E.V., Legg, M., “Field Confirmation of CFD Design for FPSO-mounted
Separator”, Offshore Technology Conference, OTC-16137, 2004, 1-6.
Lee, J.M., Khan, R.I., Phelps, D.W., “Debottlenecking and Computational Fluid
Dynamics Studies of High and Low-Pressure Production Separators”, Society of
Petroleum Engineers, SPE-115735, 2009, 124-131.
Levich, V.G., “Physiochemical Hydrodynamics”, Prentice Hall, 1962.
Lu, Y., Lee, J.M., Phelps, D., Chase, R., “Effect of Internal Baffles on Volumetric
Utilization of a FWKO - A CFD Evaluation”, Society of Petroleum Engineers, SPE-
109944, 2007, 1-6.
Lyons, W.C., Plisga, G.J. (Editors), “Standard Handbook of Petroleum and Natural Gas
Engineering”, Volume 2, Gulf Professional Publishing, 2005.
McCain, W.D., “The Properties of Petroleum Fluids”, Petroleum Publishing Company,
1973.
Monnery, W.D., Svrcek, W.Y., “Successfully Specify Three-Phase Separators”, Chem.
Eng. Progress, September, 1994, 29-40.
Monnery, W.D., Svrcek, W.Y., “Analytical Study of Liquid/Vapor Separation
Efficiency”, PTAC, 2000.
Navier, C.L.M.H., “Mémoire sur les lois du movement des fluides”, Mémoires de
VAcadémie des Sciences de VInstitut de France , 6, 1822, 389-440.
Newton, T., Connolly, D., Mokhatab, S., “Tools to Model Multiphase Separation”,
Chem. Eng. Progress, 103(6), 2007, 26-31.
CFD Simulation of Multiphase Separators 195

Patankar, S.V., Spalding, D.B., “A Calculation Procedure for Heat, Mass and Momentum
Transfer in Three-dimensional Parabolic Flows”, International Journal of Heat and
Mass Transfer, 15, 1972, 1787–1806.
Perry, R.H., Green, D.W., Maloney, J.O., “Perry’s Chemical Engineers’ Handbook”, 7th
Edition, McGraw-Hill, 1999.
Poling, B.E., Prausnitz, J.M., O’Connell, J.P., “The Properties of Gases and Liquids”, 5th
Edition, McGraw-Hill, 2001.
Rosin, P., Rammler, E., “The Laws Governing the Fineness of Powdered Coal”, J. Inst.
Fuel, 7, 1933, 29-36.
Sharratt, P.N., “Computational Fluid Dynamics and its Application in the Process
Industries”, Trans IChemE, 68-A, January, 1990, 13-18.
Shelley, S., “Computational Fluid Dynamics – Power to the People”, Chem. Eng.
Progress, 103(4), April, 2007, 10-13.
Sinnott, R.K., “Chemical Engineering Design” in “Coulson & Richardson’s Chemical
Engineering”, 2nd Edition, Butterworth-Heinemann, 1997.
Skelton, G.F., “Production”, in “Our Industry Petroleum”, Stockil, P.A. (Ed.), British
Petroleum Company Limited, 1977.
Sleicher, C.A., “Maximum Stable Drop Size in Turbulent Flow”, AIChE J., 8(4), 1962,
471-477.
Smith, H.V., “Oil and Gas Separators”, in “Petroleum Engineering Handbook”, Bradley,
H.B. (Ed), Society of Petroleum Engineers, 1987.
Stokes, G.G., “On the Theories of Internal Friction of Fluids in Motion, and of the
Equilibrium and Motion of Elastic Solids”, Transaction of the Cambridge
Philosophical Society, 8(22), 1845, 287-305.
Streeter, V.L., Wylie, E.B., “Fluid Mechanics”, 8th Edition, McGraw-Hill, 1985.
Svrcek, W.Y., Monnery, W.D., “Design Two-Phase Separators within the Right Limits”,
Chem. Eng. Progress, October, 1993, 53-60.
Swartzendruber, J., Fadda, D., Taylor, D., “Accommodating Last Minute Changes: Two
Phase Separation Performance Validated by CFD”, ASME Fluids Engineering
Division Summer Meeting and Exhibition, Proceedings of FEDSM, 2005, 713-715.
CFD Simulation of Multiphase Separators 196

Tatterson, D.F., Dallman, J.C., Hanratty, T.J., “Drop Sizes in Annular Gas-Liquid
Flows”, AIChE J., 23(1), 1977, 68-76.
Verlaan, C.C.J., Olujic, Z., De Graauw, J., “Performance Evaluation of Impingement
Gas-Liquid Separators”, Proceedings of the 4th International Conference on
Multiphase Flow, 1989, 103-115.
Vetter, O.J., Bent, M., Kandarpa, V., Salzman, D., Williams, R., “Three-Phase PVT and
CO2 Partitioning”, Society of Petroleum Engineers, SPE 16351, 1987, 297-310.
Viles, J.C., “Predicting Liquid Re-Entrainment in Horizontal Separators”, Society of
Petroleum Engineers, Journal of Petroleum Technology, 1993, 405-409.
Visser, R.C., “Offshore Production of Heavy Oil”, J. Petroleum Technology, 1989, 67-
70.
Walas, S.M., “Process Vessels”, Chapter 18 in “Chemical Process Equipment Selection
and Design”, Butterworth-Heinemann, 1990.
Watkins, R.N., “Sizing Separators and Accumulators”, Hydrocarbon Proc., 46(11), 1967.
Wilkinson, D., Waldie, B., “CFD and Experimental Studies of Fluid and Particle Flow in
Horizontal Primary Separators”, Trans IChemE, 72(A), 1994, 189-196.
Wilkinson, D., Waldie, B., Nor, M.I.M., Lee, H.Y., “Baffle Plate Configuration to
Enhance Separation in Horizontal Primary Separators”, Chem. Eng. J., 77, 2000,
221-226.
Wu, F.H., “Separators, Liquid-Vapor, Drum Design”, in “Encyclopedia of Chemical
Processing and Design”, J.J. McKetta, W.A. Cunningham (Ed.), Marcel Dekker,
1990.
CFD Simulation of Multiphase Separators 197

Appendix A: Design of Three-Phase Separators

The aim of Appendix A is to provide the approach and the main steps proposed by
Monnery and Svrcek (1994) for designing the optimum three-phase separator. In their
design approach, the economical aspect ratio of a separator is assumed to be between 1.5
and 6 with a functionality of operating pressure as presented in Table A-1.

A.1 Vertical Configuration

First, the vessel diameter is calculated based on satisfactory separation of oil droplets
from gas phase. Then, the heights of the light and heavy liquids are assumed, and the
liquid-liquid separation velocities and times are calculated. If the residence times of the
continuous liquid phases are not larger than the required separation times, then the vessel
diameter will be increased to satisfy the liquid-liquid separation requirements. Figure A-1
shows a schematic of the vertical three-phase separator for which the design procedure is
outlined here:
1. Calculate the terminal settling velocity of oil droplets using Equation 2-1 or Equation
2-4.
2. Set the vapor velocity equal to 75% of the terminal settling velocity.
3. Calculate all of the volumetric flow-rates.
4. Calculate the vessel diameter based on the vapor volumetric flow-rate and the vapor
velocity.
5. Calculate the separation velocities of both the liquid phases through each other using
Stokes’ law.
6. Set the thickness of liquid phases (assume H L = H H = 30 cm as minimum), and
calculate the separation times for the liquid droplets ( t HL and t LH ).
7. Calculate the cross-sectional areas of the liquid phases. Note, this area is the same as
the vessel cross-sectional area for heavy liquid phase ( AH = A ), but in the case of
using a baffle plate down-comer, the area allotted to baffle plate should be subtracted
from the vessel cross-sectional area to obtain the area of light liquid phase
( AL = A − AD ).
CFD Simulation of Multiphase Separators 198

Table A-1. The aspect ratio values suggested by Monnery and Svrcek (1994) for
multiphase separators.
P (kPa ) 0-1700 1700-3400 0>3400
Separator Aspect Ratio 1.5-3 3-4 4-6

Figure A-1. The Vertical Three-Phase Separator.

H L AL
8. Calculate the residence time of the light liquid: θ L = . If θ L < t HL , increase the
QLL

vessel diameter so that θ L = t HL .

CFD Simulation of Multiphase Separators 199

H H AH
9. Calculate the residence time of the heavy liquid: θ H = . If θ H < t LH , increase
QHL

the vessel diameter so that θ H = t LH .

10. Calculate the height of light liquid phase above the outlet ( H R ) based on the required
holdup time.
11. Calculate the surge height ( H S ) based on the surge time.

HT
12. Calculate the vessel height ( H T ). If is not in the range of 1.5-6.0, increase the
D
diameter (to decrease the ratio) or height (to increase the ratio) of the separator and
fix the problem.

A.2 Horizontal Configuration

An iterative design procedure is required to determine the most economical separator. At

each iteration, with an assumed diameter, the vapor disengagement area is set to provide
a satisfactory separation of liquid droplets from gas phase, and the heights of the light and
heavy liquids are assumed. Then, the lengths required by vapor-liquid separation and
retention time requirements are calculated. Similar to the vertical configuration, if the
residence times are not greater than the required separation times, then the separator size
should be increased. Note, when increasing the length (preferably) or diameter of a
separator, the separator aspect ratio should be in the acceptable range of 1.5-6.
The different common designs of horizontal three-phase separator, composed of
“simple”, “boot”, “weir”, and “bucket and weir”, are illustrated in Figure A-2 to Figure
A-5, respectively. In design procedure, first, the terminal settling velocity of oil droplets
is estimated using Equation 2-1 or Equation 2-4 and vapor velocity is set to be equal to
75% of the terminal settling velocity. Then, holdup and surge volumes ( VH and VS ) are
calculated based on given holdup and surge times and volumetric flow-rate of light liquid
( Q LL ). The other steps differ from one design to another and are outlined in the following
subsections.
CFD Simulation of Multiphase Separators 200

A.2.1 “Simple” Design

1. Pick an aspect ratio from Table A-1 and calculate the initial vessel diameter:

4(V H + VS )
D= .
3 L
0.5π  
D

πD 2
2. Calculate vessel cross-sectional area: A = .
4
3. Set H H and H L (assume H L = H H = 30 cm as minimum), and calculate
( AH + AL ) .
4. Set H V to the larger of 0.20 × D or 0.60 m, and then calculate AV .
5. Calculate the vessel length based on the liquid holdup/surge:
V H + VS
L= .
A − AV − ( AH + AL )

HV
6. Calculate the liquid dropout time: φ = .
VV

7. Calculate the actual vapor velocity using QV and AV .

8. Calculate the minimum length required for vapor-liquid separation (Lmin) using the
actual vapor velocity and the liquid dropout time.
9. If L < 0.8Lmin , increase H V , and go to step 4. Else if L < Lmin , set L = Lmin . Else if

L > 1.2 Lmin , decrease H V (if acceptable), and go to step 4. Else, L is acceptable.
10. Calculate the separation velocities of both the liquid phases through each other using
Stokes’ law.
11. Calculate the separation times of the liquid droplets ( t HL and t LH ).
L( A − AV − AH )
12. Calculate the residence time of the light liquid: θ L = . If θ L < t HL , set
QLL

t HL QLL
L= .
A − AV − AH
CFD Simulation of Multiphase Separators 201

Figure A-2. The “Simple” Design of Horizontal Three-Phase Separator.

LAH
13. Calculate the residence time of the heavy liquid: θ H = . If θ H < t LH , set
QHL

t LH QHL
L= .
AH
L L
14. If < 1.5 , decrease D (if acceptable), and go to step 2. Else if > 6 , increase D ,
D D
and go to step 2.
15. Calculate the approximate vessel weight based on thickness and surface area of shell
and heads.
16. In order to find the optimum case (corresponding to the minimum weight), change the
vessel diameter by 15 cm increments, and repeat the calculations from step 2 while
keeping the aspect ratio in the range of 1.5 to 6.0.

A.2.2 “Boot” Design

1. Pick an aspect ratio from Table A-1 and calculate the initial vessel diameter:

4(V H + VS )
D= .
3 L
0.5π  
D
CFD Simulation of Multiphase Separators 202

Figure A-3. The “Boot” Design of Horizontal Three-Phase Separator.

πD 2
2. Calculate vessel cross-sectional area: A = .
4
3. Set H V to the larger of 0.20 × D or 0.60 m, and then calculate AV .

4. Set H LLV and H LLB , and then calculate ALLV .

V H + VS
5. Calculate the vessel length based on the liquid holdup/surge: L = .
A − AV − ALLV

HV
6. Calculate the liquid dropout time: φ = .
VV

7. Calculate the actual vapor velocity using QV and AV .

8. Calculate the minimum length required for vapor-liquid separation (Lmin) using the
actual vapor velocity and the liquid dropout time.
9. If L < 0.8Lmin , increase H V , and go to step 3. Else if L < Lmin , set L = Lmin . Else if

L > 1.2 Lmin , decrease H V (if acceptable), and go to step 3. Else, L is acceptable.
10. Calculate the settling velocity of the heavy liquid through the light liquid using
Stokes’ law.
11. Calculate the settling time of the heavy liquid through the light liquid ( t HL ).
CFD Simulation of Multiphase Separators 203

L( A − AV )
12. Calculate the residence time of the light liquid: θ L = . If θ L < t HL , set
QLL

t HL QLL
L= .
A − AV
L L
13. If < 1.5 , decrease D (if acceptable), and go to step 2. Else if > 6 , increase D ,
D D
and go to step 2.
14. Calculate the approximate vessel weight based on the thickness and the surface area
of shell and heads.
15. In order to find the optimum case (corresponding to the minimum weight), change the
vessel diameter by 15 cm increments, and repeat the calculations from step 2 while
keeping the aspect ratio in the range of 1.5 to 6.0.
16. Design the heavy liquid boot:
16.1. Set H H .
16.2. Calculate the rising velocity of the light liquid out of the heavy liquid using
Stokes’ law, and use 75% of this velocity as VB in the calculations.
16.3. Calculate the heavy liquid boot diameter, DB , using QHL and VB .
16.4. Calculate the rising time of the light liquid droplets through the heavy liquid
( t LH ).

πDB 2 H H
16.5. Calculate the residence time of the heavy liquid: θ H = . If θ H < t LH ,
4QHL

increase the boot diameter so that θ H = t LH .

A.2.3 “Weir” Design

1. Pick an aspect ratio from Table A-1 and calculate the initial vessel diameter:
16(V H + VS )
D= .
3 L
0.6π  
D
CFD Simulation of Multiphase Separators 204

Figure A-4. The “Weir” Design of Horizontal Three-Phase Separator.

πD 2
2. Calculate vessel cross-sectional area: A = .
4
3. Set H V to the larger of 0.20 × D or 0.60 m, and then calculate AV .

4. Set the low liquid level ( H LLL ) in light liquid section, and then calculate ALLL .
5. Calculate the weir height: H W = D − H V . If H W < 60 cm , increase D , and go to
step 2.
V H + VS
6. Calculate L2 based on the light liquid holdup/surge: L2 = .
A − AV − ALLL

HW
7. Set the interface at (typical setting), and obtain H H and H L .
2

8. Using H H value calculate AH , and set AL = A − AV − AH .

9. Calculate the separation velocities of liquid phases using Stokes’ law.
10. Calculate the separation times of the liquid droplets ( t HL and t LH ).

t LH QHL t Q
11. Set the larger of and HL LL as the required length for liquid-liquid
AH AL

separation ( L1 ).
CFD Simulation of Multiphase Separators 205

12. Set L = L1 + L2 .
HV
13. Calculate the liquid dropout time: φ = .
VV

14. Calculate the actual vapor velocity using QV and AV .

15. Calculate the minimum length required for vapor-liquid separation (Lmin) using the
actual vapor velocity and the liquid dropout time.
16. If L < 0.8Lmin , increase H V , and go to step 3. Else if L < Lmin , set L = Lmin . Else if

L > 1.2 Lmin , decrease H V (if acceptable), and go to step 3. Else, L is acceptable.

L L
17. If < 1.5 , decrease D (if acceptable), and go to step 2. Else if > 6 , increase D ,
D D
and go to step 2.
18. Calculate the approximate vessel weight based on the thickness and the surface area
of shell and heads.
19. In order to find the optimum case (corresponding to the minimum weight), change the
vessel diameter by 15 cm increments, and repeat the calculations from step 2 while
keeping the aspect ratio in the range of 1.5 to 6.0.

A.2.4 “Bucket and Weir” Design

1. Assume residence times of light and heavy liquid phases, θ L and θ H .

2. Pick an aspect ratio from Table A-1 and calculate the initial vessel diameter:

4(Q LLθ L + Q HLθ H )

D= .
3 L
0.7π  
D

πD 2
3. Calculate vessel cross-sectional area: A = .
4
4. Set H V to the larger of 0.20 × D or 0.60 m, and then calculate AV .

QLLθ L + QHLθ H
5. Calculate L1 : L1 = .
A − AV
CFD Simulation of Multiphase Separators 206

Figure A-5. The “Bucket and Weir” Design of Horizontal Three-Phase Separator.

HV
6. Calculate the liquid dropout time: φ = .
VV

7. Calculate the actual vapor velocity using QV and AV .

8. Calculate the minimum length required for vapor-liquid separation (Lmin) using the
actual vapor velocity and the liquid dropout time.
9. If L1 < 0.8Lmin , increase H V , and go to step 4. Else if L1 < Lmin , set L1 = Lmin . Else,

L1 is acceptable.
10. Calculate H L using Stokes’ law and θ L .
11. Calculate the height difference between the light and heavy liquid weirs:
 ρ 
∆H = H L 1 − L  .
 ρH 
12. Design the light liquid bucket:
12.1. Set the top of the light liquid weir.
12.2. Assume holdup and surge times.
12.3. Assume H HLL , and calculate AHLL .
12.4. Assume H LLL , and calculate ALLL .
CFD Simulation of Multiphase Separators 207

QLL (t H + t S )
12.5. Calculate L2 : L2 = .
AHLL − ALLL

13. Set L3 as the larger of D / 12 or 30 cm .

14. Design the heavy liquid section:
14.1. Set the top of the heavy liquid weir.
14.2. Assume holdup and surge times.
14.3. Assume H HLL , and calculate AHLL .
14.4. Assume H LLL , and calculate ALLL .
QHL (t H + t S )
14.5. Calculate L4 : L4 = .
AHLL − ALLL

15. Set L = L1 + L2 + L3 + L4 .

L L
16. If < 1.5 , decrease D (if acceptable), and go to step 3. Else if > 6 , increase D ,
D D
and go to step 3.
17. Calculate the approximate vessel weight based on thickness and surface area of shell
and heads.
18. In order to find the optimum case (corresponding to the minimum weight), change the
vessel diameter by 15 cm increments, and repeat the calculations from step 3 while
keeping the aspect ratio in the range of 1.5 to 6.0.
CFD Simulation of Multiphase Separators 208

Appendix B: Droplet Breakups in the Two-Phase Separators

In this Appendix, the diagrams of droplet breakup variations versus continuous gas phase
velocity for all the pilot-plant scale separators have been presented. As explained in
Chapter Three, two modeling approaches were implemented: DPM approach and VOF-
DPM approach.

B.1 Results for DPM Approach

Figures B-1 to B-4 show the obtained diagrams for the number of droplet breakups versus
gas phase velocity in all the case studies. As represented, three different operating
pressures were involved, and three different mean droplet sizes were also tested in the
simulations. The presented variations indicate that higher velocities and operating
pressures usually intensify the number of droplet breakups. However, the horizontal
separators usually proposed a maximum breakup point so that before this point, which
was independent of the incipient velocity, increasing the velocity led to an increase in the
droplet breakup number, and after this maximum point, increasing the velocity decreased
the number of breakups. The simulation results also showed that changing the mean
droplet size when keeping the other size distribution parameters constant, has no tangible
effect on the droplet breakup number.

B.2 Results for VOF-DPM Approach

Figures B-5 to B-8 present the obtained diagrams for number of droplet breakups versus
gas phase velocity in all the case studies. The trend of variations in the horizontal
separators indicates that higher velocities usually intensified the number of droplet
breakups, and higher pressures had a stabilizing effect on the droplet breakup variations.
Moreover, increasing the separator aspect ratio was partially in favor of droplet breakups.
In vertical separators, droplet breakup variations versus gas velocity tended to be
somewhat oscillating at low pressures, but at higher pressures the number of breakups
was stabilized to a constant rate of about 20%.
CFD Simulation of Multiphase Separators 209

(a)

(b)

(c)
Figure B-1. Droplet Breakup Diagrams for Separator A at Multiple Pressures (DPM
Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa.
CFD Simulation of Multiphase Separators 210

(a)

(b)

(c)
Figure B-2. Droplet Breakup Diagrams for Separator B at Multiple Pressures (DPM
Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa.
CFD Simulation of Multiphase Separators 211

(a)

(b)

(c)
Figure B-3. Droplet Breakup Diagrams for Separator C at Multiple Pressures (DPM
Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa.
CFD Simulation of Multiphase Separators 212

(a)

(b)

(c)
Figure B-4. Droplet Breakup Diagrams for Separator D at Multiple Pressures (DPM
Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa.
CFD Simulation of Multiphase Separators 213

(a)

(b)

(c)
Figure B-5. Droplet Breakup Diagrams for Separator A at Multiple Pressures (VOF-
DPM Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa.
CFD Simulation of Multiphase Separators 214

(a)

(b)

(c)
Figure B-6. Droplet Breakup Diagrams for Separator B at Multiple Pressures (VOF-
DPM Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa.
CFD Simulation of Multiphase Separators 215

(a)

(b)

(c)
Figure B-7. Droplet Breakup Diagrams for Separator C at Multiple Pressures (VOF-
DPM Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa.
CFD Simulation of Multiphase Separators 216

(a)

(b)

(c)
Figure B-8. Droplet Breakup Diagrams for Separator D at Multiple Pressures (VOF-
DPM Approach); (a) 70 kPa, (b) 700 kPa, and (c) 2760 kPa.
CFD Simulation of Multiphase Separators 217

Appendix C: Separation Performance of Wire Mesh Demisters

As explained in Chapter Three, the demister of Gullfaks-A separator was modeled using
the Fluent Porous Media Model. The results of CFD simulations, presented in Chapter
Four, indicated that pressure and velocity profiles assigned to the demister were correct.
However, since almost all the liquid droplets had already settled from gas stream,
separation performance of the demister could not be verified. To examine this issue, a
CFD model was developed for the configuration represented in Figure C-1. This typical
demister is 0.15 m thick with physical properties as reported by Helsør and Svendsen
(2007), Table 3-10. Chapter Three provides more details of the developed CFD model for
the wire mesh demister. The oil, water, and gas properties of the Gullfaks-A case study
were used. Two particle size distributions were defined (Table C-1) to study the effect of
size distribution on the demister separation efficiency.

Figure C-1. The Model Developed for CFD Simulation of Demister Separation
Efficiency.
CFD Simulation of Multiphase Separators 218

Table C-1. The discrete phase parameters used in CFD simulations of demister
performance.
Expected for Demister Vicinity Extended Size Distribution
Discrete Phase Parameters Oil Drops Water Drops Oil Drops Water Drops
Maximum Diameter ( µm ) 150 150 2000 2000
Mean Diameter ( µm ) 60 60 800 800
-7 -7 -4
Total Mass Flow-Rate ( kg / s ) 1.9 × 10 2.3 × 10 4.5 × 10 5.5 × 10-3
Number of Tracked Particles 1000
Minimum Diameter ( µm ) 5
Spread Parameter 2.6

The original particle size distribution was assumed based on the fact that the main role of
demisters is to remove very fine droplets, 10 to 100 µm , from the gas stream (Smith,
1987). Note, while the maximum droplet size was increased from 150 µm to 2000 µm for
the case of extended droplet size distribution, the minimum droplet size was set at a
constant value of 5 µm for both distributions.
In the developed model, the gas velocity was gradually increased from 0.02 to 4.0
m/s and the demister separation efficiency was calculated using Equation C-1:

min − mout
η demister = × 100 (C-1)
min

where min and mout are mass of entered to and exited from the demister in kg,
respectively.
Figures C-2 and C-3 show the separation efficiency of demister for the two
distributions of oil and water droplets as a function of gas velocity. These figures do
show that the separation efficiency of demister has a maximum value at a very low
velocity and then drops sharply to very low values. Therefore, as was expected from
experience (Coker, 2007), the increase in gas velocity within the operating range has
resulted in an improved demister performance, until a maximum safe velocity is reached,
after which increasing gas velocity causes rapid loss of demister efficiency.
CFD Simulation of Multiphase Separators 219

Figure C-2. Separation Efficiency of the Demister versus Gas Velocity for the Original
Droplet Size Distribution.

Figure C-3. Separation Efficiency of the Demister versus Gas Velocity for the Extended
Droplet Size Distribution.
CFD Simulation of Multiphase Separators 220

Both the maximum demister efficiency and the corresponding gas velocity are higher for
the case of the extended droplet size distribution, which is consistent with the practical
experience in that large droplets are expected to be separated more efficiently than
smaller droplets. However, from a practical perspective, the maximum demister
efficiency is expected to be close to 100% for droplets with a diameter of 100 µm or less
(down to around 5 µm ), and this performance should occur for the higher gas velocities
when compared with the simulated results. Therefore, it can be concluded that although
the general behavior of mist eliminators has been simulated, the maximum demister
efficiency and the range of demister operability were predicted much more
conservatively.
Another CFD-based study of the demister performance was provided by the
droplet size distribution analysis in inlet and outlet zones. Table C-2 reports the droplet
size distribution in inlet and outlet zones of the demister operating at its best
performance. If the results given for the demister outlet zone are compared with those
given for the demister inlet zone, it is evident that the demister had no problem in
separating the larger droplets and reducing the maximum and mean droplet sizes.
Moreover, the droplet size distribution in the demister outlet zone, with higher values
assigned for the spread parameter, is narrower than that in the demister inlet zone.

Table C-2. The droplet size distribution in inlet and outlet zones of the demister.
Discrete Phase Original Size Distribution Extended Size Distribution
Parameters Demister Inlet Demister Outlet Demister Inlet Demister Outlet
dmin ( µm ) 5 5 5 5
Oil dmax ( µm ) 51 48 129 99
Drops
d ( µm ) 48 35 152 96
n 3.55 3.72 2.65 2.69
dmin ( µm ) 5 5 5 5
Water dmax ( µm ) 61 57 167 161
Drops
d ( µm ) 51 41 234 129
n 3.49 3.62 2.57 2.64
CFD Simulation of Multiphase Separators 221

To further investigate the demister separation performance issue, it is worth referring to

two relevant experimental studies by Verlaan et al. (1989) and El-Dessouky et al. (2000).
The experiments were performed using air and water at atmospheric pressure and
ambient temperature as system fluids. Figure C-4 is taken from Verlaan et al. (1989), and
shows the separation efficiency of their studied demister versus gas velocity and droplet
size. The mean droplet size was between 5 µm and 20 µm . Figure C-5 is taken from El-
Dessouky et al. (2000), and represents their demister performance versus gas velocity and
droplet size. Here, the mean droplet size was from 1000 µm to 5000 µm . Comparison
between the data of Figures C-4 and C-5 indicates that the provided demister
performance data for these two independent studies are in agreement with each other and
consistent with the practical experience noted above. In spite of different droplet sizes
and apparatus dimensions involved in experiments, the maximum separation efficiency
did occur when the air stream velocity was 3 m/s in both studies.

(a) (b)
Figure C-4. The Demister Separation Performance Reported by Verlaan et al. (1989) at;
(a) Low Gas Velocity, and (b) High Gas Velocity.
CFD Simulation of Multiphase Separators 222

Figure C-5. The Demister Separation Performance Reported by El-Dessouky et al.

(2000).

Based on the experimental apparatus of Verlaan et al. (1989), another CFD model
including air and water at atmospheric pressure was developed. The experimental
apparatus dimensions, with an internal diameter of 39.2 cm, were almost the same as the
simulated model. So, the same grid system could be used for simulation purposes. The
extended water droplet size distribution provided in Table C-1, which resulted in more
normal results as per the previous case studies, was again used. The air velocity was
gradually increased from 0.02 m/s to 4.0 m/s and the demister separation efficiency was
calculated using Equation C-1. Figure C-6 shows the separation efficiency of demister as
a function of gas velocity, which was compared with experimental data of Figure C-4 and
Figure C-5. This comparison shows that the maximum demister efficiency and the range
of demister operability have been poorly predicted. Table C-3 presents the droplet size
distribution in inlet and outlet zones of the demister operating at its best performance,
which confirms that demister has successfully eliminated the large droplets from gas
stream.
CFD Simulation of Multiphase Separators 223

Figure C-6. Separation Efficiency of the Demister versus Gas Velocity for the Air-Water
Case Study.

Table C-3. The water droplet size distribution in inlet and outlet zones of the demister.
Discrete Phase Parameters Demister Inlet Demister Outlet
d min ( µm ) 5 5
d max ( µm ) 135 103
d ( µm ) 113 74
n 2.69 2.86

The CFD simulations of mist eliminators are acceptable but there are still shortcomings.
As demonstrated in Chapter Four, the predicted pressure and velocity profiles for the
demister simulations are as expected. However, the separation performance of demisters
has been poorly predicted by the CFD simulations, hence the CFD simulations cannot be
used for evaluating the separation efficiency of demisters or their operability ranges. In
fact, based on the CFD simulations, the maximum demister efficiency and the range of
demister operability are predicted much more conservatively than what is normally
experienced. In order to explain these shortcomings, it should be noted that not all the
CFD Simulation of Multiphase Separators 224

physical properties of demister, as given in Table 3-10, can be specified and modeled via
CFD simulations. Therefore, realistic interactions between droplets and demister
filaments cannot be simulated properly. Actually, the separation efficiency of knitted
wire mesh demisters has been improved by subtle enhancements during the last several
years. For instance, as described in Chapter One, in the late 1960’s and early 1970’s,
considerable research led to the design of high-quality demisters through efficient
combination of filaments of different materials and diameters (Smith, 1987). The
resultant changes in droplet-demister interactions, however, cannot currently be
simulated by CFD. Therefore, it is not unexpected that droplet coalescence, for example,
with a rate of around 0.1% has been poorly simulated in all demister simulation case
studies. Although the overall performance of wire mesh demisters in terms of pressure
and velocity profiles and some separation features can be simulated using existing CFD
models, the separation efficiency of demisters or their operability ranges cannot be
simulated because of complicated separation mechanism involved that are not easily
modeled. Consequently, the current CFD models do under-predict the demister
efficiency.
CFD Simulation of Multiphase Separators 225

Appendix D: Fluid Flow Profiles for Various Separator Designs

Appendix D provides fluid flow profiles for the case studies performed to design the
realistic optimum separator for Gullfaks-A (discussed in section 5.4).

(a)

(b)
Figure D-1. Contours of Pressure (Pa) in the Middle of the “Simple” Separator Designed
for Processing Gullfaks-A at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 226

(a)

(b)
Figure D-2. Vectors of Velocity (m/s) in the Middle of the “Simple” Separator Designed
for Processing Gullfaks-A at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 227

(a)

(b)
Figure D-3. Contours of Density (kg/m3) in the Middle of the “Simple” Separator
Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 228

(a)

(b)
Figure D-4. Contours of Pressure (Pa) in the Middle of the “Bucket and Weir” Separator
Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 229

(a)

(b)
Figure D-5. Vectors of Velocity (m/s) in the Middle of the “Bucket and Weir” Separator
Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 230

(a)

(b)
Figure D-6. Contours of Density (kg/m3) in the Middle of the “Bucket and Weir”
Separator Designed for Processing Gullfaks-A at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 231

(a)

(b)
Figure D-7. Contours of Pressure (Pa) in the Middle of the Vertical Separator Designed
for Processing Gullfaks-A (25% Capacity) at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 232

(a)

(b)
Figure D-8. Vectors of Velocity (m/s) in the Middle of the Vertical Separator Designed
for Processing Gullfaks-A (25% Capacity) at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 233

(a)

(b)
3
Figure D-9. Contours of Density (kg/m ) in Middle of the Vertical Separator Designed
for Processing Gullfaks-A (25% Capacity) at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 234

(a)

(b)
Figure D-10. Contours of Pressure (Pa) in the Middle of the “Stabilized” Separator
Designed for Gullfaks-A at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 235

(a)

(b)
Figure D-11. Vectors of Velocity (m/s) in the Middle of the “Stabilized” Separator
Designed for Gullfaks-A at (a) 1988, and (b) the Future Conditions.
CFD Simulation of Multiphase Separators 236

(a)

(b)
3
Figure D-12. Contours of Density (kg/m ) in the Middle of the “Stabilized” Separator
Designed for Gullfaks-A at (a) 1988, and (b) the Future Conditions.