Experimental Thermal and Fluid Science 64 (2015) 81–93

Contents lists available at ScienceDirect

Experimental Thermal and Fluid Science

journal homepage: www.elsevier.com/locate/etfs

Experimental characterization of vertical gas–liquid pipe flow for

annular and liquid loading conditions using dual Wire-Mesh Sensor
Ronald E. Vieira a,⇑, Mazdak Parsi a, Carlos F. Torres b, Brenton S. McLaury a, Siamack A. Shirazi a,
Eckhard Schleicher c, Uwe Hampel c
a
Department of Mechanical Engineering, The University of Tulsa, 800 South Tucker Drive, Tulsa, OK 74104, USA
b
Thermal Science Department, University of Los Andes, Mérida, Mérida 5101, Venezuela
c
Helmholtz-Zentrum Dresden-Rossendorf (HZDR), Dresden, Saxony 01328, Germany

a r t i c l e i n f o a b s t r a c t

Article history: In gas well production, liquid is produced in two forms, droplets entrained in the gas core and liquid film
Received 19 October 2014 flowing on the tubing wall. For most of the gas well life cycle, the predominant flow pattern is annular
Received in revised form 9 February 2015 flow. As gas wells mature, the produced gas flow rate reduces decreasing the liquid carrying capability
Accepted 9 February 2015
initiating the condition where the liquid film is unstable and flow pattern changes from fully cocurrent
Available online 19 February 2015
annular flow to partially cocurrent annular flow. The measurement and visualization of annular flow and
liquid loading characteristics is of great importance from a technical point of view for process control or
Keywords:
from a theoretical point of view for the improvement and validation of current modeling approaches. In
Wire-Mesh Sensor
Gas–liquid flow
this experimental investigation, a Wire-Mesh technique based on conductance measurements was
Void fraction applied to enhance the understanding of the air-water flow in vertical pipes. The flow test section con-
Phase distribution sisting of a 76 mm ID pipe, 18 m long was employed to generate annular flow and liquid loading at
Flow visualization low pressure conditions. A 16  16 wire configuration sensor is used to determine the void fraction with-
in the cross-section of the pipe. Data sets were collected with a sampling frequency of 10,000 Hz. Physical
flow parameters were extracted based on processed raw measured data obtained by the sensors using
signal processing. In this work, the principle of Wire-Mesh Sensors and the methodology of flow para-
meter extraction are described. From the obtained raw data, time series of void fraction, mean local void
fraction distribution, characteristic frequencies and structure velocities are determined for different
superficial liquid and gas velocities that ranged from 0.005 to 0.1 m/s and from 10 to 40 m/s, respectively.
In order to investigate dependence of liquid loading phenomenon on viscosity, three different liquid
viscosities were used. Results from the Wire-Mesh Sensors are compared with results obtained from
previous experimental work using Quick Closing Valves and existing modeling approaches available in
the literature.
Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction underestimation of flow characteristics (e.g., hold up and liquid

Two-phase flow is encountered in a wide range of engineering
applications. For instance, in the petroleum industry, the common
problems associated with gas–liquid two-phase flow include calcu-
lation of flow rate, pressure loss, and liquid holdup/void fraction in
the pipeline. These parameters are essential in design of produc-
tion tubing, gathering and separation system, sizing of gas lines,
heat exchanger design and gas condensate line design [1]. Model
⇑ Corresponding author.
E-mail addresses: rev87@utulsa.edu (R.E. Vieira), mazdak-parsi@utulsa.edu
(M. Parsi), ctorres@ula.ve (C.F. Torres), brenton-mclaury@utulsa.edu (B.S. McLaury),
siamack-shirazi@utulsa.edu (S.A. Shirazi), e.schleicher@hzdr.de (E. Schleicher),
u.hampel@hzdr.de (U. Hampel).
film velocity) results in inappropriate pipe size selection and pos-
sible solids dropout and corrosion issues [2].
Gas and gas condensate lines operate in the annular flow
regime. Annular flow is characterized by a fast moving gas core
with entrained liquid droplets and a slow-moving liquid film flow-
ing around the pipe wall. The flow is associated with a wavy inter-
facial structure, which results in a high interfacial shear stress [3].
In annular vertical up-flow, the average liquid film thickness
around the pipe wall is considered uniform.
Experimental evidences have shown that the entrained liquid
fraction is responsible for a significant part of the pressure drop
in annular two-phase flow [4]. In a producing gas and gas conden-
sate well, as reservoir pressure decreases, entrained liquid forms

http://dx.doi.org/10.1016/j.expthermflusci.2015.02.007
0894-1777/Ó 2015 Elsevier Inc. All rights reserved.
82 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93

Nomenclature

A pipe cross-section area (m) V SL superficial liquid velocity (m/s)

AC cross-section gas core area (m) V SG superficial gas velocity (m/s)
AF cross-section liquid film area (m) vC gas core velocity (m/s)
aði; jÞ reference area matrix vF liquid film velocity (m/s)
d pipe diameter (m) Vði; j; kÞ output matrix of the WMS
dC hydraulic diameter of the gas core (m) V  ði; j; kÞ liquid volume fraction
dF hydraulic diameter of the liquid film (m) WeSG superficial gas Weber number (–)
f frequency (Hz) X Lockhart–Martinelli number (–)
fC core friction factor (–)
fE entrainment fraction (–) Greek Symbols
fI interfacial friction factor (–) aT total void fraction (–)
fL liquid friction factor (–) aC core void fraction (–)
FrSG superficial gas Freude number (–) a G ði; jÞ local time averaged void fraction (–)
g acceleration due to gravity (m/s2) hag ðkÞi instantaneous cross-sectionally averaged void
i, j spatial indexes (–) fraction (–)
k time index (–) hag i cross-sectionally averaged time series or mean void
kT number of measured frames (–) fraction (–)
L pipe length (m) qL liquid density (kg/m3)
PDF probability density function dL liquid film thickness (m)
Pr precision limit of the void fraction result (–) qG gas core density (kg/m3)
p absolute pressure (Pa) qL liquid density (kg/m3)
QCV Quick Closing Valve sI interfacial shear stress (Pa/m)
ReC core Reynolds number (–) sWL wall shear stress (Pa/m)
ReF liquid film Reynolds number (–) lG gas viscosity (cP)
ReSL superficial liquid Reynolds number (–) lL liquid viscosity (cP)
ReSG superficial gas Reynolds number (–) ra ði; jÞ standard deviation of the local time average void frac-
SI interfacial perimeter (m) tion value (–)
SL wetted perimeter (m) ra standard deviation of the cross-sectionally averaged
StG gas based Strouhal number (–) time void fraction data (–)
t measuring time (s) h inclination angle (radians)
V water time avg. signals with the pipe filled with water
V air time avg. of signals with the pipe filled with air

an increasing restriction on gas production. At the end of the life-
time of gas wells, the gas production rate decreases strongly. Due
to this decrease, the drag force of the gas phase exerted on the liq-
uid phase may no longer be sufficient to bring all the liquid to the
surface. Liquid starts to drain downward (flow reversal). In such a
situation, depending on the gas reservoir conditions, the liquid
could accumulate down hole, block the inflow into the production
tubing and gas production could cease. This phenomenon is called
liquid loading. As reported by Belfroid et al. [5] virtually all of the
world’s gas wells are either at risk of or suffering from liquid load-
ing and that the modeling of liquid loading behavior is still quite
immature.
Interaction between gas and liquid phases in two-phase flow
provokes complex, highly deformable interfaces and configura-
tions usually difficult to describe. The simultaneous flow of gas
and liquid in a pipe for instance can produce a large number of spa-
tial configurations of the phases due to the deformable interface
between them. These spatial configurations can become periodic
over time leading to chains of interrelated flow structures. The
chain of these interrelated flow structures is usually referred to
as phase distribution in flow assurance [4]. Examples of these peri-
odic structures are waves in churn flow and disturbance waves in
annular flow. The disturbance waves have unique characteristic
velocities and frequencies which can be used to classify them into
flow patterns [6]. Thus, information about this periodic structure is
valuable to reconstruct the overall pressure drop in any multiphase
flow system.
This study presents new, time resolved quantitative informa-
tion on gas–liquid vertical flows at medium and high superficial
gas velocities. Wire-Mesh Sensor (WMS) technology was used to
measure local time varying void fractions. In order to identify the
effect of liquid viscosity on void fraction values and periodic flow
structures in pipe flows, data has been acquired for three different
liquid viscosities (1, 10 and 40 cP). This work, therefore, con-
tributes to the on-going discussions on the liquid loading problem
in natural gas production and transportation by providing novel
experimental information to better understand the behavior of
the periodic structures in annular two-phase flow as well as the
mechanisms governing exchange and transfer of momentum
between the film and gas core in vertical annular two-phase flow.
2. Characterization of multiphase flows in vertical pipelines for
annular and liquid loading conditions
Several experimental studies on annular flow and liquid loading
in vertical pipes have been presented. For example, Alamu [7]
showed that flow structure becomes more periodic as liquid vis-
cosity increases. On the other hand, structure velocity decreased
with increase in liquid viscosity due to lower liquid phase
momentum.
Zangana [8] carried out several experimental runs to measure
pressure drop, liquid film thickness and wall shear in 127 mm ver-
tical pipe. No completely unidirectional upward flow was observed
from the results of directional wall shear stress measurements. The
change in the direction of the liquid film was also supported by the
measurements of wall shear stress, local film thickness and high
speed video images.
 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93
Yuan [9] claimed that the superficial gas velocity corresponding
to minimum pressure gradient coincides with the onset of liquid
loading. Guner [10] and Skopich et al. [11] determined experimen-
tally that the transition of churn flow to annular flow is the critical
gas velocity for liquid loading initiation. Later, Van der Meulen [12]
presented a new theoretical model for the churn-annular flow
transition boundary.
From an analysis of the output from the Wire-Mesh Sensor
equipment, flow patterns were identified by Abdulkadir et al.
[13] using both the reconstructed images as well as the character-
istic signatures of probability density function (PDF) plots of the
time series of cross-section averaged void fraction. Zhang et al.
[14] carried out an experimental study of gas–liquid two-phase
flow in a 50 mm ID vertical pipe with a combination of X-ray
tomography and Wire-Mesh Sensor as measurement modalities.
From the experimental data of both measurement techniques,
radial void fraction profiles were computed and compared. Gas
velocity radial profiles were also obtained by cross-correlating
the void fraction profiles from both modalities. Vieira et al. [15]
used a Wire-Mesh Sensor to detect the local instantaneous cross-
section distribution of the phases in gas–liquid stratified and annu-
lar flow. Data were obtained for a wide range of flow rates and for
horizontal pipe orientation. Processing of the Wire-Mesh Sensor
data yielded detailed information of the 2D and 3D void fraction
distribution for different liquid viscosities.
Extensive studies have been carried out to develop models and
predict annular flow and liquid loading behavior. Turner et al. [16]
studied the minimum gas velocity required for liquid removal in
vertical pipes. Several investigators developed correlations for
the interfacial friction factor including Henstock and Hanratty
[17] and Asali et al. [18]. Others, such as Ishi and Mishima [19]
and Oliemans et al. [20] have studied the entrainment process.
Alves et al. [21] developed a mechanistic model for annular flow
in vertical pipes based on the separated flow approach.
In this experimental study, the Wire-Mesh Sensor technique
based on conductance measurements is applied to study two-
phase annular flow and liquid loading conditions in 76 mm ID ver-
tical pipes. Liquid viscosity is an important fluid property which
has been found to significantly affect multiphase flow. Thus,
experiments and analyses are carried out with liquids having
different viscosities. Comparisons with an existing modeling
approach and data available in literature are also presented in
the paper.
3. Experimental setup and data processing
3.1. Wire-Mesh Sensor
Local time varying void fractions were obtained by using a WMS
measurement transducer developed by Prasser et al. [22] and sup-
plied by HZDR, Dresden, Germany. In this study, a 16  16 wire
configuration sensor was used as depicted in Fig. 1.
The WMS consists of two parallel wire grids positioned
orthogonally but offset by a small axial distance of 2 mm. The
working principle of the Wire-Mesh Sensor equipment is to mea-
sure the local void fraction, where the transmitting and receiver
wires virtually cross each other. One wire plane works as a trans-
mitter plane with the other as a receiver plane. By activating each
transmitter wire successively, the current at each crossing point
flowing toward the receiver wires is measured in the following
way. The receiver wires are connected to the negative input of a
trans-impedance amplifier stage. This forces the wires to be on
ground potential due to the virtual short inside the trans-impe-
dance amplifier. While the currently activated transmitter wire
provides the excitation signal, all other transmitter wires are kept
83
Fig. 1. Illustration of the dual Wire-Mesh Sensor (WMS).
to ground potential to avoid cross talk. The local instantaneous
void fractions are calculated from the measured conductance at
the crossing points and a series of 2-dimensional data sets are
obtained.
The sensor has two planes of 16 stainless steel wires evenly dis-
tributed over the circular pipe cross-section. This provides a spatial
resolution of approximately 4.76 mm (x, y) across the majority of
the pipe, which is a great advancement compared to more tradi-
tional phase fraction measurement techniques. Data was acquired
at a frequency (f) of 10,000 Hz for a 60 s experimental run period. A
stainless steel flange supports the sensor and allows fixation into
the test flow pipe section. For the experiments, a set of two WMSs
with an axial distance of 32 mm between have been utilized to
measure structure velocities by use of cross correlation algorithms.
Further details about the operation of Wire-Mesh Sensors can be
found in Prasser et al. [22].
3.2. Experimental test facility
The current study was carried out at the Erosion/Corrosion
Research Center (E/CRC) tower-boom loop facility located at The
University of Tulsa. The loop main elements are diesel compres-
sors, electrical compressor, slurry tank, test section, diaphragm
pumps, and separator. For these experiments, the pipe diameter
was 76 mm and 18 m long. A schematic of the experimental facility
is shown in Fig. 2.
For the measurements reported in this paper the test pipe sec-
tion was mounted in the vertical direction. The experiments were
carried out near atmospheric pressure. The fluids used were air and
either water or water mixed with Carboxy Methyl Cellulose (CMC)
to increase the liquid viscosity. This is a long chain polymer widely
used to control viscosity in the food industry due to its thickening
properties. Small amounts of CMC do not cause non-Newtonian
behavior of the liquid. The viscosity of the liquid solution was mea-
sured using a Hydramotion Viscolite 700 portable viscometer. The
viscometer’s precision is 0.1 cP and has a repeatability of ±0.1%.
The air required for the testing was supplied by diesel compressors
1 and 2. The outlet of the test section is connected to a separator,
where air exhausts from the top and liquid is pumped back to
the 250 gallon liquid tank for recirculation due to the considerable
amounts of liquid required to maintain onset of liquid loading and
annular flow inside the test section. The installed WMS in the flow
loop is shown in Fig. 3.
84 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93

Boom Dual Wire Mesh Sensor

Gas Test
17 m
V21
and Liquid
Liquid
76.2 mm ID Test Section
Gas

V11 V12 V21 V22 Gas &

Flow
Liquid
meters

Collecting
250 gallon tank
Gas Tank
V1
Stirrer

Pump 1.1) Pump 1.2)

1 2 3

Pump 2.1) Pump 2.2)

Diesel Compressors

Fig. 2. Schematic of large scale boom loop.

3.3. Test matrix

Seventy-seven experiments were performed at different liquid

viscosities, superficial gas and liquid velocities. The test matrix
was designed to cover the transition boundaries from annular flow
to churn/slug flow predicted by literature models.
The range of superficial gas and liquid velocities are shown in
Table 1.
The test matrix is plotted on a flow pattern map to observe the
transitions predicted by literature models. Fig. 4 presents the test
matrix for the three liquid viscosities considered. The flow pattern
map was generated using the computer program FLOPATN devel-
oped by Pereyra and Torres [25]. The transition criteria used was
calculated using Barnea [26]’s model.
Flow patterns were mostly observed by using two low speed
surveillance cameras. The effective resolution of each camera is
0.3 megapixel with a maximum video resolution of 640  480.
The maximum frame rate 30 fps. Both cameras were mounted on
the test section, before the WMS, in order to observe the hydrody-
namically developed flow during experimental runs. The videos
were captured by using the Digital Video Recorder (DVR) of the
system. From video observations, liquid film direction was
observed to be fully upwards when the superficial gas velocity is
high (V SG > 20 m/s). This region can be defined as annular flow
where the liquid is transported both with the liquid film and the
FLOW DIRECTION
droplets entrained in the gas core (Fig. 5a). A gradual decrease in
superficial gas velocity results in a periodic change of liquid film
Fig. 3. Installed Wire-Mesh Sensor (WMS). direction, namely, film reversal initiation. In the vicinity of
V SG = 15–20 m/s, the liquid film is not fully upwards (Fig. 5b). This
specific region has been indicated by previous researchers as the

Using the Kataoka and Ishii [23] correlation and previous

experimental set-ups reported in literature [20], the test section Table 1
Experimental test matrix – vertical air–water or air–water & CMC.
was installed at a distance 17 m downstream from the inlet (L/
d = 233) to promote fully developed two-phase flow. Similarly, in Test (#) VSG (m/s) VSL (m/s) Viscosity (cP)
order to avoid disturbances to the flow, the WMS was placed 1–30 10.2–40.4 0.005–0.10 1
0.9 m upstream of a standard elbow (r/d = 1.5). More details about 31–60 10.2–37.9 10
the experimental facility can be found in Vieira et al. [24]. 61–77 10.1–33.1 40
 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93 85

1 cP Test Points 10 cP Test Points

10 10

Superficial liquid velocity, m/s

Superficial liquid velocity, m/s

DISPERSED BUBBLE DISPERSED BUBBLE
ANNULAR ANNULAR
1 1

BUBBLE BUBBLE

0.1 0.1
INTERMITTENT INTERMITTENT
FLOW FLOW

0.01 0.01

0.001 0.001
0.01 0.1 1 10 100 0.01 0.1 1 10 100
Superficial gas velocity, m/s Superficial gas velocity, m/s
(a) L = 1 cP (b) L = 10 cP
40 cP Test Points
10
Superficial liquid velocity, m/s

DISPERSED BUBBLE
ANNULAR
1

BUBBLE

0.1
INTERMITTENT
FLOW

0.01

0.001
0.01 0.1 1 10 100
Superficial gas velocity, m/s
(c) L = 40 cP
Fig. 4. Flow pattern map for three different viscosities.

time index. The value of V(i, j, k) is a digital signal proportional to

the local fluid conductivity. First, the signal is normalized taking
into account the single phase reference matrix. The time history
signal of each mesh point can be normalized between 0 and 1 by:
Vði; j; kÞ  V air ði; jÞ
V  ði; j; kÞ ¼ ; ð1Þ
V water ði; jÞ  V air ði; jÞ
where Vwater and Vair are the time average values of the signals with
the pipe filled with water and air respectively at the beginning of
the test. For a constant conductivity value of the liquid during the
experiment, V  ði; j; kÞ can be interpreted as the liquid volume frac-
tion (LVF), assuming a linear relationship between the measured
signal and the total amount of liquid in the single volume element
formed by one crossing point. Since the value of Vair is always ‘‘0’’
Fig. 5. Sketch of the two investigated flow conditions: (a) fully upward flow; (b) (air has no conductivity) the gas volume fraction ag for each cross-
unstable annular flow (film reversal).
ing point is determined using the following equation:
ag ði; j; kÞ ¼ 1  V  ði; j; kÞ: ð2Þ
onset of liquid loading [9] or unstable annular flow [10] owing to
oscillations, upward and downward in the vertical direction for From the time history of the local void fraction, the local time
the entire liquid film region. Although this velocity oscillation is averaged value a
 g ði; jÞ is derived by
observed in the film, the characteristics of annular flow are pre-
1X
k
T

served (liquid film around the pipe wall and liquid droplets a g ði; jÞ ¼ ag ði; j; kÞ; ð3Þ
kT k¼1
entrained in the gas core).
where kT is the number of measured frames (equal to f  t). The
instantaneous cross-section average void fraction is calculated by
4. Data processing and analysis
X
16 X
16

The output of the WMS is a matrix V(i, j, k), where the indexes i hag ðkÞi ¼ ag ði; j; kÞ  aði; jÞ; ð4Þ
i¼1 j¼1
and j refer to the space position of the mesh points and k is the
86 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93
where a(i, j) is a matrix of weighting coefficients. These coefficients
are the ratios of the small areas related to each single crossing
point to the total area of the circular cross-section. The total sum
P P16
of all weighting coefficients 16 i¼1 j¼1 aði; jÞ equals to 1. The cross-
sectionally averaged time series of void fraction is derived by
1X Tk
ha
g i ¼ hag ðkÞi: ð5Þ
kT k¼1
The time history of the average void fraction is characterized by
signal processing techniques such as cross-correlation and prob-
ability density function (PDF) implemented in MATLAB.
Cross-correlation is a measure of similarity of two waveforms as
a function of a time-lag applied to one of them [27]. The cross cor-
relation technique was used to find the time lag (s) between the
two signals produced by each plane of WMS. The cross-sectionally
average void fraction signals’ time lag was used to calculate fre-
quencies and velocities of the periodic structures of two-phase
mixtures.
The PDF analysis has been used to characterize the two-phase
flow pattern using the information from the first plane of the WMS.
5. Precision limit and uncertainty of the experimental
measurements
The void fraction measurements were obtained from many
samples over a specified time interval of sixty seconds at a fixed
test condition. The precision limit of the result (Pr) can be calculat-
ed, assuming no correlated precision limits for the precision errors
[28], using the following equation
Pr ¼ c  StdDevðhag ðkÞiÞ; ð6Þ
where c is the coverage factor and StdDevðhag ðkÞiÞ is the standard
deviation of the sample of N readings of the instantaneous cross-
sectionally average void fraction. For N P 10, c = 2 is assumed.
The value of the standard deviation is calculated from N readings
over the specified time interval of the void fraction measurement.
The maximum precision limit Pr;max of the data for each set of tests
is presented in Table 2.
Multiphase flow parameters such as superficial liquid and gas
velocities were not measured directly from the experiment, it
was therefore necessary to determine the uncertainty on these
velocities. Given that only one experiment was performed for each
run, a reasonable estimation of the systematic uncertainty
propagation was performed on superficial liquid and gas velocities
in this study. The results of the maximum error obtained are
shown in Tables 3 and 4.
In the next section, void fractions measured in various flow con-
ditions are presented and discussed.
6. Results and discussion
6.1. Void fraction
A great deal of information can be obtained from both WMSs by
considering the time series of the cross-sectionally averaged
void fractions. An example of this is shown in Fig. 6. The
Table 2
Maximum precision limits of void fraction measurements.
Test (#) VSG (m/s) VSL (m/s) lL (cP) Pr,max (–)
1–30 10.2–40.4 0.005–0.10 1 0.07
31–60 10.2–37.9 0.005–0.10 10 0.12
61–77 10.1–33.1 0.005–0.10 40 0.12
Table 3
Uncertainty on superficial liquid velocities.
VSL (m/s) Max. Error (%)
0.005–0.10 8.5
Table 4
Uncertainty on superficial gas velocities.
VSG (m/s) Max. Error (%)
10.2–40.4 2.5
10.2–37.9 2.3
10.1–33.1 2.0
cross-sectionally averaged time series of void fraction are obtained
at superficial liquid velocity of 0.04 m/s and four different superfi-
cial gas velocities of 10, 15, 23, and 40 m/s. It is observed that both
signals (Sensors 1 and 2) are not qualitatively different from each
other for all cases. For VSG = 10 m/s (Fig. 6a), the flow is unstable.
That is, liquid transport is irregular because the pressure drop
becomes dominated by gravity. The standard deviation of the void
fraction data is ra = 0.05. Fig. 6b and c shows the characteristic
alternate regions of higher and lower void fractions which identify
unstable annular flow. The standard deviation of the void fraction
data of both conditions are ra = 0.016 and ra = 0.010, respectively.
With further increase of superficial gas velocity, the average void
fraction slightly increases while the amplitude of void fraction time
series decreases, as can be observed in Fig. 6d. The standard devia-
tion for that annular conditions is ra = 0.006.
Fig. 7 shows a comparison of cross-sectionally averaged time
series of void fraction at superficial gas and liquid velocities of
15 m/s and 0.10 m/s, respectively, for three different liquid viscosi-
ties of 1, 10 and 40 cP. From this graph, a significant decrease in
void fraction at higher liquid viscosity can be observed. Also, peri-
odic disturbances in the void fraction are observed approximately
every 8 s for lL = 40 cP which indicates the transition to slug or
pseudo-slug flow.
The cross-sectionally averaged time void fractions, ha  g i, are
extracted for the three liquid viscosities from their time series
and plotted against superficial gas velocity as shown in Fig. 8.
The mean void fraction generally increases with the gas flow rate
for all the viscosities examined. Also, it is observed that a small
void fraction rate of change (dha  g i=dV SG ) occurs on the right of
VSG = 15 m/s as compared to the left side, where larger slope is
observed. Thus, a drastic change of the void fraction rate of change
occurs around VSG = 15 m/s. Similar observations have been made
by Guner [10] and Skopich et al. [11] for vertical pipes, where
the drastic change in void fraction rate of change starts just to
the right of VSG values between 10 m/s and 15 m/s. Fig. 8c shows
significant decrease in void fraction at higher liquid viscosity as
the liquid velocity increases. The possible reason for decrease in
void fraction may be due to decrease in liquid film velocity of the
viscous phase adjacent to the wall which causes higher shear
stress, liquid holdup and less void distribution across the cross sec-
tion of the pipe.
Fig. 9a shows the contours of local time averaged void fraction,
a G ði; jÞ, obtained by the WMS for the onset of the liquid loading
condition. As can be observed, the void fraction values in the gas
core are close to 0.95, which means that liquid bridging could
occur in that region. For the onset of the liquid loading condition,
the wave structures in the interface become larger as compared
with annular flow. These waves get larger with decreasing superfi-
cial gas velocity, coalescing and instantaneously blocking the gas
flow in the core region. Fig. 9b shows the contours of void fraction
for unstable annular flow with VSG = 15 m/s; VSL = 0.04 m/s. A
 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93 87

1 1

Void fraction αg(k)

0.9 0.9

Void fraction αg(k)

0.8 0.8

0.7 0.7

0.6 Sensor 1 0.6 Sensor 1

Sensor 2 Sensor 2
0.5 0.5
0 5 10 15 20 0 5 10 15 20
Time, s Time, s
(a) Onset of Liquid Loading, VSG = 10 m/s (b) Unstable Annular Flow, VSG = 15 m/s

1 1

0.9 0.9

Void fraction αg(k)

Void fraction αg(k)

0.8 0.8

0.7 0.7

0.6 Sensor 1 0.6 Sensor 1

Sensor 2 Sensor 2
0.5 0.5
0 5 10 15 20 0 5 10 15 20
Time, s Time, s
(c) Unstable Annular Flow, VSG = 23 m/s (d) Annular Flow, VSG = 40 m/s

Fig. 6. Cross-section average void fraction time series for annular and low-liquid low: VSL = 0.04 m/s, lL = 1 cP.

1 6.2. Probability density functions

1 cP 10 cP

0.8 The time series data presented in Figs. 6 and 7 can be further
Void fraction α (k)

analyzed using the probability density function (PDF). The prob-

g

0.6 ability density function returns a probability density estimate for

0.4
40 cP
0.2
0
0 5 10
Time, s
Fig. 7. Comparison of average void fraction data for different liquid viscosities:
VSG = 15 m/s, VSL = 0.10 m/s.
relatively thin liquid film surrounding the gas core can be
observed. No liquid bridging was detected.
The contour plots of the standard deviation, ra ði; jÞ, distribution
of the time average void fraction are used to identify the amplitude
of the interface fluctuations in all flow patterns studied in this
work. Fig. 10 shows the contour plots of standard deviation of time
averaged void fraction in the cross-section area of the WMS. Due to
fluctuation of the interface, typical of the interaction between the
liquid and gas phases at lower superficial gas velocities, the signal
of the onset of liquid loading (Fig. 10a) shows high standard devia-
tion values in the liquid film. Fig. 10b, illustrates that by increasing
the superficial gas velocity while maintaining the same superficial
liquid velocity, the amplitude of the liquid film decreases while the
diameter of the gas core (ra ði; jÞ 6 5) increases. In this case, the
waves in the liquid film region are not big enough to block the pipe
cross section.
the sample in the vector hag ðkÞi (instantaneous cross-sectionally
averaged void fraction). The estimate is based on a normal kernel
μ =1 cP
L function and is evaluated at 100 equally spaced points, hag ðkÞii ,
μ =10 cP that cover the range of the selected time series in the first plane
L
μ =40 cP (sensor) of the WMS. The probability density function shows
L
how often each cross-sectionally averaged void fraction value
15 20 occurs. In this work, the probability density function was deter-
mined using the ksdensity function in MATLAB.
Fig. 11 shows the PDFs for twenty-four experimental results.
The figures are organized as follows: horizontally, the viscosity is
constant; vertically, the superficial liquid velocity is constant.
As can be seen in Fig. 11, there are typical shapes or signatures
when the superficial velocities and liquid viscosity change. For
example, Fig. 11a and b show that for lL = 1 cP there is a narrow
single peak for void fractions between 0.95 and 1.0. This peak
can be observed at superficial gas velocity of 34 m/s which is typi-
cal of annular flow. A narrow peak with high frequencies in the PDF
signature means a dominant average void fraction value character-
istic of the particular flow pattern. As the superficial gas velocities
decreases, single peaks between 0.9 and 0.95 can be recognized at
lower frequencies for superficial gas velocities of 15 m/s and 23 m/
s (unstable annular flows). At the onset of liquid loading, at super-
ficial gas velocity of 10 m/s, a broader, flatter signature with a sin-
gle peak at lower frequency between void fractions of 0.8 and
0.925 can be observed. Costigan and Whalley [29] also observed
a gradual transition from a single PDF peak at high void fraction
for annular flow to a double peaked distribution for slug flow in
air-water flows in a vertical 32 mm diameter tube.
88 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93

1.00 1.00
0.95 0.95

Mean void fraction

Mean void fraction
0.90 0.90
0.85 0.85
= 0.005 m/s = 0.005 m/s
0.80 = 0.01 m/s 0.80 = 0.01 m/s
0.75 = 0.02 m/s 0.75 = 0.02 m/s
= 0.04 m/s = 0.04 m/s
0.70 0.70
= 0.10 m/s = 0.10 m/s
0.65 0.65
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 45
Superficial gas velocity, m/s Superficial gas velocity, m/s
(a) μL = 1 cP (b) μ L = 10 cP

1.00
0.95
Mean void fraction 0.90
0.85
0.80
= 0.005 m/s
0.75 = 0.02 m/s
= 0.04 m/s
0.70
= 0.10 m/s
0.65
5 10 15 20 25 30 35 40 45
Superficial gas velocity, m/s
(c) μ L = 40 cP s

Fig. 8. Cross-section time averaged void fraction.

(a) Onset of Liquid Loading, (b) Unstable Annular Flow,

VSG = 10 m/s, VSL = 0.04 m/s VSG = 15 m/s, VSL = 0.04 m/s

Fig. 9. Contours of time averaged void fraction, lL = 1 cP.

The PDF plots for liquid viscosities of 10 cP and 40 cP liquid
show a marked shift toward lower void fraction values. For exam-
ple, all PDFs for VSL = 0.10 m/s are characterized by lower dominant
peak frequencies, with void fraction from 0.7 to 0.85. This is the
fingerprinting of unstable annular flows.
6.3. Structure frequency
Frequency analysis is carried out to establish the relationship
between superficial gas velocity and characteristic frequency. By
computing the cross-correlation of instantaneous cross-sectionally
averaged void fraction from both meshes and selecting the time
delay which corresponds to the best correlation of the cross-corre-
lation function, the structure frequency for each flow condition is
obtained. Cross-correlation technique has been discussed in
detailed by Orfanidis [27].
Fig. 12 shows a slight increase in structure frequency for cocur-
rent annular flows with increase in superficial gas velocity. Similar
results are obtained for annular flows for 10 cP and 40 cP. From
Fig. 12, it can be observed that at superficial gas velocity of 10 m/
s, the characteristic frequency for 10 cP and 40 cP is higher than
the characteristic frequency for 1 cP at the same superficial liquid
velocity. This result indicates that higher liquid viscosities tend to
increase the characteristic frequency for the onset of liquid loading.
For the analysis of oscillating structures in fluid flow dynamics
problems, a dimensionless value useful is the Strouhal number. It
represents a measure of the ratio of inertial forces due to the
unsteadiness of the flow to the inertia forces due to changes in
velocity from one point to another [30]. Azzopardi [4] showed that
multiphase regime frequency data could be correlated using Strou-
hal number and Lockhart–Martinelli parameter. This approach
brings together data whose frequencies would normally have cov-
ered two orders of magnitude. The frequency of periodic structure
in all flow patterns encountered has been correlated using gas
based Strouhal number (StG ) and Lockhart–Martinelli (X) dimen-
sionless numbers given by Eqs. (7) and (8), respectively.
 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93 89

(a) Onset of Liquid Loading, (b) Unstable Annular Flow,

VSG = 10 m/s, VSL = 0.04 m/s VSG = 15 m/s, VSL = 0.04 m/s

Fig. 10. Contours of standard deviation of time averaged void fraction, lL = 1 cP.

100 100
= 10 m/s = 10 m/s
80 = 15 m/s 80 = 15 m/s
= 23 m/s
PDF, % = 23 m/s
PDF, %

60 = 34 m/s 60 = 34 m/s

40 40

20 20

0 0
0.75 0.8 0.85 0.9 0.95 1 0.75 0.8 0.85 0.9 0.95 1
Void fraction Void fraction
(a) VSL = 0.005 m/s; μ L = 1 cP (b) VSL = 0.10 m/s; μ L = 1 cP

100 100
= 10 m/s = 10 m/s
80 = 15 m/s 80 = 15 m/s
= 23 m/s = 23 m/s
PDF, %

PDF, %

60 = 34 m/s 60 = 38 m/s

40 40

20 20

0 0
0.75 0.8 0.85 0.9 0.95 1 0.75 0.8 0.85 0.9 0.95 1
Void fraction Void fraction
(c) VSL = 0.005 m/s; μ L = 10 cP (d) VSL = 0.10 m/s; μ L = 10 cP

100 100
= 10 m/s = 10 m/s
80 = 15 m/s 80 = 15 m/s
= 27 m/s = 23 m/s
PDF, %

60 60
PDF, %

= 33 m/s = 30 m/s

40 40

20 20

0 0
0.75 0.8 0.85 0.9 0.95 1 0.6 0.7 0.8 0.9 1
Void fraction Void fraction
(e) VSL = 0.005 m/s; μ L = 40 cP (f) VSL = 0.10 m/s; μ L = 40 cP

Fig. 11. Probability density function of cross-section averaged void fraction.

90 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93

Characteristic frequency, Hz 300 μ L= 1 cP μ L = 10 cP μ L = 40 cP

μ L = 1 cP 7.0
250

Structure velocity, m/s

μ L = 10 cP 6.0
μ L = 40 cP
200 5.0 Onset of Film
Onset of Film Reversal
Reversal 4.0
150
3.0
100 Cocurrent
Cocurrent Annular Flow
Annular Flow 2.0
50
1.0
0 0.0
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 45
Superficial gas velocity, m/s Mixture velocity, m/s
Fig. 12. Effect of superficial gas velocity on periodic structure frequency for lL = 1 - Fig. 14. Structure velocity as a function of mixture velocity for lL = 1 cP, 10 cP and
cP, 10 cP and 40 cP; VSL = 0.10 m/s. 40 cP.

f wd exists a noticeable increase in structure velocities for gas flowing

StG ¼ ; ð7Þ
V SG
!0:5
qL V 2SL
X¼ : ð8Þ
qG V 2SG
The relationship between Strouhal number and the Lockhart–
Martinelli parameter is shown in Fig. 13 on a log–log plot. The fig-
ure shows that Strouhal number increases with increase in Lock-
hart–Martinelli parameter for onset of liquid loading conditions.
On the other hand, for cocurrent annular flows a smaller slope is
observed.
From Fig. 13, it can be concluded that the influence of the super-
ficial gas velocity on characteristic frequencies is significant for the
onset of liquid loading and unstable annular flows (V SG 6 15). The
dependence of structure frequency on superficial gas and liquid
velocities is not significant for cocurrent annular flows.
6.4. Structure velocity
Time series of void fraction can be examined further to extract
the velocity of periodic structures by cross-correlating both spatial
averaged signals defined by Eq. (4) from the WMSs which are
separated 32 mm apart. Cross-correlation produces transit time
in seconds. Structure velocity is obtained by dividing the separa-
tion distance between the meshes by the transit time.
Fig. 14 shows how structure velocity varies with mixture velo-
city (sum of the superficial liquid and gas velocities) for different
viscosities. The figure shows that at low mixture velocities, there
μ L = 1 cP
10
Onset of Film
Reversal
1
Strouhal number
with 1 cP liquid and 10 cP and 40 cP viscous liquid. In the region
of these unstable flows, major structures such as huge waves and
disturbance waves may exist, revealing a change in the interface
characteristics. The disturbance waves can travel at velocities near-
ly an order of magnitude greater than the film and have amplitudes
several times the mean film thickness [4]. On the other hand, at
higher mixture velocities, the effect of liquid viscosity on structure
velocities becomes indistinguishable for annular flows.
6.5. Comparison with Quick Closing Valve experimental results
Yuan [9] measured pressure gradient and liquid holdup in his
vertical experiments for superficial gas velocities ranging from
10 m/s to 30 m/s. Liquid holdup was measured by trapping fluids
between Quick Closing Valve sections in vertical pipes. Similarly,
Guner [10] measured liquid hold up for different superficial liquid
and gas velocities that ranged from 0.005 to 0.10 m/s and from 10
to 40 m/s, respectively The system also was equipped with two
QCV systems, one at the entrance and one in the developed region.
Data were obtained from the second trapping section located at the
test section, where the flow is considered not to be affected by the
entrance and outlet. Both experiments were obtained in a 76 mm
ID test pipe. Tap water and air were used by both researchers.
Fig. 15 shows the comparison.
Fig. 15a shows that the data obtained from the present study
shows a similar trend to QCV results. However, a difference of
approximately 0.05 (5%) is present for lower superficial liquid
velocities. Fig. 15b shows better agreement between the three sets
of data. The intrusive effect of the WMS could cause a liquid holdup
and thereby cause under-prediction of the void fraction measure-
μ L = 10 cP μ L = 40 cP
ment for low superficial gas velocities.
6.6. Comparison with mechanistic model

In this section, the time averaged mean void fraction data is

compared with the predictions of a mechanistic one-dimensional
two-fluid model for annular flow in vertical pipes [3,21]. A
0.1
schematic of the physical model for annular flow is given in Fig. 16.
The two fluids are the liquid film and core fluid (gas and
0.01 Cocurrent entrained droplets). The analysis is carried out utilizing a control
Annular Flow
volume with a differential axial length of dL. The pipe diameter is
d and the inclination angle is h. The liquid flows as a film around
0.001
the pipe perimeter, while the gas core, including entrained liquid
0.001 0.01 0.1 1
droplets, flows at the center. The main assumptions of the model
Lockhart-Martinelli parameter
are fully developed flow, uniform film thickness, homogeneous
Fig. 13. Correlation of structure frequency using gas based Strouhal number with no-slip flow of the gas and the entrained droplets in the core, uni-
Lockhart–Martinelli parameter. form velocity profile and isothermal-incompressible flow.
 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93 91

1.00 1.00

Mean void fraction

0.95 0.95

Mean void fraction

0.90 0.90

0.85 0.85
WMS DATA WMS DATA
0.80 QCV (Yuan, 2011) 0.80 QCV (Yuan, 2011)
QCV (Guner, 2012) QCV (Guner, 2012)
0.75 0.75
10 15 20 25 30 35 40 45 10 15 20 25 30 35 40 45
Superficial gas velocity, m/s Superficial gas velocity, m/s

(a) V SL = 0.01 m/s, L= 1 cP. (b) VSL = 0.10 m/s, L= 1 cP.

Fig. 15. Comparison with Quick Closing Valve data.

The hydraulic diameters of the liquid film and core are given,
respectively, by:
dF ¼ 4dL ðd  2dL Þ=d; ð16Þ

dC ¼ d  2dL : ð17Þ
From mass balances, the film velocity vF is determined as:
2
ð1  f E Þd
mF ¼ V SL : ð18Þ
4dL ðd  dL Þ
The core velocity mC is determined in a similar manner:
2
ðV SG þ V SL f E Þd
mC ¼ 2
: ð19Þ
ðd  2dL Þ
Assuming a homogeneous no-slip mixture of the gas-phase and
the entrained droplets in the core, the core void fraction aC is deter-
mined as:
V SG
Fig. 16. Schematic of annular flow physical model; (a) film region and (b) core aC ¼ : ð20Þ
V SG þ V SL f E
region [2].
The entrainment fraction, fE, is defined as the fraction of the liq-
uid flow rate that is entrained in the gas core as droplets. Eq. (21) is
The model is derived by applying the momentum balances to the expression for the entrainment fraction used in this work.
the liquid film and the core. The momentum (force) balances for Zhang et al. [31] created this dimensionless form based on an ori-
the liquid film and core are given, respectively, by [3] ginal expression by Oliemans et al. [20]:
 
SL SI dp  0:38  0:97
sWL þ sI   qL g sin h ¼ 0; ð9Þ fE qL lL
AF AF dL F ¼ 0:003We1:8 0:92
SG FrSG Re1:24
SG Re0:7
SL : ð21Þ
1  fE qG lG
 
SI dp The dimensionless parameters WeSG , FrSG and ReSL and ReSG are
sI   qC g sin h ¼ 0: ð10Þ
AC dL C given by:

For fully developed flow, the pressure gradient in the film and qG V 2SG d V SG
WeSG ¼ ; FrSG ¼ pffiffiffiffiffiffi : ð22Þ
core are equal. Thus, Eqs. (9) and (10) can be combined by r gD
eliminating the pressure gradient, resulting in the combined
momentum equation for annular flow, given by: qG V SG d qL V SL d
  ReSG ¼ ; ReSL ¼ : ð23Þ
SL 1 1 lG lL
sWL þ sI SI þ  ðqL  qC Þg sin h ¼ 0: ð11Þ
AF AF AC The shear stresses are given by:
The geometrical parameters are derived based on a uniform qL v 2F
film thickness, dL : sWF ¼ f L ; ð24Þ
2
2
AC ¼ pðd  2dL Þ =4; ð12Þ
qC ðv C  v F Þ2
sI ¼ f I : ð25Þ
AF ¼ pdL ðd  dL Þ; ð13Þ 2
The liquid friction factor, fL, is calculated using Blasius equation
SI ¼ pðd  2dL Þ; ð14Þ based on the film Reynolds number, which considers the hydraulic
diameter as the ratio of four times the area over the wetted
SL ¼ pd: ð15Þ perimeter:
92 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93

1.00 1.00

Mean void fraction

Mean void fraction
0.95 0.95

0.90 0.90

0.85 WMS DATA; = 0.01 m/s 0.85 WMS DATA; = 10 cP

MODEL; = 0.01 m/s MODEL; =10 cP
0.80 WMS DATA; = 0.1 m/s 0.80 WMS DATA; = 40 cP
MODEL; = 0.1 m/s MODEL; = 40 cP
0.75 0.75
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 45
Superficial gas velocity, m/s Superficial gas velocity, m/s
(a) VSL = 0.01 m/s, μL = 1 cP. (b) VSL = 0.10 m/s; μL = 10 cP and 40 cP.

Fig. 17. Comparison with mechanistic model.

qL mF 4dL ðd  dL Þ series plots of void fraction, contour plots of void fraction distribu-
ReF ¼ : ð26Þ
lL d tion, PDF of void fraction and average void fraction for various flow
regimes considered. These results shed additional light on the
Guner [10] showed that the Asali et al. [18] correlation for inter-
hydrodynamics of vertical gas–liquid flows.
facial friction factor fI provided good agreement with the experimen-
The main conclusions from this study are stated as follows:
tal data for both thick and thin liquid films in vertical flows, and it is
suggested as the friction factor correlation for this study.
 From the plots of mean void fraction versus superficial gas velo-
sffiffiffiffiffi !! city, it was observed that void fraction rate of change increases
f I dL
f I ¼ f C 1 þ 0:45Re0:2
C ReC 4 : ð27Þ for decreasing velocity values less than the observed velocity of
fC d
the initiation of liquid loading (VSG 6 15 m/s).
The core friction factor fC is calculated based on the core Rey-  The contour plots of averaged void fraction show that the cross-
nolds number as follows section distribution of the void fraction changes before and after
the transition from liquid loading to annular flows. These
qC v C ðd  dL Þ results help to gain a better understanding of how the liquid
ReC ¼ : ð28Þ
lG phase is distributed during the liquid loading initiation.
 Examination of the plots of PDF of the cross-sectionally aver-
Substituting the geometrical relationships, velocities, and shear
aged time series of void fraction leads to the conclusion that
stresses into Eq. (11) results in an implicit equation for the film
hydrodynamic changes in flow patterns and transition insta-
thickness, dL, which can be solved iteratively. Once dL is deter-
bilities are present in the onset of liquid loading conditions.
mined, the gas void fraction can be calculated with Eq. (29).
 Different characteristic frequencies have been identified for liq-
 2 uid viscosities of 1, 10 and 40 cP and that the periodicity of liq-
dL
aT ¼ aC 1  2 : ð29Þ uid loading flow structure increases as liquid viscosity
d
increases. The transition from annular to cocurrent annular flow
Fig. 17 shows void fraction predictions of the mechanistic model occurs at superficial gas velocities of 15–20 m/s. Characteristic
compared with the WMS experimental data for superficial liquid frequencies found in this work can be related to interfacial wave
velocities of VSL = 0.01 m/s and VSL = 0.10 m/s. The mechanistic characteristics in annular flows. Axial and transversal wave
model shows fair agreement for all the ranges of superficial gas information can be extracted from characteristic frequencies
velocities and viscosities, showing a maximum deviation of 34% but further work is needed in this area.
for unstable annular flow at VSG = 10 m/s, VSL = 0.10 m/s and  Comparison with Quick Closing Valve void fraction data avail-
lL = 40 cP. Although the model is based on the combined momen- able in literature for similar experimental conditions show
tum equation for annular flows, it is able to capture the trend of void similar trends.
fraction data, but the model tends to over predict void fraction val-  Although the mechanistic model of Shoham [2] based on the
ues, particularly for higher liquid velocities and lower gas velocities. combined momentum equation for annular flows captures the
trend of void fraction data, the model tends to overpredict void
7. Conclusions fraction values, particularly for higher liquid velocities.

A dual Wire-Mesh Sensor with a sampling frequency of The Wire-Mesh Sensor technique is a useful tool in the investi-
10,000 Hz was used to detect the local instantaneous cross-section gation of two-phase flows, since it can show the flow in detail with
distribution of the phases in gas–liquid flows under annular and different degrees of complexity. Based on this experimental data,
liquid loading conditions in a vertical 76 mm ID vertical pipe. more information about the behavior of gas–liquid flows can be
The fluids used were air and either water or water mixed with Car- obtained, and the experimental data can support the development
boxy Methyl Cellulose (CMC). Data were obtained for a range of of new theoretical models or the improvement of existing models.
superficial velocities that varied from 0.005 to 0.1 m/s for liquid
and from 10 to 40 m/s for gas. The signals of the sensor, that are
proportional to the liquid conductivity, are processed to obtain Acknowledgment
void fraction time series, cross-sectionally averaged time series,
local void fraction distributions, structure frequency and structure This material is based upon work supported by the member
velocities. The effects of the superficial liquid and gas velocities companies of the Erosion/Corrosion Research Center (E/CRC) and
and liquid viscosity on the average and local void fractions have collaboration with Helmholtz-Zentrum Dresden-Rossendorf
been experimentally investigated. The analysis included time (HZDR). Thanks are also due to Ed Bowers, senior technician at
 R.E. Vieira et al. / Experimental Thermal and Fluid Science 64 (2015) 81–93
E/CRC for constructing and maintaining the boom loop facility at
University of Tulsa North Campus.
References
[1] K. Brown, The Technology of Artificial Lift Methods, PennWell Publishing Co.,
Tulsa, OK, USA, 1980.
[2] C. Deng, K. Adams, T. MacFarlane, The Effects of Flow Pattern Transitions of
Three-Phase Flows on Corrosion, in: Proceedings of Corrosion Conference and
Expo (Corrosion 2008), Paper No. 08565, NACE International, 2008.
[3] O. Shoham, Mechanistic Modeling of Gas-Liquid Two-Phase Flow in Pipes,
Society of Petroleum Engineers, Richardson, Texas, USA, 2006.
[4] B.J. Azzopardi, Gas-Liquid Flows, Begell House Inc., New York, USA, 2006.
[5] S.P. Belfroid, W. Schiferli, G.J. Alberts, C.A. Veeken, E. Biezen, Prediction Onset
and Dynamic behavior of Liquid Loading Gas Well, in: SPE Annual Technical
Conference and Exhibition, SPE115567, vol. 3, 2008, pp. 1528–1536.
[6] M. Parsi, R.E. Vieira, C.F. Torres, N.R. Kesana, B.S. McLaury, S.A. Shirazi, E.
Schleicher, U. Hampel, Characterizing Slug/Churn Flow Using Wire Mesh
Sensor, in: Proceedings of the ASME 4th Joint US-European Fluids Engineering
Summer Meeting (FEDSM 2014), paper no. FEDSM2014-21678, Chicago,
Illinois, USA, August 3–7, 2014.
[7] M.B. Alamu, Identification of Periodic Structures in Gas-Liquid Flow, PhD
Dissertation, School of Chemical & Environmental Engineering, University of
Nottingham, UK, 2010.
[8] M.S. Zangana, Film Behaviour of Vertical Gas-Liquid Flow in a Large Diameter
Pipe, PhD Dissertation, School of Chemical & Environmental Engineering,
University of Nottingham, UK, 2011.
[9] M. Yuan, Liquid Loading of Gas Wells, MSc Thesis, McDougall School of
Petroleum Engineering, University of Tulsa, USA, 2011.
[10] M. Guner, Liquid Loading of Gas Wells with Deviations from 0° to 45°, MSc
Thesis, McDougall School of Petroleum Engineering, University of Tulsa, USA,
2012.
[11] A. Skopich, E. Pereyra, C. Sarica, M. Kelkar, Pipe Diameter Effect on Liquid
Loading in Vertical Gas Wells, in: SPE Production and Operation Symposium,
SPE164477, Oklahoma City, Oklahoma, USA, March 23–26, 2013.
[12] G.P. Van Der Meulen, Churn-Annular Gas-Liquid Flows in Large Diameter
Vertical Pipes, PhD Dissertation, School of Chemical & Environmental
Engineering, University of Nottingham, UK, 2012.
[13] M. Abdulkadir, D. Zhao, S. Sharaf, L. Abdulkareem, I. Lowndes, B. Azzopardi,
Interrogating the effect of 90 degree bends on air-silicone oil flows using
advanced instrumentation, Chem. Eng. Sci. 66 (2011) 2453–2467.
[14] Z. Zhang, M. Bieberle, F. Barthel, L. Szalinski, U. Hampel, Investigation of
upward cocurrent gas-liquid pipe flow using ultrafast X-ray tomography and
wire-mesh sensor, Flow Meas. Instrum. 32 (2013) 111–118.
93
[15] R.E. Vieira, N.R. Kesana, C.F. Torres, B.S. McLaury, S.A. Shirazi, E. Schleicher,
U. Hampel, Experimental investigation of horizontal gas-liquid stratified
and annular flow using wire-mesh sensor, ASME J. Fluids Eng. 136 (2014)
1–16.
[16] R.G. Turner, M.G. Hubbard, A.E. Dukler, Analysis and prediction of minimum
flow rate for the continuous removal of liquids from gas wells, J. Pet. Tech. 21
(1969) 1475–1482.
[17] W. Henstock, T.J. Hanratty, The interfacial drag and the height of the wall layer
in annular flow, AIChE J. 22 (1976) 990–1000.
[18] J.C. Asali, T.J. Hanratty, P. Andreussi, Interfacial drag and film height for vertical
annular flow, AIChE J. 31 (1985) 895–902.
[19] M. Ishii, K. Mishima, Droplet entrainment correlation in annular two-phase
flow, Int. J. Heat Mass Transf. 32 (1989) 1835–1846.
[20] R.V. Oliemans, B.F. Pots, N. Trompe, Modeling of annular dispersed two-phase
flow in vertical pipes, Int. J. Multiph. Flow 12 (1986) 711–732.
[21] I.N. Alves, E.F. Caetano, K. Minami, O. Shoham, Modeling annular flow behavior
for gas wells, SPE Prod. Eng. 4 (1991) 435–440.
[22] H. Prasser, A. Böttger, J. Zschau, A new electrode-mesh tomograph for gas/
liquid flows, Flow Meas. Instrum. 9 (1998) 111–119.
[23] I. Kataoka, M. Ishii, Entrainment and deposition rates of droplets in annular
two-phase flow, in: Proceedings of the ASME-JSME Thermal Engineering Joint
Conference, Paper No. CONF-830301-11, Honolulu, HI, USA, 1983.
[24] R.E. Vieira, N.R. Kesana, C.F. Torres, B.S. McLaury, S.A. Shirazi, E. Schleicher, U.
Hampel, Experimental investigation of the effect of 90 degrees standard elbow
on horizontal gas-liquid stratified and annular flow characteristics using dual
wire-mesh sensors, Exp. Thermal Fluid Sci. 59 (2014) 72–87.
[25] E. Pereyra, C. Torres, FLOPATN – Flow Pattern Prediction and Plotting
Computer Code, The University of Tulsa, Tulsa, Oklahoma, USA, 2005.
[26] D. Barnea, A unified model for predicting flow pattern transitions for the whole
range of pipe inclinations, Int. J. Multiphase Flow 13 (1987) 1–12.
[27] S. Orfanidis, Optimum Signal Processing: An Introduction, McGraw-Hill, New
York, USA, 2007.
[28] F. Stern, M. Muste, M. Beninati, W. Eichinger, Summary of Experimental
Uncertainty Assessment Methodology, Iowa Institute of Hydraulic Research
Technical Report No. 406, College of Engineering, The University of Iowa,
1999.
[29] G. Costigan, P.B. Whalley, Slug flow regime identification from dynamic void
fraction measurements in vertical air-water flows, Int. J. Multiphase Flow 23
(1997) 263–282.
[30] M. Abdulkadir, Experimental and Computational Fluid Dynamics (CFD) Studies
of Gas-Liquid Flow in Bends, PhD Dissertation, School of Chemical &
Environmental Engineering, University of Nottingham, UK, 2011.
[31] Q. Zhang, Q. Wang, C. Sarica, J.P. Brill, Unified model for gas-liquid pipe flow
via slug dynamics-Part 1: Model development, ASME J. Energy Resources
Technol. 125 (2003) 266–273.