Assessment of A Hybrid VOF Two-Fluid CFD Solver - Thesis - 2016

Politecnico di Milano
SCUOLA DI INGEGNERIA INDUSTRIALE E DELL’INFORMAZIONE

Corso di Laurea Magistrale in Ingegneria Aeronautica
Tesi di Laurea Magistrale
Assessment of a hybrid VOF two-fluid CFD solver for

simulation of gas-liquid flows in vertical pipelines in
OpenFOAM
Relatore Candidato
Prof. Luciano Galfetti Francesco Tocci
Matr. 804717
Anno Accademico 2015–2016

Francesco Tocci: Assessment of a hybrid VOF two-fluid CFD solver for simulation
of gas-liquid flows in vertical pipelines in OpenFOAM | Tesi di Laurea Magistrale
in Ingegneria Aeronautica, Politecnico di Milano.
© Copyright Dicembre 2016.
Politecnico di Milano:
www.polimi.it
Scuola di Ingegneria Industriale e dell’Informazione:
www.ingindinf.polimi.it
Abstract
In the framework of OpenFOAM, multiphaseEulerFoam is used for simulation

of incompressible multiphase flows in vertical pipelines. This model combines the
Eulerian-Eulerian (two-fluid) approach with the Volume of Fluid (VOF) interface
tracking model. The results from the coupled model are compared to experimental
data.
From the physical point of view the hybrid approach is not problematic since the
VOF model uses an indicator function for tracking the interface between the phases
which has the same meaning as a volume fraction variable in the two-fluid model.
The hybrid model is obtained via the addition of interface capturing on top of an
Eulerian two-fluid model. Averaged mass and momentum conservation equations
are applied across the entire domain to describe the time-dependent motion of
both phases and an interface sharpening algorithm is used to obtain sharp interface
regions.
The results from 3D transient simulations are compared to both time-dependent
and time-averaged experimental data such as phase distributions and pressure drop.
Two case studies are presented in which the vertical risers differ in length, diameter,
superficial gas and liquid velocity and flow patterns. In the first case, an annular
film of liquid is introduced at the inlet. The coupled solver is able to correctly
predict the liquid holdup fraction and pressure drop. For the second test case, slug
flow is predicted and a reasonably good agreement in the form of void fraction time
series is obtained. Taylor bubbles are detected, separated by slugs of continuous
liquid which bridge the pipe and contain small gas bubbles. By means of velocity
field views the three regions of Taylor bubble, falling film and the wake are captured.
There is a dependency of the flow behaviour on the mesh employed and the most
suitable grid topology is provided.
The aim of this work is to gain a deeper understanding of transitional regimes of
multiphase flows. This thesis highlights the application of a hybrid CFD model in
prediction of complex gas-liquid flow in vertical pipes, without a priori knowledge
of the flow pattern and thus overcoming the issue of flow regime selection.
Keywords: CFD, OpenFOAM, hybrid multiphase solver, multiphaseEulerFoam,
coupled VOF two-fluid model, vertical pipelines, slug flow, annular flow.
iii
Sommario
Nel contesto di OpenFoam, il solutore multiphaseEulerFoam è utilizzato per la

simulazione di flussi multifase, gas-liquido, in condotti verticali circolari. Questo so-
lutore accoppia una metodologia Eulerian-Eulerian (two-fluid) con il metodo Volume
of Fluid (VOF) per la risoluzione dell’interfaccia tra le fasi.
Dal punto di vista fisico l’accoppiamento non presenta problemi poichè la
funzione che distingue le diverse fasi nel modello VOF non è nient’altro che la
variabile frazione volumetrica nell’approccio two-fluid. Le equazioni mediate della
conservazione della massa, con l’ipotesi di incomprimibilità, e del bilancio di quantità
di moto sono definite sull’intero dominio per descrivere il comportamento delle fasi.
L’interfaccia è ricostruito attraverso l’aggiunta di un termine, chiamato compressione
artificiale, nell’equazione di conservazione della massa.
Simulazioni instazionarie e tridimensionali sono confrontate con dati sperimentali
mediati a loro volta nello spazio, su sezioni del condotto, e nel tempo. Sono trattati
due casi studio, differenti per dimensioni del condotto circolare e condizioni al
contorno. I due casi simulati richiedono particolare attenzione per la scelta della
mesh. Da precedenti risultati trovati in letteratura, viene impiegata la tipologia
di mesh più adatta. Per il primo caso, dove un anello di liquido viene introdotto
all’inlet, annular flow, i risultati della simulazione sono in accordo con gli esperimenti,
mostrando valori di caduta di pressione e liquid holdup del tutto simili. Il secondo
caso riguarda il cosiddetto slug flow. Il modello è in grado di simulare la formazione
di tale regime e la comparazione con i dati sperimentali è ancora positiva. Andamenti
nel tempo di void fractions mediate su diverse sezioni, le loro funzioni di densità di
probabilità e l’immagine qualitativa del campo di moto sono i termini di paragone
impiegati. La bolla di Taylor viene correttamente riprodotta così come la regione
di film di liquido che la circonda.
Lo scopo di questo lavoro è di approfondire la conoscenza per la modellazione
di flussi gas-liquido. In particolar modo la tesi sottolinea le capacità del solutore
impiegato a trattare regimi di natura diversa, grazie alla sua natura ibrida.
iv
Contents
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research objective . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Literature Review 5
2.1 Multiphase flow: basic definitions and dimensionless numbers . . . . 5
2.1.1 Velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Slip effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.3 Froude number . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.4 Morton number . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.5 Eotvos number . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.6 Dimensionless inverse viscosity number . . . . . . . . . . . . 7
2.1.7 Slippage number . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Upward gas-liquid flow in vertical pipes . . . . . . . . . . . . . . . . 7
2.2.1 Flow regimes . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Flow pattern maps . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Transition mechanism . . . . . . . . . . . . . . . . . . . . . 15
2.3 CFD methodologies for two-phase flow . . . . . . . . . . . . . . . . 16
2.3.1 Direct numerical simulation . . . . . . . . . . . . . . . . . . 16
2.3.2 Eulerian-Lagrangian method . . . . . . . . . . . . . . . . . . 18
2.3.3 Eulerian-Eulerian method . . . . . . . . . . . . . . . . . . . 19
3 Coupling Euler-Euler two-fluid with VOF 21

3.1 Hybrid approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 multiphaseEulerFoam . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.1 Multifluid momentum equations . . . . . . . . . . . . . . . . 22
3.2.2 Time step . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.3 Turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.4 Solution procedure . . . . . . . . . . . . . . . . . . . . . . . 28
4 Validation 33
4.1 Case study 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1.1 Case description . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1.2 Case setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.1.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.2 Case study 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
v
vi CONTENTS
4.2.1 Case description . . . . . . . . . . . . . . . . . . . . . . . . . 44

4.2.2 Case setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Conclusions 63
A blockMesh with m4 65
A.1 Case study 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
A.2 Case study 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
B Reynolds-averaged modelling for the mixture 73

B.1 Listings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
B.2 yPlus function object . . . . . . . . . . . . . . . . . . . . . . . . . 74
C fvSolution 75
D fvSchemes 77
E Velocity reconstruction in pEqn.H 79
Bibliography 81
List of Figures
2.1 Sketches of flow regimes for two-phase flow in vertical pipes . . . . . 9

2.2 Test section photographs of upward air-water flow . . . . . . . . . . 10
2.3 Visualization of wire-mesh sensor data from upward air-silicone oil flow 11
2.4 An image of the Taylor bubble bottom, the near wake and the film
region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Slug unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6 Weisman flow pattern map for vertical tubes, air-water . . . . . . . 14
2.7 Taitel flow pattern map for vertical tubes, air-water . . . . . . . . . 15
2.8 Schematic representation of interface-tracking and interface-capturing
methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.9 Representation of geometrical scales in a bubbly flow . . . . . . . . 19
3.1 Flow chart of multiphaseEulerFoam. . . . . . . . . . . . . . . . . . 29

3.2 Procedure to solve the phase volume fraction. . . . . . . . . . . . . 31
3.3 Solution procedure of the phase continuity equation. . . . . . . . . . 32
4.1 Model geometry for simulation of vertical upward flow through a

50.8 mm diameter pipe. . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2 Cross-Section grid of the computational domain. . . . . . . . . . . . 35
4.3 Boundary conditions at the inlet. . . . . . . . . . . . . . . . . . . . 36
4.4 CFD results with OpenFOAM for the flow pattern in the vertical
pipe at t = 10 s. Cells having a water volume fraction larger than
0.5 are coloured blue. . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.5 Time series for the void fraction in the CFD simulations with
OpenFOAM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.6 CFD results for the last 30D of the vertical pipe at t = 10 s. . . . . 41
4.7 Liquid holdup at different cross sections at t = 10 s. . . . . . . . . . 42
4.8 Mixture velocity at different cross sections at t = 10 s. . . . . . . . 43
4.9 Total liquid holdup fraction in the 50D long vertical pipe. . . . . . . 44
4.10 3D geometry of the computational flow domain showing the location
of the experimental measurement. . . . . . . . . . . . . . . . . . . . 46
4.11 Cross-sectional view of different computational grid used for mesh
independent study. . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.12 Simulation with OpenFOAM for the Taylor bubble with different
grid number. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.13 Void fraction (see Eq. (4.4)) at the measurement section (0.5 m). . . 48
4.14 Computational mesh used for simulations by Abdulkadir et al. in [5]. 49
vii
viii LIST OF FIGURES
4.15 Cross-Section grid of the computational domain in OpenFOAM. . . 49

4.16 Simulation with OpenFOAM for the for the leading Taylor bubble ris-
ing through a 0.067 m internal diameter vertical pipe: instantaneous
velocity field for the liquid phase. . . . . . . . . . . . . . . . . . . . 51
4.17 Simulation with OpenFOAM for the fully developed slug flow: in-
stantaneous void fraction. . . . . . . . . . . . . . . . . . . . . . . . 52
4.18 Simulation with OpenFOAM for the fully developed slug flow: in-
stantaneous velocity. . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.19 Simulation with OpenFOAM for a large trailing Taylor bubble and
leading train of Taylor bubbles. . . . . . . . . . . . . . . . . . . . . 54
4.20 Comparison between experimental data (red) and CFD simulation
with Star-CCM+ (black) from Abdulkadir et al. [5]. . . . . . . . . . 54
4.21 Time series for the void fraction in different planes; comparison
between experiments and CFD simulations with OpenFOAM. . . . 55
4.22 PDF of cross-sectional average void fraction for the case of slug
flow obtained from the experiments, from the CFD simulations with
Star-CCM+ (in Abdulkadir et al. [5]) and with OpenFOAM. . . . . 56
4.23 Comparison between experiments and CFD simulations with Star-
CCM+ (from [5]) and with OpenFOAM. . . . . . . . . . . . . . . . 57
4.24 Time delay of a Taylor bubble passing along two different measuring
locations in the pipe. . . . . . . . . . . . . . . . . . . . . . . . . . . 58
CCM+ (from [5]) and with OpenFOAM for the leading Taylor bubble
(Start-up). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
List of Tables
4.1 Results comparison: data with Fluent, SFC, and OLGA from [115]
and with OpenFOAM. . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.2 Number of cells used in different grids. . . . . . . . . . . . . . . . . 45
4.3 Dimensionless numbers. . . . . . . . . . . . . . . . . . . . . . . . . 51
4.4 Flow development along the vertical riser in the simulation with
OpenFOAM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
CCM+ (from [5]) and with OpenFOAM for the leading Taylor bubble:
void fraction and liquid film thickness. . . . . . . . . . . . . . . . . 61
ix
Listings
3.1 Drag specification in transportProperties dictionary . . . . . . . 25

3.2 MULES constructor in multiphaseSystem.C . . . . . . . . . . . . . 30
A.1 m4 Script to generate the blockMesh dictionary for the Case study 1. 65
A.2 m4 Script to generate the blockMesh dictionary for the Case study 2. 68
B.1 Changes in UEqns.H . . . . . . . . . . . . . . . . . . . . . . . . . . 73
B.2 Changes in createFields.H . . . . . . . . . . . . . . . . . . . . . . 73
B.3 Changes in multiphaseEulerFoam.H . . . . . . . . . . . . . . . . . 73
B.4 Changes in yP lusmixture . . . . . . . . . . . . . . . . . . . . . . . . 74
B.5 Changes in yP lusmixture . . . . . . . . . . . . . . . . . . . . . . . . 74
C.1 fvSolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
D.1 fvSchemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
E.1 Velocity reconstruction in pEqn.H . . . . . . . . . . . . . . . . . . . 79
xi
Chapter 1
Introduction
This thesis is part of a collaboration project between Delft University of Tech-

nology and Dynaflow Research Group, under the supervision of Dr. ir. Frank Bos
(Dynaflow Research Group BV, Zoetermeer) and Prof. dr. ir. Ruud Henkes (Shell,
TU Delft, 3ME).
1.1 Background
Multiphase flows remain an area where the prediction through CFD is yet
out of reach for the majority of applications. Multiphase flows present a unique
challenge due to the variety of flow regimes. They are characterized by a broad
range of scales, from the dispersed droplets at the microscale up to macroscale free
surface flows. A complication is the regime dependency limitation of the existing
multiphase CFD methodology: specific simulation methods are applicable only to a
specific class of flows. For example, free surface flows, in which the dynamics of the
interface are important to the overall physics, are typically modeled with a shared
momentum equation and a phase fraction field using a sharp interface capturing
method. However, such methods are not suited for the simulation of dispersed
flows characterized by many small fluid particles which cannot be fully resolved
on a computational mesh. In such cases, the two-fluid or Euler-Euler approach is
taken. Here the phases are treated as interpenetrating continua, each governed by
its own momentum equation and having exchange terms to account for interphase
momentum transfer. This methodological segregation of flow regimes is acceptable
for many classes of flows but presents severe limitations for problems in which both
segregated and dispersed flow features are present or where transitions between
flow types are critical to the phenomena of interest.
Recently, a hybrid CFD solver has been developed which aims to overcome the
issue of regime dependency by combining the Euler-Euler multifluid method with
sharp interface capturing in a regime flexible framework [89], [109].
In this context the interest is focused on vertical upward gas-liquid flows. Gas-
liquid flows occur in many applications. The motion of bubbles in a liquid as well as
droplets in a conveying gas stream are examples of gas-liquid flows. Bubble columns
1
2 Chapter 1. Introduction
are commonly used in several process industries. Atomization to generate small

droplets for combustion is important in power generation systems. Also droplet
formation and impaction are important in spray forming for materials processing.
Steam-water flows in pipes and heat exchangers are very common in power systems
such as fossil fuel plants and nuclear reactors.
Other important gas-liquid flows occur in pipelines. Here free gas may exist
because it is originally present at the inlet, as in many gas-condensate pipelines,
but it may also be due to the formation of gases originally dissolved in the liquid
as the pressure along the pipeline falls. Depending on the liquid and gas flow rates
and on the slope of the pipeline, one may observe a whole variety of flow regimes
such as bubbly, stratified, wavy, slug and annular. Each one of them brings its own
pressure drop and liquid hold-up fraction.
For example, in slug flow, solid surfaces such as pumps and tube walls are often
subjected to large fluctuating forces which may cause dangerous vibrations and
fatigue. It is therefore of great practical importance to be able to predict which
flow regime would occur in a given situation, the operational limits to remain in
the desired regime, and how the system would react to transients such as start-ups
and shut-downs. The experimental effort devoted to this subject has been very
considerable and the computational method described in this thesis is a promising
tool for a better understanding of these problems.
1.2 Research objective

For the design and improved operations of oil and gas production systems
simulation tools are used extensively. Most of the simulations are carried out with
steady and transient one-dimensional tools, such as PIPESIM, OLGA and LedaFlow.
For some applications, however, such as flow in bends, flow in splitters, flow in
headers to facilities etc., the one-dimensional assumption limits the prediction
accuracy. As an alternative, Computational Fluid Dynamics (CFD) can be used,
either for 2D and 3D configurations. Available CFD packages include Fluent,
START-CCM+, and OpenFOAM. The present study has focused on the verification
and validation of CFD results for multiphase flow of gas and liquid through vertical
pipe sections.
The problem originates from the failure of the VOF method as implemented
in the CFD model in [115] with Fluent to predict slug flow in the risers. The first
objective is then investigate the abilities of OpenFOAM for industrial applications
which have dispersed and segregated flow regimes, such as slug flow in the risers.
The object is initially restricted by the following constraints: the use of the VOF
model, for comparison, and the simulations are performed with the same setups
in OpenFOAM and Fluent [115] for the case study 1 and with the same setups
in OpenFOAM and Star-CCM+ [5] for the case study 2. The latter constraint
includes the use of a rather coarse grid to limit computation times.
From a first analysis, reported in 4.1, the disagreement between experiment and
CFD is attributed to the deficiencies of the VOF method to simulate flows in which
both dispersed and segregated flows are present. Therefore the constraint of using
the VOF model is removed and a hybrid solver is further developed as described
1.3. Thesis outline 3
in 3.
The aim of this thesis becomes to build up a knowledge base, through numerical
simulations and validation against experimental data, of a hybrid solver developed
in the framework of OpenFOAM, that can be used for defining multiphase CFD
procedures for industrial applications.
1.3 Thesis outline

The outline for the thesis will be as follows:
• Chapter 2: Basic theory on multiphase flow in general, description of upward

gas-liquid flows in risers and the most common CFD approaches for multiphase
flow.
• Chapter 3: Theory of the hybrid approach coupling two-fluid model with

VOF.
• Chapter 4: Results and validation of the model.

Chapter 2
Literature Review
2.1 Multiphase flow: basic definitions and dimen-

sionless numbers
A phase refers to the liquid, vapour, or solid state of matter. A multiphase
flow is the flow of a mixture of phases such as gases (bubbles) in a liquid, or liquid
(droplets) in gases, and so on. Dispersed phase flows are flows in which one phase
consists of discrete particulates, such as droplets in a gas or bubbles in a liquid. On
the other hand, separated flows are flows where the two phases are separated by
a line of contact [67]. In other words, in a separated flow one can pass from one
point to another in the same phase while remaining in the same medium. Here the
scope is limited to gas-liquid mixtures where gas and liquid simultaneously flow
through a vertical pipe. This section covers definitions and dimensionless numbers
used to characterize gas-liquid flows.
2.1.1 Velocities
The superficial gas velocity Us,gas is the gas velocity as if the gas was flowing in
the pipe without liquids, in other words the total gas throughput (Qgas in m3 /s at
operating temperature and pressure) divided by the total cross sectional area of
the pipe (A). For the superficial liquid velocity the same can be derived, and the
simple expressions are:
Qgas Qliquid
Us,gas = Us,liquid = (2.1)
A A
The sum of the Us,gas and Us,liquid is the multiphase mixture velocity, and the
expression is given by:
Um = Us,gas + Us,liquid (2.2)
The actual phase velocities are defined as:
Qgas Qliquid
Ugas = Uliquid = (2.3)
Agas Aliquid
5
6 Chapter 2. Literature Review
The difference and ratio for phase velocities is the slip velocity and the slip ratio,
respectively:
Ugas
Us = |Ugas − Uliquid | S= (2.4)
Uliquid
2.1.2 Slip effects

The liquid fraction (αliquid ) and the gas fraction (αgas ) of the pipe cross sectional
area (A) as measured under two-phase flow conditions are known as the liquid hold-
up fraction (λliquid ) and gas void fraction (λgas ), respectively. The liquid hold-up
fraction is the ratio of the pipe cross-section occupied by liquid (Aliquid ) and the
total pipe cross section (A). In a similar way the local gas void fraction is the ratio
of the pipe cross-section occupied by gas (Aliquid ) and the pipe cross section (A):
λliquid + λgas = 1 (2.5)

αliquid + αgas = 1 (2.6)
Aliquid
λliquid = Liquid hold-up fraction (2.7)
A
Agas
λgas = Gas void fraction (2.8)
A
In pipe flow the liquid volume fraction is usually defined as the fraction of the total
volumetric flow rate that consists of liquid. The liquid hold-up is equal to the liquid
volume fraction only under conditions of no-slip, when the flow is homogeneous and
the two phases travel at equal velocities. Only in no-slip conditions the gas volume
fraction is equal to the gas void fraction. Owing to the slip, the liquid hold-up will
be larger than the liquid volume fraction in the majority of the conditions in the
flow regimes and the gas void fraction will be smaller than the gas volume fraction.
The lighter gas phase will normally move much faster than the liquid phase. The
superficial velocity and the phase velocity are related by the hold-up fraction:
Us,gas = αgas Ugas (2.9)
The same relations apply for the other phase.

It should be noted that in CFD the volume fraction is normally defined differently
from the description above: in CFD the volume fraction is taken the same as the
holdup. This is also how we will use the volume fraction in the present study.
2.1.3 Froude number

The Froude number (F r) is the ratio of the inertial force and the gravitational
force for a particular phase; in other words, the ratio of the kinetic to the potential
energy of the gas or the liquid as given in Eq. (2.10)
u2
Fr = (2.10)
gD
2.2. Upward gas-liquid flow in vertical pipes 7
2.1.4 Morton number
gη 4
Mo = (2.11)
ρσ 3
where g is the gravity constant, η the viscosity [Kg/ms], ρ the density and σ the
surface tension [N/m].
2.1.5 Eotvos number
ρgD2
Eo = (2.12)
σ
where D is the pipe diameter [m].
2.1.6 Dimensionless inverse viscosity number

From the Eotvos number and the Morton number the dimensionless inverse
viscosity number (Nf ) can be derived:
1/4
Eo3

Nf = (2.13)
Mo
2.1.7 Slippage number

Abdelsalam et al. [1] proposed a new dimensionless number, the Slippage number,
for gas-liquid flow in pipes. The number is defined as the ratio of the difference in
the gravitational forces between slip and no-slip conditions to the inertial force of
the gas:
(pTP − pH )gD
SL = 2
(2.14)
pUs,gas
In gas-liquid flow, the average two-phase flow mixture density or the slip-density,
ρTP , is different from the homogenous or no-slip density of the mixture, ρH , due to
slippage between the phases. The no-slip mixture density, ρH , is simply calculated
based on the ratio of the volumetric flow rate of the phases assuming no slippage
between the phases. The slip density, ρTP , is calculated based on the actual or
measured liquid holdup.
2.2 Upward gas-liquid flow in vertical pipes

From a practical engineering point of view one of the major design difficulties
in dealing with multiphase flow is that the mass, momentum, and energy transfer
rates and processes are sensitive to the topology of the components within the flow.
There is a complicated two-way coupling between the flow in each of the phases
and the geometry of the flow [24]. The way in which the phases are distributed
is called flow pattern and its occurrence depends mainly on the combination of
different gas and liquid flow rates. Other parameters, like pipe orientation, pipe
diameter and fluid properties, play a role as well. An appropriate starting point
is a phenomenological description of the geometric distributions or flow patterns
that are observed in common multiphase flows. This section describes the flow
patterns observed in vertical pipes and identifies a number of the instabilities that
lead to transition from one flow pattern to another. The material in this section is
extracted from various sources: [3], [24], [48], [13].
2.2.1 Flow regimes

Depending on the combination of gas and liquid superficial velocities, a specific
flow regime may occur in vertical upward gas-liquid flow. We are primarily interested
in four clearly distinguishable flow regimes, namely, bubbly, slug, churn and annular
flows, which will be referred to as distinct regimes. To these, we also add three
other regimes that share some of the features of at least two of the distinct regimes:
bubbly-slug, slug-churn and churn-annular flows [88]. In past literature [100] such
regimes have sometimes been referred to as transitional regimes, because, if the
liquid flow rate in a pipe was kept constant while the gas flow rate was increased,
they would be observed at the transition between two distinct regimes. For vertical
upflow, as the amount of gas is gradually increased, the following four distinct
regimes evolve: bubbly flow, slug flow, churn flow, and annular flow (Figs. 2.1, 2.3).
Bubbly flow: In this configuration, there is a continuous liquid phase and the
gas phase is dispersed as bubbles within the liquid. The bubbles travel with
a complex motion within the flow, may be coalescing and are generally of
non-uniform size.
Slug flow: Characteristic bullet-shaped bubbles, often called Taylor bubbles, flow
up the pipe surrounded by a thin film of liquid. The liquid slug body between
the Taylor bubbles often contains a dispersion of smaller bubbles.
Churn flow: At higher velocities, the Taylor bubbles in slug flow break down
into an unstable pattern in which there is an oscillatory motion or churning
of liquid in the tube.
Annular flow: This configuration is characterised by liquid travelling as a film

on the channel walls. Part of the liquid can also be carried as drops in the
central gas core.
In Fig. 2.2 a visual identification of each flow pattern, including the transitional
regimes, is shown. Bubbly-slug flow is characterized by the presence of a relatively
large cap-shaped bubble and the churn flow shows an unstable slug since the liquid
slug might be able to bridge the pipe diameter but it is continually penetrated by
the gas [81]. In the semi-annular flow one may notice the existence of a gas core
and a relatively uniform annular liquid film on the pipe wall as well as liquid slugs.
We now confine our attention to the slug flow since this regime with its formation
and transition will be the key topic of the CFD simulations of this thesis.
Figure 2.1: Sketches of flow regimes for two-phase flow in vertical pipes. From left to
rigth: Bubbly, Slug, Churn and Annular Flow. (Taitel and Bornea, 1980)
2.2.1.1 Slug flow

Slug flow is characterised by an alternating flow of gas pockets and liquid slug
bodies. Most of the gas-phase is concentrated in large bullet-shaped gas pockets,
defined as Taylor bubbles (after Geoffrey Taylor, notable for his pioneer work on
slug flow [35]) surrounded by a thin falling film of liquid [16]. The Taylor bubbles
are separated by intermediate liquid slug bodies, which may contain small entrained
gas bubbles [8].
In large diameter pipes, there is significant entrainment of gas from the Taylor
bubble into the liquid slug region caused by the wall film plunging into the upward
flowing slug of liquid. In smaller diameter pipes the slug of liquid is bubble free. The
gas content in the slug region increases systematically with the pipe diameter [71].
Isolated Taylor bubbles rise almost uniformly in vertical pipes, occupying nearly the
entire cross-section of the tube. As it reaches the bottom of the bubble, the annular
film enters the liquid slug behind it with the possibility of creating a recirculation
region known as the bubble wake [68].
Both the shape of the bubble trailing edge and the wake flow pattern depend
on the fluid properties and the tube geometry [17], besides the flow conditions. If
the separation distance between two Taylor bubbles is small enough, the motion
and shape of the trailing bubble get largely affected by the flow in the wake of
the leading one: the nose becomes distorted and wavy, its velocity increases and
coalescence between bubbles will occur. Extensive investigations of these bubbles
and slugs have been carried out in the past and several physically based models
have been proposed for fully developed flow [39], [40], [23].
Fig. 2.4 shows an image of the Taylor bubble bottom from an experimental
investigation through PIV [53] while Figs. 2.5a and 2.5b present a numerical
Figure 2.2: Test section photographs of upward air-water flow exhibiting the visual flow
pattern features: (a) Bubbly; (b) Spherical cap; (c) Stable slug; (d) Unstable
slug; (e) Semi-annular ; (f) Annular. (Rosa, Flora and Souza, 2012)
image of the flow around a Taylor bubble and a schematic model of the slug unit
respectively [68]. The bubble velocity (UTB ), the radius of curvature at the nose
(RN ), the interaction distance above the bubble (ZA ), the length needed to have
fully developed annular liquid film (Z ∗ ), the film thickness (δ), the wall shear stress
in the stabilized film (τW ), and the wake dimensions, represented by its length (lW ),
are depicted in Fig. 2.5b.
In [101] Taitel and Barnea presented a review on the modeling of gas-liquid flow
in pipes. As in [99] the review follows an approach based on Fernandes et al. [40],
which establishes macroscopic balances to the two phases and quantifies the various
physical parameters comprising those equations. In Campos et al. [68] different
physical mechanisms inside the flow of a Taylor bubble are extensively assessed
through dimensional analysis, following the work of White and Beardmore [113].
To accurately model slug flow it is first essential to understand the motion of
single Taylor bubbles rising through stagnant liquids. Throughout its rise, the
bubble is influenced by gravitational, inertial, viscous and interfacial forces. Fluid
density, surface tension and viscosity also have an important effect on the flow
regime in a pipe [25], [60], [2]. Szalinski et al. [97] found that in the air-water
flow there is the relation between coalescence and breakup shifted towards more
coalescence than in air-silicone oil flow. Bubbles in the air-water flow tend to be
larger than in air-silicone oil flow at similar superficial velocities. These differences
in coalescence intensity are obviously due to the different viscosities of the liquid
phases. Hernandez et al. [47] studied the effect of the Morton number on Taylor
bubble behaviour: for a higher Morton number there is more distortion.
Neglecting gas expansion, several authors, [113], [31] and [98], reported F r =
f (Eo, M o) to be sufficient for the characterization of the rise of single Taylor bubble
Figure 2.3: Visualization of wire-mesh sensor data from air-silicone oil two-phase flow at
liquid superficial velocity = 0.25 m/s and different gas superficial velocities.
The gas phase is represented by bright and the liquid phase by dark colour.
Left: virtual side projection, right: axial slice images. (Szalinski et al., 2010)
in stagnant liquid conditions.

Taylor bubbles have significant variations in the shape of their rear ends [14].
In low viscosity liquids they have an approximately flat bottom and a recirculation
region below it. Small bubbles are torn off the rear end of the Taylor bubble and
either move down or are reabsorbed into the rear end. As the liquid viscosity
increases the rear of the bubble becomes rounded and there is no recirculation in
the wake as reported by [113], [108] and [65].
Numerous authors have confirmed the importance of the diameter effect on flow
regime [94], [82]: slug flows cannot be sustained in pipes with diameters larger
than some limit [50], [77]. The behaviour of Taylor bubbles can also depend on the
geometry in which they flow, particularly the inlets and outlets for the gas and
liquid.
The rise velocity of bullet-shaped bubbles was first studied analytically by Du-
mitrescu
√ [38] and Davies and Taylor [35] who determined the rise velocity to be equal
to F r gD. They proposed values of F r of 0.351 and 0.328 respectively. Viana et
al. [108] presented an equation for F r based on Eo and a dimensionless inverse viscos-
ity which they term the Buoyancy Reynolds number (Reb = (gD [ρl − ρg ]ρl )/ηl )).
p
3
Nicklin et al. [69] proposed an empirical relationship to describe the rise velocity
of single Taylor air bubble in a static water column. Nicklin’s empirical relationship,
given by Eq. (2.15), describes the translational velocity of a Taylor bubble, Ub , as
the sum of drift velocity, which is the velocity of a Taylor bubble in a stagnant
liquid, plus the contribution of the mixture superficial velocity in the preceding
slug.
(2.15)
p
Ub = C0 (Us,gas + Us,liquid ) + 0.35 gD
Figure 2.4: An image of the Taylor bubble bottom, the near wake and the film region.
Bright dots in the liquid region represent the fluorescent particles. (R. van
Hout, 2001)
A value of 1.2 for C0 was suggested by Nicklin et al. [69].

It is outside the scope of this work is to provide more details regarding empirical
correlations. This has been recently covered by Campos et al. [68], who reviewed
the literature on vertical gas-liquid slug flow with Newtonian fluids, from 1943 to
2015. A comprehensive study on the modelling of the rise behaviour of single Taylor
bubbles is carried out by Ambrose in [12].
2.2.2 Flow pattern maps

Flow pattern data are often represented on a two dimensional diagram in
terms of system variables. The most common dimensional variables used are the
superficial velocities (Sternling 1965, Wallis 1969), superficial momentum fluxs
(Hewitt and Roberts 1969) or volumetric flux (Fig. 2.6). Since variables other than
the superficial velocities are known to affect the flow, pattern maps of this kind are
specific to a particular combination of fluids and geometry. These maps have failed
to gain universal acceptance because, although they apply to the experimental
configuration for which they were developed, they may not be accurate for different
configurations [88], [13]. However, they are simple to use, and unlike the case of
single-phase Newtonian flow where the single parameter of Reynolds number brings
all flows together, it is by no means clear exactly which other variables should be
included. Hence, even for the simplest duct geometries, there exist no universal,
dimensionless flow pattern maps that incorporate the full, parametric dependence
of the boundaries on the fluid characteristics.
The commonest way of constructing a flow map is to identify the flow pattern
(a) Numerical image of the (b) Schematic representation of the main

flow around a Taylor bub- hydrodynamic features assessed in the
ble. (Morgado, 2016) slug unit. (Morgado, 2016)
Figure 2.5: Slug unit.

Figure 2.6: A flow regime map for the flow of an air-water mixture in a vertical, 25 mm di-
ameter pipe showing the experimentally observed transition regions hatched.
Adapted from Weisman (1983).
at a set of conditions covering the field, and then to sketch in boundary lines
separating the different patterns. Because of problems in correctly identifying flow
patterns, it often happens that a few experimental points lie on the wrong side
of these lines, and the lines would be better regarded as transition zones. This
should always be remembered when using maps on which only the boundary lines
appear [13]. Besides these difficulties there are a number of other troublesome
questions. In single phase flow it is well known that an entrance length of 30 to
50 diameters is necessary to establish fully developed turbulent pipe flow. The
corresponding entrance lengths for multiphase flow patterns are less well established
and it is quite possible that some of the reported experimental observations are for
temporary or developing flow patterns. Moreover, the implicit assumption is often
made that there exists a unique flow pattern for given fluids with given flow rates.
It is by no means certain that this is the case [24]. Consequently, there may be
several possible flow patterns whose occurence may depend on the initial conditions,
specifically on the manner in which the multiphase flow is generated. In vertical
upward flows, the pressure decreases with increasing elevation, but also drops due to
viscous friction; this implies a decrease in gas density and a commensurate increase
in the gas superficial velocity with increasing elevation. Consequently, the flow
pattern at a certain location in the pipe would not only depend on inlet conditions,
but also on the coupled actions of bubble breakup/coalescence and pressure drop.
For some conditions, gas-liquid flows may not reach full development even not for
very long axial pipe lengths. Among the wide variety of flow pattern maps for
vertical flow found in the literature, Taitel et al. [100] suggested physically based
mechanisms which underlie each transition and modelled the transitions based on
these mechanisms. Figs. 2.7a and 2.7b report their results.
(a) 50 mm diameter (b) 25 mm diameter
Figure 2.7: Flow pattern map for vertical tubes, air-water. (Taitel et al., 1980)
2.2.3 Transition mechanism

In this section we concentrate on the qualitative features and underlying insta-
bilities, that give rise to transitions, of the boundaries between the following flow
patterns: bubble-slug and slug-churn.
2.2.3.1 Bubble-slug transition

Transition from the condition of dispersed bubbles observed at low gas rates
to slug flow requires a process of agglomeration or coalescence. As the gas rate
is increased, the bubble density increases. This closer bubble spacing results in
an increase in the coalescence rate. However, as the liquid rate increases, the
turbulent fluctuations associated with the flow can cause breakup of larger bubbles
formed as a result of agglomeration. If this breakup is sufficiently intense to prevent
recoalescence, then the dispersed bubble pattern can be maintained [100]. Hence
in any bubbly flow, two opposing processes are at work: bubble coalescence as a
result of collisions between bubbles, and bubble break up as a result of turbulence
in the liquid phase.
At low liquid velocities, turbulence is small, so coalescence dominates and the
equilibrium bubble size is large; these larger bubbles have distorted, constantly
varying shapes, and rise in zigzag or spiral motion. At higher velocities turbulence
is increased, and the equilibrium bubble size is smaller, reducing collision. The
work of Song et al. [93] indicates that the void fraction at transition depends on
the bubble size. Kytomaa and Brennen [62] and Cheng et al. [29] have considered
a wider range of pipe diameters, 25 mm to 100 mm, and show that the critical void
fraction depends on the ratio of mean bubble size to pipe diameter.
With increases in gas flow rate, at low liquid rates, the rate of agglomeration to
larger bubbles increases sharply. This results in a transition to slug flow. Recent
work has shown that this flow pattern does not occur in larger diameter pipes
(0.15 m and 0.2 m), where there is a direct transition from bubble to churn [70].
Experiments suggest that the bubble void fraction at which this happens is around
0.25 to 0.30 [43]. Published data agree in that the void fraction in bubbly flow
rarely exceeds 0.35, whereas for void fractions less than 0.20 coalescence is rarely
observed [44]. A review of models for this transition can be found in [13].
2.2.3.2 Slug-churn transition

It is difficult to accurately identify the slug-churn transition because there is
confusion on the description of the churn flow itself [15]. We characterize the churn
flow pattern as the condition where oscillatory motion of the liquid is observed. In
slug flow, the liquid between two Taylor bubbles moves at a constant velocity and
its front as well as its tail have constant speed. In churn flow, the liquid slug is
too short to support a stable liquid bridge between two consecutive Taylor bubbles.
The falling film around the bubble penetrates deeply into the liquid slug creating a
highly agitated aerated mixture at which point the liquid slug is seen to disintegrate
and to fall in a rather chaotic fashion [100]. The liquid reaccumulates at a lower
level at the next slug where liquid continuity is restored and the slug then resumes
its upward motion. Thus, an oscillatory motion of the liquid can be observed, which
is a characteristic of churn flow. Nicklin and Davidson [69] suggested that churn
flow occurred when the gas flow is sufficiently low to cause flooding in the film
surrounding it. This mechanism of transition was also suggested by Jayanti and
Hewitt [59] and Watson [110].
The mechanistic models that have been published in the identification of the
occurence of this transition are presented in [13].
2.3 CFD methodologies for two-phase flow

Several models have been designed to simulate multiphase phenomena: three
levels of modeling can be identified based on the spatial and temporal resolution of
the models.
2.3.1 Direct numerical simulation

The direct numerical simulation of gas-liquid flow is also called an interface
resolving method [64]. The DNS techniques are based on the local instantaneous
conservation equations. These methods focus on the finest level, e.g. individual
bubbles, small vortices behind bubbles and bubble-bubble interactions. All the
closure equations for the forces acting on a bubble can be directly computed.
However, these approaches are restricted to a single bubble or to a few interacting
bubbles due to very expensive computational requirements. In general, the numerical
methods employed to solve interfaces are based on Finite Volume Methods (F V M )
and can be classified into two categories: interface-tracking and interface-capturing
(see Fig. 2.8).
Regarding the interface-tracking methods, they follow the motion of the interface
with high precision [105], [104]. The conservation equations are solved for each phase
and they are coupled across the interface, implementing the kinematic and dynamic
2.3. CFD methodologies for two-phase flow 17
Figure 2.8: Schematic representation of interface-tracking and interface-capturing meth-

ods used to determine the interface in multiphase systems simulations. (a)
Interface-Tracking method, (b) Detail of (a) near the interface, (c) Front
Tracking (FT), (d) Level-Set (LS), (e) Marker and Cell (MAC), (f) Volume
of Fluid (VOF). [26]
conditions at the interface, which is defined as a sharp surface. Furthermore,

each phase is solved on its own mesh, which deforms according to the interface
movement. This method is limited to moderate interface deformations without
topology changes, although this can be addressed through re-meshing the domain.
As for the interface-capturing methods [86], [92], they solve not only the continuity
and momentum equations in both phases, but also an additional equation for a
scalar function, used as a phase indicator. Alternatively, marker-particles might be
used to determine the interface over a fixed grid.
Moreover, different techniques can be employed in the interface-capturing meth-
ods, which can be classified into two main groups: surface and volume meth-
ods [26], [54].
Concerning surface methods, several techniques can be employed, such as Front
Tracking (FT) or Level-Set (LS) [96]. In the FT technique, the interface is fitted
using connected massless particles over the free-surface or a secondary moving grid
to locate the interface in a Lagrangian way across an Eulerian mesh. Although in
this algorithm the interface has a zero thickness, capturing interface deformations
can be troublesome. In the case of the LS technique, the interface is defined as a
zero level of a colour function to identify the interface position. Similarly as with
FT, the interface is tracked accurately with a zero thickness, although mass is not
always conserved.
Regarding volume methods, several techniques are also available, such as the
Marker and Cell (MAC) and the Volume of Fluid (VOF) techniques [52], where the
latter is the most popular one. In the VOF technique, the interface is determined
by using a scalar indicator which varies between one (fluid 1) and zero (fluid 2),
corresponding to the volume fraction occupied in each cell. As a consequence, in
the VOF method, the interface is characterized by a discontinuous change in the
volume fractions and difficulties related to its advection are well documented in the
literature [42].
There are several ways to implement this concept and in the framework of
OpenFOAM the methodology is described in extenso by Ubbink [106], Rusche [83]
and Berberovic et al. [21]. In OpenFOAM [66], the solver interFoam is developed
based on the compressive volume-of-fluid method, which incorporates an interfacial
compression flux term to mitigate the effects of numerical smearing of the inter-
face [36], [85], [111]. The interface location, its normal and curvature are known
only implicitly from the underlying indicator function.
The main features of VOF are the conservation of mass and the ability to include
phenomena such as interface breakup and coalescence. It should also be noted
that this technique is dependent on the grid resolution, which might be a source
of errors in the advection and reconstruction algorithms. This drawback induces
the so-called artificial or numerical coalescence of the disperse phase, preventing
the real definition of the interface when the separation between two interfaces is
smaller than the size of the computational cell.
Furthermore, the calculation of the surface tension force constitutes a key issue
of these methods, since its effects at the interface play a decisive role in multiphase
flows [18], [11]. In most cases, the surface tension term is accounted for in the
momentum equation by using the Continuum Surface Force (CSF) model developed
by Brackbill et al. [22].
Nevertheless, this model may introduce parasitic or spurious currents [45], [41].
The main sources of spurious currents have been identified as inaccurate interface
curvature and lack of a discrete force balance [36]. Several solvers have been
developed to reduce these parasitic currents [73], [78] although the problem has not
been completely solved so far [57].
2.3.2 Eulerian-Lagrangian method
In the Eulerian-Lagrangian approach, the continuous phase is treated in the

Eulerian framework and the averaged equations are solved, whereas the motion of
individual discrete particle is simulated by solving Newton’s equation of motion.
The trajectories of the particles are computed in the control volume. In contrast
to the DNS methods, the Eulerian-Lagrangian approach requires closure relations
to account for the interphase forces. The closure models can be obtained from
empirical relations or from more sophisticated simulations with a fine resolution.
Most of the gas-solid Eulerian-Lagrangian methods are developed by coupling
CFD with the discrete element method (DEM): also OpenFOAM is capable of
doing the Eulerian-Lagrangian simulations for gas-solid flows [117].
2.3. CFD methodologies for two-phase flow 19
Figure 2.9: Representation of short and long geometrical scales in a bubbly flow. a)
Long scale interfaces, b) short scale interfaces, c) presence of both scales
simultaneously. [84]
2.3.3 Eulerian-Eulerian method

In the Eulerian-Eulerian approach, each phase is treated as a continuous medium
interpenetrating the other phase, and is represented by the macroscopic conservation
equations, which are valid throughout the entire flow domain.
This method is commonly known as the two-fluid model or multifluid model [55].
This approach requires less computational effort than the Eulerian-Lagrangian
approach. However, the discrete character of the dispersed phase is lost due to
the averaging procedure. The model employs averaged mass and momentum
conservation equations to describe the time-dependent motion of both phases
and requires additional models for the inter-phase momentum transfer and for
the turbulence, due to the averaging process [83]. In [112] the derivation, using
conditional averaging ( [37]), of the two-phase equations is outlined followed by
the manipulation necessary to cast the equations in a form suitable for numerical
solution.
The two-fluid model is widely employed to simulate gas-liquid flows. Most of
the gas-liquid two-fluid simulations were carried out using a single mean bubble
size [116]. This assumption is usually reasonable in the homogeneous flow regime.
To account for the bubble size distribution, many attempts have been made to
couple computational fluid dynamics with a population balance model (CFD-PBM)
to simulate gas-liquid flows [28], see [19], [90].
The applicability of each model and the quality of the results strongly relies
on the nature of the problem. The VOF model is used in problems where surface
capturing is crucial, such as the importance of surface tension and adhesion phe-
nomena (jet break-up, drop formation) or where the free surface position prediction
is essential (free-surface problems in hydraulics, liquids separation). In all of these
cases the interfaces are covered by an appropriate mesh size, Fig. 2.9. In the case
of two-fluid models the interest is in the capability of predicting the behaviour of
flows with small-scale interfaces when is not possible or desirable to have a more
complete modeling. This kind of interfaces is often found in sedimentation tanks,
cyclone separators, annular flow in refineries and fine bubbles flow in heat exchangers.
The DNS is the only method that is capable to resolve general multiphase
problems [86]; however, the required huge computational resources make DNS out
of reach for most applications. On the other hand, due to their lack of generality, the
other models work only for particular cases that sufficiently satisfy the underlying
model assumptions.
This master thesis deals with the validation of a two-phase CFD code in
OpenFOAM based on the combination of an Eulerian two-fluid approach and sharp
interface capturing using the Volume of Fluid approach. Details of the model are
presented in Chapter 3.
Chapter 3
Coupling Euler-Euler two-fluid with

VOF
3.1 Hybrid approach

In a dispersed flow, where chunks of the particular fluid are smaller than the grid
cells, the interface tracking algorithm fails. Two-phase flows of practical importance,
including the flow regimes described previously in 2.2, can be too dispersed to be
resolved with the interface tracking algorithm. The two-fluid models are suitable for
two-phase problems where the length scale of the interface shape is smaller than the
grid size. However the basic assumption of those models is that each fluid behaves
as a continuum that occupies the whole domain and the information that is lost due
to the averaging is replaced by closure relations for the interfacial transfer of mass,
momentum and energy. This approach is suitable for simulation of the dispersed
flow. The equations of the two-fluid model are less accurate than the VOF model
due to the empirical closures required in the averaged equations. However, in the
case of sufficiently dispersed flow, the results of the two-fluid model are still much
closer to the laboratory and field data than the results of the VOF method, which
do not have any physical meaning when the grid becomes too coarse. Moreover
VOF fails to take into account the slippage velocity between the two phases due
to its shared momentum concept. Therefore at higher flow rates, where a large
slippage exists, the functionality of this model is questionable [74].
In complex multiphase flows in which both dispersed flow and segregated flow
regions are present one would like to couple VOF and the two-fluid model. From
the physical point of view the coupling is not problematic since the VOF model
uses an indicator function for tracking the interface between the phases, which has
the same meaning as a volume fraction variable in the two-fluid model.
One of the first studies in this field is due to Cerne et. al. [27] where a
coupling between the VOF and two-fluid models is used to solve the Rayleigh-
Taylor instability. They employed a switching routine based on the gradient of the
volume fraction across neighbouring cells to flag cells as either VOF or two-fluid
and solved the appropriate number of equations in each cell. Thus, there is a
threshold value over which the interface is treated as having long scale and captured
by VOF, and the opposite case with the two-fluid model. The challenge of such a
21
22 Chapter 3. Coupling Euler-Euler two-fluid with VOF
coupled model was to combine two mathematical models with different numbers of
equations across the same domain.
Strubelj and Tiselj [95] give a good overview of methods that have been employed
for this coupling along with details regarding difficulties in coupling the phase
momentum equations at the sharp interface, where the phase velocities should be
equal. This issue has particular importance in the treatment of the velocities, since
the VOF model has only one velocity field whereas the two-fluid model has one
velocity per phase. The transition from two velocities to one is managed by the
definition of the velocity for the centre-of-volume, in the opposite case the same
velocity is assigned to each phase, losing the interface friction.
Wardle and Weller [109] developed a hybrid multiphase method in OpenFOAM
based on the combination of an Eulerian multifluid solution framework and sharp
interface capturing. This solver, named multiphaseEulerFoam, has been included
in the release of version 2.1 of OpenFOAM.
3.2 multiphaseEulerFoam
In multiphaseEulerFoam a numerical interface sharpening is implemented
within the Eulerian framework. The advantage is that the governing equations
to be solved are the same in the whole domain; on the other hand, a numerical
sharpening on top of an Eulerian solver may not be as effective as a coupled model
with interface capturing capabilities [103].
3.2.1 Multifluid momentum equations

In order to derive the conservation equations of the model, the individual phases
should be distinguished. This is achieved by conditioning the local equations so
that contributions to the averaged conservation equation of one phase come only
from regions which contain that particular phase [51]. The idea of conditioning
is based on the work of Dopazo [37]. For the conditional averaging (sometimes
called phase-weighted averaging), the governing equations are multiplied by a phase
indicator function before standard averaging techniques are applied.
The phase indicator function Ik (x, t) is defined as:
(
1 if point (x, t) is in phase k
Ik (x, t) (3.1)
0 otherwise
The phase volume fraction is calculated as the probability of point (x, t) being
in phase k:
αk = Ik (x, t) (3.2)
where the overbar represents the ensemble average. The ensemble average is
more fundamental than time and volume averaging and it does not have the time
3.2. multiphaseEulerFoam 23
and space restrictions. As already pointed out the averaging process introduces
new variables and terms and therefore closure laws are required.
In Rusche [83] and Weller [112] this ensemble averaging method is described in
details, and this is not reproduced in the present report.
Assuming that there is no mass transfer across the phases, the conditionally
averaged phase continuity and momentum equations for incompressible, isothermal
flow are given by a set of mass and momentum equations for each of the phases k:
∂αk
+ uk · ∇αk = 0 (3.3)
∂t
∂(ρk αk uk )
+(ρk αk uk ·∇)uk = −αk ∇p+∇·(µk αk ∇uk )+ρk αk g +FD,k +Fs,k (3.4)
∂t
where ρk , αk , uk are the density, phase fraction and velocity, respectively, for
phase k and g is the gravity. The two interfacial forces are the drag force FD,k
and the surface tension force Fs,k . Generally, in the gas-liquid flows, the interfacial
forces (interphase momentum transfer) are divided into two categories: drag force
and non-drag forces. The non-drag forces are the lift force, the virtual mass force,
the turbulent dispersion force and the wall lubrication force [64]. In this work lift,
virtual mass force, wall lubrication, and turbulent dispersion are neglected, and
only the drag force, that is the dominant interfacial force, is included [74].
3.2.1.1 Interface capturing

The solver has a flexible algorithm that is able to give a sharp interface between
the phases. The interface sharpening method of Weller [111] is employed, wherein
an additional term is added to Eq. (3.3) in the following way:
∂αk
+ uk · ∇αk + ∇ · (uc αk (1 − αk )) = 0 (3.5)
∂t
The interface compression method for interface capturing is not as accurate as
other interface reconstruction methods; however, it is much easier to implement
and it is mass conserving [42]. The interface compression scheme of Weller adds
an additional term to the LHS of the volume fraction transport equation. This
additional convective term is referred to as the compression term keeping in mind
its role to compress the free surface towards a sharper one (it should be noted
that the wording compression represents just a denotation and does not relate to
compressible flow). The value for the artificial interface compression velocity, uc , is
given by:
∇α
uc = Cα |u| (3.6)
|∇α|
uc is applied in the direction normal to the interface to compress the volume fraction
field and to maintain a sharp interface. The term αk (1 − αk ) ensures that it is only
active in the interface region [109].
Since the derivation of Eq. (3.5), through Eq. (3.6), relies on the weighted
average velocity u, a strong coupling between the classical VOF and the two-fluid
model is achieved.
The term |∇α|

∇α
denotes the interface unit normal vector for the direction of
the applied compression velocity. The magnitude of the velocity |u| is used since
dispersion of the interface can occur, in the worst case, as fast as the magnitude of
the local velocity [109].
The coefficient Cα controls the interfacial compression which can be switched
on (Cα = 1) or off (Cα = 0). With Cα set to 0 for a given phase pair, there is no
imposed interface compression which will result in phase dispersion according to
the two-fluid model. In contast to this when it is set to 1, sharp interface capturing
is applied and VOF-style phase fraction capturing occurs. The implementation of
the solver is configured such that the interface compression coefficient Cα is defined
and applied independently for all phase pairs.
It is necessary to devise a method to calculate the coefficient Cα such that
interface capturing is used only in regions where meshing is sufficient to resolve the
interface. However, in the release version of multiphaseEulerFoam, Cα is simply
defined as a scalar that remains a constant value in both space and time. In this
work Cα is equal to 1. Even if the coefficient Cα is always set to 1, we should keep
in mind that the momentum equations for each phase (after conditional averaging)
include the interfacial forces (see Eq.(3.4)) and therefore the model does not become
a simple VOF model. This thesis aims to provide a preliminary validation of the
solver. Therefore further details on the implementation of Cα as a spatially and
temporally varying volumetric field variable, are not considered. Interesting studies
in this direction are [109], [84] and [89].
3.2.1.2 Surface tension
The surface tension at the gas-liquid interface generates an additional pressure

gradient (or jump), which is evaluated using the continuum surface force (CSF)
model (see [22]). This model interprets surface tension as a continuous, threee
dimensional effect across an interface, rather than as a boundary value condition
on the interface:
Fs,k = σκ∇α (3.7)
where σ is the fluid surface tension coefficient and κ is the local surface curvature
determined from:

∇α
κ = −∇ · (3.8)
|∇α|
The accurate calculation of the phase fraction distribution is crucial for a proper
evaluation of the surface curvature, which is required for the determination of
the surface tension force and the corresponding pressure gradient acrosss the free
surface. The interface region between two phases is typically smeared out over a
few grid cells and is therefore highly sensitive to the grid resolution.
Version 2.2.x of OpenFOAM is used in this work. In this version, and also
in the following versions, surface tension was omitted from the reconstruction of
the velocity field, though it is included in the construction of the pressure itself.
However it is straightforward to fix the velocity reconstruction, see E and E.1.
3.2.1.3 Interphase drag models

The interface drag represents the resistance opposed to the bubble motion in
the fluid (or, more generally, the resistance due to the relative motion between two
phases). The drag force clearly depends on the bubbles size (i.e. a larger bubble
experiences a larger drag force) and on the relative velocity between the two phases.
The drag term FD,k is given by [109]:
FD,k = αc αd K(ud − uc ) (3.9)
where the subscripts c and d denote the continuous and dispersed phase values
and K is given by:
3 |ud − uc |
K = ρc CD (3.10)
4 dd
The drag calculation implemented in the solver is generic, such that the model must
simply return the value of K. The drag coefficient CD is usually deduced from
experiments and many models were developed in order to fit different experimental
data sets. The drag coefficient is related to many factors: the bubble shape,
orientation with respect to the flow, flow parameters such as the bubble Reynolds
number, the Eotvos number, turbulent level, and so on. An interesting and
comprehensive literature study on the topic can be found in [83]. In the solver
many different drag coefficient models are implemented [66].
The model of Schiller and Naumann [87] is used for the simulations in this thesis.
According to this model the correlation for the drag coefficient is:

 24 1+0.15Re0.683
CD = Re
Re ≤ 1000 (3.11)
0.44 Re > 1000
|ud − uc |
Re = dd (3.12)
νc
where νc is the kinematic viscosity of the continuous phase and dd is the diameter
of the disperse phase.
In multiphaseEulerFoam the calculation of the drag coefficient can be done
by specifying a dispersed phase or by an independent calculation with each phase
as the dispersed phase. For the latter case the overall drag coefficient, applied to
the momentum equations, is taken as the volume fraction weighted average of the
two values. This is the blended scheme in in the code listing of Table 3.1. In this
way both drag forces are calculated and the solver automatically weights the drag
forces based on the phase fraction. Different models for each dispersed phase can be
chosen.Therefore the parameters for the drag model need to be specified pair-wise.
The phases are read in the transportProperties dictionary from left to right, see
line 3 in the code listing of Table 3.1. In this example the first phase, air, is the
dispersed phase and its diameter is used for the calculation of Eq. (3.12).
Table 3.1: Drag specification in transportProperties dictionary
1 drag
2 (
3 ( air siliconeOil )
4 {
5 type blended ;
6
7 air
8 {
9 type SchillerNaumann ;
10 residualPhaseFraction 0;
11 residualSlip 0;
12 }
13
14 siliconeOil
15 {
16 type SchillerNaumann ;
17 residualPhaseFraction 0;
18 residualSlip 0;
19 }
20
21 r e s i d u a l P h a s e F r a c t i o n 1 e −3;
22 r e s i d u a l S l i p 1 e −3;
23 }
24 );
In the listing given in Table 3.1 there are two extra parameters in the setup
of transportProperties: residualPhaseFraction and residualSlip. They are
not included in the calculation of the actual phase fractions or velocities, but only
in the calculation of the drag force itself to help the coupled formulation stability.
In the limit of a sharp interface, the velocities on either side of the interface must be
equal to meet the so-called no-slip interface condition. This is an inherent feature
of traditional VOF simulations as all phases share a single momentum equation
and thus the phase velocities are the same everywhere. Imposing a sharp interface
through the addition of interface compression on top of a two-fluid formulation
requires that an additional artificial drag is used to equalize the velocities at the
interface. That is why constants for a small residual drag and residual phase
fraction are added for each phase pair [109].
3.2.1.4 Diameter models

The multiphaseEulerFoam module contains, since its introduction in version
2.1.0, two diameter models: constant and isothermal. Other models for variable
droplet size, e.g. based on a reduced population balance method [89], are compatible
with this flexible framework.
The isothermal model is used in the work reported here, assuming the change of
state to be isothermal. Gas bubbles change their diameter as the ambient pressure
changes. Generally, the ideal gas law governs the state of gas:
pV = nRT (3.13)
under the assumption of an isotermal state:
pV = const (3.14)
Next we introduce the bubble volume:
d3 π
V = (3.15)
6
Thus we obtain the relation:

p1 d31 π p2 d32 π
= (3.16)
6 6
This leads to the isothermaldiameter model:
r
p1
d2 = 3 d1 (3.17)
p2
3.2.2 Time step

To ensure convergence for a time-marching solver a restriction on the Courant
number will influence the maximum time step. multiphaseEulerFoam uses an
adjustable time step which is based on the maximum Courant number in the
domain. The definition of the Courant number for n degrees of freedom is given in
Eq. (3.18).
n
X ui
Co = ∆t (3.18)
i=1
∆x i
The time step is directly proportional to the mesh spacing, that is, if the mesh
spacing is cut in half, the timestep must essentially be decreased by a factor 2.
Bounded discretization schemes for divergence terms and time step control are both
used to improve the numerical stability. It is generally recommended to keep the
maximum local Courant number much below unity. It is also beneficial to solve the
phase fraction equation in several subcycles within a single time step. The time
step to be used in a single time subcycle, ∆tsc , is set by dividing the global time
step, ∆t, by the preset number of subcycles ( nAlphaSubCycles, nsc ).
∆t
∆tsc = (3.19)
nsc
Besides the Courant number several other physical parameters will also have an
influence on the stability and accuracy of the solver [20]. The surface tension can
impose a restriction on the time step, related to:
r
µ∆x ρ∆x3
∆t ≤ max(10τµ , 0.1τp ) with τµ = and τp = (3.20)
σ σ
3.2.3 Turbulence
In the release version 2.2.x of OpenFOAM, turbulence can be covered by using
Large Eddy Simulation (LES) with the Smagorinsky sub-grid model for the mixture
or with RANS (see B). These models have been originally developed for turbulence
modeling in single-phase flows and thus solely accounts for turbulence within the
continuous phase or within the effective fluid (i.e. the mixture). However, for the
Euler-Euler formulation it is more rigorous to have a turbulence model per phase
and inter-phase turbulence exchange terms [83], [32]. This is more general but also
more complex. In this study turbulence is accounted through RANS models for the
mixture, [74], [4], [49], [34].
3.2.4 Solution procedure

multiphaseEulerFoam handles the requirement to describe both the properties
of the individual phases as well as the interaction terms between the phases by
two distinct libraries phaseModel and multiphaseSystem. Fig. 3.1 shows the main
parts of multiphaseEulerFoam.
The procedure starts with updating the timestep according to the Courant
number limit and then solves the coupled set of volume fraction equations with
interface sharpening for selected phase pairs (3.5).
The drag coefficients are computed and an equation for the phase velocities is
constructed and solved for preliminary values.
The Pressure Implicit with Splitting of Operators (PISO) algorithm is used to
solve the pressure-velocity coupling.
For the sake of generality in Fig. 3.1, the pressure-velocity coupling is handled
by a PIMPLE loop rather than by PISO. The PIMPLE algorithm is a combination
of PISO and SIMPLE. However, if in the PIMPLE control dictionary the number
of nOuterCorrectors is set to 1 (default settings), the PIMPLE solver will operate
in PISO mode.
The implementation of the PISO algorithm follows [56] and a description is given
by Jasak in [58]. In a PISO loop the momentum equation is solved first (momentum
predictor). Then the pressure equation can be formulated giving the first estimate
of the new pressure field (pressure solution) and the velocity field is corrected as a
consequence of the new pressure distribution. The correction is done in an explicit
manner. In PISO there is no loop over the whole set of equations.
PIMPLE performs sub-iterations (outer corrector steps) to ensure convergence
and to be able to use an acceptable time step. However the time step is already
limited to small values by the nature of the problem (interface capturing) and the
advantage of PIMPLE regarding using a larger time step can not be considered for
this work.
PISO requires that the solution must converge at every time step. We assume
that one solution step plus pressure corrections are enough to obtain convergence.
Decreasing residuals indicate that we are converging to the right solution and they
are checked at the end of every simulations.
3.2.4.1 Phase fraction

In order to ensure phase conservation for the coupled phase fractions with added
interface sharpening, limiters on the phase fraction, as well as on the sum of the
phase fractions, are incorporated prior to the explicit solution of the phase fraction
equation system [109]. These additional limiters have been incorporated in a new
multiphase implementation of the Multidimensional Universal Limiter with Explicit
Solution (MULES).
The MULES algorithm handles the boundedness property by first limiting the
flux transport and then solves for the phase fraction. Limiting the flux transport is
important since large transport of fluxes from a cell may drive the volume fraction
in a particular cell below zero during one time step. The MULES function is given in
the listing of Table 3.2:
Figure 3.1: Flow chart of multiphaseEulerFoam.

Table 3.2: MULES constructor in multiphaseSystem.C

1 MULES : : e x p l i c i t S o l v e
2 (
3 geometricOneField () ,
4 phase1 ,
5 phiAlpha ,
6 zeroField () ,
7 zeroField ()
8 );
The first argument passed to the constructor is the density; in case of incompressible
phases, it is passed as geometricOneField() which is a unit value field. The second
argument is the variable to be solved which is the phase fraction in our case. The
limited normal convective flux is the next argument which is earlier solved explicitly
by MULES:limit. The next two terms are the explicit and implicit source terms in
the continuity equation which arise when mass transfer across the phases or reaction
source terms exist. In our case we pass both the arguments as zeroFields() since
there is no mass transfer. The MULES algorithm solves for the phase fraction with the
explicit consideration of the convective flux of the phase fraction. The considered
transport mechanism is convection only [61]. The procedure is showed in Fig. 3.2.
The solution for the phase fraction is invoked by the fluidSolve() function.
The schematic of the function operation is shown in Fig. 3.3. If the number of
correctors are larger than one, then the phase fraction at the old time is stored and
fluxes for all phases are set to zero.
Figure 3.2: Procedure to solve the phase volume fraction.

Figure 3.3: Solution procedure of the phase continuity equation.

Chapter 4
Validation
Numerical simulations were conducted on the cluster of LINUX computer servers

at the Laboratory for Aero & HydroDynamics of the TU Delft. This cluster uses
the Slurm scheduler to efficiently manage workloads. One node with 12 processors
has been used for the simulations presented in this thesis.
Before proceeding it is worth noting that the assumptions made in 1.2 are
the basis for the following analysis. The starting objective of this thesis is the
comparison of the results obtained with OpenFOAM against experiments and
reference solutions that have been obtained previously with Fluent and Star-CCM+
for the industry.
4.1 Case study 1

This section will present results for the first case, which consists of annular
flow at the inlet. Section 4.1.1 will introduce the problem and the flow regime
considered. Section 4.1.2 discusses the computational details, such as the mesh and
solver settings. Results will be presented in section 4.1.3.
4.1.1 Case description

In [115] Worthen and Henkes carried out CFD simulations with ANSYS Fluent
15.0 for the splitting of two-phase, gas-liquid flow from a horizontal flowline into two
vertical risers. The same flow conditions that were used in the air-water experiments
at the Shell Technology Centre Amsterdam were simulated. In addition to the
splitter geometry, the flow through the vertical pipe alone with fixed inlet boundary
condition was simulated, as is illustrated in Fig. 4.1. The latter is the case study
that is considered in this work. The outlet is located 50D (2.54 m) downstream the
inlet.
The liquid holdup fraction at the inlet was set to 0.18 which is the value predicted
by the Shell Flow Explorer tool (SFE version 6.0). The inlet flow rates of gas and
liquid were:
Qair = 31.1 m3 /h and Qwater = 1 m3 /h (4.1)
The nominal pipe diameter was 50.8 mm which gives the superficial velocities
of:
33
34 Chapter 4. Validation
Figure 4.1: Model geometry for simulation of vertical upward flow through a 50.8 mm
diameter pipe.
Us,air = 4.26 m/s and Us,water = 0.137 m/s (4.2)

The liquid entered the domain as an annular film while the gas entered through
the core and then the two phases flow upwards and discharge through the outlet
at atmospheric pressure. The same conditions are imposed in the simulation with
OpenFOAM. In the Fluent simulations of [115] a total flow time of 9 s was simulated
while in the present study the physical time used is 14 s. This time interval allows
flow stabilization. The liquid holdup and pressure drop were only calculated over
the last 30D of pipe length to ensure a more fully developed flow. The same
quantities as considered with Fluent are calculated from the present OpenFOAM
simulations in order to compare the results.
4.1.2 Case setup

4.1.2.1 Mesh
It has been reported by Hernandez-Perez et al. [49] and confirmed in [3] and [102]
that different grid structures, applied in the cylindrical pipe geometry, can give
a different numerical accuracy. In [49] the O-grid is highly recommended for the
simulation of two-phase flow in a pipe. This grid allows refining the mesh close to
the wall and prevents a singularity at the centre of the pipe. In this grid, a Cartesian
mesh is used in the centre of the pipe combined with a cylindrical one around
it. It requires multiple blocks but generally has the best grid quality in terms of
orthogonality and mesh density. Therefore the O-grid is the type of mesh employed.
The construction of the mesh is realized with blockMesh. A parameterization of
the blockMesh dictionary is done by using m4-scripting (see A). Fig. 4.2 shows a
section of the computational domain. The computational mesh contains 563200
hexahedral cells. The number of cells along the axis of the pipe (positive z axis) is
4.1. Case study 1 35
Figure 4.2: Cross-Section grid of the computational domain.
uniform.
It was not possible to carry out a grid convergence study to check the mesh
dependency because of the high computational costs required, however we can
consider the work of Worthen & Henkes in [115] where the mesh dependency of
the solution for the same case is checked. Moreover there are experimental results
available for the comparison with the CFD results. About 55 days of real time were
needed to simulate 14 seconds of physical time by performing a parallel computation
on 12 processors with the time step limitations discussed in 4.1.2.6.
The mesh shown in Fig. 4.2 is a rather coarse grid for an interface capturing
solver, where a mesh spacing of ∼ 10 times smaller than the smallest droplet would
be required. However the scope of this work is to provide a first validation of the
solver using limited computational time.
The idea of the coupled solver is to deal with different scales: the VOF model is
appropriate for flows with large interfacial length scales (larger than grid size), and
the Eulerian-Eulerian model for the cases with small length scales. Neither of them
seems to be applicable for this problem: VOF requires a finer mesh and two-fluid is
to be discarded because we want to track the interface, at least for the large scales
of the annular liquid film. With this problem we can test multiphaseEulerFoam
capabilities.
The quality of this full hexahedral O-grid mesh is high. The maximum skewness
is equal to 0.48 and the maximum mesh non-orthogonality is equal to 38.3. Non-
orthogonality introduces misalignment between the face normal vector and the
computed gradients and affects the Laplacian terms. High values of skewness can
(a) Liquid holdup (b) Velocity
Figure 4.3: Boundary conditions at the inlet.
affect the convective terms. It introduces a first order interpolation error due to
the distance between the intersection of the line connecting two cell centres with
their common face and the centre of that face. The volume ratio plays also a role,
introducing artificial, non-physical boundaries and inconsistent interpolation.
4.1.2.2 Turbulence
Turbulence is modelled using the RANS equations with the SST K − ω model
for the mixture [115], with wall functions [76], [114] (see also B.2). The initialization
of the turbulence quantities follows the expressions provided in [46].
4.1.2.3 Boundary and initial conditions

Flow rates are prescribed at the inlet through the superficial velocities (see (4.1)
and (4.2)) and through the liquid hold up. Fig. 4.3 shows the boundary condition
for the inlet. The liquid holdup is 1 for the annular film and 0 in the gas core. The
velocity at the inlet is shown in Fig. 4.3b where the air and water velocity (not the
superficial ones, see Eqs. (2.3)) are prescribed to ensure that the flow rates in the
simulations and in the experiments are identical.
In order to have the non-uniform volume fraction and velocity conditions at the
inlet the special utility provided by OpenFOAM, called funkySetBoundaryField,
is used. At the wall the no slip condition is applied for the velocity. At the outlet
the pressure is fixed at the atmosferic value through the type fixedMean. It is
worth mentioning that the first choice was for the type totalPressure boundary
condition. This condition reverts to a fixed value condition for flow out of the
domain. However this was causing an instability when a liquid/gas mixture starts
to leave the domain. Therefore a fixedMean condition for the pressure is used at
the outlet, which extrapolates the field to the patch using the near-cell values and
adjusts the distribution to match the specified mean value.
The initial condition is that the riser contains gas and that all the velocity
components are equal to 0 m/s.
4.1.2.4 Fluid properties

The relevant fluid properties are:
ρair = 1.25 kg/m3 and µair = 0.0178 cP (4.3a)
ρwater = 999 kg/m3 and µwater = 1.3 cP (4.3b)
The air-water surface tension is specified as 0.0742 N/m.
4.1.2.5 Solver Settings

The subdirectory system contains all the files related to the solution procedure
itself. Governing equations have to be discretized in time and space to be solved
numerically and the discretization schemes adopted can be specified in fvSchemes.
For the time discretization the implicit Euler scheme is used, which is a first
order scheme. For the discretization of gradient terms the leastSquares scheme is
used with cellMDLimited as gradient limiters. It is a second-order scheme,with
least squares distance calculation using all neighbour cells. Gradient limiters will
avoid over and under shoots on the gradient computations. For laplacian terms
the Gauss linear limited corrected scheme is adopted. The discretization of
the divergence terms is specified with the Gauss scheme with Van Leer (strictly
bounded between 0 and 1, vanLeer01) and with the linear interpolation scheme.
The fvSchemes dictionary is reported in D. Using the implicit first order scheme in
time reduces the overall accuracy to first order. However the rather coarse mesh
used is already limiting the spatial discretisation accuracy. Equivalent temporal
and spatial accuracy implies that the peak error magnitudes from temporal and
spatial discretisation are the same. It would be useful to know how much of the
total error is caused by spatial and how much by temporal discretisation but it is
outside the scope of this work, keeping in mind the objective to simulate industrial
flows.
In fvSolution the linear-solver control is specified. Each discretized equation
needs to have a selected linear solver and the solution tolerances need to be
specified specified to control the accuracy of the solver. Specifying a low tolerance
might lead to long calculation times. It is therefore important to consider an
appropriate compromise between accuracy and the number of iteration loops needed
to achieve convergence. To solve the pressure equations the Geometric-algebraic
multi-grid solver (GAMG) was chosen. The principle behind GAMG is to first calculate
an intermediate solution on a coarse mesh, and then map it onto a finer grid
using the first solution as an initial guess. This is considered a relatively quick
method. The pressure and velocity fields were coupled by the PISO algorithm,
since nOuterCorrectors is set to 1. The fvSolution dictionary is reported in C.
4.1.2.6 Simulation control

The time step is adjuststed to the largest possible value, while still fulfilling the
Courant number criterion, that is the maximum Courant number being specified to
Figure 4.4: CFD results with OpenFOAM for the flow pattern in the vertical pipe at t
= 10 s. Cells having a water volume fraction larger than 0.5 are coloured
blue.
0.2. The upper limit on the time step is set to 5e−6 which decrease the maximum
Courant number to ∼ 0.03 during the simulation. This is done to increase the
stability of the simulation and improve the convergence.
4.1.3 Results
The CFD results described in [115] were obtained with Fluent using a VOF
method. An attempt to use the VOF method is made also in OpenFOAM with
similar results. The disagreement between experiment and CFD (see Table 4.1 and
Fig. 4.9) is attributed to the failure of the VOF method to predict slug flow in the
risers. Therefore the hybrid model is used in the following simulations.
In the experiments the gas flow was too low to obtain annular flow in the risers
and the flow condition observed is churn flow. The CFD results for the flow pattern
in the vertical pipe at t = 10 s are presented in Fig. 4.4. The CFD simulations
predict a churn-like flow in the riser.
Fig. 4.5 displays time series of cross-sectional averaged void fraction data (see
Eq. (4.4)) which contain prime information concerning liquid structures within the
flow. All time series exhibit cylic fluctuations that are in the form of sudden drops
in cross-sectional averaged void fraction. The presence of the drops in the time
series is due to the passage of cyclic liquid structures. The amplitude of these drops
is controlled by the liquid volume of the passing structure. In other words, the
more the liquid content of the passing structure, the larger the amplitude of the
drops in the time series. According to Parsi et al. [75], the cyclic liquid structures
can be liquid continuous (liquid slug body) or non-continues liquid structures
(huge wave, disturbance wave). In order to determine the time series of the void
fraction, the following procedure is followed (which is similar to the one used by
Hernandez-Perez. [48] and Abdulkadir [3]): a cross-sectional plane is defined across
the measurement location and an area-weighted average value of the void fraction
is calculated. The area-weighted average of the void fraction is computed by:
n
1 1X
Z
αA = αi Ai (4.4)
A A i=1
where the sum is extended to all the faces of the cells that lie on the measurement
section.
Fig. 4.6 demonstrates the ability of the CFD model in OpenFOAM to reproduce
a liquid structure briefly bridging the pipe cross-section as well as large interfacial
waves of different sizes compressing the gas core.
The instantaneous (at t = 10 s) liquid holdup and mixture velocities are shown
in Fig. 4.7 and Fig. 4.8 for different sections.
For this case the experimental results available are the pressure drop and the
liquid holdup fraction. The liquid holdup was not measured but it was estimated
based on the measured pressure drop across the riser, assuming that the pressure
drop was due solely to gravity (i.e., frictional pressure drop was assumed negligible).
The liquid holdup and pressure drop were time-averaged after a time at which
the flow was considered to have been stabilized to obtain a single value for each
and compare them with [115]. In the work of Worthen & Henkes the quantities
were time-averaged over the time between 2 s and 9 s. From Fig. 4.9 we can see
that for the case with multiphaseEulerFoam the time to reach fully developed flow
is larger than with Fluent and therefore the average is computed over a flow time
from 5 s to 14 s.
The value of the pressure drop per unit length is computed by defining two
cross-sectional planes: at 20D and at the outlet. An area-weighted average value
of the pressure is calculated for the two sections (p20D and pOutlet respectively).
Finally the pressure drop is given by:
p20D − pOutlet
∆pl = (4.5)
l
where l is the distance between the two sections. The total liquid holdup fraction is
simply defined as the volume average of the phase fraction of water over the entire
pipe. Table 4.1 gives a summary of the predicted and measured pressure drop per
unit length and for the liquid holdup fraction for upward flow through a 50.8 mm di-
ameter vertical pipe. The results provided by this work with multiphaseEulerFoam
are indicated as OpenFOAM.
The prediction by OpenFOAM using multiphaseEulerFoam agrees better than
any other method with the measurements, indicating that multiphaseEulerFoam
does properly capture the multiphase flow pattern. The large differences between
the liquid holdup fraction and pressure drop as found in the CFD simulation
with multiphaseEulerFoam and the CFD simulation with VOF, show the capability
of this hybrid model to overcome the limitations of the VOF model that was used
in Fluent and OpenFOAM.
Fluent and OpenFOAM with VOF overpredicts the dispersion between the
phases, which gives a too low liquid holdup fraction, which in turn also gives a too
low hydrostatic head and thus a too low pressure drop.
0.8
Void Fraction
0.6
0.4
0.2
0
7.3 8.3 9.3 10.3 11.3 12.3 13.3 14.3
Time [s]
(a) 20D from inlet
0.8
Void Fraction
0.6
0.4
0.2
0
7.3 8.3 9.3 10.3 11.3 12.3 13.3 14.3
Time [s]
(b) 30D from inlet
0.8
Void Fraction
0.6
0.4
0.2
0
7.3 8.3 9.3 10.3 11.3 12.3 13.3 14.3
Time [s]
(c) 40D from inlet
Figure 4.5: Time series for the void fraction in the CFD simulations with OpenFOAM.
(a) Liquid holdup
(b) Streamlines for the gas phase on the yz-plane
Figure 4.6: CFD results for the last 30D of the vertical pipe at t = 10 s.
(a) 10D (b) 20D
(c) 30D (d) 40D
Figure 4.7: Liquid holdup at different cross sections at t = 10 s.

(a) 10D (b) 20D
(c) 30D (d) 40D
Figure 4.8: Mixture velocity at different cross sections at t = 10 s.

0.3
0.25
Total liquid holdup fraction
0.2
0.15
0.1
Experiment
0.05 Hybrid model with OpenFOAM
VOF with OpenFOAM
VOF with Fluent
0
0 2 4 6 8 10 12 14
Time [s]
Figure 4.9: Total liquid holdup fraction in the 50D long vertical pipe.
Table 4.1: Results comparison: data with Fluent, SFC, and OLGA from [115] and with
OpenFOAM.
Pressure Drop Pa/m Liquid Holdup Fraction Flow regime

Experiments 2400 0.24 churn/slug
OpenFOAM Hybrid model 2370 0.21 churn
OpenFOAM VOF 865 0.074 intermittent
Fluent VOF 860 0.072 intermittent
OLGAS 7.2.2 1250 0.19 slug
Shell FC 1050 0.18 slug
multiphaseEulerFoam better captures the separation between the phases,

which gives a relatively high liquid holdup fraction and high hydrostatic head (high
pressure drop), in line with the experiments.
4.2 Case study 2

This section will present results for the second case, being slug flow in a vertical
pipe. Section 4.2.1 will introduce the problem and the flow regime considered.
Section 4.2.2 discusses the computational details, such as the mesh and solver
settings. Results will be presented in section 4.2.3.
4.2.1 Case description

This second test case is based on the work of Abdulkadir et al. [3], [5], [6], [8],
which compares the results obtained from experiments and CFD studies of slug flow
in a vertical riser. In [3] a series of two experimental investigations were carried
out for a 6 m vertical pipe with a 0.067 m internal diameter using an air-silicone
Table 4.2: Number of cells used in different grids.
Grid name Cross-section z-direction Total

a 352 180 63360
b 513 207 106191
c 737 250 184250
d 1148 300 344400
oil mixture. The details of this experimental facility can be found in Azzopardi et
al. [6], [9]. The experimental test section consists of a transparent acrylic pipe and
the flow patterns were recorded using electrical capacitance tomography (ECT) and
wire mesh sensors (WMS). A detailed description of the ECT and WMS technology
can be found in [6], [10], [91] and [7]. A ring of two measurement electrodes (also
known as twin-plane sensors) were placed around the circumference of the riser at
a given height above the injection portals at the bottom of the riser section. The
use of the twin-plane sensors enabled the determination of the rise velocity of any
observed Taylor bubbles and liquid slugs. The twin-plane ECT sensors were placed
at a distance of 4.4 m (ECT-Plane 1) and 4.489 m (ECT-Plane 2) from the base of
the riser. The capacitance WMS (WMS-Plane 3) was placed at 4.92 m away from
the mixing section, at the base of the riser. In addition to the physical experiments,
CFD simulations were carried out by Abdulkadir et al. in [3] using Star-CD and
Star-CCM+ with a VOF method. In this case the VOF method in Star-CCM+
seems to provide good results. However in OpenFOAM the hybrid model is again
used since a preliminary study with VOF failed in simulating the slug flow.
4.2.2 Case setup

The Solver settings and Simulation controls in OpenFOAM will not be reported
again, since they are identical to the previous case (4.1.2.5 and 4.1.2.6), except for
the maximum time step which is set to 5e−5.
4.2.2.1 Mesh
Fig. 4.10 shows the geometry for the computational flow domain. A mesh
sensitivity analysis has been carried out in [5] in order to identify the minimum
mesh density that is required to ensure that the solution is almost independent
of the mesh resolution. In this work four different sets of grid are developed for
a computational domain of 1 m length. This length is sufficient to carry out a
test on the performance of the mesh with quite reasonably computational effort [5].
A comprehensive mesh dependency study on the 6 m long vertical pipe, with
successively refined grids, would require huge computer times which is out of the
scope of this thesis. The meshes that were investigated are shown in Fig. 4.11 and
Table 4.2 lists the details of the grids employed.
The simulation results of the Taylor bubble by mean of the aforementioned four
different sets of grid are shown in Fig. 4.12. From Fig. 4.12, the Taylor bubble
shape can be clearly observed in all four meshes. However, the interface of the higher
Figure 4.10: 3D geometry of the computational flow domain showing the location of the
experimental measurement.
(a) 63360 cells (b) 106191 cells
(c) 184250 cells (d) 344400 cells
Figure 4.11: Cross-sectional view of different computational grid used for mesh inde-
pendent study.
(a) (b) (c) (d)
Figure 4.12: Simulation with OpenFOAM for the Taylor bubble with different grid
number.
1
344400 cells
184250 cells
0.9 106191 cells
63360 cells
0.8
0.7
0.6
Void Fraction
0.5
0.4
0.02
0.3
0.01
0.2
0
0.68 0.7 0.72
0.1
0
0.65 0.7 0.75 0.8 0.85
Time [s]
Figure 4.13: Void fraction (see Eq. (4.4)) at the measurement section (0.5 m).
Figure 4.14: Computational mesh used for simulations by Abdulkadir et al. in [5].
Figure 4.15: Cross-Section grid of the computational domain in OpenFOAM.
mesh resolution is sharper than in the other cases. In order to evaluate whether or
not the employed mesh resolution is adequate to obtain accurate solutions, the time
for the Taylor bubble to reach a section located at a distance of 0.5 m downstream
the inlet is compared in Fig. 4.13 for the four meshes [3]. From Fig. 4.12 and
Fig.4.13 it can be concluded that the mesh c is adequate. The change in the results
produced is very small when the number of cells is increased to the finer mesh d.
Fig. 4.14 shows the mesh used in [3]. The mesh used in this study is given
by Fig. 4.15 and the same considerations of 4.1.2.1 hold also for this case.
In [48], [75], [107] and [5] CFD similations with similar coarse meshes are described.
The number of hexahedral cells is 1105500, the maximum orthogonality is 32.7
and the maximum skewness is 0.52. The experimental data were obtained over an
interval of 60 s while the CFD covered 17 s both with OpenFOAM in the present
work and with Star-CCM+ in [5]. About 36 days of real time were needed to
simulate 17 seconds of physical time by performing a parallel computation on 12
processors with the upper limit on the time step set to 5e−5.
4.2.2.2 Turbulence
It is important to consider the turbulence in the numerical simulation as the
bubbles rising through the liquid create a developing film around themselves and a
wake behind them [98], [80]. The k-ε model with wall function is used as suggested
by the multiphase flow study of Ramos-Banderas et al. [79] and as was also used
in [4] and [30]. The flow inlet values for the turbulent kinetic energy k, and
dissipation rate ε, are estimated using the equations proposed by Launder and
Spalding [63].
4.2.2.3 Boundary and initial conditions

At the flow inlet, the mixture superficial velocity, Um given by (2.2) is specified.
The volume fraction of each phase is specified at the inlet as a homogeneous mixture:
Us,liquid Us,gas
αliquid = αgas = (4.6)
Um Um
The superficial velocites are:
Us,air = 0.344 m/s and Us,silicone oil = 0.05 m/s (4.7)
At the outlet at the top of the riser, a fixed value for the pressure is specified.
As in 4.1.2.3 the type fixedMean is used. At t = 0 the riser is completely filled
with liquid at zero velocity.
4.2.2.4 Fluid properties

The relevant fluid properties are:
ρair = 1.18 kg/m3 and µair = 0.000018 kg/ms (4.8a)

ρsilicone oil = 900 kg/m3 and µsilicone oil = 0.0053 kg/ms (4.8b)
The air-silicone oil surface tension is specified as 0.02 N/m. The dimensionless
numbers Eo, Nf , and M o are presented in Table 4.3. The regime of this case
corresponds to Eo > 100 and Nf > 300. According to Wallis et al. [44] when
Eo > 100 the surface tension plays little role in determining the slug ascent velocity.
In a later critical review of the literature Fabre et al [39] concluded that the viscosity
acts essentially to develop the liquid velocity profile far ahead of the bubbles, but
has no influence near the front where the inertia dominates. This condition is
satisfied provided Nf > 300 and the rise velocity of the bubble is determined solely
by liquid inertia [44].
Table 4.3: Dimensionless numbers.
Dimensionless numbers
Eotvos number Eo = 1981.67
Dimensionless inverse viscosity Nf = 9311.72
Morton’s number M o = 1.035 × 10−6
4.2.3 Results
In Fig. 4.16 the vector plot of the velocity field around the leading Taylor bubble
rising in the stagnant silicone oil is shown.
Figure 4.16: Simulation with OpenFOAM for the for the leading Taylor bubble rising
through a 0.067 m internal diameter vertical pipe: instantaneous velocity
field for the liquid phase.
As shown in Fig. 4.16 the bubble has a round nose and fills almost the cross
sectional area of the pipe. The liquid ahead of the bubble moves around the bubble
as a thin liquid film moving downwards in the annular space between the pipe wall
and the bubble surface. At the rear of that bubble, the liquid film plunges into the
liquid plug behind the bubble and produces a highly agitated mixing zone in the
bubble wake. This recirculation zone contains small bubbles shed from the bubble
tail due to the turbulent jet of the liquid film. A similar observation was reported
in [40].
Figure 4.17: Simulation with OpenFOAM for the fully developed slug flow: instanta-
neous void fraction.
Fig. 4.17 depicts the fully developed slug flow. Taylor bubbles can clearly be
viewed followed by liquid slug bodies with dispersed small bubbles. The oscillatory
motion of the liquid phase can be observed in Fig. 4.18 where the downward flow of
silicone oil around the air bubbles is shown. The flow in the liquid slug body shown
in Fig. 4.17 can be divided into two main parts: immediately below the rear of the
bubble, where there is the formation of a recirculation and mixing region, also called
wake region and the main body of the liquid slug body where the flow is gradually
recovering its original and undisturbed state. For a fully developed continuous slug
flow, the length of the liquid slug bodies between any pair of consecutive bubbles
remains constant and long enough, so that all the bubbles are not interacting with
each other, and are rising at the same translational velocity. The intermittence
evident in the developing slug flow pattern indicate the extremely complicated
nature of the hydrodynamics involved.
The void fraction is a fundamental quantity in the description and analysis of
Figure 4.18: Simulation with OpenFOAM for the fully developed slug flow: instanta-
neous velocity.
two phase flows; in the case of slug flows the void fraction at any point varies with
time. Therefore monitoring the time trace of cross-sectional average void fraction
is desirable.
A fully developed flow is defined as one where the flow pattern does not change
with the downstream distance, and it also does not change over time. Time series of
void fraction and probability density function (PDF) of void fraction obtained from
the CFD simulation are used to assess the change in the flow characteristics [72], [33].
The flow development over time is shown in Table 4.4 for different test section.
For large numbers of discrete samples, the PDF can be estimated by a normal-
ized histogram calculation. Here the number of occurrences in the histogram is
normalized by the total number of samples, thus representing a probability. For
the histogram calculation, 200 bins in the phase fraction interval 0 to 1 were used.
Costigan and Whalley [33] proposed that twin peaked probability density function of
recorded void fractions represent slug flow. The low void fraction peak corresponds
to the liquid slug body while the high void fraction peak corresponds to the Taylor
bubble.
The PDF of the time series of void fraction in Table 4.4 shows that the results
obtained from 1.0 m are initially affected by entrance effects. At a distance of 4.0
m from the inlet the PDF of void fraction has taken the shape of slug flow showing
a double peak.
It can be concluded that at 4.0 m the flow is fully developed. The experimental
measuring instruments used by Abdulkadir et al [5] were located at 4.4 m (ECT-
Plane 1), 4.489 m (ECT-Plane 2) and 4.92 m (WMS-Plane 3).
Fig. 4.19 shows a typical plot of a large trailing Taylor bubble (start-up) and
leading train of Taylor bubbles (stationary) from the simulation with OpenFOAM.
A qualitative comparison between CFD simulations with Star-CCM+ and
experiment is given in Fig. 4.20 as reported by Abdulkadir et al. in [5].
The comparison between CFD simulations with OpenFOAM and experiments
is shown in Fig. 4.21. The experimental data were kindly provided by Professor
0.8
Void Fraction
0.6
0.4
0.2
0
1.2 2.2 3.2 4.2
Time [s]
Figure 4.19: Simulation with OpenFOAM for a large trailing Taylor bubble and leading
train of Taylor bubbles.
1 1
Experiment Star-CCM+
0.8 0.8
Void Fraction
Void Fraction
0.6 0.6
0.4 0.4
0.2 0.2
0 0
1 2 3 4 5 7.2 8.2 9.2 10.2 11.2 12.2
Time [s] Time [s]
(a) WMS-Plane 3 (b) WMS-Plane 3

1 1
0.8 0.8
Void Fraction
Void Fraction
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 1 2 3 4 5 6.2 7.2 8.2 9.2 10.2 11.2
Time [s] Time [s]
(c) ECT-Plane 2 (d) ECT-Plane 2

1 1
0.8 0.8
Void Fraction
Void Fraction
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 1 2 3 4 5 6.1 7.1 8.1 9.1 10.1 11.1
Time [s] Time [s]
(e) ECT-Plane 1 (f ) ECT-Plane 1
Figure 4.20: Comparison between experimental data (red) and CFD simulation with
Star-CCM+ (black) from Abdulkadir et al. [5].
Barry Azzopardi and Professor Mukhtar Abdulkadir.

From the time traces in Fig 4.21 we can clearly distinguish the slug flow regime
with a reasonably good agreement between the CFD and experimental results. The
slug flow data have alternating periods of high and low void fraction. High void
fraction marks the gas bubble passage, and low void fraction marks the passage of
the liquid slug body with same entrained dispersed gas bubbles. However, the void
fraction in the liquid slug body is seen to have lower values in the CFD simulations
than in the experiments.
The PDFs corresponding to the void fraction time series in Fig. 4.21 and in
Fig. 4.20 (for the CFD simulation with Star-CCM+) are shown in Fig. 4.22.
1
Experiment
OpenFOAM
0.8
Void Fraction
0.6
0.4
0.2
0
11 12 13 14 15 16 17
Time [s]
(a) WMS-Plane 3
1
Experiment
OpenFOAM
0.8
Void Fraction
0.6
0.4
0.2
0
11 12 13 14 15 16 17
Time [s]
(b) ECT-Plane 2
1
Experiment
OpenFOAM
0.8
Void Fraction
0.6
0.4
0.2
0
11 12 13 14 15 16 17
Time [s]
(c) ECT-Plane 1
Figure 4.21: Time series for the void fraction in different planes; comparison between
experiments and CFD simulations with OpenFOAM.
Again, we can notice the higher values of the void fraction in the liquid slug
0.08
Experiment
OpenFOAM
Star-CCM+
0.06
Probability density function

0.04
0.02
0
0 0.2 0.4 0.6 0.8 1
Void Fraction
(a) WMS-Plane 3
0.08
Experiment
OpenFOAM
Star-CCM+
0.06
0.04
0.02
0
0 0.2 0.4 0.6 0.8 1
Void Fraction
(b) ECT-Plane 2
0.08
Experiment
OpenFOAM
Star-CCM+
0.06
0.04
0.02
0
0 0.2 0.4 0.6 0.8 1
Void Fraction
(c) ECT-Plane 1
Figure 4.22: PDF of cross-sectional average void fraction for the case of slug flow
obtained from the experiments, from the CFD simulations with Star-CCM+
(in Abdulkadir et al. [5]) and with OpenFOAM.
body in the CFD results with both Star-CCM+ and OpenFOAM, as compared to
the experiments. There is a good agreement between the CFD simulations and
the experiments in predicting the same flow pattern, being slug flow: there is a
twin-peaked PDF, with one peak being characteristic for the liquid slug body and
the other peak being characteristic for the Taylor bubble.
By cross-correlating the void fraction signals from ECT-Plane 1 and ECT-Plane
2 (in Fig. 4.23), the transit time between the two section can be measured; this
together with the distance between the measurement sections enable us to calculate
the velocity of the slug unit for the fully developed flow. Fig. 4.24 shows that the
time delay between the planes in the CFD predictions with OpenFOAM and the
experiments is 0.08 s, while in the CFD simulations with Star-CCM+ is equal to
0.075 s. Dividing the distance of 0.089 m between the Plane 1 and the Plane 2 by
the delay time of 0.075 s gives a slug velocity of 1.19 m/s. The slug velocity given
by the simulations with Star-CCM+ is slightly lower and equal to 1.11 m/s.
Fig. 4.25 represent a comparison between CFD simulations and experiments as
the large leading Taylor bubble reaches the ECT-Plane 1 and the ECT-Plane 2.
From Fig. 4.25 the void fraction in the leading Taylor bubble can be obtained.
The liquid film thickness could not be measured directly using the ECT planes
in [5] and the expression proposed by Fernandes et al. [40] is used. This is given by:
D √
δ= (1 − εT B ) (4.9)
2
In Eq. (4.9) δ represents the liquid film thickness, D represents the internal pipe
diameter and εTB is the void fraction in the Taylor bubble. The void fraction and
the liquid film thickness obtained are summarised in Table 4.5.
The comparison between CFD simulations and experiments is again reasonably
good.
Experiment
1
Plane 1
Plane 2
0.8
Void Fraction
0.6
0.4
0.2
0
11 12 13 14 15 16
Time [s]
(a) Time series for the void fraction in the experiments

OpenFOAM
1
Plane 1
Plane 2
0.8
Void Fraction
0.6
0.4
0.2
0
11 12 13 14 15 16
Time [s]
(b) Time series for the void fraction in the CFD simulations
with OpenFOAM
Star-CCM+
1
Plane 1
Plane 2
0.8
Void Fraction
0.6
0.4
0.2
0
7.8 8.8 9.8 10.8 11.8 12.8
Time [s]
(c) Time series for the void fraction in the CFD simulations
with Star-CCM+ (in [5])
Figure 4.23: Comparison between experiments and CFD simulations with Star-CCM+
(from [5]) and with OpenFOAM.
1
Experiment: t = 0.075 s
OpenFOAM: t = 0.075 s
Star-CCM+: t = 0.08 s
Correlation Coefficient
0.5
-0.5
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Time delay [s]
Figure 4.24: Time delay of a Taylor bubble passing along two different measuring
locations in the pipe.
Table 4.4: Flow development along the vertical riser in the simulation with OpenFOAM.
Distance from the inlet [m] Void fraction traces PDF of void fraction
0.08
1
0.8 0.06
Void Fraction
0.6
PDF
0.04
0.4
0.2 0.02
0
11 12 13 14 15 16 17
1.0 (15 pipe diameters)
0
0 0.2 0.4 0.6 0.8 1
Time [s] Void Fraction
0.08
1
0.8 0.06
Void Fraction
0.6
PDF
0.04
0.4
0.2 0.02
0
11 12 13 14 15 16 17
0
0 0.2 0.4 0.6 0.8 1
0.08
1
0.8 0.06
Void Fraction
0.6
PDF
0.04
0.4
0.2 0.02
0
11 12 13 14 15 16 17
0
0 0.2 0.4 0.6 0.8 1
0.08
1
0.8 0.06
Void Fraction
0.6
PDF
0.04
0.4
0.2 0.02
0
11 12 13 14 15 16 17
0
0 0.2 0.4 0.6 0.8 1
0.08
1
0.8 0.06
Void Fraction
0.6
PDF
0.04
0.4
0.2 0.02
0
11 12 13 14 15 16 17
0
0 0.2 0.4 0.6 0.8 1
0.08
1
0.8 0.06
Void Fraction
0.6
PDF
0.04
0.4
0.2 0.02
0
11 12 13 14 15 16 17
0
0 0.2 0.4 0.6 0.8 1
0.08
1
0.8 0.06
Void Fraction
0.6
PDF
0.04
0.4
0.2 0.02
0
11 12 13 14 15 16 17
0
0 0.2 0.4 0.6 0.8 1
1
Experiment
OpenFOAM
Star-CCM+
0.8
Void Fraction
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8
Time [s]
(a) ECT-Plane 1
1
Experiment
OpenFOAM
Star-CCM+
0.8
Void Fraction
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8
Time [s]
(b) ECT-Plane 2
Figure 4.25: Comparison between experiments and CFD simulations with Star-CCM+
(from [5]) and with OpenFOAM for the leading Taylor bubble (Start-up).
Table 4.5: Comparison between experiments and CFD simulations with Star-CCM+
(from [5]) and with OpenFOAM for the leading Taylor bubble: void fraction
and liquid film thickness.
Experiments OpenFOAM Star-CCM+

ECT-Plane 1 0.77 0.79 0.81
Void fraction in the Taylor bubble
ECT-Plane 2 0.76 0.80 0.82
ECT-Plane 1 4.10 3.72 3.35
Liquid film thickness [mm]
ECT-Plane 2 4.30 3.54 3.16
Conclusions
Two-phase gas-liquid upward flow in vertical pipes or risers offers a challenge

to CFD simulation methods. This is particularly because different flow regimes
may occur, such as bubbly flow, slug flow, churn flow, or annular flow. VOF is
suggested to be the only available multiphase flow model applicable to segregated
flow regimes. However, further investigations have revealed that VOF method
in its original form in OpenFOAM and Fluent is not suitable for the flow condi-
tions leading to slug or churn flow i.e. where a large slippage between the two
phases exists. This shortcoming in the VOF method in OpenFOAM and Fluent
is explained by the overpredicted dispersion between the phases. This gives a too
low liquid holdup fraction, which in turn also gives a too low hydrostatic head
and thus a too low pressure drop. The focus then turns to find the improvement
possibilities in the VOF method. The approach of coupling VOF with a two-fluid
model is adopted and further developed in this thesis with the hybrid multiphase
solver multiphaseEulerFoam which is available in the OpenFOAM CFD package.
VOF keeps a sharp interface between segragated flow structures (such as a liquid
film in annular flow that is separated from the gas core) and the Euler-Euler (or
the two fluid model) is able to represent dispersed regions, such as in bubbly flow
or in the liquid slug body that has entrained gas bubbles. The results highlight
the application of the presented CFD solver to predict flow regimes for gas-liquid
vertical flows, without a priori knowledge on the flow pattern.
The hybrid model is applied in two examples for which experimental data exist.
Case study 1
The main goal to simulate vertical upward air-water churn flow through a pipe with
50.8 mm internal diameter is achieved. From a comparison of the results obtained
from the CFD simulations in OpenFOAM and the experiments, the following
conclusions can be drawn:
• The prediction by OpenFOAM using multiphaseEulerFoam captured the

overall physics of the flow quite successfully. Time series of cross-sectional
averaged void fraction and contour plots of liquid holdup and velocity showed
that the hybrid model was able to distinguish the flow pattern in the pipe,
being churn flow.
• Very good agreement was detected for the pressure drop and liquid holdup
fraction. The pressure drop was 2400 Pa/m in the the experiments and 2370
Pa/m in the CFD simulations. The liquid holdup fraction in the experiments
was estimated to be 0.24 based on the measured pressure drop whereas for
61
62 Conclusions
the simulations the time averaged liquid holdup fraction was 0.21. The
experimentally determined holdup was higher than the CFD prediction and
this can be partly due to the assumption of zero frictional pressure drop.
Case study 2
A detailed simulation of the slug flow in a 6 m vertical pipe with a 0.067 m internal
diameter with air and silicone oil has been successfully carried out. The results
were compared with the experimental measurements and the CFD simulations with
Star-CCM+ reported in literature. The main findings are:
• The CFD simulations with OpenFOAM have shown to be able to qualitatively

capture the distinctive characteristics of the slug flow. Three regions were
observed: the Taylor bubble, the falling film and the wake region. The Taylor
bubble can be seen moving vertically upwards whereas the liquid film is
moving downwards. In the wake region there were some entrained bubbles
that were carried upwards.
• The slug flow pattern can be considered fully developed at 4 m, 60 pipe

diameters.
• A reasonably good agreement between the CFD simulations and the experi-
ments was obtained for the time series of cross-sectional averaged void fraction
over three monitoring planes. The slug flow data had alternating periods of
high and low void fraction. High void fraction marked the gas bubble passage,
and low void fraction marked the passage of the liquid slug body with same
entrained dispersed gas bubbles. However, the void fraction in the liquid
slug body was seen to have lower values in the CFD simulations than in the
experiments.
• The agreement of the PDFs of cross-sectional average void fraction between

the CFD simulations and the experiments was quite reasonable. There was
a twin-peaked PDF, with one peak being characteristic for the liquid slug
body and the other peak being characteristic for the Taylor bubble. The void
fraction in the liquid slug body showed lower values in the CFD simulations
than in the experiments.
• A perfect agreement was found for the velocity of the slug unit for the
fully developed flow between the CFD simulations with OpenFOAM and the
experiments. The CFD simulations with Star-CCM+ predicted a slightly
lower value.
• A satisfactory agreement with the experimental data was observed regarding

the formation and the shape of the leading Taylor bubble and for the liquid
film thickness. The peak of the void fraction in the leading Taylor bubble was
larger in the Star-CCM+ and OpenFOAM results than with the experiments.
The length of the leading bubble was correctly predicted with OpenFOAM
whereas it was lower in the Star-CCM+ results.
Appendix A
blockMesh with m4
m4 is a macro processor that can be used for the parameterization of the blockMesh
dictionary. The scripts that generate the meshes for the two cases are listed below.
A.1 Case study 1
Table A.1: m4 Script to generate the blockMesh dictionary for the Case study 1.
1 //Run u s i n g :
2 //m4 −P bloc kM esh Dic t . m4 > b loc kMe sh Dic t
3
4 //m4 d e f i n i t i o n s :
5 m4_changecom ( // ) m4_changequote ( [ , ] )
6 m4_define ( c a l c , [ m4_esyscmd ( p e r l −e ’ u s e Math : : T r i g ; p r i n t f ( $1 ) ’ ) ] )
7 m4_define (VCOUNT, 0 )
8 m4_define ( v l a b e l , [ [ // ] Ve rt ex $1 = VCOUNT m4_define ( $1 , VCOUNT) m4_define ( [VCOUNT
] , m4_incr (VCOUNT) ) ] )
9
10 // Mathematical c o n s t a n t s :
11 m4_define ( pi , 3 . 1 4 1 5 9 2 6 5 3 6 )
12
13 // Geometry
14 m4_define ( rOut , 0 . 0 2 5 4 )
15 m4_define ( r I n t , c a l c ( 0 . 9 0 5 5 1 1 8 1 1 0 2 3 6 2 2 ∗ rOut ) )
16
17 // Grid p o i n t s :
18 m4_define ( r N u m b e r O f C e l l s 1 s t , 1 0 ) // i n t h e f i r s t O−g r i d
19 m4_define ( rNumberOfCells2nd , 7 ) // i n t h e s e c o n d O−g r i d
20 m4_define ( tNumberOfCells , 2 0 )
21 m4_define ( zABnumberOfCells , 3 2 0 )
22
23 m4_define ( rGrading1 , 1 . 8 )
24 m4_define ( rGrading2 , 2 . 7 )
25 m4_define ( zGrading , 1 )
26
27 // Plane A:
28 m4_define ( zA , 0 )
29 m4_define ( rA , r I n t )
30 m4_define ( rOutA , rOut )
31 m4_define ( rRelA , 0 . 6 9 0 5 3 7 0 8 4 3 9 8 9 7 7 )
32 m4_define ( rRelAc , 0 . 8 5 )
33
34 // Plane B :
35 m4_define ( zB , 2 . 5 4 )
36 m4_define ( rB , r I n t )
37 m4_define ( rOutB , rOut )
38 m4_define ( rRelB , 0 . 6 9 0 5 3 7 0 8 4 3 9 8 9 7 7 )
39 m4_define ( rRelBc , 0 . 8 5 )
40
63
64 Appendix A. blockMesh with m4
41 /∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗\
42 | ========= | |
43 | \\ / F ield | OpenFOAM: The Open S o u r c e CFD Toolbox |
44 | \\ / O peration | Version : 2.2.2 |
45 | \\ / A nd | Web : h t t p : / /www. openfoam . o r g |
46 | \\/ M anipulation | |
47 \∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗/
48
49 FoamFile
50 {
51 version 2.0;
52 format ascii ;
53
54 root "" ;
55 case "" ;
56 instance "" ;
57 local "" ;
58
59 class dictionary ;
60 object bl oc kMe sh Dic t ;
61 }
62
63 // ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ //
64
65 convertToMeters 1 ;
66
67 vertices
68 (
69
70 // Plane A:
71 ( c a l c ( rRelA ∗rA∗ c o s ( p i / 4 ) ) −c a l c ( rRelA ∗rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A0)
72 ( c a l c ( rRelA ∗rA∗ c o s ( p i / 4 ) ) c a l c ( rRelA ∗rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A1)
73 ( c a l c (−rRelA ∗rA∗ c o s ( p i / 4 ) ) c a l c ( rRelA ∗rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A2)
74 ( c a l c (−rRelA ∗rA∗ c o s ( p i / 4 ) ) −c a l c ( rRelA ∗rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A3)
75 ( c a l c ( rA∗ c o s ( p i / 4 ) ) −c a l c ( rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A4)
76 ( c a l c ( rA∗ c o s ( p i / 4 ) ) c a l c ( rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A5)
77 ( c a l c (−rA∗ c o s ( p i / 4 ) ) c a l c ( rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A6)
78 ( c a l c (−rA∗ c o s ( p i / 4 ) ) −c a l c ( rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A7)
79 ( c a l c ( rOutA∗ c o s ( p i / 4 ) ) −c a l c ( rOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A8)
80 ( c a l c ( rOutA∗ c o s ( p i / 4 ) ) c a l c ( rOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A9)
81 ( c a l c (−rOutA∗ c o s ( p i / 4 ) ) c a l c ( rOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l ( A10 )
82 ( c a l c (−rOutA∗ c o s ( p i / 4 ) ) −c a l c ( rOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l ( A11 )
83
84
85 // Plane B :
86 ( c a l c ( rRelB ∗rB∗ c o s ( p i / 4 ) ) −c a l c ( rRelB ∗rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B0 )
87 ( c a l c ( rRelB ∗rB∗ c o s ( p i / 4 ) ) c a l c ( rRelB ∗rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B1 )
88 ( c a l c (−rRelB ∗rB∗ c o s ( p i / 4 ) ) c a l c ( rRelB ∗rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B2 )
89 ( c a l c (−rRelB ∗rB∗ c o s ( p i / 4 ) ) −c a l c ( rRelB ∗rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B3 )
90 ( c a l c ( rB∗ c o s ( p i / 4 ) ) −c a l c ( rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B4 )
91 ( c a l c ( rB∗ c o s ( p i / 4 ) ) c a l c ( rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B5 )
92 ( c a l c (−rB∗ c o s ( p i / 4 ) ) c a l c ( rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B6 )
93 ( c a l c (−rB∗ c o s ( p i / 4 ) ) −c a l c ( rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B7 )
94 ( c a l c ( rOutB∗ c o s ( p i / 4 ) ) −c a l c ( rOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B8 )
95 ( c a l c ( rOutB∗ c o s ( p i / 4 ) ) c a l c ( rOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B9 )
96 ( c a l c (−rOutB∗ c o s ( p i / 4 ) ) c a l c ( rOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B10 )
97 ( c a l c (−rOutB∗ c o s ( p i / 4 ) ) −c a l c ( rOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B11 )
98
99 );
100
101 // D e f i n i n g b l o c k s :
102 blocks
103 (
104 // B l o c k s between p l a n e E and p l a n e F :
105 // b l o c k 0 − p o s i t i v e x O−g r i d b l o c k 1 s t b e l t
106 hex (A5 A1 A0 A4 B5 B1 B0 B4 ) AB
107 ( r N u m b e r O f C e l l s 1 s t tNumberOfCells zABnumberOfCells )
108 s i m p l e G r a d i n g ( r G r a d i n g 1 1 zGrading )
109 // b l o c k 1 − p o s i t i v e y O−g r i d b l o c k 1 s t b e l t
110 hex (A6 A2 A1 A5 B6 B2 B1 B5 ) AB
A.1. Case study 1 65
113 // b l o c k 2 − n e g a t i v e x O−g r i d b l o c k 1 s t b e l t
114 hex (A7 A3 A2 A6 B7 B3 B2 B6 ) AB
117 // b l o c k 3 − n e g a t i v e y O−g r i d b l o c k 1 s t b e l t
118 hex (A4 A0 A3 A7 B4 B0 B3 B7 ) AB
121 // b l o c k 4 − c e n t r a l O−g r i d b l o c k
122 hex (A0 A1 A2 A3 B0 B1 B2 B3 ) AB
123 ( tNumberOfCells tNumberOfCells zABnumberOfCells )
124 s i m p l e G r a d i n g ( 1 1 zGrading )
125 // b l o c k 5 − p o s i t i v e x O−g r i d b l o c k 2nd b e l t
126 hex (A9 A5 A4 A8 B9 B5 B4 B8 ) AB
127 ( rNumberOfCells2nd tNumberOfCells zABnumberOfCells )
129 // b l o c k 6 − p o s i t i v e y O−g r i d b l o c k 2nd b e l t
130 hex ( A10 A6 A5 A9 B10 B6 B5 B9 ) AB
133 // b l o c k 7 − n e g a t i v e x O−g r i d b l o c k 2nd b e l t
134 hex ( A11 A7 A6 A10 B11 B7 B6 B10 ) AB
137 // b l o c k 8 − n e g a t i v e y O−g r i d b l o c k 2nd b e l t
138 hex (A8 A4 A7 A11 B8 B4 B7 B11 ) AB
141
142
143 );
144
145 edges
146 (
147 // Plane A:
148 a r c A0 A1 ( c a l c ( rRelAc ∗ rRelA ∗rA ) 0 zA )
149 a r c A1 A2 ( 0 c a l c ( rRelAc ∗ rRelA ∗rA ) zA )
150 a r c A2 A3 (− c a l c ( rRelAc ∗ rRelA ∗rA ) 0 zA )
151 a r c A3 A0 ( 0 −c a l c ( rRelAc ∗ rRelA ∗rA ) zA )
152 a r c A4 A5 ( rA 0 zA )
153 a r c A5 A6 ( 0 rA zA )
154 a r c A6 A7 (−rA 0 zA )
155 a r c A7 A4 ( 0 −rA zA )
156 a r c A8 A9 ( rOutA 0 zA )
157 a r c A9 A10 ( 0 rOutA zA )
158 a r c A10 A11 (−rOutA 0 zA )
159 a r c A11 A8 ( 0 −rOutA zA )
160
161
162 // Plane B:
163 a r c B0 B1 ( c a l c ( rRelBc ∗ rRelB ∗rB ) 0 zB )
164 a r c B1 B2 ( 0 c a l c ( rRelBc ∗ rRelB ∗rB ) zB )
165 a r c B2 B3 (− c a l c ( rRelBc ∗ rRelB ∗rB ) 0 zB )
166 a r c B3 B0 ( 0 −c a l c ( rRelBc ∗ rRelB ∗rB ) zB )
167 a r c B4 B5 ( rB 0 zB )
168 a r c B5 B6 ( 0 rB zB )
169 a r c B6 B7 (−rB 0 zB )
170 a r c B7 B4 ( 0 −rB zB )
171 a r c B8 B9 ( rOutB 0 zB )
172 a r c B9 B10 ( 0 rOutB zB )
173 a r c B10 B11 (−rOutB 0 zB )
174 a r c B11 B8 ( 0 −rOutB zB )
175
176 );
177
178 // D e f i n i n g p a t c h e s :
179 patches
180 (
181 patch i n l e t
182 (
183 (A1 A5 A4 A0)

184 (A2 A6 A5 A1)
185 (A3 A7 A6 A2)
186 (A0 A4 A7 A3)
187 (A3 A2 A1 A0)
188
189 (A5 A9 A8 A4)
190 (A6 A10 A9 A5)
191 (A7 A11 A10 A6)
192 (A4 A8 A11 A7)
193
194 )
195
196 patch o u t l e t
197 (
198 ( B0 B4 B5 B1 )
199 ( B1 B5 B6 B2 )
200 ( B2 B6 B7 B3 )
201 ( B3 B7 B4 B0 )
202 ( B0 B1 B2 B3 )
203
204 ( B4 B8 B9 B5 )
205 ( B5 B9 B10 B6 )
206 ( B6 B10 B11 B7 )
207 ( B7 B11 B8 B4 )
208
209 )
210
211 wall wallPipe
212 (
213 (A8 A9 B9 B8 )
214 (A9 A10 B10 B9 )
215 ( A10 A11 B11 B10 )
216 ( A11 A8 B8 B11 )
217
218 )
219 );
220
221 mergePatchPairs
222 (
223 );
224
225 // ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ //
A.2 Case study 2
Table A.2: m4 Script to generate the blockMesh dictionary for the Case study 2.
1 //Run u s i n g :
2 //m4 −P b lo c k M e sh D i c t 3 . m4 > b lo ckM es hDi ct
3
4 //m4 d e f i n i t i o n s :
5 m4_changecom ( // ) m4_changequote ( [ , ] )
6 m4_define ( c a l c , [ m4_esyscmd ( p e r l −e ’ u s e Math : : T r i g ; p r i n t f ( $1 ) ’ ) ] )
7 m4_define (VCOUNT, 0 )
8 m4_define ( v l a b e l , [ [ // ] Ve rt ex $1 = VCOUNT m4_define ( $1 , VCOUNT) m4_define ( [VCOUNT
] , m4_incr (VCOUNT) ) ] )
9
10 // Mathematical c o n s t a n t s :
11 m4_define ( pi , 3 . 1 4 1 5 9 2 6 5 3 6 )
12
13 // Geometry with c e n t r a l s q u a r e , 1 s t c y l i n d e r , 2nd c y l i n d e r , 3 rd c y l i n d e r
14
15 m4_define ( rOutOut , 0 . 0 3 3 5 ) // r a d i u s p i p e
16 m4_define ( rOut , c a l c ( 0 . 7 9 7 0 1 4 9 2 5 ∗ rOutOut ) ) // r a d i u s 2nd c y l i n d e r
17 m4_define ( r I n t , c a l c ( 0 . 7 5 5 2 2 3 8 8 ∗ rOutOut ) ) // r a d i u s 1 s t c y l i n d e r
18
19 // Grid p o i n t s :
20 m4_define ( r N u m b e r O f C e l l s 1 s t , 5 ) // f i r s t O−g r i d
21 m4_define ( rNumberOfCells2nd , 1 ) // s e c o n d O−g r i d
22 m4_define ( rNumberOfCells3rd , 8 ) // t h i r d O−g r i d
23
24 m4_define ( tNumberOfCells , 1 1 ) // c e n t r a l s q u a r e
25 m4_define ( zABnumberOfCells , 1 5 0 0 ) // a x i s d i r e c t i o n
26
27 m4_define ( rGrading1 , 1.5)
28 m4_define ( rGrading2 , 1)
29 m4_define ( rGrading3 , 2.5)
30 m4_define ( zGrading , 1)
31
32 // Plane I n l e t − A:
33 m4_define ( zA , 0 )
34 m4_define ( rA , r I n t )
35 m4_define ( rOutA , rOut )
36 m4_define ( rOutOutA , rOutOut )
37 m4_define ( rRelA , 0.720762613)
38 m4_define ( rRelAc , 0 . 8 5 )
39
40 // Plane O u t l e t − B :
41 m4_define ( zB , 6 )
42 m4_define ( rB , r I n t )
43 m4_define ( rOutB , rOut )
44 m4_define ( rOutOutB , rOutOut )
45 m4_define ( rRelB , 0.720762613)
46 m4_define ( rRelBc , 0 . 8 5 )
47
48 /∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗\
49 | ========= | |
52 | \\ / A nd | Web : h t t p : / /www. openfoam . o r g |
54 \∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗/
55
56 FoamFile
57 {
58 version 2.0;
59 format ascii ;
60
61 root "" ;
62 case "" ;
63 instance "" ;
64 local "" ;
65
67 object bl oc kMe sh Dic t ;
68 }
69
70 // ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ //
71
72 convertToMeters 1 ;
73
74 vertices
75 (
76
77 // Plane A:
78 ( c a l c ( rRelA ∗rA∗ c o s ( p i / 4 ) ) −c a l c ( rRelA ∗rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A0)
79 ( c a l c ( rRelA ∗rA∗ c o s ( p i / 4 ) ) c a l c ( rRelA ∗rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A1)
80 ( c a l c (−rRelA ∗rA∗ c o s ( p i / 4 ) ) c a l c ( rRelA ∗rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A2)
81 ( c a l c (−rRelA ∗rA∗ c o s ( p i / 4 ) ) −c a l c ( rRelA ∗rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A3)
82 ( c a l c ( rA∗ c o s ( p i / 4 ) ) −c a l c ( rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A4)
83 ( c a l c ( rA∗ c o s ( p i / 4 ) ) c a l c ( rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A5)
84 ( c a l c (−rA∗ c o s ( p i / 4 ) ) c a l c ( rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A6)
85 ( c a l c (−rA∗ c o s ( p i / 4 ) ) −c a l c ( rA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A7)
86 ( c a l c ( rOutA∗ c o s ( p i / 4 ) ) −c a l c ( rOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A8)
87 ( c a l c ( rOutA∗ c o s ( p i / 4 ) ) c a l c ( rOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l (A9)
88 ( c a l c (−rOutA∗ c o s ( p i / 4 ) ) c a l c ( rOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l ( A10 )
89 ( c a l c (−rOutA∗ c o s ( p i / 4 ) ) −c a l c ( rOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l ( A11 )
90 ( c a l c ( rOutOutA∗ c o s ( p i / 4 ) ) −c a l c ( rOutOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l ( A12 )

91 ( c a l c ( rOutOutA∗ c o s ( p i / 4 ) ) c a l c ( rOutOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l ( A13 )
92 ( c a l c (−rOutOutA∗ c o s ( p i / 4 ) ) c a l c ( rOutOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l ( A14 )
93 ( c a l c (−rOutOutA∗ c o s ( p i / 4 ) ) −c a l c ( rOutOutA∗ s i n ( p i / 4 ) ) zA ) v l a b e l ( A15 )
94
95
96 // Plane B :
97 ( c a l c ( rRelB ∗rB∗ c o s ( p i / 4 ) ) −c a l c ( rRelB ∗rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B0 )
98 ( c a l c ( rRelB ∗rB∗ c o s ( p i / 4 ) ) c a l c ( rRelB ∗rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B1 )
99 ( c a l c (−rRelB ∗rB∗ c o s ( p i / 4 ) ) c a l c ( rRelB ∗rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B2 )
100 ( c a l c (−rRelB ∗rB∗ c o s ( p i / 4 ) ) −c a l c ( rRelB ∗rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B3 )
101 ( c a l c ( rB∗ c o s ( p i / 4 ) ) −c a l c ( rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B4 )
102 ( c a l c ( rB∗ c o s ( p i / 4 ) ) c a l c ( rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B5 )
103 ( c a l c (−rB∗ c o s ( p i / 4 ) ) c a l c ( rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B6 )
104 ( c a l c (−rB∗ c o s ( p i / 4 ) ) −c a l c ( rB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B7 )
105 ( c a l c ( rOutB∗ c o s ( p i / 4 ) ) −c a l c ( rOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B8 )
106 ( c a l c ( rOutB∗ c o s ( p i / 4 ) ) c a l c ( rOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B9 )
107 ( c a l c (−rOutB∗ c o s ( p i / 4 ) ) c a l c ( rOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B10 )
108 ( c a l c (−rOutB∗ c o s ( p i / 4 ) ) −c a l c ( rOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B11 )
109 ( c a l c ( rOutOutB∗ c o s ( p i / 4 ) ) −c a l c ( rOutOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B12 )
110 ( c a l c ( rOutOutB∗ c o s ( p i / 4 ) ) c a l c ( rOutOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B13 )
111 ( c a l c (−rOutOutB∗ c o s ( p i / 4 ) ) c a l c ( rOutOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B14 )
112 ( c a l c (−rOutOutB∗ c o s ( p i / 4 ) ) −c a l c ( rOutOutB∗ s i n ( p i / 4 ) ) zB ) v l a b e l ( B15 )
113
114 );
115
116 // D e f i n i n g b l o c k s :
117 blocks
118 (
119 // B l o c k s between p l a n e A and p l a n e B :
120 // b l o c k 0 − p o s i t i v e x O−g r i d b l o c k 1 s t b e l t
121 hex (A5 A1 A0 A4 B5 B1 B0 B4 ) AB
124 // b l o c k 1 − p o s i t i v e y O−g r i d b l o c k 1 s t b e l t
125 hex (A6 A2 A1 A5 B6 B2 B1 B5 ) AB
128 // b l o c k 2 − n e g a t i v e x O−g r i d b l o c k 1 s t b e l t
129 hex (A7 A3 A2 A6 B7 B3 B2 B6 ) AB
132 // b l o c k 3 − n e g a t i v e y O−g r i d b l o c k 1 s t b e l t
133 hex (A4 A0 A3 A7 B4 B0 B3 B7 ) AB
136 // b l o c k 4 − c e n t r a l O−g r i d b l o c k
137 hex (A0 A1 A2 A3 B0 B1 B2 B3 ) AB
138 ( tNumberOfCells tNumberOfCells zABnumberOfCells )
139 s i m p l e G r a d i n g ( 1 1 zGrading )
140 // b l o c k 5 − p o s i t i v e x O−g r i d b l o c k 2nd b e l t
141 hex (A9 A5 A4 A8 B9 B5 B4 B8 ) AB
144 // b l o c k 6 − p o s i t i v e y O−g r i d b l o c k 2nd b e l t
145 hex ( A10 A6 A5 A9 B10 B6 B5 B9 ) AB
148 // b l o c k 7 − n e g a t i v e x O−g r i d b l o c k 2nd b e l t
149 hex ( A11 A7 A6 A10 B11 B7 B6 B10 ) AB
152 // b l o c k 8 − n e g a t i v e y O−g r i d b l o c k 2nd b e l t
153 hex (A8 A4 A7 A11 B8 B4 B7 B11 ) AB
156 // b l o c k 9 − p o s i t i v e x O−g r i d b l o c k 3 rd b e l t
157 hex ( A13 A9 A8 A12 B13 B9 B8 B12 ) AB
158 ( rNumberOfCells3rd tNumberOfCells zABnumberOfCells )
160 // b l o c k 1 0 − p o s i t i v e y O−g r i d b l o c k 3 rd b e l t
161 hex ( A14 A10 A9 A13 B14 B10 B9 B13 ) AB

164 // b l o c k 1 1 − n e g a t i v e x O−g r i d b l o c k 3 rd b e l t
165 hex ( A15 A11 A10 A14 B15 B11 B10 B14 ) AB
168 // b l o c k 1 2 − n e g a t i v e y O−g r i d b l o c k 3 rd b e l t
169 hex ( A12 A8 A11 A15 B12 B8 B11 B15 ) AB
172
173
174 );
175
176 edges
177 (
178 // Plane A:
179 a r c A0 A1 ( c a l c ( rRelAc ∗ rRelA ∗rA ) 0 zA )
180 a r c A1 A2 ( 0 c a l c ( rRelAc ∗ rRelA ∗rA ) zA )
181 a r c A2 A3 (− c a l c ( rRelAc ∗ rRelA ∗rA ) 0 zA )
182 a r c A3 A0 ( 0 −c a l c ( rRelAc ∗ rRelA ∗rA ) zA )
183 a r c A4 A5 ( rA 0 zA )
184 a r c A5 A6 ( 0 rA zA )
185 a r c A6 A7 (−rA 0 zA )
186 a r c A7 A4 ( 0 −rA zA )
187 a r c A8 A9 ( rOutA 0 zA )
188 a r c A9 A10 ( 0 rOutA zA )
189 a r c A10 A11 (−rOutA 0 zA )
190 a r c A11 A8 ( 0 −rOutA zA )
191 a r c A12 A13 ( rOutOutA 0 zA )
192 a r c A13 A14 ( 0 rOutOutA zA )
193 a r c A14 A15 (−rOutOutA 0 zA )
194 a r c A15 A12 ( 0 −rOutOutA zA )
195
196
197 // Plane B:
198 a r c B0 B1 ( c a l c ( rRelBc ∗ rRelB ∗rB ) 0 zB )
199 a r c B1 B2 ( 0 c a l c ( rRelBc ∗ rRelB ∗rB ) zB )
200 a r c B2 B3 (− c a l c ( rRelBc ∗ rRelB ∗rB ) 0 zB )
201 a r c B3 B0 ( 0 −c a l c ( rRelBc ∗ rRelB ∗rB ) zB )
202 a r c B4 B5 ( rB 0 zB )
203 a r c B5 B6 ( 0 rB zB )
204 a r c B6 B7 (−rB 0 zB )
205 a r c B7 B4 ( 0 −rB zB )
206 a r c B8 B9 ( rOutB 0 zB )
207 a r c B9 B10 ( 0 rOutB zB )
208 a r c B10 B11 (−rOutB 0 zB )
209 a r c B11 B8 ( 0 −rOutB zB )
210 a r c B12 B13 ( rOutOutB 0 zB )
211 a r c B13 B14 ( 0 rOutOutB zB )
212 a r c B14 B15 (−rOutOutB 0 zB )
213 a r c B15 B12 ( 0 −rOutOutB zB )
214
215 );
216
217 // D e f i n i n g p a t c h e s :
218 patches
219 (
220 patch i n l e t
221 (
222 (A1 A5 A4 A0)
223 (A2 A6 A5 A1)
224 (A3 A7 A6 A2)
225 (A0 A4 A7 A3)
226 (A3 A2 A1 A0)
227
228 (A5 A9 A8 A4)
229 (A6 A10 A9 A5)
230 (A7 A11 A10 A6)
231 (A4 A8 A11 A7)
232
233 (A9 A13 A12 A8)
234 ( A10 A14 A13 A9)
235 ( A11 A15 A14 A10 )
236 (A8 A12 A15 A11 )
237
238 )
239
240 patch o u t l e t
241 (
242 ( B0 B4 B5 B1 )
243 ( B1 B5 B6 B2 )
244 ( B2 B6 B7 B3 )
245 ( B3 B7 B4 B0 )
246 ( B0 B1 B2 B3 )
247
248 ( B4 B8 B9 B5 )
249 ( B5 B9 B10 B6 )
250 ( B6 B10 B11 B7 )
251 ( B7 B11 B8 B4 )
252
253 ( B8 B12 B13 B9 )
254 ( B9 B13 B14 B10 )
255 ( B10 B14 B15 B11 )
256 ( B11 B15 B12 B8 )
257
258 )
259
260 wall wallPipe
261 (
262 ( A12 A13 B13 B12 )
263 ( A13 A14 B14 B13 )
264 ( A14 A15 B15 B14 )
265 ( A15 A12 B12 B15 )
266
267 )
268 );
269
270 mergePatchPairs
271 (
272 );
273
274 // ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ //
Appendix B
Reynolds-averaged modelling for the

mixture
For the release version of multiphaseEulerFoam 2.2.x, RANS is not a standard

option. The changes required in order to use a RANS model for the mixture are
listed below.
B.1 Listings
B.1.0.1 In UEqns.H:
Table B.1: Changes in UEqns.H

1 phaseModel& phase = i t e r ( ) ;
2 c o n s t v o l S c a l a r F i e l d& a l p h a = phase ;
3 v o l V e c t o r F i e l d& U = phase .U( ) ;
4
5 v o l S c a l a r F i e l d n u E f f ( t u r b u l e n c e −>nut ( ) + i t e r ( ) . nu ( ) ) ;
6
7 UEqns . s e t
B.1.0.2 In createFields.H:
Table B.2: Changes in createFields.H

1 autoPtr<i n c o m p r e s s i b l e : : t u r b u l e n c e M o d e l > t u r b u l e n c e
2 (
3 i n c o m p r e s s i b l e : : t u r b u l e n c e M o d e l : : New(U, phi , f l u i d )
4 );
B.1.0.3 In multiphaseEulerFoam.H:
Table B.3: Changes in multiphaseEulerFoam.H

1 w h i l e ( p im pl e . l o o p ( ) )
2 {
3 t u r b u l e n c e −>c o r r e c t ( ) ;
4 fluid . solve () ;
5 rho = f l u i d . rho ( ) ;
6 #i n c l u d e " zonePhaseVolumes .H"
7
71
72 Appendix B. Reynolds-averaged modelling for the mixture
8 //#i n c l u d e "TEqns .H"

9 #i n c l u d e "UEqns .H"
10
11 // −−− P r e s s u r e c o r r e c t o r l o o p
12 w h i l e ( p im pl e . c o r r e c t ( ) )
13 {
14 #i n c l u d e "pEqn .H"
15 }
16
17 #i n c l u d e "DDtU .H"
18 }
B.2 yPlus function object

yPlusRAS is a postProcessing function object that calculates y + [76]for incom-
pressible cases employing RANS turbulence.
One challenge in CFD is how to treat the thin near-wall sublayer, where
viscous effects become important. In the near-wall region additional numerical
difficulties/costs arise and fine grids are needed near the wall. However if a wall-
function is applied, there is no need to resolve the viscous sublayer (y + < 5) and
the buffer layer, and thus the traditional industrial solution is to use wall-functions.
For multiphase flow cases the utility yPlusRAS does not work, since it is looking
in transportProperties for single phase transport properties. The code is changed
in order to deal with the mixture properties.
Table B.4: Changes in yP lusmixture
1 #i n c l u d e "fvCFD .H"
2
3 //#i n c l u d e " i n c o m p r e s s i b l e / s i n g l e P h a s e T r a n s p o r t M o d e l / s i n g l e P h a s e T r a n s p o r t M o d e l .H"
4 #i n c l u d e " i n c o m p r e s s i b l e T w o P h a s e M i x t u r e .H"
5 #i n c l u d e " i n c o m p r e s s i b l e /RAS/RASModel/RASModel .H"
6 #i n c l u d e " nutWallFunction / n u t W a l l F u n c t i o n F v P a t c h S c a l a r F i e l d .H"
7
8 #i n c l u d e " f l u i d T h e r m o .H"
9 #i n c l u d e " c o m p r e s s i b l e /RAS/RASModel/RASModel .H"
10 #i n c l u d e " mutWallFunction / m u t W a l l F u n c t i o n F v P a t c h S c a l a r F i e l d .H"
11
12 #i n c l u d e " w a l l D i s t .H"
Table B.5: Changes in yP lusmixture

1 #i n c l u d e " c r e a t e P h i .H"
2
3 // s i n g l e P h a s e T r a n s p o r t M o d e l l a m i n a r T r a n s p o r t (U, p h i ) ;
4 i n c o m p r e s s i b l e T w o P h a s e M i x t u r e t w o P h a s e P r o p e r t i e s (U, p h i ) ;
5
6 autoPtr<i n c o m p r e s s i b l e : : RASModel> RASModel
7 (
8 // i n c o m p r e s s i b l e : : RASModel : : New(U, phi , l a m i n a r T r a n s p o r t )
9 i n c o m p r e s s i b l e : : RASModel : : New(U, phi , t w o P h a s e P r o p e r t i e s )
10 );
Appendix C
fvSolution
Table C.1: fvSolution

1 /∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗− C++ −∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗\
2 | ========= | |
5 | \\ / A nd | Web : www.OpenFOAM. o r g |
7 \∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗/
8 FoamFile
9 {
10 version 2.0;
11 format a s c i i ;
13 l o c a t i o n system ;
14 object fvSolution ;
15 }
16
17 PIMPLE
18 {
19 nCorrectors 3;
20 nNonOrthogonalCorrectors 1 ;
21 nOuterCorrectors 1;
22 nAlphaCorr 2 ;
23 nAlphaSubCycles 5 ;
24 turbOnFinalIterOnly f a l s e ;
25 }
26
27 solvers
28 {
29 p
30 {
31 s o l v e r GAMG;
32 agglomerator faceAreaPair ;
33 mergeLevels 1 ;
34 cacheAgglomeration true ;
35 nCellsInCoarsestLevel 1000;
36 t o l e r a n c e 1 e −8;
37 relTol 0;
38 smoother G a u s s S e i d e l ;
39 nPreSweeps 0 ;
40 nPostSweeps 2 ;
41 nFinestSweeps 2 ;
42 minIter 3;
43 maxIter 1500;
44 }
45
46 pFinal
47 {
48 $p ;
49 tolerance 1 e −8;
73
74 Appendix C. fvSolution
50 relTol 0;
51 }
52
53 pcorr
54 {
55 $pFinal ;
57 relTol 0;
58 }
59
60 U
61 {
62 solver PBiCG ;
63 preconditioner DILU ;
64 t o l e r a n c e 1 e −8;
65 relTol 0;
66 minIter 3;
67 }
68
69 UFinal
70 {
71 $U ;
73 relTol 0;
74 minIter 3;
75 }
76
77 k
78 {
79 $U ;
80 }
81
82 kFinal
83 {
84 $UFinal ;
85 }
86
87 omega
88 {
89 $U ;
90 }
91
92 omegaFinal
93 {
94 $UFinal ;
95 }
96 }
97
98 relaxationFactors
99 {
100 fields
101 {
102 p 1;
103 }
104
105 equations
106 {
107 U 1;
108 k 1;
109 omega 1;
110 alpha . a i r 1;
111 a l p h a . water 1;
112 }
113 }
Appendix D
fvSchemes
Table D.1: fvSchemes

1 /∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗− C++ −∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗\
2 | ========= | |
5 | \\ / A nd | Web : www.OpenFOAM. o r g |
7 \∗−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−∗/
8 FoamFile
9 {
10 version 2.0;
11 format a s c i i ;
13 l o c a t i o n system ;
14 o b j e c t fvSchemes ;
15 }
16
17 ddtSchemes
18 {
19 d e f a u l t Euler ;
20 }
21
22 gradSchemes
23 {
24 d e f a u l t cellMDLimited l e a s t S q u a r e s 1 ;
25 }
26
27
28 divSchemes
29 {
30 " d i v \ ( phi , a l p h a . ∗ \ ) " Gauss vanLeer01 ;
31 " div \( phir , alpha . ∗ , alpha . ∗ \ ) " Gauss vanLeer01 ;
32
33 " d i v \ ( phiAlpha . ∗ ,U. ∗ \ ) " Gauss limitedLinearV 1;
34 d i v ( Rc ) Gauss linear ;
35 " d i v \ ( p h i . ∗ ,U. ∗ \ ) " Gauss limitedLinearV 1;
36 d i v ( phi , k ) Gauss limitedLinear 1;
37 d i v ( phi , omega ) Gauss limitedLinear 1;
38 d i v ( ( muEff ∗ dev (T( grad (U) ) ) ) ) Gauss linear ;
39 }
40
41 laplacianSchemes
42 {
43 d e f a u l t Gauss l i n e a r l i m i t e d 1 ;
44 }
45
46 interpolationSchemes
47 {
48 default linear ;
49 }
75
76 Appendix D. fvSchemes
50
51 snGradSchemes
52 {
53 default limited 1;
54 }
55
56 fluxRequired
57 {
58 d e f a u l t no ;
59 p;
60 pcorr ;
61 " alpha . ∗ " ;
62 }
Appendix E
Velocity reconstruction in pEqn.H
Surface tension is included in the construction of the pressure equation itself,

but it has been omitted from the reconstruction of the velocity field. Therefore the
surface tension force needs to be added to the fvc::reconstruct(). In the original
code line 6 is missing.
Table E.1: Velocity reconstruction in pEqn.H
1 phase .U( ) =
2 HbyAs [ p h a s e i ]
3 + fvc : : reconstruct
4 (
5 rAlphaAUfs [ p h a s e i ] ∗ ( g & mesh . S f ( ) )
6 + rAlphaAUfs [ p h a s e i ] ∗ f l u i d . s u r f a c e T e n s i o n ( phase ) ∗mesh .
magSf ( ) / phase . rho ( )
7 + rAlphaAUfs [ p h a s e i ] ∗ mSfGradp/ phase . rho ( )
8 );
77
Bibliography
[1] A. Abdelsalam, S. Cem, and P. Eduardo. ‘New dimensionless number for

gas-liquid flow in pipes’. In: International Journal of Multiphase Flow 81
(2016), pages 15–19. doi: 10.1016/j.ijmultiphaseflow.2015.12.008
(cited on page 7).
[2] A. Abdulahi. ‘Investigating the effect of liquid viscosity on two-phase gas-
liquid flows’. PhD thesis. University of Nottingham, 2014 (cited on page 10).
[3] M. Abdulkadir. ‘Experimental and computational fluid dynamics (CFD)
studies of gas-liquid flow in bends’. PhD thesis. University of Nottingham,
2011 (cited on pages 8, 34, 38, 44, 45, 49).
[4] M. Abdulkadir et al. ‘Comparison of experimental and computational fluid
dynamics (CFD) studies of slug flow in a vertical 90° bend’. In: Journal of
Computational Multiphase Flows 5.4 (2013), pages 265–282. doi: 10.1260/
1757-482X.5.4.265 (cited on pages 27, 50).
[5] M. Abdulkadir et al. ‘Comparison of experimental and Computational Fluid
Dynamics (CFD) studies of slug flow in a vertical riser’. In: Experimental
Thermal and Fluid Science 68 (2015), pages 468–483. doi: 10 . 1016 / j .
expthermflusci.2015.06.004 (cited on pages 2, 44, 45, 49, 50, 53, 54,
56–59, 61).
[6] M. Abdulkadir et al. ‘Detailed analysis of phase distributions in a vertical riser
using wire mesh sensor (WMS)’. In: Experimental Thermal and Fluid Science
59 (2014), pages 32–42. doi: 10.1016/j.expthermflusci.2014.07.010
(cited on pages 44, 45).
[7] M. Abdulkadir et al. ‘Experimental investigation of phase distributions
of two-phase air-silicone oil flow in a vertical pipe’. In: World Academy
of Science, Engineering and Technology 61 (2010), pages 52–59 (cited on
page 45).
[8] M. Abdulkadir et al. ‘Experimental study of the hydrodynamic behaviour of
slug flow in a vertical riser’. In: Chemical Engineering Science 106 (2014),
pages 60–75. doi: 10.1016/j.ces.2013.11.021 (cited on pages 9, 44).
[9] M. Abdulkadir et al. ‘Interrogating the effect of 90° bends on air-silicone oil
flows using advanced instrumentation’. In: Chemical Engineering Science
66.11 (2011), pages 2453–2467. doi: 10.1016/j.ces.2011.03.006 (cited
on page 45).
79
80 Bibliography
[10] L.A. Abdulkareem et al. ‘Characteristics of air-oil slug flow in inclined

pipe using tomographic techniques’. In: ASME/JSME 2011 8th Thermal
Engineering Joint Conference, AJTEC 2011. 2011 (cited on page 45).
[11] A. Albadawi et al. ‘Influence of surface tension implementation in volume of
f luid and coupled volume of fluid with Level Set methods for bubble growth
and detachment’. In: International Journal of Multiphase Flow 53 (2013),
pages 11–28 (cited on page 18).
[12] S. Ambrose. ‘The rise of Taylor bubbles in vertical pipes’. PhD thesis.
University of Nottingham, 2015 (cited on page 12).
[13] B.J. Azzopardi. Gas-Liquid Flows. Begell House, 2006 (cited on pages 8, 12,
14, 16).
[14] B.J. Azzopardi, L. Pioli, and L.A. Abdulkareem. ‘The properties of large
bubbles rising in very viscous liquids in vertical columns’. In: International
Journal of Multiphase Flow 67 (2014), pages 160–173 (cited on page 11).
[15] B.J. Azzopardi and E. Wren. ‘What is entrainment in vertical two-phase
churn flow?’ In: International Journal of Multiphase Flow 30.1 (2004),
pages 89–103. doi: 10.1016/j.ijmultiphaseflow.2003.11.001 (cited on
page 16).
[16] B.J. Azzopardi et al. ‘Characteristics of air/water slug flow in an intermediate
diameter pipe’. In: Experimental Thermal and Fluid Science 60 (2015),
pages 1–8. doi: 10 . 1016 / j . expthermflusci . 2014 . 08 . 004 (cited on
page 9).
[17] B.J. Azzopardi et al. ‘Persistence of frequency in gas-liquid flows across a
change in pipe diameter or orientation’. In: International Journal of Multi-
phase Flow 67.S (2014), pages 22–31. doi: 10.1016/j.ijmultiphaseflow.
2014.03.010 (cited on page 9).
[18] M.W. Baltussen, J.A.M. Kuipers, and N.G Deen. ‘A critical comparison of
surface tension models for the volume of fluid method’. In: 10th International
Conference on CFD in Oil & Gas, Metallurgical and Process Industries
SINTEF, Trondheim, Norway. 2014 (cited on page 18).
[19] J. Becker. ‘Coupling of population balance modeling and computational
fluid dynamics applied to turbulent emulsification processes in complex
geometries’. PhD thesis. Chemical and Process Engineering. Université
Claude Bernard - Lyon, 2013 (cited on page 19).
[20] J.C. Beerens. ‘Lubricated transport of heavy oil Simulation of multiphase
flow with OpenFOAM’. Master’s thesis. Delft University of Technology, 2013
(cited on page 27).
[21] E. Berberović et al. ‘Drop impact onto a liquid layer of finite thickness:
Dynamics of the cavity evolution’. In: Phys. Rev. E 79 (3 Mar. 2009). doi:
10.1103/PhysRevE.79.036306 (cited on page 18).
Bibliography 81
[22] J.U. Brackbill, D.B. Kothe, and C. Zemach. ‘A continuum method for
modeling surface tension’. In: Journal of Computational Physics 100.2 (1992),
pages 335–354. doi: 10.1016/0021-9991(92)90240-Y (cited on pages 18,
24).
[23] N. Brauner and A. Ullmann. ‘Modelling of gas entrainment from Taylor
bubbles. Part A: Slug flow’. In: International Journal of Multiphase Flow
30.3 (2004), pages 239–272 (cited on page 9).
[24] C. E. Brennen. Fundamentals of Multiphase Flows. Cambridge University
Press, New York, NY, USA, 2005 (cited on pages 7, 8, 14).
[25] R. A. S. Brown, G. A. Sullivan, and G. W. Govier. ‘The upward vertical flow
of air-water mixtures: III. Effect of gas phase density on flow pattern, holdup
and pressure drop’. In: The Canadian Journal of Chemical Engineering 38.2
(1960), pages 62–66. doi: 10.1002/cjce.5450380207 (cited on page 10).
[26] J.C. Cano-Lozano et al. ‘The use of volume of fluid technique to analyze
multiphase flows: Specific case of bubble rising in still liquids’. In: Applied
Mathematical Modelling 39.12 (2015), pages 3290–3305 (cited on page 17).
[27] G. Cerne, S. Petelin, and I. Tiselj. ‘Coupling of the Interface Tracking and
the Two-Fluid Models for the Simulation of Incompressible Two-Phase Flow’.
In: Journal of Computational Physics 171.2 (2001), pages 776–804. doi:
10.1006/jcph.2001.6810 (cited on page 21).
[28] P. Chen, J. Sanyal, and M.P. Dudukovic. ‘{CFD} modeling of bubble columns
flows: implementation of population balance’. In: Chemical Engineering
Science 59.22–23 (2004), pages 5201–5207 (cited on page 19).
[29] H. Cheng, J.H. Hills, and B.J. Azzopardi. ‘Effects of initial bubble size on
flow pattern transition in a 28.9 mm diameter column’. In: International
Journal of Multiphase Flow 28.6 (2002), pages 1047–1062 (cited on page 15).
[30] A. Clarke and R.I. Issa. ‘A numerical model of slug flow in vertical tubes’.
In: Computers & Fluids 26.4 (1997), pages 395–415. url: http://www.
sciencedirect.com/science/article/pii/S0045793096000163 (cited
on page 50).
[31] R. Collins et al. ‘The motion of a large gas bubble rising through liquid
flowing in a tube’. In: Journal of Fluid Mechanics 89.3 (Apr. 1, 2006),
[32] M. Colombo and M. Fairweather. ‘Multiphase turbulence in bubbly flows:
{RANS} simulations’. In: International Journal of Multiphase Flow 77 (2015),
pages 222–243. url: http://www.sciencedirect.com/science/article/
pii/S0301932215002013 (cited on page 27).
[33] G. Costigan and P.B. Whalley. ‘Slug flow regime identification from dynamic
void fraction measurements in vertical air-water flows’. In: International
Journal of Multiphase Flow 23.2 (1997), pages 263–282. url: http://www.
sciencedirect.com/science/article/pii/S030193229600050X (cited
on page 53).
82 Bibliography
[34] Y. Dai, D. Dakshinammorthy, and M. Agrawal. ‘CFD Modeling of Bubbly,

Slug and Annular Flow Regimes in Vertical Pipelines’. In: OTC-24245-MS
Offshore Technology Conference. 2013 (cited on page 27).
[35] R. M. Davies and G. Taylor. ‘The Mechanics of Large Bubbles Rising through
Extended Liquids and through Liquids in Tubes’. In: Proceedings of the
Royal Society of London A: Mathematical, Physical and Engineering Sciences
200.1062 (1950), pages 375–390. doi: 10.1098/rspa.1950.0023. eprint:
http://rspa.royalsocietypublishing.org/content/200/1062/375.
full.pdf (cited on pages 9, 11).
[36] S. S. Deshpande, L. Anumolu, and M. F. Trujillo. ‘Evaluating the perfor-
mance of the two-phase flow solver interFoam’. In: Computational Science
& Discovery 5 (2012) (cited on page 18).
[37] C. Dopazo. ‘On conditional averages for intermittent turbulent flows’. In:
Journal of Fluid Mechanics 81 (1977), pages 433–438 (cited on pages 19,
22).
[38] D. T. Dumitrescu. ‘Strömung an einer Luftblase im senkrechten Rohr’. In:
ZAMM - Journal of Applied Mathematics and Mechanics / Zeitschrift für
Angewandte Mathematik und Mechanik 23.3 (1943), pages 139–149. doi:
10.1002/zamm.19430230303. url: http://dx.doi.org/10.1002/zamm.
19430230303 (cited on page 11).
[39] J. Fabre and A. Line. ‘Modeling of Two-Phase Slug Flow’. In: Annual Review
of Fluid Mechanics 24.1 (1992), pages 21–46. doi: 10.1146/annurev.fl.
24.010192.000321 (cited on pages 9, 50).
[40] R. C. Fernandes, R. Semiat, and A. E. Dukler. ‘Hydrodynamic model for gas-
liquid slug flow in vertical tubes’. In: AIChE Journal 29.6 (1983), pages 981–
989. doi: 10.1002/aic.690290617 (cited on pages 9, 10, 51, 58).
[41] S. Ganesan, G. Matthies, and L. Tobiska. On Spurious Velocities in Incom-
pressible Flow Problems with Interfaces. Preprint. 2005 (cited on page 18).
[42] V.R. Gopala and B.G.M. van Wachem. ‘V olume of fluid methods for
immiscible-fluid and free-surface flows’. In: Chemical Engineering Jour-
nal 141.1-3 (2008), pages 204–221. doi: 10.1016/j.cej.2007.12.035
[43] P. Griffith and G. A. Synder. The Bubbly-Slug Transition in a High Velocity
Two-Phase Flow. Technical report. MIT Report, 1964 (cited on page 16).
[44] P. Griffith and G. B. Wallis. ‘Two-Phase Slug Flow’. In: J . Heat Trans.
83.307 (1961) (cited on pages 16, 50).
[45] D.J.E. Harvie, M.R. Davidson, and M. Rudman. ‘An analysis of parasitic
current generation in volume of fluid simulations’. In: Applied Mathematical
Modelling 30.10 (2006), pages 1056–1066. doi: http://dx.doi.org/10.
1016/j.apm.2005.08.015 (cited on page 18).
[46] A. Hedlund. Evaluation of RANS turbulence models for the simulation of
channel flow. Technical report. Uppsala Universitet, Department of Engi-
neering Sciences, 2014 (cited on page 36).
Bibliography 83
[47] L.A. Hernández-Pérez V.and Abdulkareem and B.J. Azzopardi. ‘Effects of

physical properties on the behaviour of Taylor bubbles’. In: WIT Transactions
on Engineering Sciences. Volume 63. 2009, pages 355–366. doi: 10.2495/
MPF090301 (cited on page 10).
[48] V. Hernandez-Perez. ‘Gas-liquid two-phase flow in inclined pipes’. PhD thesis.
University of Nottingham, 2007 (cited on pages 8, 38, 50).
[49] V. Hernandez-Perez, M. Abdulkadir, and B.J. Azzopardi. ‘Grid generation
issues in the CFD modelling of two-phase flow in a pipe’. In: Journal of
Computational Multiphase Flows 3.1 (2011), pages 13–26. doi: 10.1260/
1757-482X.3.1.13 (cited on pages 27, 34).
[50] T. Hibiki and M. Ishii. ‘One-dimensional drift–flux model for two-phase
flow in a large diameter pipe’. In: International Journal of Heat and Mass
Transfer 46.10 (2003), pages 1773–1790 (cited on page 11).
[51] D. P. Hill. ‘The computer simulation of dispersed two-phase flow’. PhD thesis.
Imperial College of Science, Technology and Medicine, London, 1998 (cited
on page 22).
[52] C. Hirt and B. Nichols. ‘Volume of fluid (VOF) method for the dynamics of
free boundaries’. In: J. Comput. Phys. 39 (1981), pages 201–225 (cited on
page 17).
[53] R. van Hout et al. ‘Experimental investigation of the velocity field induced
by a Taylor bubble rising in stagnant water’. In: International Journal of
Multiphase Flow 28.4 (2002), pages 579–596 (cited on page 9).
[54] J.M. Hyman. Numerical methods for tracking interfaces. url: http://math.
lanl.gov/~mac/papers/numerics/H84.pdf (cited on page 17).
[55] M. Ishii and T. Hibiki. ‘Thermo-fluid dynamics of two-phase flow’. In:
Springer Verlag, 2010 (cited on page 19).
[56] R.I. Issa, A.D. Gosman, and A.P. Watkins. ‘The computation of compressible
and incompressible recirculating flows by a non-iterative implicit scheme’.
In: Journal of Computational Physics 62.1 (1986), pages 66–82. url: http:
//www.sciencedirect.com/science/article/pii/0021999186901002
(cited on page 28).
[57] D. Jamet, D. Torres, and J.U. Brackbill. ‘On the Theory and Computation
of Surface Tension: The Elimination of Parasitic Currents through Energy
Conservation in the Second-Gradient Method’. In: Journal of Computational
Physics 182.1 (2002), pages 262–276 (cited on page 18).
[58] H. Jasak. ‘Error Analysis and Estimation for the Finite V olume Method
with Applications to Fluid Flows’. PhD thesis. Imperial College of Science,
Technology and Medicine, London, 1996 (cited on page 28).
[59] S. Jayanti and G.F. Hewitt. ‘Prediction of the slug-to-churn flow transition
in vertical two-phase flow’. In: International Journal of Multiphase Flow
18.6 (1992), pages 847–860 (cited on page 16).
84 Bibliography
[60] O. T. Kajero, B. J. Azzopardi, and L. Abdulkareem. ‘Experimental in-

vestigation of the effect of liquid viscosity on slug flow in small diameter
bubble column’. In: EPJ Web of Conferences 25 (2012), page 01037. doi:
10.1051/epjconf/20122501037 (cited on page 10).
[61] C. M. Kumar. ‘CFD Simulation of Multicomponent gas flow through porous
media’. Master’s thesis. Bergische Universitat Wuppertal, 2013 (cited on
page 30).
[62] H.K. Kytömaa and C.E. Brennen. ‘Small amplitude kinematic wave propa-
gation in two-component media’. In: International Journal of Multiphase
Flow 17.1 (1991), pages 13–26 (cited on page 15).
[63] B.E. Launder and D.B. Spalding. ‘The numerical computation of turbulent
flows’. In: Computer Methods in Applied Mechanics and Engineering 3.2
(1974), pages 269–289. url: http://www.sciencedirect.com/science/
article/pii/0045782574900292 (cited on page 50).
[64] Y. Liu. ‘Two-Fluid Modeling of Gas-Solid and Gas-Liquid Flows: Solver
Development and Application’. PhD thesis. Technische Universitat Munchen,
2014 (cited on pages 16, 23).
[65] E. W. Llewellin et al. ‘The thickness of the falling film of liquid around a
Taylor bubble’. In: Proceedings of the Royal Society of London A: Mathemat-
ical, Physical and Engineering Sciences 468.2140 (2012), pages 1041–1064.
doi: 10.1098/rspa.2011.0476 (cited on page 11).
[66] OpenCFD Ltd. OpenFOAM 2.3.1 User’s Guide. 2014 (cited on pages 18,
25).
[67] E. Michaelides, T. C. Clayton, and J. D. Schwarzkopf. ‘Basic Concepts and
Definitions’. In: Mechanical Engineering Series. CRC Press, Sept. 2005, pages
(cited on page 5).
[68] A.O. Morgado et al. ‘Review on vertical gas–liquid slug flow’. In: Inter-
national Journal of Multiphase Flow 85 (2016), pages 348–368 (cited on
pages 9, 10, 12).
[69] D.J. Nicklin. ‘Two-phase bubble flow’. In: Chemical Engineering Science
17.9 (1962), pages 693–702 (cited on pages 11, 12, 16).
[70] A. Ohnuki and H. Akimoto. ‘Experimental study on transition of flow pattern
and phase distribution in upward air–water two-phase flow along a large
vertical pipe’. In: International Journal of Multiphase Flow 26.3 (2000),
[71] N.K. Omebere-Iyari and B.J. Azzopardi. ‘The effect of pipe diameter in
vertical gas/liquid upflow’. In: 4th Japanese-European Two-Phase Flow
Group Meeting, Kanbaikan, Kyoto, 24–28 September 2006. 2006 (cited on
page 9).
[72] C. J. Owen and N. Zuber. ‘The interrelation between void fraction fluctua-
tions and flow patterns in two-phase flow’. In: International Journal of Mul-
tiphase Flow 2.3 (1975), pages 273–306. url: http://www.sciencedirect.
com/science/article/pii/0301932275900154 (cited on page 53).
Bibliography 85
[73] Z. Pan, J. A. Weibel, and S. V. Garimella. ‘Spurious Current Suppression in

VOF-CSF Simulation of Slug Flow through Small Channels’. In: Numerical
Heat Transfer Part A - Applications 67 (Jan. 2015), pages 1–12. doi: 10.
1080/10407782.2014.916109 (cited on page 18).
[74] M. Parsi et al. ‘Assessment of a hybrid CFD model for simulation of complex
vertical upward gas-liquid churn flow’. In: Chemical Engineering Research
and Design 105 (2016), pages 71–84. doi: 10.1016/j.cherd.2015.10.044
(cited on pages 21, 23, 27).
[75] M. Parsi et al. ‘On the effect of liquid viscosity on interfacial structures
within churn flow: Experimental study using wire mesh sensor’. In: Chemical
Engineering Science 130 (2015), pages 221–238 (cited on pages 38, 50).
[76] S. B. Pope. Turbulent flows. Cambridge: Cambridge University Press, 2000
[77] C.C.T. Pringle et al. ‘The existence and behaviour of large diameter Taylor
bubbles’. In: International Journal of Multiphase Flow 72 (2015), pages 318–
323. doi: 10.1016/j.ijmultiphaseflow.2014.04.006 (cited on page 11).
[78] A. Q. Raeini, M. J. Blunt, and B. Bijeljic. ‘Modelling two-phase flow in
porous media at the pore scale using the volume of fluid method’. In: Journal
of Computational Physics 231.17 (2012), pages 5653–5668 (cited on page 18).
[79] A. Ramos-Banderas et al. ‘Dynamics of two-phase downwards flows in
submerged entry nozzles and its influence on the two-phase flow in the
mold’. In: International Journal of Multiphase Flow 31.5 (2005), pages 643–
665. url: http : / / www . sciencedirect . com / science / article / pii /
S0301932205000182 (cited on page 50).
[80] N. Ratkovich, S.K. Majumder, and T.R. Bentzen. ‘Empirical correlations
and CFD simulations of vertical two-phase gas-liquid (Newtonian and non-
Newtonian) slug flow compared against experimental data of void fraction’.
In: Chemical Engineering Research and Design 91.6 (2013), pages 988–998.
doi: 10.1016/j.cherd.2012.11.002 (cited on page 50).
[81] E.S. Rosa, B.F. Flora, and M.A.S.F. Souza. ‘Design and performance pre-
diction of an impedance void meter applied to the petroleum industry’. In:
Measurement Science and Technology 23.5 (2012), page 055304 (cited on
page 8).
[82] S.Z. Rouhani and M.S. Sohal. ‘Two-phase flow patterns: A review of research
results’. In: Progress in Nuclear Energy 11.3 (1983), pages 219–259 (cited
on page 11).
[83] H. Rusche. ‘Computational Fluid Dynamics of Dispersed Two-Phase Flows at
High Phase Fractions’. PhD thesis. Imperial College of Science, Technology
and Medicine, London, 2001 (cited on pages 18, 19, 23, 25, 27).
[84] M.D. Santiago. ‘An Extended Mixture Model for the Simultaneous Treatment
of Short and Long Scale Interfaces’. PhD thesis. Universidad Nacional del
Litoral, 2013 (cited on pages 19, 24).
86 Bibliography
[85] M.D. Santiago. Description and utilization of interFoam multiphase solver

(cited on page 18).
[86] R. Scardovelli and S. Zaleski. ‘Direct numerical simulation of free-surface
and interfacial flow’. In: Annual Review of Fluid Mechanics 31.1 (1999),
pages 567–603. doi: 10.1146/annurev.fluid.31.1.567 (cited on pages 17,
20).
[87] L. Schiller and A. Naumann. ‘A drag coefficient correlation’. In: Z. Ver.
Dtsch. Ing. 77.12 (1933), pages 318–320 (cited on page 25).
[88] H. Shaban and S. Tavoularis. ‘Identification of flow regime in vertical upward
air–water pipe flow using differential pressure signals and elastic maps’. In:
International Journal of Multiphase Flow 61 (2014), pages 62–72 (cited on
pages 8, 12).
[89] O.Y. Shonibare and K.E. Wardle. ‘Numerical investigation of vertical plung-
ing jet using a hybrid multifluid-VOF multiphase CFD solver’. In: Interna-
tional Journal of Chemical Engineering 2015 (2015). doi: 10.1155/2015/
925639 (cited on pages 1, 24, 26).
[90] L.F.L.R. Silva, R.B. Damian, and P.L.C. Lage. ‘Implementation and analysis
of numerical solution of the population balance equation in {CFD} packages’.
In: Computers & Chemical Engineering 32.12 (2008), pages 2933–2945 (cited
on page 19).
[91] M.J. da Silva et al. ‘High-resolution gas-oil two-phase flow visualization with
a capacitance wire-mesh sensor’. In: Flow Measurement and Instrumentation
21.3 (2010), pages 191–197. doi: 10.1016/j.flowmeasinst.2009.12.003
(cited on page 45).
[92] M. van Sint Annaland, N.G. Deen, and J.A.M. Kuipers. ‘Numerical simu-
lation of gas bubbles behaviour using a three-dimensional volume of fluid
method’. In: Chemical Engineering Science 60.11 (2005), pages 2999–3011
(cited on page 17).
[93] C.H. Song, H.C. No, and M.K. Chung. ‘Investigation of bubble flow develop-
ments and its transition based on the instability of void fraction waves’. In:
International Journal of Multiphase Flow 21.3 (1995), pages 381–404 (cited
on page 15).
[94] P.L. Spedding et al. ‘Vertical Two-Phase Flow’. In: Chemical Engineering
Research and Design 76.5 (1998), pages 612–619 (cited on page 11).
[95] L. Štrubelj and I. Tiselj. ‘Two-fluid model with interface sharpening’. In:
International Journal for Numerical Methods in Engineering 85.5 (2011),
pages 575–590. doi: 10.1002/nme.2978 (cited on page 22).
[96] M. Sussman, P. Smereka, and S. Osher. ‘A Level Set Approach for Computing
Solutions to Incompressible Two-Phase Flow’. In: Journal of Computational
Physics 114.1 (1994), pages 146–159 (cited on page 17).
[97] L. Szalinski et al. ‘Comparative study of gas-oil and gas-water two-phase
flow in a vertical pipe’. In: Chemical Engineering Science 65.12 (2010),
pages 3836–3848. doi: 10.1016/j.ces.2010.03.024 (cited on page 10).
Bibliography 87
[98] T. Taha and Z.F. Cui. ‘{CFD} modelling of slug flow in vertical tubes’. In:
Chemical Engineering Science 61.2 (2006), pages 676–687 (cited on pages 10,
50).
[99] Y. Taitel and D. Barnea. ‘Two-Phase Slug Flow’. In: Advances in Heat
Transfer. Volume 20. Elsevier, 1990, pages 83–132 (cited on page 10).
[100] Y. Taitel, D. Bornea, and A.E. Dukler. ‘Modelling flow pattern transitions
for steady upward gas-liquid flow in vertical tubes.’ In: AICHE. J. 26.3 ,
May 1980 (1980), pages 345–354 (cited on pages 8, 14–16).
[101] J.R. Thome. Encyclopedia of Two-phase Heat Transfer and Flow I: Funda-
mentals and Methods. World Scientific, 2015 (cited on page 10).
[102] P. Tkaczyk. ‘CFD simulation of annular flows through bends’. PhD thesis.
University of Nottingham, 2011 (cited on page 34).
[103] P. D. Tomaselli and E. D. Christensen, editors. Investigation on the Use of
a Multiphase Eulerian CFD solver to simulate breaking waves. ASME 34th
International Conference on Ocean, Offshore and Artic Engineering. 2015
(cited on page 22).
[104] G. Tryggvason et al. ‘A front-tracking method for the computations of mul-
tiphase flow’. In: Journal of Computational Physics 169.2 (2001), pages 708–
759 (cited on page 16).
[105] Ž. Tuković and H. Jasak. ‘A moving mesh finite volume interface tracking
method for surface tension dominated interfacial fluid flow’. In: Computers
& Fluids 55 (2012), pages 70–84 (cited on page 16).
[106] O. Ubbink. ‘Numerical prediction of two fluid systems with sharp interfaces’.
PhD thesis. Imperial College of Science, Technology and Medicine, London,
[107] G.P. Van Der Meulen. ‘Churn-Annular Gas-Liquid Flows in Large Diame-
ter Vertical Pipes’. PhD thesis. University of Nottingham, 2012 (cited on
page 50).
[108] F. Viana et al. ‘Universal correlation for the rise velocity of long gas bubbles in
round pipes’. In: Journal of Fluid Mechanics 494 (Oct. 22, 2003), pages 379–
398. doi: 10.1017/S0022112003006165 (cited on page 11).
[109] K.E. Wardle and H.G. Weller. ‘Hybrid multiphase CFD solver for coupled
dispersed/segregated flows in liquid-liquid extraction’. In: International
Journal of Chemical Engineering (2013). doi: 10.1155/2013/128936 (cited
on pages 1, 22–26, 28).
[110] M.J. Watson and G.F. Hewitt. ‘Pressure effects on the slug to churn transi-
tion’. In: International Journal of Multiphase Flow 25.6–7 (1999), pages 1225–
[111] H.G. Weller. ‘A new approach to VOF-based interface capturing methods for
incompressible and compressible flow’. In: Technical Report TR/HGW/04
(2008) (cited on pages 18, 23).
88 Bibliography
[112] H.G. Weller. ‘Derivation, Modelling and Solution of the Conditionally Aver-
aged Two-Phase Flow Equations’. In: Technical Report TR/HGW/02 (2005)
[113] E.T. White and R.H. Beardmore. ‘The velocity of rise of single cylindrical air
bubbles through liquids contained in vertical tubes’. In: Chemical Engineering
Science 17.5 (1962), pages 351–361 (cited on pages 10, 11).
[114] D.C. Wilcox. Turbulence Modeling for CFD. DCW Industries, Incorporated,
[115] R. A. Worthen and R. A. W. M. Henkes. ‘CFD for the Multiphase Flow Split-
ting From a Single Flowline Into a Dual Riser’. In: 17th International Con-
ference on Multiphase Production Technology, 10-12 June, Cannes, France.
2015 (cited on pages 2, 33–36, 38, 39, 44).
[116] D. Zhang, N.G. Deen, and J.A.M. Kuipers. ‘Numerical simulation of the
dynamic flow behavior in a bubble column: A study of closures for turbu-
lence and interface forces’. In: Chemical Engineering Science 61.23 (2006),
[117] J. Zhao and T. Shan. ‘Coupled CFD–DEM simulation of fluid–particle
interaction in geomechanics’. In: Powder Technology 239 (2013), pages 248–

Assessment of A Hybrid VOF Two-Fluid CFD Solver - Thesis - 2016

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Assessment of A Hybrid VOF Two-Fluid CFD Solver - Thesis - 2016

Uploaded by

Copyright:

Available Formats

Politecnico di Milano

SCUOLA DI INGEGNERIA INDUSTRIALE E DELL’INFORMAZIONE

Tesi di Laurea Magistrale

Assessment of a hybrid VOF two-fluid CFD solver for

Anno Accademico 2015–2016

In the framework of OpenFOAM, multiphaseEulerFoam is used for simulation

Nel contesto di OpenFoam, il solutore multiphaseEulerFoam è utilizzato per la

3 Coupling Euler-Euler two-fluid with VOF 21

4.2.1 Case description . . . . . . . . . . . . . . . . . . . . . . . . . 44

B Reynolds-averaged modelling for the mixture 73

E Velocity reconstruction in pEqn.H 79

2.1 Sketches of flow regimes for two-phase flow in vertical pipes . . . . . 9

3.1 Flow chart of multiphaseEulerFoam. . . . . . . . . . . . . . . . . . 29

4.1 Model geometry for simulation of vertical upward flow through a

4.15 Cross-Section grid of the computational domain in OpenFOAM. . . 49

3.1 Drag specification in transportProperties dictionary . . . . . . . 25

This thesis is part of a collaboration project between Delft University of Tech-

are commonly used in several process industries. Atomization to generate small

1.2 Research objective

1.3 Thesis outline

• Chapter 2: Basic theory on multiphase flow in general, description of upward

• Chapter 3: Theory of the hybrid approach coupling two-fluid model with

• Chapter 4: Results and validation of the model.

2.1 Multiphase flow: basic definitions and dimen-

The actual phase velocities are defined as:

2.1.2 Slip effects

λliquid + λgas = 1 (2.5)

Us,gas = αgas Ugas (2.9)

The same relations apply for the other phase.

2.1.3 Froude number

2.1.4 Morton number

2.1.5 Eotvos number

2.1.6 Dimensionless inverse viscosity number

2.1.7 Slippage number

2.2 Upward gas-liquid flow in vertical pipes

2.2.1 Flow regimes

Annular flow: This configuration is characterised by liquid travelling as a film

2.2.1.1 Slug flow

in stagnant liquid conditions.

A value of 1.2 for C0 was suggested by Nicklin et al. [69].

2.2.2 Flow pattern maps

(a) Numerical image of the (b) Schematic representation of the main

Figure 2.5: Slug unit.

(a) 50 mm diameter (b) 25 mm diameter

2.2.3 Transition mechanism

2.2.3.1 Bubble-slug transition

2.2.3.2 Slug-churn transition

2.3 CFD methodologies for two-phase flow

2.3.1 Direct numerical simulation

Figure 2.8: Schematic representation of interface-tracking and interface-capturing meth-

conditions at the interface, which is defined as a sharp surface. Furthermore,

2.3.2 Eulerian-Lagrangian method

In the Eulerian-Lagrangian approach, the continuous phase is treated in the

2.3.3 Eulerian-Eulerian method

Coupling Euler-Euler two-fluid with

3.1 Hybrid approach

3.2.1 Multifluid momentum equations

3.2.1.1 Interface capturing

The term |∇α|

3.2.1.2 Surface tension

The surface tension at the gas-liquid interface generates an additional pressure

3.2.1.3 Interphase drag models

FD,k = αc αd K(ud − uc ) (3.9)