You are on page 1of 9

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/272682590

CFD Simulations of Low Liquid Loading


Multiphase Flow in Horizontal Pipelines
CONFERENCE PAPER AUGUST 2014
DOI: 10.1115/FEDSM2014-21856

CITATIONS

READS

69

5 AUTHORS, INCLUDING:
Carlos F Torres

Mazdak Parsi

University of the Andes (Venezuela)

DNV GL, Katy, TX, USA

52 PUBLICATIONS 119 CITATIONS

15 PUBLICATIONS 49 CITATIONS

SEE PROFILE

SEE PROFILE

Eduardo Pereyra

Cem Sarica

University of Tulsa

University of Tulsa

43 PUBLICATIONS 95 CITATIONS

158 PUBLICATIONS 961 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate,


letting you access and read them immediately.

SEE PROFILE

Available from: Carlos F Torres


Retrieved on: 08 March 2016

Proceedings of the ASME 2014 4th Joint US-European Fluids Engineering Division Summer Meeting
FEDSM2014
August 3-7, 2014, Chicago, Illinois, USA

FEDSM2014-21856

CFD SIMULATIONS OF LOW LIQUID LOADING MULTIPHASE FLOW IN


HORIZONTAL PIPELINES
Hamidreza Karami
The University of Tulsa
Tulsa, OK, 74104
Mazdak Parsi
The University of Tulsa
Tulsa, OK, 74104

Carlos F. Torres
The University of Tulsa
Tulsa, OK, 74104
Eduardo Pereyra
The University of Tulsa
Tulsa, OK, 74104

ABSTRACT
Low Liquid Loading is a very common occurrence in wet gas
pipelines where very small amounts of liquid flow along with
the gas, mainly due to condensation of hydrocarbon gases and
water vapor. The effects of low liquid loading on different flow
characteristics, and flow assurance issues such as pipe corrosion
prove the necessity of analyzing the flow behavior in more
depth. In this study, CFD simulations are conducted for a
horizontal pipe where liquid and gas are supplied at separate
constant rates at the inlet. The liquid is introduced at the
bottom to help shorten the developing section. The simulations
are conducted with Ansys Fluent v14.5 using Volume Of Fluid
(VOF) as the multiphase model. The analysis targets, mainly,
the shape of the interface, velocity fields in both liquid and gas
phases, liquid holdup, and shear stress profile. On the other
hand, experiments are conducted in a 6-inch ID low liquid
loading facility with similar testing condition. Experiments are
conducted with water or oil as the liquid phase for a liquid
volume fraction range of 0.0005 - 0.0020 of the inlet stream.
For all cases, several flow parameters are measured including
liquid holdup and interface wave characteristics. A comparison
is conducted between CFD simulation results, model
predictions, and experimental results, and a discussion of the
sources of discrepancy is presented. Overall, the results help
understand the low liquid loading flow phenomenon.

Cem Sarica
The University of Tulsa
Tulsa, OK, 74104

INTRODUCTION
Low liquid loading flow is the flow condition wherein the
liquid flow rate is very small as compared to the gas flow rate.
It is widely encountered in gas condensate pipelines. Meng et
al. (2001) defined it as the flow conditions when liquid flow
rate is less than 1100 m3 per MMsm3 gas flow rate. Even
though the pipeline is fed with single phase gas, condensation of
the heavier components of gas phase along with traces of water
results in three phase flow. The presence of these liquids in the
pipeline, although in very small amounts, can influence
different flow characteristics, such as pressure distribution.
Many issues like hydrate formation, pipe corrosion, pigging
frequency, and downstream facility design associated with
pressure and holdup are also affected. Therefore, understanding
of the flow characteristics of low liquid loading gas-oil-water
flow is of great importance in transportation of wet gas.
Limited amount of studies have been conducted on low
liquid loading flows. Most of the existing experimental studies
utilized small diameter pipes. The predictions of the existing
models for pressure gradient, liquid holdups, wetted-wall
perimeter, liquid entrainment, and etc. are not satisfactory for
low liquid loading flow (Fan,2005; Dong, 2007; Gawas 2013).
Computational fluid dynamics (CFD) tools are very helpful
means to solve governing equations for fluid flow under
different conditions and complement experimental work.
Several investigators have tried to use CFD to simulate
multiphase flow in pipeline, and in particular, low liquid
loading stratified wavy flow. In one of the earliest attempts,
Shoham and Taitel (1984) combined a 2-D momentum equation
and an eddy viscosity turbulence model, and presented the

Copyright 2014 by ASME

computations for the liquid phase. However, in the gas phase,


only bulk flow calculations were made to estimate gas and
interfacial shear stress values. Later, Newton and Behnia
(2000) tried to predict pressure gradient and liquid holdup using
another two-dimensional simulation and with the assumption of
smooth interface. They also presented some wall and interfacial
shear stress values. Mouza et al. (2001) used a flat interface
and separate conduits for the two phases, coupled at the
interface to simulate the stratified flow. They tried to simulate
the waves at the interface with a static equivalent roughness,
which led to some uncertainty. In their method, the film
thickness is an input to obtain velocity and shear stress profiles.
They conducted the simulations with CFX using the
homogeneous model.
Bartosiewicz et al. (2008) also used VOF approach to
obtain transition from stratified smooth to stratified wavy flow,
wave celerity, and critical wave number. They observed a good
match between simulation results, experimental data, and linear
inviscid theory. They also compared VOF and multi-field
approaches and found some discrepancies in wave amplitude
and wave growth. This study uses the capability of Volume Of
Fluid (VOF) model to capture the gas-liquid interface in
transient flow.In this study, a horizontal pipe is considered. The
liquid and gas are introduced separately with constant rate inlet
boundaries. The liquid inlet is placed at the bottom to help
shorten the flow developing section. The analysis targets are
mainly the shape of the interface and the liquid holdup.
EXPERIMENTAL DESIGN
In this study, the experimental data used are from the tests
conducted in a 6 in. ID flow loop. Figure 1 shows the
schematic of the facility.

Figure 1. Low Liquid Loading 6-in. ID Facility Schematic

pressure in the flow loop. The facility has been previously


utilized by several researchers to investigate two or three phase
low liquid loading flow.
The test fluids are air and water. Air is provided using two
different compressors and water phase is provided, using tap
water from the Tulsa city water supply. Three differential
pressure transmitters, with uncertainty of 0.2 in. of water are
used to get the pressure drop values.
Five quick-closing valves (QCV) are used to bypass the
flow and at the same time trap the liquid in the test sections.
The liquid trapped in the QCV is pigged out with a specially
designed pigging system and is drained into graduated cylinders
to measure the oil and water volumes. The uncertainty of liquid
holdup measurements was analyzed by some pigging efficiency
calibration tests. After two pigging operations, at least 98% of
the liquid is drained out of the section. An approximated value
of 100 ml. was added to the measured liquid volume in all the
tests to take into account the residual liquid in the section.
A set of conductivity probes was used to conduct wave
characteristics analysis, mainly for two-phase flow of water and
air. Two probes were positioned 6 inches apart in the facility.
The cross correlation between the wave signals from the two
probes gives the wave celerity of the interface. Both static and
dynamic calibration tests were conducted for the probes, and
the results were repeatable, and very close for both probes. The
resulting calibration curve was used for further wave
characteristics analysis and estimation of liquid film thickness at
the bottom of the pipe.
The liquid superficial velocity is kept either at 1 or 2 cm/s,
and the gas superficial velocity is varied between 10 m/s and
22.5 m/s. These superficial velocity values ensure the low
liquid loading stratified flow.
The experimental results obtained for different parameters
are used to benchmark CFD simulation results and document
discrepancies. In particular, liquid holdup and liquid film
thickness are considered and analyzed in this work.
CFD SIMULATION DESIGN
The purpose of the CFD simulations was to obtain liquid
holdup for different flowing conditions, and compare the
outputs with experimental data. For this objective, symmetry in
tangential direction condition is assumed. Half a pipe is
considered with an inner diameter of 6 in. and a length of 15 ft.
In order to shorten the developing section, inlet plate was
divided into two zones. Liquid was entered from the bottom
zone, and the gas was entered from the other inlet zone. Figure
2 shows the schematic of the section geometry and meshing.

The flow loop consists of two parallel sections, with 6-in.


(0.15 m) ID pipes. Each section is 56.4 m long. Acrylic
visualization sections about 8 m long are provided at the end of
each section. The inclination angle can change from 0,
horizontal case, to 2 in inclined case. Two back pressure
valves installed at the outlet of the separator control the

Copyright 2014 by ASME

values at the outlet were compared. Figures 3, 4, and 5 are


showing the acquired results for the two experimental cases,
using coarse, medium and fine mesh, respectively.

Figure 2. Test Section Geometry Schematic


The entrance velocities are corrected by the area of the
entrance for each phase. The equations used to obtain liquid
and gas velocity values at their inlet zones are:
vSL Apipe
.
(1)
v L,inlet
2 Azone,1

vG ,inlet

vSG Apipe
2 Azone, 2

Figure 3. Coarse Mesh Results for Two Cases

(2)

The outlet boundary is a pressure boundary with given


pressure value of 9 psig in order to simulate the experimental
conditions. The liquid phase is water with given default fluid
properties, and the gas phase is air. The multiphase flow model
used was Volume Of Fluid model (VOF) with two Eulerian
phases, in an explicit time integration fashion. As a viscous
model, standard k- (2 eqns.) is used with default model
constants and standard wall functions.
Gravitational acceleration is activated in negative ydirection with given default value of 9.81 m/s2. Transient runs
are obtained with a total run time of between 15 to 20 seconds
for all cases. The time step value is chosen to vary using
Courant-Friedrich criterion, which is one of the most common
ways to check the stability of an explicit scheme, with one as
the global Courant number.
As the initial condition, water volume fraction is chosen to
be 0.002, which is close to the no-slip condition. This is
expected to shorten the stabilization period.
Standard
initialization with Pressure Implicit with Splitting of Operator
(PISO) scheme for pressure-velocity coupling was chosen.
The pipe geometry and meshing is designed using Gambit,
a commercial software for grid generation. In order to analyze
the effects of meshing quality, three different meshing types are
implemented, namely, coarse, medium, and fine meshing. The
total number of grid blocks in the coarse meshing is 96000;
while for the medium mesh, this number increases to 224250
grid blocks. The fine meshing design includes the highest
number of grid blocks with 864000 grids.
Two different experimental conditions are tried with each
mesh quality. In order to decide if a meshing quality is working
well for an experimental condition, the resulting liquid holdup

Figure 4. Medium Mesh Results for Two Cases

Figure 5. Fine Mesh Results for Two Cases


Using the coarse mesh, the lower gas flow rate case with
vSG value of 10 m/s is simulated smoothly. This means that a
continuous liquid film is simulated at the bottom of the pipe,
along with some minor fluctuations in liquid volume fraction.

Copyright 2014 by ASME

These oscillations could be due to the interfacial wave structure.


However, for the case of higher gas flow rate, as shown in Fig.
3, a continuous film cannot be simulated, and very large
oscillations in liquid film are an indicator that liquid is flowing
in a separated surging manner. Overall, the coarse mesh was
not able to capture the whole fluid behavior, especially for the
cases of higher gas flow rate.
The cases with medium and fine mesh are showing
smoother curves for liquid volume fraction. This shows that in
both cases a continuous liquid film can be formed in the
simulation. Fine mesh simulations take significantly longer
times. Therefore, medium mesh is used as the main simulating
tool in further analysis. However, results with both mesh
qualities are presented in the next section.
Different discretization methods can produce different
results in a computational work. In this study, the least squares
cell based method is used for gradient spatial discretization.
The momentum is discretized in a second order upwind fashion,
and first order upwind method is used for turbulent kinetic
energy and dissipation rate discretization.
Volume fraction spatial discretization method is an
important parameter in a multiphase VOF CFD simulation.
Two methods are tried for this option. Geo-reconstruct method
and compressive method are the tested options. Three
experimental cases are simulated using both methods for
volume fraction discretization. Figure 6 shows the liquid
volume fraction results using each method along with the
experimental results for the three cases.

RESULTS AND ANALYSIS


Several different parameters can be acquired as the output
of a CFD simulation. For a multiphase low liquid loading flow,
some of these parameters are liquid holdup, liquid film
thickness at the bottom of the pipe, velocity profiles at liquid
and gas phases, pressure drop, wave structure, and liquid
droplet entrainment. The entrainment phenomenon is one of the
most important ones since it significantly affects other
parameters. It has been studied by several researchers,
especially in recent years. However, simulating the entrainment
process with CFD requires a very fine mesh design that comes
with an excessive run time. The mesh design applied in this
study is not able to capture the entrainment phenomenon. This
is considered as one of the main sources of uncertainty of the
results, especially for higher gas flow rate cases where
entrainment becomes more significant.
The results obtained for different parameters are analyzed
separately, and the simulation predictions are compared to
experimental results, where applicable. These parameters
include liquid holdup, liquid film thickness, wave
characteristics, velocity profile, and wall shear stress.
Liquid Holdup
The liquid holdup is the areal fraction of the liquid phase in
the test section. It is measured as an area weighted average of
the liquid phase volume fraction in the outlet plate. Due to the
waves formed at the interface, the volume fraction value
changes with time. Therefore, the average value of the volume
fraction over a period of time was considered to be the liquid
holdup for the given case.
Figure 7 shows the contours of volume fraction for two
cases, one with very low vSG value of 10 m/s at the right, and
one with the highest vSG value of 22.5 m/s at the left. Both
cases have vSL value of 2 cm/s. However, the increase in gas
flow rate has increased the drag force significantly, dropping the
liquid level in the section and decreasing the liquid holdup.

vSG = 22.5
m/s

vSG = 10
m/s

Figure 6. Simulation Results with Two Different Volume


Fraction Discretization Methods
The geo-reconstruct method seems to over-predict and the
compressive method seems to under-predict the value of liquid
volume fraction. On the other hand, if the fluctuations in liquid
fraction are interpreted as the result of the interfacial wave
structure, geo-reconstruct method does a better job predicting
this phenomenon. Especially for the higher gas flow rate case
with vSG value of 15 m/s, the wavy structure is clearer. Overall,
geo-reconstruct method seems to be a better option for a
volume fraction spatial discretization method. This method has
been used for the simulations conducted in this study.

Figure 7. Liquid Fraction Contours for Lowest and Highest


Gas Flow Rate Cases

Copyright 2014 by ASME

One of the most common models used in an ordinary


stratified wavy flow in pipelines is the model developed by
Taitel & Dukler (1976). This model combines the momentum
balance equations for gas and liquid phases and formulates all
of the variables as functions of the liquid film thickness.
Wettability effects are ignored, and the interface is assumed to
be flat. The liquid holdup is calculated from the film thickness
with a simple geometrical relationship. The results of this
model are used for comparison purposes.
The CFD simulation results are shown for five cases, using
medium mesh design, along with three cases using fine mesh
design. Figure 8 shows the liquid holdup results from the
experiments, compared with the CFD simulations and Taitel &
Dukler model.
The medium and fine meshes are typically used for low and
high gas flow rate cases, respectively. The experimental results
have been predicted fairly well using Taitel & Dukler model.
The CFD simulations results are in an acceptable range. A
small over-prediction is observed for medium mesh simulation
results, especially for higher gas flow rate cases. The
entrainment phenomenon cannot be simulated accurately, unless
the grids are small enough to capture very small droplets ripped
off the liquid surface. This can be the reason for the
discrepancy.

geometry, especially for a large diameter pipeline. However,


when the vSG value increases and the flow pattern gets closer to
annular flow, the turbulence and surface tension effects become
dominant and the interface curvature increases. For the
experimental range of this study, the flat interface seems to be a
good option for the geometry design.
In order to obtain film thickness values from CFD
simulations, a line is created at the center of the pipeline
connecting the bottom to the top of the pipe. The liquid phase
volume fraction on this line is averaged over the run time.
Since the entrainment cannot be detected in the simulation, it is
assumed that the whole liquid fraction is flowing in the film.
Multiplying this fraction with pipe diameter gives the liquid
film thickness. Figure 9 shows the comparison between the
experimental results, model prediction, and CFD simulation
results with the two given mesh size. Taitel & Dukler model
tends to under-predict the film thickness and CFD simulations
tend to over-predict it. The method used for the estimation of
film thickness, which includes all of the liquid on the centerline
as part of the film, can be the reason for over-prediction.
Overall, considering different sources of uncertainty involved
with the simulation, the results are within an acceptable range.

Figure 9. Liquid Film Thickness Results from Experiments,


Simulation, and Modeling
Figure 8. Liquid Holdup Results from Experiments,
Simulation, and Modeling
Liquid Film Thickness
In the experimental work, the liquid film thickness is
obtained by time-averaging the voltage signals from a set of
conductivity probes. These probes are placed at the bottom of
the pipeline and their signals are indicators of the thickness of
the water layer. The results of the tests conducted with different
vSG values, and vSL of 2 cm/s are used for the following analysis.
The interface is assumed to be flat and the entrainment
phenomenon is neglected in Taitel & Dukler model. This
means that the liquid holdup is related with a simple
geometrical relationship to the film thickness. The multiphase
stratified wavy flow can be modeled with flat interface

Interfacial Wave Characteristics


The waves in the liquid-gas interface are of significant
importance in a stratified wavy flow. By increasing the gas
flow rate, these waves can change from 2-D waves to 3-D
waves, and then to roll waves. One of the most interesting
observations in the CFD simulations is the structure of the
waves. Figure 10 shows a snapshot of these waves in the
simulation.
Different wave characteristics, such as the wave celerity,
frequency, amplitude, and length can be used to identify the
interfacial wave. In order to increase the accuracy of prediction
for these parameters, a very high quality mesh is needed. A
comparison between the simulation and experimental results is
not completed at this stage. The difference in definition of the

Copyright 2014 by ASME

wave can affect the results for the wave amplitude and wave
frequency. However, the wave celerity experimental results
have been compared to the interfacial velocity from CFD
simulations in the next section.

presented for different values of superficial gas velocity.


Finding the exact position of maximum velocity requires a very
fine mesh. However, for all of the cases, the point with
maximum velocity has a relative vertical position of about 0.75.

Figure 10. Wave Structure along the Flow in a CFD


Simulation
Velocity Profile
The velocity profile in a single-phase fluid flow in the
pipeline is symmetric with a parabolic shape. But for the
multiphase flow, and especially in horizontal pipelines, the
gravitational effects are dominant and the symmetry vanishes.
The velocity magnitude in the liquid phase is much smaller, and
the point with maximum velocity is deviated from the center of
the pipeline. Unfortunately, there are no experimental data
available for comparison purposes at this stage. However, the
simulation results will be discussed here.
Figure 11 shows the velocity contours for two cases, one
with very low vSG value of 10 m/s at the right, and one with the
highest vSG value of 22.5 m/s at the left. Both cases have vSL
value of 2 cm/s.

Figure 11. Velocity Profile Contours for Lowest and Highest


Gas Flow Rate Cases
The scales are different at the two plots. However, they
both show a similar behavior. The position of the point with
maximum velocity has shifted towards the top of the pipeline.
And also, the velocity magnitude is much smaller in the liquid
phase, compared to the gas phase.
Figure 12 shows the velocity profiles obtained with respect
to the vertical position on the centerline. These plots are

Figure 12. Velocity Profile vs. Relative Vertical Position for


Different vSG Values
In all these curves, at a point around the interface position,
the rate of velocity change increases sharply. This point can be
considered as the interface between liquid and gas phases, and
the velocity at this point can be identified as the velocity of
liquid particles at the interface. Figure 13 shows a comparison
between this value and the value of wave celerity, acquired
experimentally. The wave celerity, which is the speed of wave
propagation, is naturally higher than the velocity of the liquid at
the interface. However, the trends are very similar for both
cases.

Figure 13. Simulation Interfacial Velocity and Experimental


Wave Celerity
Wall Shear Stress
Similar to velocity profile, the wall shear stress value is
uniform in a single phase fluid flow in pipeline. But
introducing the second phase results in deviations from the
uniform behavior. Having the liquid film at the bottom and gas
phase on top Taitel & Dukler (1976) proposed three different
average shear parameters into the simplified momentum balance

Copyright 2014 by ASME

equationsThese are liquid-wall shear stress (WL), gas-wall shear


stress (WG), and liquid-gas interfacial shear stress (I).
In this section, local values of WL and WG are obtained
from CFD simulations, and the results are compared with
outputs of Taitel & Dukler model. There are no experimental
data to compare with these predictions at this stage. Figure 14
shows the shear stress profiles obtained from CFD with respect
to the vertical position on the centerline.

Figure 14. Shear Stress vs. Relative Vertical Position for


Different vSG Values
For all cases, the wall shear stress predicted by CFD is
almost constant in the gas phase, with a slight decreasing rate,
as the vertical position decreases. Then, it increases sharply to
a maximum value in the interface, and has a slight decrease in
the liquid film area. This trend can be easily justified,
considering that shear stress is a function of the fluid density
and velocity. The decreasing rate in both phases is due to the
velocity decrease, and the jump in the interface is due to the
jump in density from gas to liquid phase.
In Taitel & Dukler, the WL and WG values are assumed to
be constant along the pipeline. Figure 15 shows WL and WG
values predicted by Taitel & Dukler compared with the CFD
simulation outputs. For the CFD results, the shear stress at the
very bottom of the pipe is used as represent WL and the value at
the top of the pipeline is used to represent WG.

The values of WG are matching very well from the model


to the CFD simulation. However, the values of WL are underpredicted by the model, compared to the CFD simulation. The
uncertainties in prediction of shear stress in the liquid are very
high, considering the very thin liquid film in stratified low
liquid loading flow. The rate of shear stress change inside the
liquid film is very high, and in order to increase the accuracy of
the simulation, meshing should be very fine. Having an
accurate estimation of the shear stress profile, especially close
to the interface, is a very helpful in any effort to model stratified
wavy flow.
CONCLUSIONS
A set of CFD simulations are conducted for a piece of
pipeline where liquid and gas are supplied at separate constant
rates at the inlet. The results obtained for different parameters
are compared to some experimental data from a 6-inch ID
facility. In addition, Taitel & Dukler (1976) is used for the
purpose of model comparison.
Three different mesh sizes are investigated. The medium
and fine mesh simulation outputs are used in further analysis.
Geo-reconstruct method is selected as the discretization method
for the volume fraction.
The results for liquid holdup and film thickness are in an
acceptable agreement with the experimental results. The
interfacial wave structure, velocity profiles and deviation from
symmetric conditions are also observed. The values of wall
shear stress in the liquid and gas phases are compared with the
predictions from Taitel & Dukler. The WG values are very
close, but the WL values show some discrepancy. This
discrepancy can also be related to unsteadiness of the liquid
film due to the complex wave structure.
NOMENCLATURE
WL
Liquid-Wall Shear Stress (Pa)
WG
Gas Phase Wall Shear Stress (Pa)
I
Liquid-Gas Interfacial Shear Stress (Pa)
Apipe
Pipe Cross-Sectional Area (m2)
Azone,1
Zone-1 (Liquid Inlet) Area (m2)
Azone,2
Zone-2 (Gas Inlet) Area (m2)
vSL
Liquid Superficial Velocity (m/s)
vSG
Gas Phase Superficial Velocity (m/s)
vL
Liquid Actual In-Situ Velocity (m/s)
vG
Gas Actual In-Situ Velocity (m/s)
ACKNOWLEDGMENTS
The authors wish to thank the Tulsa University Fluid Flow
Projects (TUFFP) members for their support of this research.

Figure 15. Comparison of Shear Stress Values between CFD


Simulation and Taitel & Dukler Model

REFERENCES
Bartosiewicz, Y., Lavieville, J., and Seynhaeve, J.: A first
assessment of the NEPTUNE_CFD code: Instabilities in a

Copyright 2014 by ASME

stratified flow comparison between the VOF method and a twofield approach, International Journal of Heat and Fluid Flow
29 (2008) 460478.
Dong, H.-K.: "Low liquid loading gas-oil-water flow in
horizontal pipes", U. of Tulsa, Tulsa, 2007.
Fan, Y.: "An investigation of low liquid loading gas-liquid
stratified flow in near-horizontal pipes", U. of Tulsa, Tulsa,
2005.
Gawas, K.: Low liquid loading in gas-oil-water pipe
flow, PhD Dissertation, The University of Tulsa, 2013.
Meng, W., Chen, X.T., Kuoba, G.E., Sarica, C., and Brill,
J.P.: "Experimental study of low-liquid-loading gas-liquid flow
in near-horizontal pipes". SPE Production and Facilities, 16,
240-249, 2001.
Mouza, A. A., Paras, S. V., and Karabelas, A. J.: CFD
code application to wavy stratified gas-liquid flow, Trans
IChemE, Vol 79, Part A, July 2001.
Newton, C.H. and Behnia, M.: Numerical calculation of
turbulent stratified gas-liquid pipe flows, Int J Multiphase
Flow, 26:327337, 2000.
Shoham, O. and Taitel, Y.: Stratified turbulent-turbulent
gas-liquid flow in horizontal and inclined pipes, AIChE J,
30(3): 377385, 1984.
Taitel, Y. and Dukler, A.E.: "A model for predicting flow
regime transitions in horizontal and near horizontal gas-liquid
flow", AIChE J., l22, 1976.

Copyright 2014 by ASME

You might also like