Comparison of Experimental and Computational Fluid Dynamics (CFD) studies of slug flow

in a vertical riser

Abdulkadir, M.1* , Hernandez-Perez, V.2 , Lo, S.3 , Lowndes, I. S.2 & Azzopardi, B. J.2

1 Department of Chemical Engineering, Federal University of Technology, Minna, Niger State, Nigeria

2 Process and Environmental Engineering Research Division , Faculty of Engineering, University of Nottingham
University Park, Nottingham, NG7 2RD, United Kingdom

3 CD-adapco, Trident Park, Didcot, OX11 7HJ, United Kingdom

*Corresponding author’s e-mail: mukhau@futminna.edu.ng

Abstract: This paper presents a comparison of the results obtained from experiments and CFD studies of slug

flow in a vertical riser. A series of two experimental investigations were carried out on a 6 m vertical pipe with a

0.067 m internal diameter charged with an air–silicone oil mixture. For the first set of experiments, the riser was

initially full of air, and then liquid and gas flows set to liquid and gas superficial velocities = 0.05 and 0.344

m/s, respectively, electrical capacitance tomography (ECT) and wire mesh sensor (WMS) transducers were

employed. In the second one, the riser was initially full of (static) liquid, and then liquid and gas flows set to

liquid and gas superficial velocities = 0.05 and 0.344 m/s, respectively, only ECT was used. A characterization

of the observed slug flow regimes was carried out. This includes the evaluation of the instantaneous distribution

of the phases over the pipe cross -section, the Probability Density Function (PDF) of void fraction, time series of

cross-sectional void fraction, Power Spectral Density (PSD), structure velocity of the Taylor bubble, lengths of

the liquid slug and Taylor bubble and void fractions in the liquid slug and Taylor bubble. The simulation results

were validated both qualitatively and quantitatively against experimental data. A reasonably good agreement

was observed between the results of the experiment and CFD.

Keywords: CFD, ECT, VOF, Slug flow, air– silicone oil, riser, PDF, void fraction, PSD, Taylor bubble length, velocity

1. Introduction:

Slug flow in a vertical riser is a very common flow regime under normal operating conditions

of a two-phase flow facility, such as an oil production riser. One feature of slug flow is the

acceleration of the liquid phase to form fast moving liquid slugs, which can carry a

significant amount of liquid with high kinetic energy. This is potentially hazardous to the

1
structure of the flow transport system and processing equipment due to the strong oscillating

pressure produced by the mechanical momentum of the slugs.

In oil production, the presence of liquid slugs in the riser gives an irregular output in terms of

gas and liquid flow at either the outlet to the system or the next processing stage. This can

pose challenges to the design and operation of such flow systems. The pressure drop

experienced for slug flow regimes is substantially higher when compared to other flow

regimes, and consequently the maximum possible length of a liquid slug that might be

encountered needs to be known. Often, slug catching devices are used to collect the slugs,

and avoid any damage to the downstream equipment. For the design of such slug catchers, it

is important to know what kind of slugs to anticipate. For that reason, it is important to study

the behaviour of slug flow in great detail for the optimal, efficient and safe design and

operation of two-phase gas– liquid flow systems.

A considerable amount of research has been devoted to the study of this two-phase flow

regime. The important question of when and how these slugs are formed has received much

attention from research workers: [1–5] among others. A critical review of this topic is given

by Fabre and Line [6]. However, there remains much to be investigated and understand about

that flow pattern. In particular, deeper investigation is needed to attain a thorough

understanding of the internal structure of slug flow. Moreover, reports on slug flow behaviour

with fluids which are relevant to the industry are limited. Empirical correlations and

mechanistic models have been presented in the literature. These are mainly one-dimensional

approaches that cannot fully characterize the flow. The limitations of one-dimensional

models may be addressed by the use of Computational Fluid Dynamics (CFD). The

applications of CFD to investigate multiphase flow are highly dependent on the flow pattern

under study, as different closure models are needed for different flow regimes. These models

2
require to be validated to gain confidence in their use. The validation of CFD models requires

experimental data that characterize the important flow parameters over a range of liquid and

gas flow rates.

In this work, different slug flow characteristics (e.g. void fraction in liquid slug and Taylor

bubble, lengths of liquid slug and Taylor bubble, slug frequency, structure velocity) are

determined using the results of experiments and the solutions to the CFD models. To validate

the CFD models the results were compared against the corresponding experimental data.

2. Experimental Methodology:

The experimental investigations were carried out on an inclinable pipe flow rig within the

Chemical Engineering Laboratories at the University of Nottingham. The details of this

experimental facility may be found in Azzopardi et al. [7] and [8–10]. In brief: the

experimental test section of the facility consists of a transparent acrylic pipe of 6 m length

and 0.067 m internal diameter. The test pipe section may be rotated on the rig to allow it to

lay at any inclination angle of between -5 to 90o to the horizontal. For the experiments

reported in this paper the rig test pipe section was mounted as a vertical riser (an inclination

of 900 to the horizontal). It is worthy of mention that full-experimentation in risers of this

magnitude and other larger ones is expensive and therefore a more cost-effective approach

for exploring the behaviour of two-phase flow in these risers is by using validated CFD

model simulations.

The resultant flow patterns obtained from two sets of experimental campaigns involving air–

silicone oil flow rates were recorded using electrical capacitance tomography (ECT) and wire

mesh sensor (WMS). A detailed description of theory behind the ECT technology can be

found in [7], and [11–12]. In this study, a ring of two measurement electrodes were placed

around the circumference of the riser at a given height above the injection portals at the

3
bottom of the 6 m riser section. The use of two such circumferential rings of sensor

electrodes, located at a specified distance apart (also known as 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 and 4.489 m upstream of the air –

silicone oil mixer, and injection portal, located at the base of the riser. A flow chart of the

various experimental measurements and the calculated parameters that characterise the flow

are presented in Table 1.

The capacitance WMS placed at 4.92 m away from the mixing section, described in detail by

da Silva et al. [13], can image the dielectric components of the two-phase mixture in the pipe

by measuring, rapidly and continually, the capacitances of the passing fluid at the crossing

points in the mesh. This capacitance signal is a measure of the amount of silicone oil, and

thus indicates the local phase composition in the grid cell.

The physical experiments were conducted as a series of two campaigns: (1) pipe initially full

of air, and then liquid and gas flows set to liquid and gas superficial velocities of 0.05 and

0.344 m/s, respectively and (2) the second involved pipe initially full of (static) liquid, and

then liquid and gas flows set to liquid and gas superficial velocities of 0.05 and 0.344 m/s,

respectively (the same as for the subsequent CFD simulation models).

4
 Table 1: Flowchart of the collection and processing of the experimental measurements used to obtain the parametric
characterisation of the slug flow regime

Direct physical measurement Data processing method Parametric Output 1 Parametric Output 2

Instrument Data

ECT
PDF of void fraction Flow pattern, Lengths of liquid
Time series  gs ,  TB , frequency slug and Taylor
of void bubble, symbols of
fraction PSD – Power parameter set
Spectral Density Frequency, symbols of
parameter set
Cross-correlation
Structure velocity,
symbol
Image reconstruction
Contours of phase 3D structures
distribution

2.1 Gas-liquid mixing section

In the design of the physical experimental rig, it was ensured that the mixing section of the air

and silicone oil phases took place in such a way as to reduce flow instability. Flow stability

was achieved by using a purpose built mixing device, to provide maximum time for the two-

phase flow to develop. The mixing device is made from PVC pipe as shown in Figure 2. The

silicone oil enters the mixing chamber from one side and flows around a perforated cylinder

through which the air is introduced through a large number of 3 mm diameter orifices. This

arrangement ensures that the gas and liquid flows were well mixed at the entry to the test

section. The inlet volumetric flow rates of the liquid and air were determined by a set of

rotameters located above a set of valves on the two inlet feed flow pipes.

The introduction of the air and liquid flows at the inlet to the CFD models was defined as a

velocity-inlet boundary at which the mixture velocity and the liquid volume fraction are

specified. The mixture velocity profile is assumed to be uniform. This approach requires no

5
additional experimental knowledge about the formation of the liquid slugs to formulate the

numerical simulation.

Figure 2: Air– silicone oil mixing section

3. CFD Model:

Parallel to the execution of the physical experiments, the construction and solution of CFD

models were carried out. The aim of the numerical simulations was to investigate the

potential application of the multiphase flow models, built in the commercial CFD codes Star-

CD and Star-CCM+. The Star-CD code employs the Finite Volume method to numerically

discretize the computational flow domain. In the present work, isothermal motion of an

incompressible two-phase flow is considered. The condition of slug two-phase flow has been

simulated with the Volume of Fluid (VOF) method of Hirt and Nichols [14].The movement

of the modelled gas–liquid interface is tracked based on the distribution of,  G1 , the volume

fraction of gas in a computational cell; where  G  0 , is a liquid cell and  G  1 in a gas

phase cell, [14]. Therefore, the gas-liquid interface exists in the cell where  G lies between 0

and 1.

6
3.1 Computational domain:

In order for the simulation to produce meaningful results, it was important to ensure that the

geometry of the flow domain faithfully represented the experimental arrangement. Hence, a

full 3-Dimensional flow domain, as shown in Figure 1, was considered based on the fact that

the flow simulated has been found to be axisymmetric according to the conclusions of the

previous experimental studies of Azzopardi et al. [7] and [10] and In the present work, three

CFD measurement sections were located at positions similar to those of the experimental

work, namely, at distances of 4.4 m, 4.489 m and 4.92 m above the base of the riser. Here,

the locations 4.4 m and 4.489 m represent the two electrical capacitance tomography (ECT)

planes, whilst 4.92 m the wire mesh sensor (WMS). Air and silicone oil are injected at the

inlet section of the pipe, then the two-phase mixture flows upwards through the vertical riser

pipe, finally discharges through the outlet at atmospheric pressure. The relevant fluids

properties are shown in Table 2.

Figure 1: 3-D geometry of the computational flow domain showing the location of the recording sections that correspond to
the locations of the experimental measurement transducers.

7
Table 2: Properties of the fluids

Fluid Density(kg/m3 ) Viscosity(kg/ms) Surface tension (N/m)

Air 1.18 0.000018

0.02
Silicone oil 900 0.0053

3.2 Grid generation:

The model riser flow geometry was built and meshed with Star-CD, then imported into Star-

CCM+, where the computation and post-processing of the results were performed. The

geometries of the mesh employed is the butterfly grid (O-grid), which has been successfully

employed by [10], and [15–16]. Figure 3 shows the mesh for the riser used for the CFD

simulation. It uses a Cartesian mesh at the centre of the pipe combined with a cylindrical one

around it. According to Hernandez-Perez [15], the O-grid (butterfly grid) allows for a good

representation of the boundary layer and it is adequately stretched along the longitudinal axis.

It was essential in this work to have a reasonably fine grid close to the wall, with a thickness

of 0.0000123 m. A growth function was used to build the grid with these properties where the

first grid was fixed to a distance of 10 -6 m to ensure a y+<1 and to properly resolve the

boundary layer close to the wall surface. The y+ is a non-dimensional wall distance that

describes how coarse or fine a mesh is for a wall-bounded flow. In this particular case, a fine

grid is required to properly capture the sub layer between the laminar flow (near the wall

region) (y+  5 ), the transition (5  y   30 ) and the turbulent flow (in the bulk region)

(y+  30 ).

8
 Figure 3: Computational mesh used for simulations

3.3 Governing equations

Slug flow was modelled using the Volume of Fluid (VOF) method with a High Resolution

Interface Capturing Scheme (HRIC) based on the Compressive Interface Capturing Scheme

for Arbitrary Meshes (CISCAM) introduced by [18] and enhanced by [19].

The continuity and momentum equations represented respectively by equations (1) and (2)

for two-phase flow through the flow domain are:

 u i
 0 (1)
t xi

u j u i u j P   u i u j 
       g j  F j (2)
t xi x j xi  x j xi 


where, ui and x i denote, respectively, the velocity component and the co-ordinate in the

direction i (i =1, 2 or 3), t, being the time; and through the resolution of the momentum

equation shared by the two considered fluids, P, g and F indicate, respectively, the pressure,

the gravitational acceleration and the external force per unit volume.

9
The tracking of the phase indicator function and the identification of the location and shape

of the interface between the gas–liquid phases are accomplished by solving the volume

fraction continuity equation for each phase, expressed as equation (3):

 q u i ( q )
 0 (3)
t xi

where,

and the volume fractions of all phases as shown in equation (4) sum to unity in each control

volume:


q 1
q 1 (4)

where, q is the liquid or gas phase.

The properties of the qth phase are used in the transport equations when the computational

cell is completely controlled by the qth phase. At the interface between the phases, the

mixture properties are determined based on the volume fraction weighted average, and the

density and viscosity can be expressed as:

   2 2  (1   2 ) 1 (5)

   2 2  (1   2 )1 (6)

where, the phases are represented by the subscripts 1 and 2 and if the volume fraction of the

phase 2 is known, the  and  in each cell can be determined.

In the present work, the Continuum Surface Force (CSF) model proposed by Brackbill et al.

[20] was used to model the surface tension. With this model, the addition of surface tension

to the VOF model calculation results in a source term in the momentum equation.

10
3.4 Turbulence model

In order to simulate turbulence, the standard k-ε model, Launder and Spalding [21] was used

for this study as suggested by the multiphase flow studies of Ramos-Banderas et al. [22] and

[23]. The model is described by the following elliptic equations required as closure for the

Reynolds Averaged Navier Stokes (RANS) equations:

k    t k     
uj      t u j  ui  u j    (7)
 x j  x j   k  x j   xi   x j  xi 

    t     u j   ui  u j  
uj     C1  t   C 2  (8)
 
 xj  xj   xj  k  xi   x j  xi  k

In the above equations, k is the turbulent kinetic energy; ε is the dissipation rate of k.  k ,   ,

C1 and C2 are constants whose values are 1.0, 1.3, 1.44 and 1.92 respectively, ui is the i

component of the fluid velocity u, x j is the j spatial coordinate. The fluid viscosity must be

corrected for turbulence in the Navier-Stokes equations by employing an effective viscosity

eff    t where  is the dynamic viscosity and  t is the turbulent viscosity.

The numerical solution of these sets of equations (1-8) was performed using the software

package Star-CCM+. A second order discretization scheme was used to determine the fluxes

at the control volume faces required by the VOF model.

3.5 Boundary and initial conditions:

All solid boundary walls were assumed to possess a no slip boundary condition, where v = 0

relative to the wall and the standard wall function approach based on the Launder and

Spalding [21] was used. At the flow inlet at the base of the riser, the mixture superficial

velocity, UM,, defined as the sum of liquid and gas superficial velocities (USL+U SG) is

specified. Also specified are the homogeneous volume fraction for the liquid (USL/UM) and

gas (USG/U M). The flow inlet values for turbulent kinetic energy, k , and its dissipation

11
rate,  , are estimated using the following equations proposed by Launder and Spalding [29]:

3 2 2 (9)
kin  I U in
2

 in  2kin 3 / 2 / d (10)

0.16
I (11)
Re1/ 8
Where d is the internal pipe diameter, and I the turbulence intensity for fully developed pipe

flow.

The volume fraction and density of each phase were both specified at the riser inlet as a

homogeneous mixture. It is worthy of mention that the volume fraction of gas at the riser

inlet is different from void fraction. The latter cannot be calculated analytically based on the

fact that it is a function of different operational (liquid and gas properties, flow pattern, etc)

and geometric properties (i.e. pipe diameter, pipe inclination, etc). At the flow outlet at the

top of the riser, the remaining variables are transported out of the computational domain with

zero average static pressure so that the mass flow balance is satisfied.

^
At t = 0 seconds, all velocity components are set to 0 m/s ( t  t 0 n .v  V0 ) and the initial

condition was the riser full of (static) liquid. This initial condition eases the convergence

process. In addition, an initial guess for the turbulent kinetic energy and the dissipation rate

were applied in the simulation.

A surface average monitor were located at 4.4 m, 4.489 m and 4.92 m corresponding to ECT-

plane 1, ECT-plane 2 and WMS, respectively at three stations of the riser to avoid any inlet

and outlet effects and to ensure that the slug flow is fully developed. This surface monitor

determines the void fraction of air in these sections.

The operating conditions were specified as being standard atmospheric pressure (101.3 kPa)
12
and temperature 20o C. Gravity effects are accounted for and the acceleration due to gravity to

be -9.81 m/s2 on the vertical.

3.6 Solution algorithm:

In order to numerically solve the system of governing partial differential equations,

discretization of the equations has been carried out using a Finite Volume Method (FVM)

with an algebraic segregated solver and co-located grid arrangement, as implemented in Star-

CCM+[17]. In this grid arrangement, pressure and velocity are both stored at cell centres.

Details of the discretization (FVM) can be in Versteeg and Malalasekera [24]. Since Star-

CCM+ uses a segregated solver for the VOF model, the continuity and momentum equations

need to be linked. Various techniques are reported in the literature. However, the Semi-

Implicit Method for Pressure-Linked Equations (SIMPLE) algorithm, (Patankar and Spalding

[25]), is applied as it produces a fast and convergent solution. In addition, the iterative solver

was further improved by the use of an Algebraic Multigrid (AMG) technique to yield a better

convergence rate.

All simulations in this work are performed under time dependent conditions. Under

relaxation factors of 0.3, 0.7 and 0.8 respectively, were applied on pressure, momentum and

turbulence kinetic energy parameters, as recommended by [17]. The residuals were set to 10-4

to ensure a converged solution.

3.7 Mesh independence study:

In order to identify the minimum mesh density to ensure that the solution is independent of

the mesh resolution, a mesh sensitivity analysis has been carried out in the construction and

analysis of the CFD model. In the mesh independence study, a computational domain of 1m

length was used as this length is sufficient to carry out a test on the performance of the mesh

with quite reasonably computational effort. Six 3-Dimensional meshes were investigated in

13
the present study as shown in Figure 4. The mesh sensitivity study is performed with a

t
constant ratio  1  104 and the mesh sizes of 24,000, 36,000, 54,600, 76,800, 84,000 and
x

102,600 cells. The meshes were tested with an inlet flow condition (mixture superficial

velocity, UM = USL+U SG = 0.05+0.344 = 0.394 m/s and homogeneous void fraction = U SG/UM

= 0.87). An initial condition of flow domain full of (static) liquid was used.

Since slug flow is characterized by void fraction fluctuation, one aspect that is interesting to

look at is the time trace of cross-sectional average void fraction.

In order to determine the time series of the void fraction, the following procedure similar to

that used by Hernandez-Perez [15] was performed: 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 dividing the

summation of the product of the air volume fraction and facet area by the total area of the

surface as follows:

1 1 n
A
A    Ai
A i1
(12)

Finally the value of average void fraction in the cross-sectional plane is recorded for every

time step.

The velocity of the Taylor bubble, UN is given by the relation of Nicklin et al. [3]:

U N  1.2(U SL  U SG )  0.35 gD (13)

A calculation was performed to compare the performance of these meshes. The time

calculated for the bubble to reach the measurement section (0.5 m) turned out to be 0.66

14
seconds. The plot of the time history of the void fraction for the six meshes is shown in Table

3.

From Table 3, it can be observed that meshes 5 (84,000 cells) and 6 (102,600 cells) perform

well as the time the Taylor bubble got to the measurement location is closer to the one

predicted by the theoretical expression. Here, the % error is 1.4.

The % error is evaluated as follows:

X analytical  X simulated
% error  (14)
X analytical

Where X is the time average of the variable for which the error is computed. The purpose of

this is to compare the predictions once the code has reached a steady-state.

It can be concluded that for a given flow condition, the residence time of a fluid particle in a

cell (t=x/u) decreases as the mesh density increases and that the finer the mesh is, the

narrower the error between predicted (simulated) and analytical solution becomes.

An insight into the effect of mesh density can also be obtained from the probability density

function (PDF) of void fraction that was successfully employed by Hernandez-Perez [15] for

his mesh independence studies. An examination of the plots of the PDF of void fraction

shows that when the mesh is too coarse a refinement in the mesh can have a remarkable

influence on the results, as depicted in Table 3. Therefore, it can be concluded that the mesh 5

with 84000 cells is adequate, as the change in the results produced is very small when the

number of cells is increased to 102600, and it requires less computational effort than the

102600 cells.

15
Figure 4: Cross-sectional view of different sizes of computational grid used for mesh independent study (a) 26400 cells (b)
36000 cells (c) 54,600 cells (d) 76,800 cells (e) 84,000 cells (f) 102,600 cells. Liquid and gas superficial velocities = 0.05
and 0.344 m/s, respectively. An initial condition of riser full of (static) liquid was used.

Table 3: The results obtained from the CFD mesh independence studies. Liquid and gas superficial velocities = 0.05 and
0.344 m/s, respectively. An initial condition of riser full of (static) liquid was used.

Number of Time series of void PDF of void fraction Time the Taylor
cells fraction bubble arrived the
measurement location
(seconds)
0.12
26400 0.737
0.09
PDF

0.06

0.03

0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.12
36000 0.724
0.09
PDF

0.06

0.03

0
0 0.2 0.4 0.6 0.8 1
Void fraction

16
 0.12
54600 0.696
0.09

PDF
0.06

0.03

0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.12
76800 0.671
0.09

PDF
0.06

0.03

0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.12
84000 0.669
0.09
PDF
0.06

0.03

0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.12
102600 0.669
0.09
PDF

0.06

0.03

0
0 0.2 0.4 0.6 0.8 1
Void fraction

3.8 Flow development:

A fully developed flow is defined as one when the flow pattern does not change with the

distance downstream. Flow development in the vertical riser was studied using CFD and the

results are presented and discussed. The advantage of the CFD simulation compared to the

physical experiment is the possibility to record the void fraction time series at many

measurement sections along the pipe. Also, due to physical limitations in the length of the rig,

the question that we are going to address here is whether a sufficient pipe length (often

quoted in terms of pipe diameter) had been provided so that observations taken at the end of

the pipe could be considered to be a true representation of a fully developed flow situation.

17
 Table 4: Interrogating flow development in a vertical 67 mm internal diameter and 6 m long riser. Riser initially full of
(static) liquid, and the liquid and gas flows set to liquid and gas superficial velocities of 0.05 and 0.344 m/s, respectively

Distance from Time averaged void fraction Probability density function

the mixing (PDF) of void fraction
section of the
riser (m)
0.06

0.04

PDF
1.0 (15 pipe 0.02
diameters)
0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.06

0.04

PDF
1.15 (17 pipe 0.02

diameters)
0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.06

0.04

PDF
2.0 (30 pipe 0.02
diameters)
0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.06

0.04
PDF
2.1 (31.3 pipe 0.02

diameters)
0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.06

0.04
PDF

2.8 (41.8 pipe 0.02

diameters) 0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.06

0.04
PDF

3.0 (45 pipe 0.02

diameters)
0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.06

0.04
PDF

4.0 (60 pipe

0.02
diameters)
0
0 0.2 0.4 0.6 0.8 1
Void fraction

18
 4.4 (66 pipe
diameters)

0.06

0.04

PDF
4.489 (67 0.02
pipe
diameters) 0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.06

0.04

PDF
4.92 (73 pipe 0.02

diameters) 0
0 0.2 0.4 0.6 0.8 1
Void fraction
0.06

0.04

PDF
5.5 (82 pipe
0.02
diameters)
0
0 0.2 0.4 0.6 0.8 1
Void fraction

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 flow characteristics with distance.

Table 4 shows simulation results of time varying void fraction and PDF of void fraction

derived from the eleven measurement locations at liquid and gas superficial velocities of 0.05

and 0.344 m/s, respectively. The simulations were performed within a flow domain of 6 m

long vertical pipe (the same length as the one used in the experiment) with the measurement

sections located as indicated in the table.

It can be observed from the time series of void fraction shown in Table 4 that the length of

the large bubbles (Taylor bubbles) increases with axial distance. This can be explained by the

occurrence of bubble coalescence. The PDF of the time series of void fraction at 1.0 m, just

downstream of the two-phase mixing section, shows a single peak at low void fraction with a

19
broadening tail down to higher void fraction. It also shows that the results obtained from 1.0

m are initially affected by entrance effects. This is further reinforced by the time trace of void

fraction. With the time series of void fraction showing a maximum void fraction of 0.78

while the PDF of void fraction depicting a single peak at about 0.16, void fraction with a tail

down to 0.8. The flow patterns begin to change to slug flow at a distance of about 2.8 m (42

pipe diameters) from the mixing section. At a distance of 2.8 m from the mixing section, both

the time series and PDF of void fraction have taken the shape of slug flow. Though, it

becomes more apparent at 4.0 m from the mixing section.

It is worthy of mention that at a distance of 4.0 to 5.5 m as depicted in Table 4, the PDF of

void fraction show the traditional features of slug flow; a double peak. One peak at lower

void fraction represents liquid slug whilst the one at higher void fraction, Taylor bubble. On

the other hand, the time series of void fraction also show large bubbles separated by smaller

ones. It can be concluded that between, 4.0 to 5.5 m, that flow is fully developed based on the

fact that the flow remains quite similar, i.e. not changing with distance from 4.0 to 5.5 m.

This corresponds to approximately 60 to 82 pipe diameters. It is in view of this development

that we decided to locate our experimental measuring instruments at 4.4 (66 pipe diameters),

4.489 (67 pipe diameters) and 4.92 m (73 pipe diameters) corresponding to the ECT plane1,

ECT-plane 2 and WMS.

4. Results and discussion:

The study will begin by providing a qualitative comparison between CFD simulations and

experiment based on different methods of initially introducing fluid into the riser. For the

CFD, the riser was initially full of (static) liquid, and then liquid and gas flows set to liquid

and gas superficial velocities of 0.05 and 0.344 m/s, respectively whilst for experiment, the

riser was initially full of air, and then liquid and gas flows set to same flow rates as for the

20
CFD. The number of cells used for the CFD calculation is 500,000. The results of the

comparison showed that the method of introducing the fluid into the riser ceases to be an

issue once the flow reaches steady-state, fully developed. And that the comparison between

CFD and experiment when steady-state is reached is reasonably good. Thereafter, a detailed

quantitative comparison between CFD and experiments was made based on same method of

initially introducing full (static) liquid into the riser. It is worth mentioning that only the ECT

is used here. WMS was not used here based on the fact that it has a single plane (velocity

cannot be determined) and as such cannot be used to characterize slug flow. It is worth

mentioning however, that a dual WMS can be used for such a task. The liquid and gas

superficial velocities = 0.05 and 0.344 m/s, respectively for both CFD and experiment. Here,

again the comparison is reasonably good.

4.1 Qualitative comparison between CFD and experiment:

As a starting point, the raw experimental data will be plotted in the form of time series of

void fraction, PDF of void fraction and PSD of void fraction, see Figure 5. The data is

collected at three measurement locations, ECT-plane 1, ECT-plane 2 and WMS. These

locations correspond respectively to 4.4 m, 4.489 m and 4.92 m from the two-phase flow

mixer. The data is obtained after an interval of 60 seconds.

It can be observed from the figure that as the flow reaches steady–state, the shape of the PDF

and PSD of void fraction for both the CFD and experiment are similar. Both CFD and

experimental PDF predict slug flow as the flow pattern, according to the definition of

Costigan and Whalley [26]. According to them, slug flow is a flow pattern characterised by a

PDF graph with two peaks, one at lower void fraction (liquid slug) and the other one at higher

void fraction, Taylor bubble.

21
Figure 5: Comparison between experimental data and CFD simulation results at liquid and gas superficial velocities of 0.05
and 0.344 m/s, respectively. The initial conditions are riser full of (static) liquid and riser full of air, for CFD and
experiment, respectively. Locations 4.4 m, 4.489 m and 4.92 m corresponds to ECT-plane 1, ECT-plane 2 and WM S,
respectively.

The contours of phase distribution reported in Figures 6 (a-d) and Figures 7 (a-d) for the

Taylor bubble obtained from both CFD and experiment show that the CFD results are in

better agreement with those obtained from the WMS. On the contrary, the comparison

between the CFD and ECT is poor.

It is worth mentioning that it is difficult to measure experimentally the velocity for these

conditions due to the presence of the bubbles and the highly turbulent flow field. However,

this has been successfully modelled and is represented in Figure 8, by means of velocity

22
vectors. From the figure, three regions can be observed from the velocity vectors: the Taylor

bubble, falling film and the wake region. Interestingly, the Taylor bubble can be seen moving

vertically upwards whilst the liquid film on the other hand is moving downwards. A similar

observation was reported by [4] and [5]. The falling film with some entrained bubbles drop

into the wake region and a vortex region is created. Furthermore, the liquid film and some of

the entrained bubbles are subsequently carried upwards by the incoming gas phase. This

behaviour is similar to that observed by Fernandes et al. [4] and [27] who worked on slug

flow in a vertical pipe using air–water as the model fluid. They claimed that the bubbles in

the liquid slug rise due to entrainment in the wake of the Taylor bubble and that much of this

entrained gas is swept around a vortex in the Taylor bubble wake and may coalesce with the

trailing Taylor bubble.

Figure 6: Comparison of contours of phase distribution at liquid and gas superficial velocities of 0.05 and 0.344 m/s,
respectively for between (a) CFD and (b) WM S and for (c) CFD and (d) ECT. For the CFD and WM S comparison, the
liquid and gas phases are represented by red and blue colours, respectively. On the contrary, blue represents gas phase for t he
ECT. The initial conditions are riser full of (static) liquid and riser full of air, for CFD and experiment, respectively.

23
Figure 7: Contours of phase distribution (cross-sectional void fraction of gas) for the Taylor bubble obtained at liquid and
gas superficial velocities of 0.05 and 0.344 m/s, respectively from (a) CFD and (b) WM S and for (c) CFD and (d) ECT. For
the CFD and WM S comparison, the liquid and gas phases are represented by red and blue colours, respectively. On the
contrary, blue represents gas phase for the ECT. The initial conditions are riser full of (static) liquid and riser full of air, for
CFD and experiment, respectively.

Figure 8: Velocity field around the (a) Taylor bubble (b) Wake region of the Taylor bubble at liquid and gas superficial
velocities of 0.05 and 0.344 m/s, respectively obtained from CFD. The initial conditions are riser full of (static) liquid

24
4.2 Quantitative comparison between CFD and experiment:

The experimental data was obtained over an interval of 60 seconds whilst for the CFD, 16

seconds. Readings were taken when the Taylor bubble arrived at the measurement sections.

Figure 9 shows a typical plot of a large trailing Taylor bubble (start–up) and leading train of

Taylor bubbles (steady–state).

Figure 9: A plot showing a combination of a large trailing Taylor bubble (start –up) and leading train of smaller Taylor
bubbles (steady-state) at liquid and gas superficial velocities = 0.05 and 0.344 m/s, respectively). The initial condition for
both CFD and experiment is riser full of (static) liquid.

A detailed methodology for the determination of these parameters can be found in

Abdulkadir et al. [28]. A comparison will finally be made between CFD and experiment

based on static pressure. The errors between experimental measurement and predictions are

listed in Tables 5 and 6. The error % is evaluated as follows:

X exp erimental  X simulated

Error   100 (14)
X exp erimental

Where X is the time average of the variable for which the error is computed. The purpose is

to compare the predictions once the code has reached a steady-state.

25
Table 5a: Comparison between the CFD and experiments for the large trailing Taylor bubble (Start -up) at liquid and gas
superficial velocities of 0.05 and 0.344 m/s, respectively). The initial condition for both CFD and experiment is riser full of
(static) liquid.

CFD EXPERIMENT %
ERROR
ECT - PLANE 1(4.4 m)
Void fraction 1 1
0.8

Void fraction
0.8
0.6 0.6
Parameters 0.4
0.4
0.2 0.2

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Time (seconds)
Time (seconds)
Velocity of the 0.89 0.84 5.95
back of the
Taylor bubble
(m/s)
Velocity of the 0.89 0.84 5.95
front of the
Taylor bubble
(m/s)
Length of 6.68 6.38 4.70
Taylor bubble
(m)
Void fraction 0.8 0.77 3.90
in the Taylor
bubble
Liquid film 3.54 4.10 13.66
thickness
(mm)

Table 5b: Comparison between the CFD and experiments for the large trailing Taylor bubble (Start-up) at liquid and gas
superficial velocities of 0.05 and 0.344 m/s, respectively). The initial condition for both CFD and experiment is riser full of
(static) liquid.

CFD EXPERIMENT %
error
ECT –PLANE 2 (4.489 m)
Parameters 1 1
0.8
Void fraction

0.8
Void fraction

0.6 0.6

0.4 0.4
0.2
0.2
0
0
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8 Time (seconds)
Time (seconds)
Velocity 0.89 0.82 8.54
26
 of the back
of the
Taylor
bubble
(m/s)
Velocity 0.89 0.82 8.54
of the
front of
the Taylor
bubble
(m/s)
Length of 6.68 6.23 7.22
slug unit
(m)
Void 0.80 0.76 5.26
fraction in
the Taylor
bubble
Liquid 3.54 4.30 21.47
film
thickness
(mm)

Tables 5a and 5b presents a summary of the quantitative comparison between CFD and

experiment in terms of different characteristics of slug flow in the riser. It can be concluded

that the best degree of agreement between CFD and experiments in terms of slug flow

characterization for the large trailing Taylor bubble is the void fraction in the Taylor bubble

while the least is the liquid film thickness.

The velocity of the back and front of the Taylor bubble from the CFD compares very well

with experiment. The length of the Taylor bubble for the CFD also compares well with the

experiment. The void fraction in the Taylor bubble for the CFD and experiment are also

compared, for this case the CFD prediction is quite accurate. The liquid film thickness was

also determined from the CFD and experiment. For the CFD, the liquid film thickness

obtained is 3.54 mm while 4.10 mm for the experiment which means CFD under predicted

the liquid film thickness by 13.66 %.

27
As the large Taylor bubble reaches the ECT-plane 2 (Table 5b), a similar comparison of the

slug flow characterisation was also carried out. The velocity of the large trailing Taylor

bubble from CFD also compares well with experiment. As expected, the length of the Taylor

bubble also dropped for the experiment but remains unchanged for the CFD. The values of

the void fraction in the Taylor bubble and liquid film thickness for the experiment changed

from (0.77 and 4.10 mm) to (0.76 and 4.30 mm) but remain unchanged for the CFD.

For the leading Taylor bubble (Table 6b), it can be concluded that the best degree of

agreement in terms of comparison between CFD and experiment is the length of the Taylor

bubble while the least, void fraction in the liquid slug.

Table 6a: Comparison between the CFD and experiments for the leading Taylor bubble (steady–state)/ (fully developed) at
liquid and gas superficial velocities of 0.05 and 0.344 m/s, respectively). The initial condition for both CFD and experiment
is riser full of (static) liquid.

CFD EXPERIMENT
ECT - PLANE 1
1
Void fraction

0.8
0.6
0.4
0.2
0
7.8 8.8 9.8 10.8 11.8 12.8
Time (seconds)
ECT –PLANE 2
1
Void fraction

0.8
0.6
0.4
0.2
0
7.8 8.8 9.8 10.8 11.8 12.8
Time (seconds)

Table 6b: Comparison between the CFD and experiments for the leading Taylor bubble (steady–state)/ (Fully developed) at
liquid and gas superficial velocities of 0.05 and 0.344 m/s, respectively). The initial condition for both CFD and experiment
is riser full of (static) liquid.
28
Table 6b: Comparison between the CFD and experiments for the leading Taylor bubble (Fully developed)

Slug (CFD) Experiment % error

characteristics

Parameter Plane 1 Plane 2 Plane 1 Plane 2 Plane 1 Plane 2

Void fraction 0.14 0.13 0.17 0.16 17.6 18.75

in liquid slug

Void fraction 0.60 0.56 0.65 0.62 7.69 9.68

in Taylor
bubble

Frequency 1.8 2.40 2.0 2.0 9.6 20

Translational 1.48 1.59 6.9

velocity of
the Taylor
bubble

Length of the 0.82 0.80 2.5

slug unit (m)

Length of the 0.5 0.49 2.04

Taylor bubble
(m)

Length of the 0.32 0.31 3.23

liquid slug
(m)

Peak of time 0.77 0.74 0.76 0.78 1.3 5.13

series of void
fraction

The maximum height of the peak of the void fraction from the time trace of void fraction and

slug frequency for the CFD compares well with those from experiment. The time of passage

of the Taylor bubble from ECT-plane 1 to 2 based on CFD and an experiment is 0.1 seconds.

Both CFD and experiment predict the flow pattern as slug flow, same flow pattern as for

plane 1. However, the appearance of slug flow according to Table 6a is more obvious than for
29
plane 1. This may be due to the fact that at 4.489 m from the mixing section (plane 2), the

flow is more fully developed. A 20 % error is observed from the comparison between slug

frequency obtained from CFD and experiment. This may be due to the fact that the

experimental measurements were taken over 60 seconds whilst for the CFD 16 seconds.

The translational velocity of the leading Taylor bubble has been calculated for the CFD as

well as for the experimental study as shown in Figure 10. The figure illustrates the procedure

to calculate the translational velocity of the Taylor bubble for both the CFD and experiment.

The results show that translational velocity of the Taylor bubble for the CFD compares well

with the experiment.

The lengths of both the liquid slug, Taylor bubble and slug unit are also obtained from CFD

which all compared well with experiment. A comparison between the CFD simulation and

the experiments is also made based on the void fractions in both the liquid slug and the

Taylor bubble. The values obtained are reasonably good as shown in Table 6b.

Figure 10: Time delay of a Taylor bubble passing through two different measuring locations along the pipe. The liquid and
gas superficial velocities are 0.05 and 0.344 m/s, respectively (a) CFD and (b) Experiment. The initial condition for both
CFD and experiment is riser full of (static) liquid. VTB represents the structure velocity of the Taylor bubble.

30
A comparison is also made between experiment and CFD based on static pressure. The value

obtained from experiment is 3.42  10 4 Pa whilst for the CFD as shown on the pressure

contour plot (Figure 11) is 3.37  10 4 Pa. The simulation under predicts the experiment by 1.5

%. The value obtained from experiment was evaluated as follows:

PStatic   m gh (15)

Where  m is the mixture density and is obtained based on the knowledge of the cross-

sectional void fraction and h is the height of the riser.

Figure 11: Static pressure contour plot for liquid and gas superficial velocities of 0.05 and 0.344 m/s, respectively obtained
from CFD. The initial condition for both CFD and experiment is riser full of (static) liquid.
31
 Conclusions:

A comparison between the results of slug flow characterization obtained from CFD

simulation and experiments has been successfully carried out for a 67 mm internal diameter

vertical riser with air and silicone oil as the model fluids and the following conclusions can

be drawn:

1) The qualitative comparison between CFD and experiment based on different methods of

introducing fluid into the riser liquid and gas superficial velocities of 0.05 and 0.344 m/s

respectively did not yield any significant difference once the flow reaches steady-state. At

steady-state, both the CFD and experiment predict similar behaviours.

2) The slug flow pattern can be considered fully developed at 4.0 m (60 pipe diameters).

3) A reasonably good agreement between CFD and experiment was obtained. CFD simulation

can be used to characterize slug flow parameters with a good level of confidence. However,

further parametric studies are required to close some of the gaps between CFD and

experimental results.

4) This work confirms the results reported in the literature for the characteristics of slug flow.

5) The best degree of agreement in terms of the slug flow characterization for the large trailing

Taylor bubble between CFD and experiment is the void fraction in the Taylor bubble whilst

the least is the liquid film thickness. On the other hand, the length of the Taylor bubble and

the void fraction in the liquid slug, respectively, represent the best and the least degree of

agreement for the leading Taylor bubble between CFD and experiment.

6) The comparison between CFD and experiment based on static pressure is qualitatively good.

32
Nomenclature:
A Area [m2 ]

F Frequency [H

VTB Structure velocity [m/s]

LSU Length of the slug unit [m]

LS Length of the liquid slug [m]

LTB Taylor bubble length [m]

g Gravitational acceleration [ m / s 2 ]

k Kinetic energy of turbulence [ m 2 / s 2 ]

n number of phases [-]

t Time [ s ]

u Velocity [ m / s ]

 Dynamic viscosity [ kg / m.s ]

 Material density [ kg / m 3 ]

 Surface tension [ N / m ]

i, j Space directions

q Phase index

ACKNOWLEDGEMENTS
M. Abdulkadir would like to express sincere appreciation to the Nigerian government
through the Petroleum Technology Development Fund (PTDF) for providing the funding for
his doctoral studies.
This work has been undertaken within the Joint Project on Transient Multiphase Flows and
Flow Assurance, sponsored by the UK Engineering and Physical Sciences Research Council
(EPSRC); Advantica; BP Exploration; CD-adapco; Chevron; ConocoPhillips; ENI;
ExxonMobil; FEESA; IFP; Institutt for Energiteknikk; Norsk Hydro; PDVSA (INTERVEP);
Petrobras; PETRONAS; Scandpower PT; Shell; SINTEF; Statoil and TOTAL. The Authors
wish to express their sincere gratitude for their supports.

33
References:

[1] A. Clarke, R.I. Issa, A numerical model of slug flow in vertical tubes, Comp. and Fluids,
26 (1997) 395 - 415.

[2] Z.S. Mao, A.E. Dukler, 1985. Brief communication: Rise velocity of a Taylor bubble in a
train of such bubbles in a flowing liquid, Chemical Engineering Science, 40, 2158 – 2160.
[3] D.J. Nicklin, J.O. Wilkes, J.F. Davidson, 1992. Two-phase flow in vertical tubes,
Transaction of Institution of Chemical Engineers, 40, 61 – 68.
[4] R.C. Fernandes, R. Semiat, A.E. Dukler, 1983. Hydrodynamics model for gas-liquid slug
flow in vertical tubes, AIChE Journal, 29, 981- 989.
[5] Z.S. Mao, A.E. Dukler, 1991. The motion of Taylor bubbles in vertical tubes. II.
Experimental data and simulations for laminar and turbulent flow, Chemical Engineering
Science, 46, 2055 - 2064.
[6] J. Fabre, A. Line, 1992. Modelling of two-phase slug flow, Annual Review of Fluid
Mechanics, 24, 21-46.
[7] B.J. Azzopardi, L.A. Abdulkareem, S. Sharaf, S, M. Abdulkadir, V. Hernandez-Perez, A.
Ijioma, 2010. Using tomography to interrogate gas-liquid flow, In: 28th UIT Heat Transfer
Congress, Brescia, Italy, 21-23 June.
[8] M. Abdulkadir, V. Hernandez-Perez, S. Sharaf, I.S. Lowndes, B.J. Azzopardi, 2010.
Experimental investigation of phase distributions of an air-silicone oil flow in a vertical
pipe. World Academy of Science Engineering and Technology (WASET), 61, 52–59.
[9] M. Abdulkadir, D. Zhao, S. Sharaf, L. Abdulkareem, I.S. Lowndes, B.J. Azzopardi, 2011.
Interrogating the effect of 90o bends on air-silicone oil flows using advanced instrumentation,
Chemical Engineering Science, 66, 2453 - 2467.
[10] V. Hernandez-Perez, M. Abdulkadir, B.J. Azzopardi, 2011. Grid generation issues in the
CFD modelling of two-phase flow in a pipe, The Journal of Computational Multiphase Flow,
3, 13 - 26.
[11] S.M. Huang, 1995. Impedance sensors-dielectric systems, In R. A. Williams, and M. S.
Beck (Eds.), Process Tomography, Cornwall: Butterworth-Heinemann Ltd.
[12] K. Zhu, S. Madhusudana Rao, C. Wang, S. Sundaresan, 2003. Electrical capacitance
tomography measurements on vertical and inclined pneumatic conveying of granular solids,
Chemical Engineering Science, 58, 4225 – 4245.
[13] M.J. da Silva, S. Thiele, L. Abdulkareem, B.J. Azzopardi, U. Hampel, 2010. High-
resolution gas-oil two-phase flow visualization with a capacitance wire-mesh sensor, Flow
Measurement and Instrumentation, 21, 191 - 197.
[14] C.W. Hirt, B.D. Nichols, 1981. Volume of Fluid (VOF) Method for the Dynamics of
Free Boundaries, Journal of Computational Physics, 39, 201.
[15] V. Hernandez-Perez, 2008. Gas-liquid two-phase flow in inclined pipes, PhD thesis,
University of Nottingham.
[16] S. Lo, D. Zhang, 2009. Modelling of break-up and coalescence in bubbly two-phase
flows, The Journal of Computational Multiphase Flow, 1, 23 – 38.
[17] STAR-CD, 2009. Version 4.10, STAR-CCM+ Documentation, CD-adapco.
34
[18] O. Ubbink, 1997. Numerical prediction of two fluid systems with sharp interfaces, PhD
thesis, University of London.
[19] S. Muzaferija, M. Peric, 1999. Computation of free surface flows using interface-
tracking and interface-capturing methods, Chap.2 in O. Mahrenholtz and M. Markiewicz
(eds.), Nonlinear Water Wave Interaction, Computational Mechanics Publication, WIT Press,
Southampton.
[20] J.U. Brackbill, D.B. Kothe, C. Zemach, 1992. A continuum method for modelling
surface tension, Journal of Computational Physics, 100, 335 - 354.
[21] B. Launder, D. Spalding, 1974. The numerical computation of turbulent flows,
Computational Methods in App. Mech. and Eng., 3, 269 - 289.
[22] A. Ramos-Banderas, R.D. Morales, R. Sanchez-Perez, L. Garcia-Demedices, G. Solorio-
Diaz, 2005. Dynamics of two-phase downwards flow in submerged entry nozzles and its
influence on the two-phase flow in the mold, International Journal of Multiphase Flow, 31,
643 - 665.
[23] M. Cook, M. Behnia, 2001. Bubble motion during inclined intermittent flow,
International Journal of Multiphase Flow, 22, 543 - 551.
[24] H.K. Versteeg, W. Malalasekera, W., 2007. An Introduction to Computational Fluid
Dynamics: the Finite Volume Method, 2 nd ed. Pearson Educational Limited.
[25] S.V. Patankar, D.B. Spalding, A calculation procedure for heat, mass and momentum
transfer in three dimensional parabolic flows, Int. J. of Heat and Mass Transfer 15 (1972)
1787.
[26] G. Costigan, P.B. Whalley, 1997. Slug flow regime identification from dynamic void
fraction measurements in vertical air-water flows, International Journal of Multiphase Flow,
23, 263 – 282.
[27] L. Shemer, A. Gulitski, D. Barnea, 2004. Velocity field in the Taylor bubble wake
measurements in pipes of various diameters, 24nd European Two-phase Flow Group
Meeting, Geneva.
[28] Abdulkadir, M., Hernandez–Perez, V., Lowndes, I.S., Azzopardi, B. J., Dzomeku, S.,
2014. Experimental study of the hydrodynamic behaviour of slug flow in a vertical riser.
Chemical Engineering Science, 106, 60–75.

35
 Figure captions:

Figure 1 3-D geometry of the computational flow domain showing the location of the
recording sections that correspond to the locations of the experimental measurement
transducers.

Figure 2 Air-silicone oil mixing section

Figure 3 Computational mesh used for simulations

Figure 4: Cross-sectional view of different sizes of computational grid used for mesh
independent study (a) 26400 cells (b) 36000 cells (c) 54,600 cells (d) 76,800 cells (e) 84,000
cells (f) 102,600 cells. Liquid and gas superficial velocities = 0.05 and 0.344 m/s,
respectively. An initial condition of riser full of (static) liquid was used.

Figure 5: Comparison between experimental data and CFD simulation results at liquid and
gas superficial velocities of 0.05 and 0.344 m/s, respectively. The initial conditions are riser
full of (static) liquid and riser full of air, for CFD and experiment, respectively. The time
difference observed in the CFD is due to the different times recorded for the Taylor bubble to
arrive the measurement locations. Locations 4.4 m, 4.489 m and 4.92 m corresponds to ECT-
plane 1, ECT-plane 2 and WMS, respectively.

Figure 6: Comparison of contours of phase distribution at liquid and gas superficial

velocities of 0.05 and 0.344 m/s, respectively for between (a) CFD and (b) WMS and for (c)
CFD and (d) ECT. For the CFD and WMS comparison, the liquid and gas phases are
represented by red and blue colours, respectively. On the contrary, blue represents gas phase
for the ECT. The initial conditions are riser full of (static) liquid and riser full of air, for CFD
and experiment, respectively.

Figure 7: Contours of phase distribution (cross-sectional void fraction of gas) for the Taylor
bubble obtained at liquid and gas superficial velocities of 0.05 and 0.344 m/s, respectively
from (a) CFD and (b) WMS and for (c) CFD and (d) ECT. For the CFD and WMS
comparison, the liquid and gas phases are represented by red and blue colours, respectively.
On the contrary, blue represents gas phase for the ECT. The initial conditions are riser full of
(static) liquid and riser full of air, for CFD and experiment, respectively.

Figure 8: Velocity field around the (a) Taylor bubble (b) Wake region of the Taylor bubble
at liquid and gas superficial velocities of 0.05 and 0.344 m/s, respectively. The initial
conditions are riser full of (static) liquid and riser full of air, for CFD and experiment,
respectively.

Figure 9: A plot showing a combination of a large trailing Taylor bubble (start–up) and
leading train of smaller Taylor bubbles (steady-state) at liquid and gas superficial velocities =
0.05 and 0.344 m/s, respectively). The initial condition for both CFD and experiment is riser
full of (static) liquid.
36
Figure 10: Time delay of a Taylor bubble passing through two different measuring locations
along the pipe. The liquid and gas superficial velocities are 0.05 and 0.344 m/s, respectively
(a) CFD and (b) Experiment. The initial condition for both CFD and experiment is riser full
of (static) liquid.

Figure 11: Static pressure contour plot for liquid and gas superficial velocities of 0.05 and
0.344 m/s, respectively obtained from CFD. The initial condition for both CFD and
experiment is riser full of (static) liquid.

Table captions:

Table 1 Table of flowchart for experimental measurement used to obtain the parametric
characterisation of the slug flow regime

Table 2 Properties of the fluids

Table 3: The results obtained from the CFD mesh independence studies. Liquid and gas
superficial velocities = 0.05 and 0.344 m/s, respectively. An initial condition of riser full of
(static) liquid was used.

Table 4: Interrogating flow development in a vertical 67 mm internal diameter and 6 m long

riser. Riser initially full of (static) liquid, and the liquid and gas flows set to liquid and gas
superficial velocities of 0.05 and 0.344 m/s, respectively

Table 5a: Comparison between the CFD and experiments for the large trailing Taylor bubble
(Start-up) at liquid and gas superficial velocities of 0.05 and 0.344 m/s, respectively). The
initial condition for both CFD and experiment is riser full of (static) liquid.

Table 5b: Comparison between the CFD and experiments for the large trailing Taylor bubble
(Start-up) at liquid and gas superficial velocities of 0.05 and 0.344 m/s, respectively). The
initial condition for both CFD and experiment is riser full of (static) liquid.

Table 6a: Comparison between the CFD and experiments for the leading Taylor bubble
(steady–state)/ (fully developed) at liquid and gas superficial velocities of 0.05 and 0.344 m/s,
respectively). The initial condition for both CFD and experiment is riser full of (static) liquid.

Table 6b: Comparison between the CFD and experiments for the leading Taylor bubble
(steady–state)/ (Fully developed) at liquid and gas superficial velocities of 0.05 and 0.344
m/s, respectively). The initial condition for both CFD and experiment is riser full of (static)
liquid.

37