Journal Pre-proof

Experiments and CFD modelling for two phase flow in a vertical annulus

Raj Kiran, Ramadan Ahmed, Saeed Salehi

PII: S0263-8762(19)30482-4
DOI: https://doi.org/10.1016/j.cherd.2019.10.012
Reference: CHERD 3847

To appear in: Chemical Engineering Research and Design

Received Date: 12 February 2019

Revised Date: 10 September 2019
Accepted Date: 11 October 2019

Please cite this article as: Kiran R, Ahmed R, Salehi S, Experiments and CFD modelling for
two phase flow in a vertical annulus, Chemical Engineering Research and Design (2019),
doi: https://doi.org/10.1016/j.cherd.2019.10.012

This is a PDF file of an article that has undergone enhancements after acceptance, such as
the addition of a cover page and metadata, and formatting for readability, but it is not yet the
definitive version of record. This version will undergo additional copyediting, typesetting and
review before it is published in its final form, but we are providing this version to give early
visibility of the article. Please note that, during the production process, errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal
pertain.

© 2019 Published by Elsevier.

 Experiments and CFD modelling for two phase flow in a vertical annulus

Raj Kiran, Ramadan Ahmed, Saeed Salehi, University of Oklahoma

*Corresponding author:, Dr. Saeed Salehi, University of Oklahoma, Petroleum Engineering, Norman
70508, Norman, OK 73069, United States, Phone: +13377064415, E-mail: saeads@gmail.com

Highlights
 Extensive description of experimental study for high superficial gas velocity (9.2 - 47.2
m/s).
 Pressure gradient, liquid holdup, and flow regime data for experimental work.

of
 Detailed explanation of CFD modeling for the annulus geometry using ANSYS Fluent.
 Mesh sensitivity, pressure gradient, and void fraction characterization using CFD
modeling.

ro
 Computational approach using probability density function to mimic the void fraction
and flow regime in high superficial gas velocity range.


-p
Comparison of experimental and numerical results for pressure gradient.
re
Abstract
Simulations of two-phase (air and water) flow in a pipe are among the widely discussed topics;
however, with the increased understanding of multiphase flow in pipes, the application of
lP

computational fluid dynamics (CFD) in other complex flow geometries involved in oilfield
operations is becoming more common. This study is aimed to investigate and better understand
two-phase flow characteristics in the annulus using computational fluid dynamics and
experimental approaches. The experimental study included two sets of five tests with increasing
na

superficial gas velocity (9.2 to 47.2 m/s) at a constant liquid flow rate. Experiments were
conducted in concentric annulus test section (35 mm × 82.5 mm) that had an overall length of 5.5
m. Two flow patterns (churn and annular) were observed during the experiment.
ur

Using CFD simulation, pressure drop, void fraction, and flow regime are determined. The VOF
multiphase model and two turbulence models (realizable k-ε and SST k-ω models) were
implemented, and comparative study was conducted to understand the relevance of each method
in high gas velocity scenarios for flow in the annulus. The simulated macroscopic behavior of the
Jo

flow shows consistent pressure gradient patterns and mimics the void fraction behavior.
Probability density functions were implemented on time series evolution of void fraction to
identify the flow regime for the CFD results. The simulation results show a reasonable agreement
with the experimental data within a mean error of 20%.

Keywords: Two-phase flow in annulus, Annular and churn flow regimes, CFD simulation,
Void fraction, Pressure gradient, Probability density function.

1
Nomenclature
Ag Gas inlet area
AN Annular
Atotal Total area of the inlet cross-section
BB Bubble
CFD Computational fluid dynamics
CH Churn
DB Dispersed bubble
Hl Liquid holdup
ID Inner diameter
OD Outer diameter
p Static pressure

of
PDF/pdf Probability density function
RNG Re-normalized group
SL Slug

ro
SST Shear stress transport
Ui’, Uj’ Reynold stresses
VFD Variable frequency drives
VOF Volume of fluid
Vsgin
Vsg
Vsl
Simulation inlet gas velocity
Superficial gas velocity
Superficial liquid velocity
-p
re
1 Introduction
lP

Computational fluid dynamics, though a computationally intensive and sophisticated area of

research, but still one of the most commonly used methods for understanding and analyzing the
flow characteristics in complex geometries. Two-phase flow in the annulus has been subjected to
analytical and experimental investigations in the past; however, there is barely any study using
computational fluid dynamics. Advances in CFD motivate the research in this area and is the main
na

motivation for this study. The distribution of phases in annulus has similar flow characteristics to
that of pipe flow. However, the configurations of the velocity envelop are different.

Past modeling works for vertical flow in the wellbore of oil and gas wells has mostly been focused
ur

on pipe flow. Studies on flow in the annulus have been mainly focused on experimental work and
analytical/mechanistic modeling. Also, past experiments were limited to low-velocity range (less
than 22 m/s) mainly due to experimental complexities (i.e. challenges in maintaining constant gas
Jo

and liquid rates during experiments, controlling gas expansion, and preventing extreme water
hammering effects occurring during gas and liquid rate manipulation). The use of low-velocity
experiments and simplistic modeling efforts are good for building up the understanding and
establish applicable models for low-speed flows. However, in this modern age of advanced control
systems and the tremendous increase in computational resources, it is possible to perform high-
velocity experiments and CFD modeling. The use of CFD provides a better understanding of two-
phase flow characteristics. In the pursuit of exploring the distinctions of flow in the annulus, a
comprehensive effort has been made in this study. We started with the theoretical background of
CFD modeling and then simulated experimental conditions. The principal objective of this paper
2
is to analyze and investigate the multiphase flow features in the annulus. The following steps were
taken to accomplish the objective in a systematic manner:
 Experimental measurements were obtained for high gas velocity (9.2 – 47.2 m/s). The
pressure-drop and liquid holdup measurements were gathered. Flow regime classification
was performed using visual observations and mechanistic model predictions (Section 2).
 CFD modeling encompassed the simulation of multiphase flow in the annulus and a
comprehensive effort has been made to understand the flow dynamics using the numerical
results.
 For the CFD simulation, a sensitivity analysis was performed to determine the effect of
mesh size and grid structure on CFD results and minimize numerical errors. The detail of
this analysis is presented in Section 3.
 The flow characteristics resolution presents a unique challenge for simulation which needs

of
a different strategy to mimic the flow features such as pressure drop, flow pattern, and void
fraction. Segregated inlet mechanism can imitate the real-time flow properties as discussed
in Section 3.
 A computational approach using probability density function was implemented to study

ro
flow pattern and compare the flow regimes detected by visual observations and mechanistic
models. The difference between high and low gas velocity flow structures is also
investigated and presented in Section 3.5.

-p
 The turbulence characteristics of two-phase flow in annulus are strongly associated with
the superficial gas and liquid velocities. Section 3.6 describes the details of turbulent
characteristics.
re
2 Experimental Study
2.1 Related Studies
lP

Multiphase flow features are mainly defined based on the spatial distribution of distinct phases in
the flowing domain and are depicted by flow regime maps. Each flow regime has distinct
characteristics based on fluid properties, flow geometry, and superficial gas and liquid velocities
(Mandhane et al., 1974; Taitel and Duckler, 1976; Barnea, 1987; Cheng et al., 1998). The pipe
na

geometry and the associated intricacies of flow features have been studied in detail (Ohnuki and
Akimoto, 2000; Jayanti and Hewitt, 1992). Several similar studies (Caetano et al., 1992a and b;
Das et al., 1999a and b; Julia et al., 2009) were performed in past to investigate flow in the annulus.
However, two-phase flow in annulus still needs more investigation on several accounts such as
ur

high fluid velocity cases. Similar to vertical pipe flows, bubble flow pattern forms at low
superficial gas velocities in the annulus in which gaseous phase is distributed as discrete bubbles
within the liquid continuum (Kelessidis and Dukler, 1989; Caetano et al., 1992a; Ozar et al., 2008;
and Julia et al., 2011). With the increase in the gas velocity, the bubbles coalesce and form bullet
Jo

shape slugs/bubbles. Further increment in the gas velocity destroys the slug geometry and
consequently makes the flow chaotic, which is often classified as churn flow regime. The churn
flow can be oscillatory in nature. It is worthy to note that the churn flow characteristics in the
annulus are not yet properly understood. Different studies (Ozar et al., 2008; Dukler and Taitel
1986) have stated various mechanisms for the formation of churn flow in vertical pipes. Dukler
and Taitel (1986) attributed the entrance effect for the evolution of churn flow regime. Hewitt et
al. (1985) observed the transportation of liquid phase through large waves and simultaneously
picking up the liquid from a falling film. Besides the churn flow regime, as the gas velocity further
increases, the liquid starts flowing adjacent to the walls while the core of the annulus is
3
predominantly filled with the gas phase. This flow pattern is commonly referred to as the annular
flow regime (Kelessidis and Dukler, 1989; and Caetano et al., 1992a). The basic characteristics of
flow patterns in annuli are similar to that observed in pipes. However, the superficial velocity (both
phases) criteria for the establishment of these regimes (flow regime map) in the annulus differ
from the criteria developed for the pipe flow. Furthermore, Das et al. (1999a and 1999b) suggested
the existence of asymmetric bubbles in slug flow in the annulus which is not observed in pipe flow.

Caetano (1986) studied the two-phase upward flow in the concentric and eccentric annulus using
water/kerosene as liquid phase and air as a gaseous phase. The study covered a broader range of
gas and liquid flow rates; however, it only presented experimental data in the gravity dominated
regime in which pressure gradient decreases with superficial gas velocity at a constant liquid rate.
In addition, the existence of asymmetric Taylor bubbles in the slug flow regime was reported, but
not considered in the model development (Caetano, 1992b). Later, the asymmetric Taylor bubbles

of
were incorporated (Das et al. 1999a; 1999b) in the characterization of the slug flow regime using
the Drift flux approach. With this brief description of past experimental studies, the current work
is presented in the next sections.

ro
2.2 Experimental Apparatus
Figure 1 shows the newly constructed multiphase flow loop. The loop consisted of a vertical

-p
annulus test section, centrifugal pumps, compressors, Coriolis flow meters, high frame speed
camera, pressure and temperature sensors, quick closing and modulating control valves,
differential pressure cells, data acquisition system with advanced flow control capabilities. The
re
annulus test section was comprised of concentric pipes with 82.5 mm outer diameter, and 35 mm
inner diameter. The height of the test section was 5.5 m. Water was used as the liquid and air as
the gas for test fluid. Considering vibrations associated with high-velocity experiments, the test
lP

section was mounted on steel frame attached to a vertical wall. It was also insulated to minimize
the heat loss to the environment. The differential pressure transmitters were used to measure the
liquid holdup and pressure gradient in the test section. The liquid and gas phases were mixed at
the inlet point located at the bottom of the test section.
na

During the test, the mixture traveled upward and established a steady state condition after passing
the midpoint. Hence, two differential pressure transmitters with a measuring accuracy of 0.05%
were installed at the upper section of the annulus to measure the pressure gradient for fully
developed flow as depicted in Figure 1. Another similar transmitter was installed at the bottom of
ur

the annulus to measure liquid holdup. The liquid holdup was determined by measuring the
hydrostatic pressure after trapping the fluid in the test section by invoking the quick closing valves
when a steady state flow condition was established. Five pressure gauges and three temperature
Jo

sensors were installed on the test section to characterize the pressure and temperature distributions
in the test section. The centrifugal pumps were employed to pump water from the tank and were
equipped with Variable Frequency Drives (VFD) which automatically controlled the flow rate
based on the input from the data acquisition system. Coriolis flowmeters were used to measure
and control liquid and gas flow rates. Two fast reacting butterfly valves were installed at inlet and
outlet of the test section. The bypass and check valves were used to return water to the tank and
prevent air from entering the water line, respectively. Also, a fully transparent polycarbonate
viewing port was installed on the test section which allowed the visualization of the flow pattern
in the test section during the experiments.
4
2.3 Experimental Procedure
During the experiment, first air control valves were opened to start air injection. Then, centrifugal
pumps were turned on to inject water. Using the data acquisition system, air and water injection
rates were gradually increased to achieve the desired test condition. The gas and liquid flow rates
were maintained at the desired level until steady state condition was established in the test section.
The establishment of a steady state flow condition was verified by data acquisition measurements
and visual monitoring of the flow pattern using a high-speed camera. At steady state, test
parameters and flow pattern videos were recorded. After this, holdup valves were closed quickly
to trap the fluid in the test section. The liquid holdup was measured using the differential pressure
transducer as mentioned earlier. Finally, the fluid trapped in the test section was slowly vented out
of the system using a modulating control valve (back pressure valve).

of
2.4 Experimental Result
At high superficial gas velocities, mainly annular flow regime exists. Since this study focuses on

ro
high superficial gas velocity, the flow regime predominantly encountered during the experiments
was annular flow pattern. In this regime, the fluid is streamlined towards the direction of the flow.
The flow becomes smoother as the gas phase that entrains small liquid droplets flows in the core

-p
and liquid films form on the walls. As the gas velocity increases, the gas erodes the film and
entrains more droplets, and a further increase in gas velocity destroys the film completely, resulting
in a mist flow. In such cases, the flow becomes gas dominated. The flow patterns detected during
this study are incorporated into the existing flow pattern map (Figure 2), which is developed by
re
theoretical treatment of pipe flow (Taitel et al. 1980). The hydraulic diameter of the annulus is
used instead of pipe diameter. The superimposition of the experimental data with the flow pattern
map shows that the churn and annular flow regimes were encountered during the experiment which
lP

was further confirmed by the visual images.

The pressure gradient and liquid holdup (or void fraction) are important parameters for the
multiphase flow characterization. Pressure gradient and liquid holdup measurements are presented
na

in Table 1 to demonstrate the influence of superficial gas velocity on flow pattern. At low liquid
velocities (less than 14 m/s), the churn flow regime established while for other cases with higher
superficial gas velocity, the flow regime was annular. This classification is based on visual
observation and confirmed by superimposing the data on the flow pattern map (Figure 2). The
ur

CFD modeling is also used to characterize the flow patterns and the details are presented in Section
3.5. In this study, the gas flow rate was varied at a constant liquid flow rate of 77.5 and 155 liters
per minute (LPM) which corresponds to 0.3 and 0.6 m/s respectively. It is worth noting that the
pressure and temperature corrections were incorporated in the superficial gas velocity calculation.
Jo

The experimental results show an increasing trend of pressure gradient and a decreasing trend of
the liquid holdup with superficial gas velocity. In the case of an increase in superficial liquid
velocity, the pressure gradient, and liquid holdup increased. This experimental data is compared
with results of CFD analysis for validation and further investigation of flow behavior in the
annulus.

5
3 CFD Modeling
3.1 Theoretical Background
The fluid dynamics of two-phase flow is very important in petroleum operations (Barati and Liang
2014; Kiran and Salehi 2018). The movement of two distinct phases together complicates the flow
characteristics (Shirdel and Sepehrnoori, 2017). Different modeling approaches (empirical,
analytical, mechanistic, and numerical methods) have been developed for characterizing
multiphase flows (Caetano et al., 1992b). Though the numerical method is computationally
intensive, it has a broader potential for characterization of the flow. In the numerical techniques,
there are several approaches which have been developed in the past. For multiphase flow analysis,
there are two types of approaches: Eulerian-Lagrangian and Eulerian-Eulerian. The Eulerian-
Lagrangian models are applied for discrete phase modeling in which continuous phase treated as
a continuum, and dispersed phase flow is solved by tracking many bubbles or droplets. The

of
dispersed phase can exchange momentum, mass, and energy with the continuous phase. This
method is not suited for flows in which the volume fraction of the second phase is considerable
(Fluent, 2016a). On the other hand, the Eulerian-Eulerian approach treats distinct phases as

ro
interpenetrating continua using the concept of phasic volume fraction. The Eulerian-Eulerian
approach has three modeling formulations: mixture model, the volume of fluid (VOF) model, and
Eulerian Model.

-p
In this study, the VOF method is used due to its comparative simplicity and lesser computational
effort. The VOF model is a surface tracking technique implemented for immiscible fluids in which
the fluid interface is studied. In this method, a single momentum equation is used, and the volume
re
fraction of each phase in each cell is tracked. The equations of motion are solved by applying the
boundary conditions, and the volume fraction of each phase is tracked eventually. It can be applied
for the steady or transient tracking of any gas-liquid interface, and the motion of large bubbles in
lP

the liquid. The governing equations used in this model are continuity and momentum equations.
The continuity equation for the secondary phase is solved to characterize the development of an
interface in the spatial and temporal domain.
na

Furthermore, turbulence structures evolve over time. Two of the widely used turbulent models are
k-ε and k-ω models. k refers to the turbulent kinetic energy, ε stands for turbulent dissipation rate,
and ω is the specific dissipation rate. There are several models for k-ε. In this study, the realizable
model is implemented. The realizable turbulence model imposes mathematical constraints on
ur

Reynolds stresses and follows the schwarz inequality which makes the k-ε turbulent model more
robust for boundary layers under adverse pressure gradients and flow separation conditions.

̅̅̅̅̅̅
𝑈𝑖′ 𝑈𝑗′ > 0
Jo

(8)

2
̅̅̅̅̅̅
(𝑈 ′ ′ ̅̅̅̅̅̅̅
2 2
𝑖 𝑈𝑗 ) ≤ 𝑈𝑖 𝑈𝑗 (9)

where Ui' and Uj' are fluctuating velocities. Wilcox (1993) introduced the k-ω turbulence models
to ease the limitations of the k-ε model (further details are presented in Sections 3.2 and 3.4). In
the shear stress transport (SST) k-ω turbulence model, ε equation is replaced by substituting ε for
kω in the transport equation, whereas the conservation equation for k is considered the same. In
addition, Menter et al. (2003) suggested improved model parameters and incorporated blending
6
functions and limiters to enhance numerical stability. They also proposed a revised eddy viscosity
correlation for k-ω turbulence model.

3.2 Related Studies

Numerous CFD simulation studies have been used to characterize the intricacies of multiphase
flow especially in pipe geometry (Zabaras et al., 2013; Waltrich et al., 2015; Chen, 2004; Sanati,
2015). These studies included flow regime identification, pressure loss, and void fraction
predictions using different multiphase flow simulation approaches. In two-phase flow, slug and
churn flow regimes occur at the intermediate superficial gas velocities (Figure 2). Lun et al. (1996)
carried out a sensitivity analysis on grid density for slug flow characterization and identified
critical regions in which gas-liquid interface is susceptible to the mesh density. Annular flow
regime occurs at high superficial gas velocities (Figure 2). A two-phase flow simulation study

of
(Dai et al. 2013) used the Eulerian-Eulerian and multi-fluid VOF method to characterize different
flow regimes in vertical pipelines. The study compared gas holdup measurements with simulation
results and found a reasonable match. However, the existence of annular flow regime at 1 m/s of

ro
superficial gas velocity was reported, which might need further investigation as it comes under the
slug flow regime in the flow pattern map. Two-phase flow simulation studies in annulus geometry
are very limited. A recent study (Sorgun et al. 2013) encompassed the modeling of different flow
patterns including dispersed bubbly, dispersed annular, plug, slug, and wavy annular flows using

-p
the mixture model in ANSYS CFX. It was inferred that the pressure gradient would be higher for
concentric annuli in comparison to the flow in fully eccentric annuli.
re
The main difference between the CFD modeling of flow in pipe and annulus is geometric variation
which affects grid structure and boundary conditions. Pipe geometry has single wall boundary
that has no slip velocity, high turbulent dissipation rate, and minimum turbulent kinetic energy.
lP

However, in the case of flow in annulus, there are two walls: (inner/tubing and outer/casing walls)
as shown in Figure 3 (a). The change in geometry results in the variation of flow characteristics
such as a change in the flow pattern, pressure gradient, and void fraction. In this study, a CFD
model is developed for flow in annulus and the results are compared with experimental data.
na

3.3 Computational Domain

The geometry considered in the CFD study (Figure 3a) included a long vertical pipe of length more
than 40 times the hydraulic diameter (L > 40Dh). The inner diameter of the outer pipe was 82.5
ur

mm, and the outer diameter of the inner pipe was 35 mm. In the actual experimental condition,
the test section had a length of 5.5 m. The shorter length was considered in the CFD analysis to
reduce the computational effort. Past studies (Lien et al., 2004; Laufer, 1954; and Nikuradse,
1933) suggest that the flow is fully developed in the axial direction after the entrance length of 30
Jo

to 40 times the hydraulic diameter. Therefore, 2 m length is sufficient to simulate the experimental
condition. Moreover, at a specific cross-section, which was 0.5 m before the outlet, the time-series
of area average void fraction was recorded in the simulation data. The specific cross-sectional
plane was chosen to avoid the impacts of both the entrance and exit regions on the results and
provide enough entrance length to establish a fully developed flow condition.

7
3.4 Numerical Schemes and Boundary Conditions
The current study includes ten simulation cases as reported in Section 2.4. In the experimental
investigation, the liquid flow rate was kept constant, and the gas flow rate was varied. The motive
behind this design of experiment was to capture the intricacies of the flow features with increasing
gas velocity. To replicate similar experimental conditions and get further insight of flow dynamics,
the volume of fluid (VOF) method of CFD was chosen to track the interface between the phases
using ANSYS Fluent (Hirt and Nichols, 1981). The VOF model showed good numerical stability,
better convergence, and reasonable accuracy when compared with experimental measurements.
The continuum surface force model was adopted for surface tension modeling which enables
accurate modeling without restrictions on the dynamic evolution of interfaces (Brackbill et al.,
1991). Considering high superficial gas velocities, air is chosen as the primary phase and water
as the secondary phase. As the effect of temperature was neglected; hence, the energy equation

of
was not considered in the simulation. The implicit body force was enacted in the model to satisfy
the equilibrium condition between body force and pressure gradient terms of the momentum
equation, which also handles the convergence issues in case of partial equilibrium.

ro
In the CFD analysis of two-phase flows, turbulence models are of critical importance. In this
study, we used two turbulence modules available in ANSYS Fluent: (a) realizable k-ε and (b) Shear
Stress Transport (SST) k-ω model. The realizable turbulence model for the k-ε model is the

-p
improved version of other turbulence models (Standard and Re-Normalization Group) for certain
kind of flows such as complex shear flow, boundary layer separation. Similarly, the SST model
has two hybrid equations developed by combining the advantages from k-ε and other k-ω models
re
(Fluent, 2016b). The k-ε realizable and k-ω models have been used in a number of studies (Parsi
et al., 2016; De Sampaio et al., 2008; Berthelsen and Ytrehus, 2005) for modeling two-phase pipe
flows. However, Wilcox (1993) suggested that the k-ε models are ill-behaved and inconsistent with
lP

the turbulent boundary layer physical structures. The SST k-ω turbulence model produces a large
turbulence level with strong acceleration; and hence, it is well-suited for high-velocity flow
(Menter, 1994). In this study, simulations were performed using both models (realizable k-ε and
SST k-ω models). The boundary layer characteristics were also compared for annular flow regime
for the first set.
na

The annulus flow geometry is symmetric with respect to X-X and Y-Y planes as depicted in Figure
3 (b). Apart from that, experimental observation of asymmetric Taylor bubbles only applies in the
slug flow regime which is not the scope of this study; hence, applying the symmetricity to reduce
ur

the computational effort doesn't affect the results. Therefore, only a quarter section of the annulus
was considered for grid generation using standard mesh module of ANSYS to reduce the
computational expenses. Results obtained from a full annulus and quarter section were found
Jo

comparable and consistent with single phase flow measurements. To replicate more realistic
conditions and achieve faster steady-state phase distribution, the inlet cross-section was divided
into three inlet zones. The three inlets were assigned (Figure 3b) with velocity-inlet boundary
conditions (two of them in the vicinity of the peripheral boundary for water inlet, and one for a gas
inlet in the core). Similar segregated inlet flow boundary conditions have also been successfully
employed by pipe flow studies (Parsi et al., 2016; Lopez et al., 2016) in the past. This arrangement
makes the simulation of the two-phase flow computationally faster and efficient.
The inlet velocity was calculated based on the area of the inlet zone. For instance, the velocity of
the gas in the middle inlet zone was calculated using the following equation:
8
 𝐴𝑡𝑜𝑡𝑎𝑙
𝑉𝑠𝑔𝑖𝑛 = 𝑉𝑠𝑔 (5)
𝐴𝑔
where Vsgin is the inlet gas velocity for the simulation model. Vsg is the experimental superficial
gas velocity. Ag is a gas inlet area depicted in Figure 3 (b). Atotal is the total area of the inlet cross-
section. It should be noted that the volume fraction of another phase for the particular inlet is
considered to be zero for that particular inlet spatial domain. For instance, the velocity of the
liquid phase is zero for the middle/core section (gas inlet) and vice-versa. The outer and inner walls
were assigned as the stationary boundary with the no-slip condition. The pipe roughness height
was considered to be 0.000015 m, while the roughness constant was taken as 0.5. The turbulence
was specified in terms of intensity and hydraulic diameter. The intensity was assumed to be 5%,
while the hydraulic diameter was 47.6 mm in this case. In addition, both planes of symmetry were
defined as symmetric boundary conditions in ANSYS Fluent. At the outlet, pressure boundary

of
conditions were implemented. The outlet was open to atmosphere, and hence the pressure at the
outlet boundary was considered to be 0 Pa (gauge pressure).

ro
A pressure-based solver was chosen from the options offered in the FLUENT package, where a
finite volume methodology is used to discretize the governing equations. SIMPLE (Semi-Implicit
Method for Pressure-Linked Equations) scheme was utilized for pressure-velocity coupling

-p
calculations. For the spatial discretization, different methods were used: (a) Least Square Cell-
Based for gradient (b) Third order MUSCL (Monotone Upstream-Centered Schemes for
Conservation Laws) for momentum, (c) QUICK (Quadratic Upstream Interpolation for
re
Convective Kinematics) for volume fraction (d) second-order upwind for turbulence kinetic energy
and turbulence dissipation rate. In addition, the first order implicit scheme was used for the
transient formulation to reduce the computational efforts.
lP

3.5 Grid Distribution

The success of any CFD modeling study relies significantly on mesh characteristics. The closest
geometry to the annulus is a circular pipe. Several studies are available in the literature regarding
pipe flow. One of the main reasons for deviations in the flow characterization is the variation in
na

the mesh construction. Parsi et al. (2015) reported that the butterfly shape of mesh geometries is
suitable for the two-phase flow simulation in pipes. Similar observations have been reported by
other studies (Hernandez-Perez et al. 2011; Abdulkadir et al., 2015). However, grid sensitivity
study has not been conducted for any modeling work in the annulus. In this investigation, before
ur

performing the main simulation study, the grid suitability was evaluated using the first
experimental dataset. Three types of grids were generated which included (a) structured uniform,
(b) structured non-uniform, and (c) unstructured. The structured uniform grid had hexagonal
Jo

elementals with same thickness in radial direction, while the structured non-uniform grids had
variable thickness in radial direction. In structured non-uniform case, the thickness was minimum
in the vicinity of pipe walls (outer and inner) and maximum in the core region of the annulus. The
non-uniform case comprised of prism and hexagonal elements and had non-uniform distribution.
The unstructured grid showed divergence; hence, such grid distribution system was discarded. A
different number of grid elements were considered in the study as summarized in Table 2. Four
set of variations in grid sizes were considered to perform sensitivity analysis. Type 1 considers
600 elements (20 X 30) in the horizontal cross-section with 50 divisions in vertical direction. In
this type of mesh, A and B are 20 and 30 divisions, respectively. A refer to divisions along the
9
edges to the plane of symmetry and B stands for the number of divisions along the peripheral edge.
Similarly, other types of meshes were also considered.

Different mesh types were tested for Case 1 (Vsg = 9.2 m/s and Vsl = 0.3 m/s). Pressure drop
measurements are summarized in Table 1. For the equal number of grids; non-uniform grid
distribution displayed a lower error than the uniform grids distribution. Therefore, non-uniform
grid distribution was selected, and an additional simulation was carried out with increased number
of elements of 1200 in the cross-section. For non-uniform grid distribution, discrepancies between
measurements and simulation reduced with the increase in the number of mesh elements.

3.6 Results and Discussion

The standard initialization method was used to initialize inlet flow conditions. After the

of
initialization, the whole flow domain was patched with the liquid phase. The time step was 0.001
s with 100 iterations allowed for each step to meet the convergence criteria. The standard
convergence criterion of 0.001 was selected for residuals of continuity, velocity (u, v, w), kinetic
energy (k) and dissipation rate (ε). Furthermore, several other parameters such as net mass flux,

ro
volumetric average pressure, void fraction, and turbulence kinetic energy were tracked with
flowtime. The monitoring of these additional parameters was essential to ensure complete
development of the flow in the relevant section. It was inferred that the volumetric average of

-p
pressure and water volume fraction stabilizes and attains approximately constant value in case of
fully developed flow (shown in Figures 4). The value of volumetric average pressure stabilizes to
a constant value after 0.5 seconds of the simulation time which signifies the establishment of fully
re
developed flow in the system for the second case. In addition, the time-series value of void fraction
was used in generating the probability density function for flow identification as described in
Section 3.6.
lP

The CFD simulation results for all ten cases as described in experimental results are presented in
Figure 5. Case (a) represents the experimental and simulation data for first set ((constant superficial
liquid velocity of 0.3 m/s) while (b) represents the second set of data (constant superficial liquid
velocity of 0.6 m/s). It is worth mentioning, that the CFD simulated pressure gradients are lower
na

than the experimental data at comparatively lower superficial gas velocities (less than 19 m/s) and
higher at velocities greater than 28.75 m/s. The pressure gradient predictions from stress transport
(SST) k-ω model and realizable k-ε model were similar. It was also observed with increase in
slippage between both phases, the error in pressure drop calculation increases especially in case of
ur

higher superficial liquid velocity. Similar observations have been reported for the volume of fluid
(VOF) modeling in other studies (Parsi et al., 2015; Waltrich et al., 2015).
Jo

Apart from the pressure gradient, the liquid holdup and void fraction are important characteristic
in multiphase flow simulation. The spatial distribution varies from one flow regime. In this study,
the flow regime area is based on high superficial gas velocity, where churn and annular flow
regimes become dominant. The annular flow regime is characterized by the formation of a thin
film on the boundary walls and gas core in the middle part of the annulus. There are several
methods to characterize flow regimes. One of the common visualization techniques is looking at
the flow structure. The zero-void fraction near the wall can be treated as a film.

10
To quantify the distribution of liquid holdup, the area-weighted average of liquid holdup fraction
distribution was captured for the horizontal cross-sectional plane at an axial distance of 1.5 m from
the inlet point. The profile of liquid holdup for the annular flow regime with the distance along the
radial direction is depicted in Figure 6. The liquid holdup fraction is 1 closer to the wall, while it
is zero in the core of the annulus corresponding to superficial gas velocity from Case 1, 2, and 3.
However, with an increase in the gas velocity, the liquid holdup fraction in the film region also
decreases. Considering this, it can be inferred that the void fraction in the film increases with
increase in the superficial gas velocity. Similar observations have been made experimentally for a
void fraction of pipe flow using wire mesh sensor (Parsi et al., 2015). Also, the videos captured
during the experiment investigation showed a similar pattern.

In the past, studies have heavily relied on the probability density function to characterize the flow
regimes and calculate the void fraction in the spatial domains. In this study, the probability density

of
functions were determined and compared. Hence, a cross-sectional plane was set at 1.5 m (axial
distance) from the inlet point. The average void fraction data corresponding to this cross-sectional
plane was recorded for 10 seconds of simulation time. To plot the probability density function

ro
(PDF), the initial 0.5 seconds of the data was neglected. This initial 0.5 second of the simulation
corresponds to the time for the system to reach a fully developed flow condition. This decision
was based on the average volume of void fraction and average volumetric pressure of the system

-p
as depicted in Figure 4.

PDFs can be generated by several methods. Two of them are histogram and kernel distribution
re
method. Histogram creates a discrete PDFs, while the kernel distribution method produces a
smooth and continuous PDF. The kernel distribution is a nonparametric representation of PDF
using a random variable which is defined by the smoothing function based on the shape of the
lP

curve for the PDF generation (Bowman and Azzalini 1997; Bendat and Peirsol 2011). The kernel
smoothing function can be employed to estimate the probability density for the void fraction in the
current cases. Experimental studies suggest that the single peak void fraction in the range of 0.8-1
is the characteristics of the annular flow pattern in the pipe while distributed single peak at void
fraction close to 1 characterizes the churn flow (Matsui, 1986; Omebere-Iyari et al., 2008; Ali et
na

al., 2014). The five simulated cases show similar trends. Case 1 belongs to churn regime as per
the flow pattern map, and our visual observation (Figure 2). The other cases were in the annular
flow regime (Figure 2).
ur

The plots of PDFs based on simulation results distinctly show different patterns for the churn and
annular case. The churn flow regime shows single distributed peaks with a void fraction around
0.8 as depicted in Figure 7 (a). In the churn flow regime, the realizable k-ε model predicts a higher
Jo

value of void fraction than the shear stress transport (SST) k-ω model and closer to experimental
data. The realizable k-ε model showed the shrinking wide tail with increase in superficial gas
velocity and eventually it becomes the single peak distribution in case of 37.6 and 47.2 m/s. It is
also interesting to note that with an increase in the gas velocity, the peak starts to shift toward the
maximum value of 1, which corresponds to the case of gas dominated flow. The four PDFs for the
void fractions are depicted in Figures 7 (b), 7 (c), 7 (d), and 7 (e). The PDFs were plotted for both
turbulence models: (a) Realizable k-ε and (b) shear stress transport (SST) k-ω model. The SST k-
ω turbulence model shows (Figure 7 (b)) a higher void fraction than the realizable k-ε model and
closer to the measured void fraction. The void fractions for this annular flow regime (Case 2) are
11
0.823 and 0.90 for realizable k-ε and SST k-ω models, respectively. The measured void fraction,
in this case, is 0.891. Therefore, it can be concluded that the SST k-ω model provides a reasonably
better phase distribution profile in the multiphase flow simulation of the annulus in the annular
flow regime.

The liquid holdup fraction comparison for both sets of the simulated cases with the corresponding
data from the experimental study is presented in Figure 8. It can be observed from simulation
results that at high superficial liquid velocity (0.58 m/s), the VOF method overpredicts the liquid
holdup fraction.

Turbulence parameters such as turbulent kinetic energy and dissipation rate are other important

of
characteristics of the multiphase flows. The turbulence parameters are dependent on the interfacial
roughness which increases the flow resistance. Figure 9 shows the turbulence kinetic energy
profile for the simulated cases. The presence of two peaks in the turbulent kinetic energy profile

ro
signifies the presence of liquid-gas interface near the wall. Berthelsen and Ytrehus (2005) pointed
out that the interfacial roughness causes a deviation in the symmetric profile of the parameter,
which is more pronounced with an increase in gas velocity. In addition to the roughness, wall

-p
shear stress variation between inner and outer walls results in non-symmetric profiles of these
parameters. It can be seen in Figure 9; the turbulence kinetic energy profiles are not symmetric.
Near vicinity of walls, the turbulent kinetic energy decreases while the dissipation rate increases
re
as we approach the walls from the core. Also, as depicted in the void fraction plot (Figure 6), the
gas phase erodes the liquid film with an increase in the gas flow rate. Hence, it becomes evident
that the turbulence kinetic energy near the wall increases with the superficial gas velocity. A
similar case exists for the turbulence dissipation. The turbulent dissipation rate shows a distinct
lP

characteristic in the liquid wall film region. It decreases in the vicinity of the wall corresponding
to the liquid film thickness. However, it is worth noting that this reduction is less than 0.5 m2/s3,
while the maximum turbulent dissipation rate is 99.14 m2/s3 in this case.
na

4 Conclusions
This paper presents experimental and CFD simulation studies on two-phase flow characteristics in
a concentric annulus. Ten experiments were conducted varying superficial gas and liquid
ur

velocities. The pressure gradient and void fraction increased with superficial gas velocity. In the
case of an increase in superficial liquid velocity, pressure gradient increased, and void fraction
decreased. The test data is plotted on a theoretical flow pattern map and compared with the visual
Jo

observations. The visual observation and theoretical flow pattern were in close agreement.

Furthermore, computational fluid dynamics studies were conducted to understand the flow
dynamics in high superficial gas velocity regime. Two turbulence models realizable k-ε and shear
stress transport (SST) k-ω models, coupled with the volume of fluid (VOF) model were used to
predict the pressure drop and void fraction in the flow domain. The realizable k-ε model performs
better at low superficial gas velocities (less than 19 m/s) in predicting void fraction, while the SST
k-ω model showed better accuracy at high velocities (greater than 28 m/s).

12
Also, the probability density functions (PDFs) were generated to identify the flow regimes. The
churn flow regime showed distributed single peak in PDF at the higher end of void fraction. The
annular flow regime showed a single peak in PDFs and displayed similar characteristics as
previously reported for pipe flow. Hence, it is worthy to note that the PDFs of void fractions from
CFD simulations can be used to identify the annular and churn flow regimes, which is otherwise
difficult to quantify by looking at void fraction distribution in the annulus at the end of the
simulation. Moreover, at high superficial gas velocities (more than 47 m/s), flow becomes gas
dominated. The gas phase interpenetrates and erodes the liquid film of the wall, and consequently,
the film gets destroyed with an increase in superficial gas velocity.

Acknowledgments
The authors would like to thank Bureau of Ocean Energy Management (BOEM) Gulf of Mexico

of
OCS Region for the financial support under the research contract award M16PS00059. In addition,
the author would like to acknowledge Jeff McCaskill for helping and facilitating the experimental
work at the University of Oklahoma.

ro
-p
re
lP
na
ur
Jo

13
References
1. Abdulkadir, M., Hernandez-Perez, V., Lo, S., Lowndes, I.S. and Azzopardi, B.J., 2015.
Comparison of experimental and Computational Fluid Dynamics (CFD) studies of slug
flow in a vertical riser. Experimental Thermal and Fluid Science, 68, pp.468-483.
2. Ali, S.F. and Yeung, H., 2014. Two‐ phase flow patterns in large diameter vertical pipes.
Asia‐ Pacific Journal of Chemical Engineering, 9(1), pp.105-116.
3. Barati, R. and Liang, J.T., 2014. A review of fracturing fluid systems used for hydraulic
fracturing of oil and gas wells. Journal of Applied Polymer Science, 131(16).
4. Barnea, D., 1987. A unified model for predicting flow-pattern transitions for the whole
range of pipe inclinations. International Journal of Multiphase Flow, 13(1), pp.1-12.
5. Bendat, J.S. and Piersol, A.G., 2011. Random data: analysis and measurement procedures
(Vol. 729). John Wiley & Sons.

of
6. Berthelsen, P.A. and Ytrehus, T., 2005. Calculations of stratified wavy two-phase flow in
pipes. International Journal of Multiphase Flow, 31(5), pp.571-592.
7. Bowman, A.W. and Azzalini, A., 1997. Applied smoothing techniques for data analysis:

ro
the kernel approach with S-Plus illustrations (Vol. 18). OUP Oxford.
8. Brackbill, J.U., Kothe, D.B. and Zemach, C., 1992. A continuum method for modeling
surface tension. Journal of computational physics, 100(2), pp.335-354.
9. Caetano, E.F., Shoham, O. and Brill, J.P., 1992a. Upward vertical two-phase flow through

-p
an annulus—part i: single-phase friction factor, Taylor bubble rise velocity, and flow
pattern prediction. Journal of Energy Resources Technology, 114(1), pp.1-13.
10. Caetano, E.F., Shoham, O. and Brill, J.P., 1992b. Upward vertical two-phase flow through
re
an annulus—Part II: Modeling bubble, slug, and annular flow. Journal of Energy Resources
Technology, 114(1), pp.14-30.
11. Chen, P. 2004. Modeling the Fluid Dynamics of Bubble Column Flows. PhD Dissertation,
lP

Department of Chemical Engineering, Washington University.

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.469.8152&rep=rep1&type=pdf
12. Cheng, H., Hills, J.H. and Azzorpardi, B.J., 1998. A study of the bubble-to-slug transition
in vertical gas-liquid flow in columns of different diameter. International Journal of
na

Multiphase Flow, 24(3), pp.431-452.

13. De Sampaio, P.A., Faccini, J.L. and Su, J., 2008. Modelling of stratified gas–liquid two-
phase flow in horizontal circular pipes. International Journal of Heat and Mass Transfer,
51(11-12), pp.2752-2761.
14. Dai, Y., Dakshinammorthy, D. and Agrawal, M., 2013, May. CFD Modeling of Bubbly,
ur

Slug and Annular Flow Regimes in Vertical Pipelines. In Offshore Technology

Conference. Offshore Technology Conference.
15. Das, G., Das, P.K., Purohit, N.K. and Mitra, A.K., 1999a. Flow pattern transition during
Jo

gas liquid upflow through vertical concentric annuli—part I: experimental investigations.

Journal of fluids engineering, 121(4), pp.895-901.
16. Das, G., Das, P.K., Purohit, N.K. and Mitra, A.K., 1999b. Flow pattern transition during
gas liquid upflow through vertical concentric annuli—part II: mechanistic models. Journal
of fluids engineering, 121(4), pp.902-907.
17. Dukler, A.E. and Taitel, Y., 1986. Flow pattern transitions in gas-liquid systems:
measurement and modeling. Multiphase Science and Technology, 2(1-4).
18. Fluent, A.N.S.Y.S., 2016a. Theory Guide 17.2. Ansys Inc. USA.

14
19. Fluent User Guide (Version 16.2). 2016b. ANSYS Fluent. Retrieved from
http://www.ansys.com/Products/Fluids/ANSYS-Fluent (Accessed on November 16, 2018)
20. Hernandez-Perez, V., Abdulkadir, M. and Azzopardi, B.J., 2011. Grid generation issues in
the CFD modelling of two-phase flow in a pipe. The Journal of Computational Multiphase
Flows, 3(1), pp.13-26.
21. Hewitt, G.F., 1985. Experimental and modelling studies of annular flow in the region
between flow reversal and the pressure drop minimum. Physico-Chemical Hydrodynamics,
6, pp.43-50.
22. Hirt, C.W. and Nichols, B.D., 1981. Volume of fluid (VOF) method for the dynamics of
free boundaries. Journal of computational physics, 39(1), pp.201-225.
23. Jayanti, S. and Hewitt, G.F., 1992. Prediction of the slug-to-churn flow transition in vertical
two-phase flow. International Journal of Multiphase Flow, 18(6), pp.847-860.
24. Julia, J.E., Ozar, B., Dixit, A., Jeong, J.J., Hibiki, T. and Ishii, M., 2009. Axial development

of
of flow regime in adiabatic upward two-phase flow in a vertical annulus. Journal of Fluids
Engineering, 131(2), p.021302.
25. Julia, J.E., Ozar, B, Jeong, J., Hibiki, T., & Mamoru Ishii, M. (2011). Flow regime

ro
development analysis in adiabatic upward two-phase flow in a vertical annulus.
International Journal of Heat and Fluid Flow 32 pp 164–175.
26. Kelessidis, V.C. and Dukler, A.E., 1989. Modeling flow pattern transitions for upward gas-

-p
liquid flow in vertical concentric and eccentric annuli. International Journal of Multiphase
Flow, 15(2), pp.173-191.
27. Kelessidis, V.C. and Dukler, A.E., 1990. Motion of large gas bubbles through liquids in
re
vertical concentric and eccentric annuli. International journal of multiphase flow, 16(3),
pp.375-390.
28. Kiran, R. and Salehi, S., 2018, June. Mathematical modeling and analysis of riser gas
lP

unloading problem. In ASME 2018 37th International Conference on Ocean, Offshore and
Arctic Engineering (pp. V008T11A063-V008T11A063). American Society of Mechanical
Engineers.
29. Laufer, J., 1954. The structure of turbulence in fully developed pipe flow. NACA Report
1174.
na

30. Lien, K., Monty, J.P., Chong, M.S. and Ooi, A., 2004, December. The entrance length for
fully developed turbulent channel flow. In 15th Australian fluid mechanics conference
(Vol. 15, pp. 356-363)
31. López, J., Pineda, H., Bello, D. and Ratkovich, N., 2016. Study of liquid–gas two-phase
ur

flow in horizontal pipes using high speed filming and computational fluid dynamics.
Experimental Thermal and Fluid Science, 76, pp.126-134.
32. Lun, I., Calay, R.K. and Holdo, A.E., 1996. Modelling two-phase flows using CFD.
Jo

Applied Energy, 53(3), pp.299-314.

33. Mandhane, J.M., Gregory, G.A. and Aziz, K., 1974. A flow pattern map for gas—liquid
flow in horizontal pipes. International Journal of Multiphase Flow, 1(4), pp.537-553.
34. Matsui, G., 1986. Automatic identification of flow regimes in vertical two-phase flow
using differential pressure fluctuations. Nuclear Engineering and Design, 95, pp.221-231.
35. Menter, F.R., 1994. Two-equation eddy-viscosity turbulence models for engineering
applications. AIAA journal, 32(8), pp.1598-1605.
36. Menter, F.R., Kuntz, M. and Langtry, R., 2003. Ten years of industrial experience with the
SST turbulence model. Turbulence, heat and mass transfer, 4(1), pp.625-632.
15
37. Nikuradse, J., 1933. Gesetzmäßigkeiten der turbulenten Strömung in glatten Rohren
(Nachtrag). Forschung im Ingenieurwesen, 4(1), pp.44-44.
38. Ohnuki, A. and Akimoto, H., 2000. Experimental study on transition of flow pattern and
phase distribution in upward air–water two-phase flow along a large vertical pipe.
International journal of multiphase flow, 26(3), pp.367-386.
39. Omebere-Iyari, N.K., Azzopardi, B.J., Lucas, D., Beyer, M. and Prasser, H.M., 2008. The
characteristics of gas/liquid flow in large risers at high pressures. International journal of
multiphase flow, 34(5), pp.461-476.
40. Ozar, B., Jeong, J. J., Dixita, A., Juliá, J.E., Hibikia, T., & Ishiia, M. (2008). Flow
structure of gas--liquid two-phase flow in an annulus. Chemical Engineering Science 63
(2008) pp 3998-4011
41. Parsi, M., Vieira, R.E., Torres, C.F., Kesana, N.R., McLaury, B.S., Shirazi, S.A.,
Schleicher, E. and Hampel, U., 2015. Experimental investigation of interfacial structures

of
within churn flow using a dual wire-mesh sensor. International Journal of Multiphase Flow,
73, pp.155-170.
42. Parsi, M., Agrawal, M., Srinivasan, V., Vieira, R.E., Torres, C.F., McLaury, B.S., Shirazi,

ro
S.A., Schleicher, E. and Hampel, U., 2016. Assessment of a hybrid CFD model for
simulation of complex vertical upward gas–liquid churn flow. Chemical Engineering
Research and Design, 105, pp.71-84.

-p
43. Sanati, A. 2015. Numerical Simulation of Air–water Two–phase Flow in Vertical Pipe
using k-ε Model. International Journal of Engineering & Technology, 4 (1) (2015) 61-70.
44. Shirdel, M., & Sepehrnoori, K. (2017, February 1). Development of transient mechanistic
re
three-phase flow model for wellbores. Society of Petroleum Engineers.
doi:10.2118/180928-PA
45. Sorgun, M., Osgouei, R. E., Ozbayoglu, M. E., and Ozbayoglu, A. M. 2011. Gas-liquid
lP

flow through horizontal eccentric annuli: CFD and experiments compared. In the
Proceedings of the ASME-JSME-KSME 2011 Joints Fluids Engineering Conference,
Shizuoka, Japan.
46. Taitel, Y. and Dukler, A.E., 1976. A model for predicting flow regime transitions in
horizontal and near horizontal gas‐ liquid flow. AIChE journal, 22(1), pp.47-55.
na

47. Taitel, Y., Bornea, D. and Dukler, A.E., 1980. Modelling flow pattern transitions for steady
upward gas‐ liquid flow in vertical tubes. AIChE Journal, 26(3), pp.345-354.
48. Waltrich, P. J., Hughes R., Tyagi M., Kam S., Williams W., Cavalcanti de Sousa P.
Zulqarnain M., Lee W., and Capovilla S. M. 2015. Experimental investigation of two-phase
ur

flows in large-diameter pipes and evaluation of flow models applied to worst-case

discharge calculations, BOEM Report M15PC00007, Craft & Hawkins Department of
Petroleum Engineering, Louisiana State University, Baton Rouge.
Jo

49. Wilcox, D.C., 1993. Comparison of two-equation turbulence models for boundary layers
with pressure gradient. AIAA journal, 31(8), pp.1414-1421.
50. Zabaras, G. J, Schoppa, W., Menon, R., and Wicks, M. 2013. Gaps and advancements for
deepwater production and remote processing: large diameter riser laboratory gas-lift tests.
Offshore Technology Conference. doi:10.4043/23968-MS
https://www.onepetro.org/conferencepaper/OTC-23968-MS

16
 of
ro
-p
re
lP

Figure 1. Schematic of the experimental flow loop (not to scale)

na
ur
Jo

17
 of
ro
-p
re
Figure 2. Flow pattern map for flow in the annulus and current experimental data (The flow pattern
map is drawn by modifying the criteria proposed by Taitel et al. (1980). The pipe diameter is replaced
by hydraulic diameter, which is difference between the outer and inner diameter of the concentric
lP

pipes)
*BB: Bubble, SL: Slug, DB: Dispersed bubble, CH: Churn, AN: Annular
na
ur
Jo

18
 of
ro
-p
Figure 3. (a) The geometry of annulus in a wellbore (b) Inlet flow geometry for the CFD model (Not
to scale)
re
3
0.95
lP
Pressure (KPa)

Void Fraction
2 0.90
na

1 0.85
ur

0 0.80
0 2 4 6 8 10
Time (s)
Jo

Figure 4. Simulated volumetric average pressure and void fraction (at 1.5 m from the inlet) in the
annulus with time

19
 (a) (b)
Figure 5. Measured and predicted pressure gradients for superficial liquid velocity of (a) 0.3 m/s
and (b) 0.6 m/s

of
ro
-p
re
lP
na
ur

Figure 6. Simulated liquid holdup fraction distribution profile in the radial direction in the cross-
Jo

sectional plane 1.5 m from the inlet

20
 of
ro
-p
re
lP
na
ur
Jo

Figure 7. Probability density function (PDF) percentage for void fraction corresponding to
superficial gas (Vsg) and liquid (Vsl) velocity

21
 (a) (b)
Figure 8. Measured and predicted void fractions at superficial liquid velocity of (a) 0.3 m/s and (b)

of
0.6 m/s

ro
-p
re
lP
na

Figure 9. Turbulence kinetic energy profiles in the radial direction for different superficial gas
velocities
ur
Jo

22
Table 1. Measurement for the liquid flow rate and superficial liquid velocity variation

Liquid
Superficial
Superficial Pressure Holdup
Liquid
Case Gas velocity Gradient Fraction
velocity
(Hl)
m/s m/s Pa/m
1 9.2 0.3 2164 0.20
2 18.5 0.3 2645 0.10
3 28.7 0.3 2761 0.04
4 37.6 0.3 3258 0.03
5 47.2 0.3 4241 0.03
6 10.4 0.6 2986 0.17
7 20.0 0.6 3916 0.11
8 28.9 0.6 4545 0.08

of
9 39.0 0.6 5276 0.04
10 47.2 0.6 6374 0.03

ro
-p
re
lP
na
ur
Jo

23
 Table 2. Number of elements used in different grid structures
Number of elements in Total number of Grid
cross-section (AXB) elements Distribution
Mesh 1 600 (20 X 30) 30000 System
Uniform
Mesh 2 900 (30 X 30) 45000 Uniform
Mesh 3 600 (20 X 30) 30000 Non-uniform
Mesh 4 1200 (40 X 30) 96000 Non-uniform

of
ro
-p
re
lP
na
ur
Jo

24