Ocean Engineering 115 (2016) 135148

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Numerical modelling of two-phase oilwater ow patterns in a subsea

pipeline
Hassan Pouraria a, Jung Kwan Seo b,n, Jeom Kee Paik a,b,c
a
Department of Naval Architecture and Ocean Engineering, Pusan National University, Busan 609-735, Republic of Korea
b
The Korea Ship and Offshore Research Institute (The Lloyd's Register Foundation Research Centre of Excellence), Pusan National University, Busan 609-735,
Republic of Korea
c
Department of Mechanical Engineering, University College London, Torrington Place, London WC1E 7JE, UK

art ic l e i nf o a b s t r a c t

Article history: The concurrent ow of oil and water in pipelines is a common occurrence in offshore oil production
Received 28 January 2015 systems. The internal structure of the oilwater ow, known as the ow pattern, and the distribution of
Accepted 3 February 2016 water have a great inuence on the design of the pipeline. However, due to the complex nature of oil
Available online 20 February 2016
water ows, predicting ow patterns under different operating conditions is a challenging task. The
Keywords: present study is an attempt toward using a computational uid dynamics model for predicting the ow
Two-phase ow patterns under different working conditions. To this end, an EulerianEulerian model along with
CFD appropriate closure laws was employed to model two-phase oilwater ows through a horizontal pipe.
Flow pattern The standard kepsilon model was adopted to account for the turbulence effects. Furthermore, a brief
Phase distribution
description of the main interfacial forces namely the drag, lift, and turbulent dispersion forces has been
Water wetting
presented. The numerical model was used to predict the ow patterns under different operating con-
Subsea pipeline
ditions ranging from low to high velocities and water cuts, respectively. A comparison among the
obtained numerical results and published experimental data showed reasonable agreement.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction
The ow of two immiscible liquids is encountered in a diverse
range of processes, such as those of the petroleum industry. The
simultaneous ow of oil and water in pipelines is a common
occurrence in offshore oil production systems. In the early stages
of a wells lifetime the amount of water is negligible. However, as
the well ages the water production increases. Furthermore, water
injection for enhanced oil recovery is commonly used to maintain
the reservoir pressure. In contrast to the early days of the offhore
oil industry, when the amount of produced water was negligible,
these days many wells are mature and are producing large
amounts of water. From an economical point of view, operation of
a well might be reasonable even for water cuts as high as 90%
(Elseth, 2001; Kumara et al., 2009). The internal structure of the
oilwater ow, known as the ow pattern, and the distribution of
water have a great inuence on the design of the pipeline. The
distribution of oil and water in the pipeline signicantly affects the
corrosion rate, the pressure drop, wax deposition, etc. An effective
n
Corresponding author. Tel.: 82 51 510 2415.
E-mail address: seojk@pusan.ac.kr (J.K. Seo).
design can be made only if the ow pattern and the phase dis-
tributions under different conditions are known. (Xu, 2007).
Liquidliquid ows in a horizontal pipeline can be classied
into two major groups and several sub-groups, based on the
interface structure. At relatively low velocities two immiscible
liquids are separated by a clearly dened interface. This ow
regime is referred to as a stratied ow pattern. However, at
relatively high velocities there is no clear interface and one uid is
in the form of drops in the continuum of the other. This ow
pattern is usually called a dispersed ow. At intermediate velo-
cities a combination of the dispersed and stratied ow patterns is
observed, where both phases retain their continuity at the top and
bottom of the pipe, with a dispersed region present in the middle
of the pipe cross section. Apart from these, formation of slug ow
and annular ow was reported by several investigators. Experi-
mental studies showed that by increasing the oil viscosity the
extent of core annular ow increases. The transition boundaries
among these ow patterns depend on many factors. A great deal of
effort has gone into studying ow patterns under different con-
ditions. Various experimental studies have been carried out to get
reliable ow pattern maps through which we can identify the ow
regime inside the pipe. However, because of the diversity of the oil
properties, the available ow pattern maps lack a general agree-
ment. Previous experimental studies show that the observed ow

http://dx.doi.org/10.1016/j.oceaneng.2016.02.007
0029-8018/& 2016 Elsevier Ltd. All rights reserved.
136 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148

Fig. 1. A schematic of the test pipe (a) and the mixing unit (b).

patterns and the phase distribution critically depend on the uid

Table 1
properties, mixture velocity, inlet water cut,and the geometry
Description of the properties of the
(diameter and inclination angle) as well as wetting property of uids and the geometry used in the
pipe (Arirachakaran et al., 1989, Angeli and Hewitt, 2000a; Lovick experiment (Elseth, 2001)
and Angeli, 2004a; Xu, 2007; Kumara et al., 2009, Strazza et al.,
Experimental condition
2011, Yusuf et al., 2012, Hanazadeh et al., 2015).
As mentioned earlier, the prediction of the ow pattern and Pipe diameter 0.0563 m
phase distribution is essential for the proper design of the wells Pipe roughness 0.00001 m
and the pipeline. Prior knowledge of the type of ow pattern is Pipe length 15 m
Oil density 790 kg/m3
needed to choose an appropriate method for predicting the pres-
Oil viscosity 0.00164 Pa s
sure drop (Xu, 2007). In addition, predicting phase distribution is Water density 1000 kg/m3
essential for wax deposition management in subsea pipelines, Water viscosity 0.00102 Pa s
because wax deposition is a ow pattern-dependent phenomenon Interfacial tension 0.043 N/m

(Matzain et al., 2002; Sarica and Panacharoensawad, 2012). Fur-

thermore, predicting the distribution of water inside the pipe is of
the utmost importance when modelling sweet corrosion. Corro-
sion problems are associated with the continuous water phase
being in contact with the wall at the bottom of the pipe. Hence, it
is crucial to predict whether a free water layer exists at the bottom
of the pipe (water wetting) or the oil is a continuous phase (oil
wetting) containing water drops. The incorrect prediction of water
distribution leads to signicant mistakes in predicting the corro-
sion rate. Furthermore, it may lead to use of the wrong type of
inhibitor, a larger amount of inhibitor, or the use of corrosion
resistant material, thereby, increasing the operational and capital
costs (Nyborge, 2005; Nesic, 2007; Cai et al., 2012).
The computational uid dynamics (CFD) technique is a pow-
erful tool for simulating ow-eld characteristics in a multiphase
pipeline. Using CFD we can gain a deeper insight into the under-
lying physics and thus foster an understanding of the different
phenomena. In addition, CFD can provide phase distribution and
velocity proles with high resolution, which can signicantly
reduce uncertainty at the design stage. However, extensive eva-
luation with experimental results is essential before using the
model predictions for scale-up and optimisation of facilities.
Thus far, few numerical studies have been reported on pre-
dicting the hold-up distribution in oilwater ow. Parvini et al.
(2010) used the EulerianEulerian approach to model dispersed oil
in water ow in vertical pipes. Their results suggested that the
interfacial lift force is more important than the turbulent disper-
sion and virtual mass forces. Hamad et al. (2013) also investigated
dispersed oilwater ow in vertical pipes and there was a good
agreement with the experimental data. Recently, Burlutsky and
Turangan (2015) used the EulerianLagrangian approach to model
dispersed oilwater ow in vertical pipes. Their numerical results
 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148 137

Fig. 2. Schematic diagram of the geometry used and a section of the three-dimensional grid.

also indicated that lift force has a dominant effect on distribution results predicting the ow pattern and the distribution of the
of the dispersed phase. water volume fraction were compared with the corresponding
Due to the inuence of gravity on horizontal pipelines, pre- experimental data.
dicting the ow pattern and phase distribution is even more
complex than that of vertical pipes. Limited numerical studies
have been carried out on modelling oilwater ows in horizontal 2. Numerical model
pipes, with prior knowledge of the type of ow pattern. Walvekar
et al. (2009) used the EulerianEulerian model to study the oil in 2.1. EulerianEulerian approach
water dispersion through horizontal pipes. The model was able to
predict the water volume fraction distribution at high velocities, In the EulerEuler approach, each phase is assumed to coexist
but it failed to predict the distribution at low velocities where the at every point in space in the form of interpenetrating continua. It
effect of gravity becomes important. Monzon (2006) carried out a solves all the phases present and coupling between the phases is
numerical simulation using Fluent 6.2. He used the Eulerian obtained through the pressure and interphase exchange coef-
Eulerian method for a dispersed oilwater ow in horizontal pipes cients (Guha et al., 2008). For each phase k, the conservation
and the numerical results of the distribution of phase fraction equation is written as a function of the volume fraction of the
agreed fairly well with the experiment. Gao et al. (2003) used the phase k. In this method, both the phases can be averaged over a
volume of uid (VOF) model to investigate the stratied oilwater xed volume. The volume fraction of each phase is calculated as
ow in a horizontal pipe. The results agreed well with the follows:
experimental data. However, the application of such interface Z
1
tracking methods is limited to ows with sharp interfaces that do X r U dVr; 1
V
not develop to the dispersed ow pattern. Generally, when mod-
elling the oilwater ow there is a lack of prior knowledge of the where V is the averaging volume and X(r) is a phase indicator
type of ow pattern. As a matter of fact the type of ow pattern function.
itself is one of the unknowns that should be predicted before
design. Hence, a more general model is required to predict the 2.2. Governing equations
ow pattern and phase volume distribution. Two-uid models are
capable of simulating all types of two-phase ows, ranging from The continuity equation for phase k without mass transfer
those with large interfacial lengths, such as stratied ows, to between phases and momentum equation are written as
dispersed ows with very small interfacial length scales. These  
: k k v k 0; 2
models have been successfully applied to simulation of stratied
and dispersed gasliquid ows (Yao et al., 2004; Prosperetti and  
!! !
U k k v k v k  k P U k k k g
Tryggvason, 2007). The capability of the two-uid model to
  ! ! ! 
identify different ow patterns in liquidliquid ow has also been ! !
K pk v p  v k F lift; F vm;k F td;k ; 3
reported (Sathe et al., 2010).

In the present study, numerical modelling of oilwater ow where k is the strainstress tensor of the phase k.
through a horizontal pipeline was carried out using the Eulerian    
k k k ! !T !
2
Eulerian approach. The two-uid and standard kepsilon models v  v k k k  k U vk I; 4
3
were used to predict the ow pattern and water distribution under
comparatively wide operating conditions, ranging from low to where mk and k are the shear and bulk viscosity of phase k,
high velocities and low to high water cuts. Furthermore, numerical respectively.
138 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148

Fig. 3. Distribution of the water phase fraction across the pipe for an input water Fig. 4. Distribution of the water phase fraction across the pipe for an input water
cut of 10% at a mixture velocity of 1 m/s. (a) Vertical distribution of the water cut of 20% at a mixture velocity of 1 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction. of the water volume fraction.

! !
Phase k designates both phases. Kpk denotes the interface Re k j v p  v k jdp
9
momentum exchange coefcient given by mk

k p p f The lift force is computed as follows, due to the moving of

K pk ; 5 droplets in the shear ow. This force is perpendicular to the
p
direction of the droplet's motion:
where p is the particulate relaxation time which is calculated as !    
! ! !
follows: F lif t;k  C L k p v k  v p   v k : 10

p d2p For a spherical bubble or droplet, in a concurrent ow, this

p :6
18k force acts towards the pipe wall. However, the direction of the lift
force may change due to the substantial deformation of the bubble
The drag function, f, was quantied using the SchillerNau- or droplet. In the present study the lift law introduced by
mann model, where C d is given as follows: Tomiyama et al. (2002) was used to specify the lift coefcient. The
main feature of this model is its capability of predicting the sign of
C D Re
f ; 7 the lift coefcient due to the dispersed droplet distortion. The
24
modied version of Tomiyama's lift model, proposed by Frank and
8 Shi (2004) is as follows:
< 241 0:15Re0:687
Cd for Re 41000 8     
Re 8 > 0:288 tanh 0:121Rep ; f EO0 EO0 r 4
: C d 0:44 Re r1000 < min
 0
f EO 4 o EO0 r 10 11
>
:
where the relative Reynolds number is as follows:  0:27 10 o EO0
 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148 139

Fig. 5. Distribution of the water phase fraction across the pipe for an input water
cut of 50% at a mixture velocity of 1 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction. Fig. 6. Distribution of the water phase fraction across the pipe for an input water
cut of 80% at a mixture velocity of 1 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of water volume fraction.
where
  model (Lopez De Bertodano, 1998; Moraga et al., 2003).
f EO0 0:00105EO03  0:0159EO02  0:0204EO0 0:474 12
where EO0 is the modied Eotvos number based on the deformable F td;k  F td;p  C TD k kk p ; 17
bubble or droplet, dh. where k is the density of the continuous phase, kk is the turbulent
 
g k  p dh
2 kinetic energy in the continuous phase and p is the gradient of
0
EO ; 13 dispersed phase volume fraction. In the present study the default
value of 1 was used for the turbulent dispersion coefcient (C TD ).
 13 The solution of the two-uid model described above requires
2
dh dp 1 0:163EO0:757 ; 14 the prior determination of the drop size of the secondary phase.
  The drop size is one of the critical parameters affecting the CFD
g k  p d p
2
solution. Mechanistically, the maximum diameter of the droplet is
EO ; 15 determined as a result of the balance of turbulent stresses that try

to break a drop and the interfacial force that tends to restore the
where is surface tension, g is gravity and dp is droplet diameter.
original shape (Sathe et al., 2010; Cai et al., 2012). In this study, the
The virtual mass force is as follows:
maximum diameter of the dispersed phase is calculated according
! ! ! !
dk v k dp v p to the equation proposed by Brauner (2001), as follows:
F vm;k C vm;k k k  : 16
dt dt !  0:6    0:4  
k U 2k D m p 0:6
dmax 2:22D f 18
Due to the small difference between the density of oil and k 1  p 1  p
water, the virtual mass force is neglected in this study. The tur-
bulent dispersion force is computed using Lopez de Bertodano's where D is the internal diameter of the pipe and p denotes the
140 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148
Fig. 7. Distribution of the water phase fraction across the pipe for an input water
cut of 10% at a mixture velocity of 1.5 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement (b) contours
of the water volume fraction.
volume fraction of the dispersed phase at the inlet. Furthermore, k,
p and m indicate the continuous phase, the dispersed phase and
the mixture, respectively. According to the correlation proposed by
Angeli and Hewitt (2000b), the Sauter mean diameter is 0.48 of
the maximum diameter. In this study, a uniform droplet size equal
to the Sauter mean diameter was assumed throughout the com-
putational domain.
The ANSYS FLUENT 15.0 package was used to solve all transport
equations. The Phase Coupled SIMPLE (PC-SIMPLE) algorithm was
used for the pressure-velocity coupling, which is an extension of
the SIMPLE algorithm for multiphase ows.
A rst-order upwind scheme was used to determine the face
uxes in the momentum and phase transport equations. The
resulting algebraic linear system, from integrating the governing
transport equations over control volumes, was solved using the
GaussSeidel iterative method and the Algebraic Multigrid (AMG)
method. To ensure the stability and convergence of the iterative
calculation, lower under-relaxation factors were chosen.
Fig. 8. Distribution of the water phase fraction across the pipe for an input water
cut of 20% at a mixture velocity of 1.5 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction.
2.3. Turbulence model
In dispersed ow, the drops interact with the turbulent eddies
of the continuous phase in various ways, according to the size of
the eddies and the drops (Angeli and Hewitt, 2000b; Elgolbashi,
1991). Hence, using an appropriate turbulence model can improve
the numerical solution of dispersed ows. Two-equation turbu-
lence models are the simplest complete models for investigating
turbulent ow. The k model is robust and has been used and
validated extensively in industrial internal ow applications
(Roudsari et al., 2012). In the present study the standard k
model, was used to model turbulent ow in two-phase oilwater
ow. By using a per-phase approach, in which a set of k and
transport equations are solved for each phase, more realistic
values of inter-phase turbulent momentum transfer are expected.
The per-phase standard k model equations used in this study are
the standard form mentioned in Fluent User's Guide (2011).
 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148 141

Fig. 10. Distribution of the water phase fraction across the pipe for an input water
Fig. 9. Distribution of the water phase fraction across the pipe for an input water
cut of 80% at a mixture velocity of 1.5 m/s. (a) Vertical distribution of the water
cut of 50% at a mixture velocity of 1.5 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction.
of the water volume fraction.

ow pattern was initiated so that the oil and water entered the
3. Experiment
test pipe without any premixing.
A separator was installed downstream of the test pipe where
In the present study the experimental study of Elseth (2001)
oil and water were separated. The oil and water were then stored
was used to examine the capability of CFD model in predicting the
in two storage tanks. The stored oil and water were circulated in
type of ow patterns. A schematic layout of the test pipe is shown
the ow loop by using the pumps capable of delivering different
in Fig. 1a.
volumetric ow rates. By changing the ow rates of oil and water
The test pipe had an outer diameter of 60.3 mm and an inner
pumps different mixture velocities and water cuts could be
diameter of 56.3 mm. The test pipe consisted of one entry section
established. Table 1 shows a brief description of the property of
and two different sections for ow eld measurements, one sec-
uids and the pipe.
tion for Laser Doppler Anemometry (LDA) measurements and
As mentioned before, the type of ow pattern depends on a
another one for gamma densitometer. Except for a short section
number of parameters such as uid properties (viscosity, density,
made of a transparent plexiglass for LDA measurements, rest of the
surface tension), the geometry of pipe (pipe diameter, inclination
pipe was made of stainless steel. The LDA method was used to
angle), wetting property of pipe and operating pressure and
measure local velocities and velocity uctuations, while the temperature. In the present experiment all of these parameters
gamma densitometer was used to measure the local volume were kept constant and only the effects of mixture velocity and
fractions across the pipe section. The entry region of the pipe was water cut was investigated.
made long enough to establish a fully developed ow before
reaching the measurement sections. 4. Geometry and boundary conditions
A mixing unit as shown in Fig. 1b was located upstream of the
test pipe. The oil and water were introduced at the top and bot- In this study, we present the numerical results of oilwater
tom, respectively. By using a plate within a mixing unit a separated turbulent ow in a horizontal pipe. The numerical simulations
142 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148
Fig. 11. Distribution of the water phase fraction across the pipe for an input water
cut of 10% at a mixture velocity of 2 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction.
were performed for different water cuts and mixture velocities, to
examine the capability of the used CFD model in predicting the
ow patterns and local water volume fraction distributions inside
the pipe. The density and viscosity of the oil were 790 kg/m3 and
1.64 cp, respectively. A brief description of the property of uids
and the used geometry is presented in Table 1 (Elseth, 2001). All
the numerical results presented in this study were obtained at the
pipe cross section located 11.3 m downstream of the inlet where
the experimental data were obtained.
At the inlet, oil and water were introduced separately according
to the experimental work of Elseth (2001). The velocity inlet
boundary condition was applied at the inlet of the pipe. The inlet
velocity of oil and water were changed to satisfy the mixture
velocity and water cut used in the experiment. The pressure outlet
boundary condition with a gauge pressure of zero was imposed at
the outlet. As symmetric ow patterns were observed in the oil
water ow, a symmetry boundary condition (Fig. 2) was applied to
reduce the computational time. A no-slip boundary condition was
applied at the wall to model the two-phase oilwater ow inside
the horizontal pipe (Fig. 2).
Fig. 12. Distribution of the water phase fraction across the pipe for an input water
cut of 20% at a mixture velocity of 2 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction.
The choice of appropriate mesh size for two-uid models is not
straightforward. On the one hand, the mesh size should be larger
than the diameter of the dispersed phase. On the other hand, the
grid size should be ne enough to resolve the boundary layer and
account for the high gradients in ow eld (Prosperetti and
Tryggvason, 2007). In the present study, Gambit 2.3.16 (Gambit
User's Guide, 2006) was used to generate different three-
dimensional symmetric grids. By comparing the numerical
results of different grids and considering the abovementioned
criteria, a grid comprised of 304,000 hexahedral cells, with
renement near the wall, was found to be appropriate.
5. Results and discussions
Fig. 3(a) shows the numerical results and the experimental data
for the vertical distribution of the water volume fraction across the
pipe cross-section for a low mixture velocity of 1 m/s. The input
water cut for this case is 0.1. Both the numerical results and
experimental data show a dispersed water-in-oil ow pattern.
 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148
Fig. 13. Distribution of the water phase fraction across the pipe for an input water
cut of 50% at a mixture velocity of 2 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction.
Fig. 3(b) shows the contours of the water volume fraction obtained
using the CFD model. For this condition, a continuous oil ow with
a zero water volume fraction at the top, and dispersion of water in
oil at the bottom of the pipe, is observed. In spite of a high water
fraction at the bottom of the pipe, the oil phase retains its con-
tinuity. Fig. 4 shows the local water volume fraction at the same
mixture velocity of 1 m/s. However, the water cut was increased to
0.2. Figs. 4(a) and 4(b) show that a segregated layer of water forms
at the bottom of the pipe. This ow pattern is considered as
stratied ow. As mentioned previously, predicting the existence
of the free water layer and its thickness is of the utmost impor-
tance for corrosion risk assessment. Comparison of the numerical
results and the experimental data, as shown in Fig. 4(a), indicates
that the CFD model is capable of predicting water wetting with
reasonable accuracy. By further increasing the water cut, the
thickness of the water layer increases while the thickness of the oil
layer decreases. However, both phases retain their continuity.
Fig. 5 shows this fact more clearly. As seen in this gure a thick
143
Fig. 14. Distribution of the water phase fraction across the pipe for an input water
cut of 80% at a mixture velocity of 2 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction.
layer of water and oil are owing at the bottom and top of the
pipe, respectively. For a certain mixture velocity, there is a critical
water cut beyond which the oil phase will lose its continuity and a
dispersion of oil in water is formed. According to the experimental
data, for a low mixture velocity of 1 m/s, this critical water cut is
equal to 0.8. However, the CFD model suggested a critical water
cut equal to 0.75. Fig. 6 shows the ow pattern for water cut of
0.8 and mixture velocity of 1 m/s. As seen in this gure, the
experimental data revealed the existence of a thin layer of oil at
the top of the pipe, while the CFD model shows a dense dispersion
of oil in water.
To study the effect of mixture velocity on the ow pattern and
water volume fraction distribution, the numerical simulations were
carried out for a higher mixture velocity of 1.5 m/s. Fig. 7(a) shows
the numerical and experimental results for vertical distribution of
the water volume fraction across the pipe cross section for an input
water cut equal to 0.1. An oil continuous ow with dispersed water
droplets at the bottom is observed for this condition. Fig. 7(b) shows
144 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148
Fig. 15. Distribution of the water phase fraction across the pipe for an input water
cut of 20% at a mixture velocity of 2.5 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction.
the distribution of the water volume fraction more clearly. Fig. 8
illustrates the predicted distribution of the local water phase hold-
up for the same velocity but with a higher water cut of 0.2. An
increase in water cut results in a higher water fraction at the bottom
of the pipe. However, in contrast to Fig. 4, a free water layer is not
observed at the bottom of the pipe. Comparison of Figs. 4 and 8
indicates that, for a constant water cut, an increase in mixture
velocity results in an increase in dispersion. Furthermore, compar-
ison of the obtained numerical results and the experimental data, as
depicted in Fig. 8(a), conrms the accuracy of the numerical solu-
tion. By increasing the mixture velocity, the level of turbulence and
mixing increases. Hence, the critical water cut required for water
wetting increases. Fig. 9(a) depicts the water hold-up distribution
for the same velocity with 50% water cut. According to both the
experimental data and the CFD results, the ow pattern is stratied.
However, the predicted interface between two continuous phases is
not as sharp as the observed interface in the experiment. Fig. 9
(b) shows the contours of the water volume fraction obtained using
Fig. 16. Distribution of the water phase fraction across the pipe for input water cut
of 30% at a mixture velocity of 2.5 m/s. (a) Vertical distribution of the water fraction
as predicted by the CFD model and obtained by measurement, (b) contours of the
water volume fraction.
the present CFD model. Fig. 10(a) shows the distribution of the
water phase as predicted by the CFD model and the published
experimental data for an 80% water cut. With further increase of the
water cut at the same mixture velocity, a dispersed oil in water ow
with a free water layer at the bottom of the pipe is observed. Fig. 10
(b) shows the contours of the water volume fraction in the pipe
cross section obtained using the CFD model.
Numerical simulations were performed for a higher velocity of
2m/s (Figs. 1114). As shown in Figs. 11 and 12, at low water cuts,
water in oil dispersion is observed at the bottom of the pipe. These
gures also reveal that an increase in mixture velocity results in
increased dispersion. Figs. 11(a) and 12(a) show that there is good
agreement between the predicted results and the measured
experimental data. Fig. 13(a) illustrates the experimental and
numerical results for a 50% water cut. The experimental data
clearly show a stratied ow pattern with a mixing layer between
the pure water and the oil layers. However, the numerical results
show greater dispersion. As seen in Fig. 13(b), there is no clear
 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148 145

Fig. 17. Distribution of the water phase fraction across the pipe for an input water cut of 50% at a mixture velocity of 2.5 m/s. (a) Vertical distribution of the water fraction as
predicted by the CFD model and obtained by measurement, (b) contours of the water volume fraction.

interface between the two phases. By further increasing the water
cut, the thickness of the oil layer decreases and nally the oil
phase loses its continuity. As a result, a dispersion of oil in water is
observed at the top of the pipe cross section. Fig. 14(a) and (b)
shows such a ow pattern observed for an 80% water cut. As seen
in Fig. 14(a), the numerical results agree very well with the
experimental data.
Figs. 1518 show the distribution of the water volume fraction
across the pipe cross section at a constant mixture velocity of 2.5 m/s.
A comparison of the predicted in-situ hold-up distribution and the
experimental data for a 20% water cut is shown in Fig. 15(a).
As expected, water in oil dispersion is observed for this condition.
Furthermore, this gure shows very good agreement between
the numerical results and the experimental data. Fig. 15(b) shows the
contours of water volume fraction as obtained by the present CFD
model. By further increasing the water cut to 30% (Fig. 16(a) and (b)),
the water fraction at the bottom of the pipe increases. Nevertheless,
due to high velocity and turbulence a free water layer cannot form at
the bottom of the pipe.
According to Fig. 17(a), the experimental data show a stratied
ow pattern, while in numerical results it is not easy to identify
the ow pattern. The distribution of the water volume fraction at
the pipe cross section is shown in Fig. 17(b). The observed dis-
crepancy between the numerical results and the experimental
data could be due to the higher level of dispersion, predicted by
the turbulent dispersion model proposed by Lopez De Bertodano,
(1998). In addition, the assumption of uniform droplet size
throughout the pipeline could be the reason for the observed
deviation. In real ow-eld a variation in droplet size is seen due
to the variation of the volume fraction, velocity and turbulence
(Angeli and Hewitt, 2000b; Lovick and Angeli, 2004b; Ngan, 2010).
Hence, using more advanced methods, such as the population
balance model, would improve the accuracy of the numerical
results.
Fig. 18(a) illustrates the vertical distribution of the water volume
fraction for 80% water cut and a mixture velocity of 2.5 m/s.
From this gure, it is obvious that water is the continuous phase
while the dispersed oil droplets are distributed homogeneously in
146 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148

Fig. 18. Distribution of the water phase fraction across the pipe for an input water cut of 80% at a mixture velocity of 2.5 m/s. (a) Vertical distribution of the water fraction as
predicted by the CFD model and obtained by measurement, (b) contours of the water volume fraction.

Fig. 19. Distribution of the water phase fraction across the pipe for an input water Fig. 20. Distribution of the water phase fraction across the pipe for an input water
cut of 35% at a mixture velocity of 3 m/s. (a) Vertical distribution of the water cut of 50% at a mixture velocity of 3 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction. of the water volume fraction.
 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148
Fig. 21. Distribution of the water phase fraction across the pipe for an input water
cut of 80% at a mixture velocity of 3 m/s. (a) Vertical distribution of the water
fraction as predicted by the CFD model and obtained by measurement, (b) contours
of the water volume fraction.
it. Fig. 18(b) shows the contours of the water fraction as predicted
by the CFD model.
The highest mixture velocity considered in this study was 3 m/s.
The distribution of the water fraction for different water cuts has been
shown in Figs. 1921 . For a mixture velocity of 3 m/s and a water cut
equal to 0.35, a dispersed ow is observed (Fig. 19(a) and (b)).
A reasonable agreement is observed between the numerical and
experimental data of the water volume fraction distribution. How-
ever, it is not clear whether the dispersed ow is water continuous or
oil continuous. The same difculty was reported in the experiment,
making it difcult to identify the continuous phase (Elseth, 2001).
Fig. 20 shows the distribution of water at a higher water cut equal to
0.5. According to the experimental data a stratied ow was not
established inside the pipe. Both the CFD model and experiment
show a dispersed ow. A comparison among the observed ow pat-
terns, for 50% water cut at different mixtures velocities, indicates that
an increase in mixture velocity results in a higher entrainments
of droplets at the interface.
Fig. 21(a) and (b) shows the water phase distribution at a high
water cut of 0.8. Due to the high velocity and water cut, a dis-
persed oil in water ow pattern is observed inside the pipeline. A
comparison of the numerical results for high water cuts indicates
147
that as we increase the mixture velocity more homogeneous dis-
persion occurs. According to the experimental data shown in
Fig. 21(a), the dispersed oil drops tend to move away from the
wall, which results in a sudden increase in the water volume
fraction near the wall. The same behaviour was also reported in
previous experimental studies (Soleimani et al., 2000a; Soleimani,
2000b). However, the numerical model does not show such fea-
tures. As mentioned before, the observed discrepancy between the
numerical and experimental data could be attributed to the
assumption of the uniform size of the droplet throughout the
pipeline.
Due to the large length scales in most industrial problems, it is
impossible to use interface resolving methods. For such conditions
using the EulerianEulerian approach is the best choice. However,
an appropriate closure law, responsible for small-scale interactions
(drag, lift and turbulent dispersion), must be provided. The dis-
crepancy between the present CFD results and the corresponding
experimental data may be attributed to the fact that the adopted
closure laws need to be improved. Hence, further studies are
required to examine the effect of using different closure laws.
6. Conclusions
Numerical modelling of two-phase oilwater ow through a
horizontal pipeline was carried out using a two-uid model and
the standard kepsilon turbulence model. The used CFD model
and closure laws were presented, accordingly.
The CFD model was used for modelling oilwater ow in
comparatively wide operating conditions, ranging from low to
high velocities and low to high water cuts.
Comparison of the CFD results with the corresponding
experimental data revealed the capability of the CFD model in
predicting the ow pattern for different conditions.
It was observed that at low velocities gravity is the dominant
force determining the type of ow pattern, while at high velocities
turbulent dispersion force is predominant.
The distribution of the in-situ water volume fraction, obtained
using the CFD model, was compared with the experimental data in
a quantitative manner for all of the simulations. The obtained
results agreed fairly well with the experiment.
According to the present numerical results, the CFD model used
in this study was capable of predicting the type of wetting at the
bottom of the pipe with reasonable accuracy. It was observed that
as we increased the velocity of ow, the critical water cut below
which we had oil wetting at the bottom of pipe increased.
The discrepancy between the present CFD results and the cor-
responding experimental data may be attributed to the fact that
the adopted closure laws need to be improved. Hence, further
studies are required to examine the effect of using different clo-
sure laws. Furthermore, the capability of CFD model in predicting
more complex ow patterns such as annular ows should be
investigated.
Acknowledgements
This research was conducted under the project to establish the
foundations of industrial technology which is funded by the
Ministry of Trade, Industry & Energy (MOTIE, Korea) (Grant no.:
N0000003, N0001154).
References
Angeli, P., Hewitt, G.F., 2000a. Flow structure in horizontal oilwater ow. Int.
J. Multiph. Flow 26, 11171140.
148 H. Pouraria et al. / Ocean Engineering 115 (2016) 135148
Angeli, P., Hewitt, G.F., 2000b. Drop size distributions in horizontal oil-water dis-
persed ows. Chem. Eng. Sci. 55, 31333143.
Arirachakaran, S., Oglesby, K.D., Malinowsky, M.S., Shoham, O., Brill, J.P., 1989. An
analysis of oil/water ow phenomena in horizontal pipes. In: Proceedings of
SPE Production Operating Symposium, SPE Paper 18836, Oklahoma City.
Brauner, N., 2001. The prediction of dispersed ows boundaries in liquidliquid and
gas-liquid systems. Int. J. Multiph. Flow 27, 885910.
Burlutsky, E., Turangan, C.K., 2015. A computational uid dynamics study on oil-in-
water dispersion in vertical pipe ows. Chem. Eng. Res. Des. 93, 4854.
Cai, J., Li, C., Tang, Z., Ayello, F., Richter, S., Nesic, S., 2012. Experimental study of
water wetting in oil-water two-phase ow-horizontal ow of model oil. Chem.
Eng. Sci. 73, 334344.
Elgolbashi, S., 1991. Particle-laden turbulent ow: direct simulation and closure
models. Appl. Sci. Res. 48, 301314.
Elseth, G., 2001. An Experimental Study of Oil/water Flow in Horizontal Pipes (Ph.D.
thesis). The Norwagian University of Science and Technology, Tronheim.
Fluent Users Guide, 2011. Release 14. Ansys Inc, USA.
Frank, T., Shi, J., Burns, A.D., 2004. Validation of Eulerian multiphase ow models
for nuclear safety applications. In : Proceedings of the 3rd International Sym-
posium on Two-phase Flow Modeling and Experimentation, Pisa, Italy, Sep-
tember 2004, pp. 2224.
Gambit Users Guide, 2006. Release 2. 3. 16. AnsysInc, USA.
Gao, H., Gu, H.Y., Guo, L.J., 2003. Numerical study of stratied oilwater two-phase
turbulent ow in a horizontal tube. Int. J. Heat Mass Transf. 46, 749754.
Guha, D., Ramachandran, P.A., Dudukovic, M.P., 2008. Evaluation of large eddy
simulation and EulerEuler CFD models for solid ow dynamics in a stirred
tank reactor. AIChE J. 54, 766778.
Hamad, F.A., He, S., Khan, M.K., Bruun, H.H., 2013. Development of kerozen-water
two-phase up-ow in a vertical pipe downstream of a 90 bend. Can. J. Chem.
Eng. 91, 354367.
Hanazadeh, P., Hojati, A., Karimi, A., 2015. Experimental investigation of oilwater
two phase ow regime in an inclined pipe. J. Pet. Sci. Eng. 136, 1222.
Kumara, W.A.S., Halvorsen, B.M., Melaaen, M.C., 2009. Pressure drop, ow pattern
and local volume fraction measurement of oilwater ow in pipes. Meas. Sci.
Technol. 20, 118.
Lopez De Bertodano, M.A., 1998. Two-uid model for two-phase turbulent jets.
Nucl. Eng. Des. 179, 6574.
Lovick, J., Angeli, P., 2004a. Experimental studies on the dual continuous ow
pattern in oilwater ows. Int. J. Multiph. Flow 30, 139157.
Lovick, J., Angeli, P., 2004b. Droplet size and velocity proles in liquidliquid hor-
izontal ows. Chem. Eng. Sci. 59, 31053115.
Matzain, A., Apte, M.S., Zhang, H.Q., Volk, M., Brill, J.P., Greek, J.L., 2002. ASME:
J. Energy Resour. Technol. 124, 180186.
Monzon, C.F.T., 2006. Theoretical Modelling of Oilwater Flow in Horizontal and
Near Horizontal Pipes (Ph.D. thesis). University of Tulsa, USA.
Moraga, F.J., Larrenteguy, A.E., Drew, A.E., Lahey., R.T., 2003. Assessment of turbu-
lent dispersion models for bubbly ows in the low stokes number limit. Int.
J. Multiph. Flow 29, 655673.
Nesic, S., 2007. Key issues related to modelling of internal corrosion of oil and gas
pipelines a review. Corros. Sci. 49, 43084338.
Ngan, K.H., 2010. Phase Inversion in Dispersed Liquidliquid Pipe Flow (Ph.D.
thesis). University College London, UK.
Nyborge, R., 2005. Controlling internal corrosion in oil and gas pipelines. Bus. Brief.:
Explor. Prod.: Oil Gas Rev. 2, 7074.
Parvini, M., Dabir, B., Mohtashami, S., 2010. Numerical simulation of oil dispersions
in vertical pipe ow. J. Jpn. Pet. Inst. 53, 4254.
Prosperetti, A., Tryggvason, G., 2007. Computational Methods for Multiphase Flows.
Cambridge University Press, Cambridge/Sao Paulo.
Roudsari, S.F., Turcotte, G., Dhib, R., Mozaffari, F.E., 2012. CFD modelling of the
mixing of water in oil emulsion. Comput. Chem. Eng. 45, 124136.
Sarica, C., Panacharoensawad, E., 2012. Review of parafn deposition research
under multiphase ow conditions. Energy Fuels 26, 39683978.
Sathe, M.J., Deshmukh, S.S., Joshi, J.B., Koganti, S.B., 2010. Computational uid
dynamics simulation and experimental investigation: study of two-phase
liquidliquid ow in a vertical Taylor-Couette contactor. Ind. Eng. Chem. Res.
49, 1428.
Soleimani, A., Lawrence, C.J., Hewitt, G.F., 2000a. Spacial distribution of oil and
water in horizontal pipe ow. SPE J. 5, 394401.
Soleimani, A., 2000b. Phase Distribution and Associated Phenomena in Oilwater
Flows in Horizontal Tubes (Ph.D. thesis). Imperial College, London.
Strazza, D., Grassi, B., Demori, M., Ferrari, V., Poesio, P., 2011. Core-annular ow in
horizontal and slightly inclined pipes: existence, pressure drops, and hold-up.
Chem. Eng. Sci. 66, 28532863.
Tomiyama, A., Tamai, H., Zun, I., Hosokawa, S., 2002. Transverse migration of single
bubbles in simple shear ows. Chem. Eng. Sci. 57, 18491858.
Walvekar, R.G., Choong, T.S.Y., Khalid, H.M., Chuah, T.G., 2009. Numerical study of
dispersed oil-water turbulent ow in horizontal tube. J. Pet. Sci. Eng. 65,
123128.
Xu, X.X., 2007. Study on oil-water two-phase ow in pipelines. J. Pet. Sci. Eng. 59,
4358.
Yao, W., Bestion, D., Coste, P., 2004. A three dimensional two-uid modelling of
stratied ow with condensation for pressurized thermal shock investigations.
Nucl. Technol. 152, 129142.
Yusuf, N., Al-Wahaibi, Y., Al-Wahaibi, T., Al-Ajmi, A., Olawale, A.S., Mohammed, I.A.,
2012. Effect of oil viscosity on the ow structure and pressure gradient in
horizontal oilwater ow. Chem. Eng. Res. Des. 90, 10191030.