Experimental Thermal and Fluid Science 59 (2014) 7287

Contents lists available at ScienceDirect

Experimental Thermal and Fluid Science

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

Experimental investigation of the effect of 90 standard elbow on

horizontal gasliquid stratied and annular ow characteristics using
dual wire-mesh sensors
Ronald E. Vieira a,, Netaji R. Kesana a, Brenton S. McLaury a, Siamack A. Shirazi a, Carlos F. Torres b,
Eckhard Schleicher c, Uwe Hampel c
a

Erosion/Corrosion Research Center, The University of Tulsa, 800 South Tucker Drive, Tulsa, OK 74104, United States
Thermal Science Department, University of Los Andes, Mrida, Mrida 5101, Venezuela
c
Helmholtz-Zentrum Dresden-Rossendorf (HZDR), Dresden, Saxony 01328, Germany
b

a r t i c l e

i n f o

Article history:
Received 17 December 2013
Received in revised form 25 June 2014
Accepted 2 August 2014
Available online 12 August 2014
Keywords:
Wire-mesh sensor
Gasliquid ow
Elbow
Void fraction
Phase distribution
Flow visualization

a b s t r a c t
Fluid owing through pipelines often encounters ttings such as elbows. Although it is true that twophase ow patterns observed in elbows are qualitatively the same as those seen in straight pipes, the
presence of a pipe elbow can modify relative positions and local velocities of the two phases as they
are subjected to forces in addition to those encountered in a straight pipe. That redistribution can affect
pressure drop values, chemical inhibitor concentration and distribution to the top of the pipe, as well as
the erosion pattern occurring from solid particles such as sand that is entrained in oil and gas transportation pipelines. In this work, a wire-mesh sensor technique based on conductance measurements of void
fraction was applied to investigate two-phase pipe ow through a standard elbow. The horizontal ow
test section, consisting of a 76.2 mm ID, 18 m long pipe, was employed to generate stratied-wavy and
annular ow conditions. Two 16  16 wire-mesh conguration sensors were positioned either 0.9 m
upstream or 0.6 m downstream of the bend. The experiments were conducted at supercial liquid velocities equal to 0.03 m/s and 0.2 m/s and supercial gas velocities that ranged from 9 m/s to 34 m/s. The
effects of liquid viscosity on the measured parameters are also investigated using two different viscosities
of 1 and 10 cP. Stratiedslug transition, stratied wavy and annular ow patterns were observed visually in the clear section placed upstream of the wire-mesh sensors. Analysis of time series void fraction
data from the dual wire-mesh sensors allows the determination of mean void fraction, local time average
void fraction distribution, liquid phase distribution around the tube periphery, interfacial structure velocities, as well as probability density function characteristic signatures within the cross-section of pipe
before and after the elbow. The results indicate that the distribution of gas and liquid phases and interfacial velocities are signicantly altered even 20 diameters downstream of the elbow.
2014 Elsevier Inc. All rights reserved.

1. Introduction
Pipe ttings such as valves, bends, elbows and tees signicantly
affect multiphase ow distribution including generation of secondary ows, uctuations in void fractions and pressure losses and
velocities of gas and liquid phases. Elbows are often used in oil
and gas production systems, and they cause redistribution of gas
and liquid which can affect distribution of corrosion inhibitors
within and downstream of the bends. Elbows are also a location
Corresponding author.
E-mail addresses: rev87@utulsa.edu (R.E. Vieira), nrk301@utulsa.edu (N.R.
Kesana), brenton-mclaury@utulsa.edu (B.S. McLaury), siamack-shirazi@utulsa.edu
(S.A. Shirazi), ctorres@ula.ve (C.F. Torres), e.schleicher@hzdr.de (E. Schleicher),
u.hampel@hzdr.de (U. Hampel).
http://dx.doi.org/10.1016/j.expthermusci.2014.08.001
0894-1777/ 2014 Elsevier Inc. All rights reserved.

susceptible to the impact of particles along the outer radius. The

requirements for optimal design and evaluation of Computational
Fluid Dynamics (CFD) codes and two-phase ow models such as
VOF (volume of uid) that are being developed to predict details
of multiphase ows force the need for quantitative information
about ow through and downstream of elbows.
Previous research available in literature concerning gasliquid
ows through elbows, only dealt with an accurate calculation of
pressure drop [1,2]. However, in applications such as erosioncorrosion of pipe bends, more detailed information of multiphase
ow such as velocities of liquid and gas phases and interfaces
are typically required. Computational Fluid Dynamics (CFD) codes
are successfully used to predict pipe erosion and even corrosion
through elbows and piping systems involving a single-phase car-

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

73

Nomenclature
A
cross-sectional area of the pipe (m2)
Br
systematic uncertainty or bias limit
Cxy, Cxx, Cyy covariance functions between x(t) and y(t)
c
coverage factor
D
pipe diameter (m)
E[]
expected value of []
L
pipe length (m)
N
number of samples, ()
n
time level ()
ID
internal diameter
Pr
precision limit

rier uid. Erosion is a gradual removal of the material from the

pipe wall due to the repeated impacts of solid particles entrained
in the production uids. Previous investigations have proved that
erosion is more prominent in pipe elbows than straight pipes
[3,4]. Also, it is known that the mechanism of erosion in multiphase ows depends on the ow pattern [5,6] and distribution
of gas and liquid phases.
Similarly, owing uids can signicantly affect corrosion rates
in a number of ways including accelerated mass transfer of reactants and corrosion products [7], distribution of corrosion inhibitors, uid shear or impingement of solid particles in the uid
which disrupt protective layers [8]. Also, if CO2 is present in the
transported uids, steel pipelines can be corroded as this process
is probably enhanced by slug ow turbulence [9]. The ability to
model and predict pipeline degradation processes allows service
intervals to be better timed so as to reduce unnecessary checks
while not being subject to costly downtime due to equipment failure. Also, modeling ability can be applied at the design stage to
reduce the susceptibility of parts to wear.
The problem of determining the ow characteristics in the
downstream section of an elbow also is important in design and
hydrodynamic analysis of the uid transportation systems. When
ow enters the curved portion, the heavier density phase is subjected to a large centrifugal force, which causes the liquid to move
away from the center of curvature. This redistribution can signicantly affect erosion/corrosion processes. In order to predict complex erosion and corrosion patterns within pipe bends, predictions
of gas and liquid velocities are becoming a growing target for many
investigators dealing with erosion and corrosion issues. This along
with relative success of CFD codes to predict details of multiphase
ow, including gas and liquid velocities have motivated investigators to conduct local measurements in multiphase ow. Thus,
these investigation efforts were concentrated to obtain experimental data that are required for evaluation of CFD models as they are
being used to calculate details of multiphase ows. Reliable CFD
calculations can help investigators to predict erosion and corrosion
in multiphase ow through bends and other geometries as data
and more information become available.
In this experimental study, a dual wire-mesh sensor (WMS)
technique based on conductance measurements has been utilized
to gather data and investigate details of two-phase ows in a pipe
before and after a 90 horizontal-to-horizontal elbow. The wiremesh sensor allows detailed measurement of the two-phase ow
due to its outstanding spatial and temporal resolution. From the
measurements, specic parameters of interest such as cross-section
time-averaged void fraction, local time-averaged void fraction distribution, gasliquid interfacial characteristics, probability density
functions and periodic structure of interface velocities have been
extracted that can be used by investigators to evaluate and improve
capabilities of CFD codes to predict details of multiphase ows.

Rxy
StdDev
t

ag
rxy
qxy
s
lL
lx
ly

cross-correlation function
standard deviation
time (s)
void fraction ()
averaged product
correlation coefcient function
temporal lag obtained from cross-correlation (s)
liquid viscosity
mean value of the random process x(t)
mean values of the random process y(t)

2. Background on characterization of multiphase ows in

elbows
In horizontal pipes carrying gasliquid two-phase ows, gravity
introduces asymmetry to the ow regimes generated. The density
difference between the phases causes the liquid to travel preferentially along the bottom of the tube. In stratied ow, the liquid
travels along the bottom of the pipe while the gas passes over it.
At low velocities, the interface between the gas and liquid is
smooth. At higher gas velocities, the shearing action of the gas at
the interface generates small two-dimensional waves (stratiedwavy ow). In stratied gasliquid horizontal pipe ow, growing
long wavelength waves may reach the top of the pipe and form
slug ow or evolve into roll-waves [10]. This ow pattern subregion is called in this work stratiedslug transition and is characterized by the presence of liquid disturbances which have the
appearance of slugs. These disturbances can touch the top of the
tube momentarily, but do not block the entire pipe section [10].
At higher gas rates, the liquid slug is pierced by a gas core and
the ow becomes essentially annular. At extremely high gas ow
rate horizontal annular ows, the liquid lm variation from top
to bottom can be small, but at low gas ows the lm is signicantly
thicker at the bottom of the pipe.
Investigations into the conditions of ow regimes experienced
in the pipe sections upstream and downstream of bends have been
restricted to single-phase ow, and only a few authors have published results for two phase gasliquid ows. For instance, Gardner
and Neller [11] carried out visual and experimental studies for
bubble/slug ow using a pipe of 76 mm diameter in a vertical
90 elbow with radii of curvature of 305 and 610 mm. A traversing
probe was used to measure local time average of gas (air) concentration. Various features of the ow were described and used to
interpret the balance between centrifugal and gravitational forces.
Usui et al. [12,13] measured the averaged void fraction over the
bend using quick closing valves. Das et al. [14] and Bandyopadhyay
et al. [15] reported an experimental investigation for gas-nonNewtonian liquid ow through 12.7 mm bends. They developed
an empirical correlation for predicting the frictional pressure drop
across the piping components. Ribeiro et al. [16] measured drop
sizes upstream and downstream of a 90, 32 mm internal diameter
bend in a horizontal plane. Their results show that the downstream
drop sizes are 5075% of those upstream caused by thicker lms
resulting from the initial deposition and gathering of the lm
induced by the secondary ow in the gas phase. They concluded
that the ratio of downstream/upstream entrained fraction
decreases with increasing upstream entrainment fraction. The fractions of drops of different sizes which deposit on the bend have
been studied by James et al. [17]. Azzi et al. [18] concluded that
in a two-phase ow, there are additional effects of dissipation
due to separation and remixing of the gas and liquid phases, and

74

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

that the gravitational force inuences the ow behavior in both

vertical and horizontal ows. Pressure drop data were reported
by Spedding and Benard [19] for two phase airwater ow through
a vertical to horizontal 90 elbow bend set in 26 mm ID pipe.
It may be concluded from an analysis of the results of most of
the research studies reported in the literature that multiphase ow
studies in 90 elbows were carried out in small diameter pipes
with airwater as the model uids. Emphasis was on the determination of the pressure drop, phase distribution and droplet deposition downstream the bends. The change of ow structure before
and after the bends was mainly obtained by visualization, and
the local void fraction and interfacial structures were not measured
and analyzed. For that reason, it is important to study the effect of
90 pipe elbows on two-phase ow for the optimal, efcient and
safe design and operation of ow systems.
Recently, a few local measurements were conducted across an
elbow using a wire-mesh sensor. Flow patterns were identied
by Abdulkadir et al. [20] using both the reconstructed images of
WMS as well as the characteristic signatures of probability density
function (PDF) plots of the time series of cross-sectional averaged
void fraction. In another study, Kesana [21] identied different slug
characteristics such as the liquid slug body length distribution and
frequency of the slugs from void fraction data obtained using dual
wire-mesh sensors in the horizontal ow orientation. Also, the differences in the characteristics of slug ow and pseudo-slug ow
conditions in the upstream and downstream locations of an elbow
were examined.
The current investigation focuses on the two-phase stratied
slug transition, stratied-wavy and annular ow characteristics
before and after a horizontalhorizontal stainless steel pipe elbow
with an internal diameter of 76.2 mm, 114 mm radius of curvature,
using a dual wire-mesh sensor. The main objective is to provide
information and data that may be used to validate CFD simulations
of multiphase ow through elbows.
3. Experimental facility, program and data analysis procedures
A dual wire-mesh sensor (Fig. 1) is used to take cross-sectional
phase distribution measurements. The sensor and the corresponding electronics were supplied by the HZDR Dresden, Germany. In
this study, a dual 16  16 wire conguration sensor was used as
depicted in Fig. 2.

Fig. 1. Dual wire mesh sensor (WMS).

Fig. 2. Detail of the 16  16 wire mesh sensor (WMS).

Each wire-mesh sensor consists of two parallel wire layers perpendicular to the pipe axis. The wires are made of stainless steel
316L. The spacing between the layers is 1.5 mm; each layer is built
of sixteen 0.125 mm diameter parallel steel wires. The wires in the
receiver layer are perpendicular to those in the transmitter. Therefore, a rectangular mesh with 256 junctions is obtained; only 192
are considered to be located within the pipe cross section. The
operation principle of the wire-mesh sensor is based on the difference in the electrical conductivity of the two phases (water and
air). Each wire in the transmitter plane is periodically activated
by a multiplexer circuit by a short voltage pulse. An additional
multiplexer circuit is used to connect consecutively each one of
the receiver wires during the single pulse supplied to the transmitter wire. Each output pulse depends on the local instantaneous
conductivity at the crossing of the transmitter and the receiver
wires.
Calibration of the sensor was performed before each experimental run. The output voltages at each junction in a pipe lled
with water and in an empty pipe were recorded. Following Prasser
et al. [22], the variation of the output voltage between these limiting values was assumed to vary linearly with the local instantaneous liquid holdup, thus enabling conversion of the recorded
voltages into void fractions, using the calibration coefcients
determined before the beginning of each experimental run. The
raw data were recorded as a sequence of voltages in each junction.
The voltages were further translated into the local instantaneous
void fraction. The resolution in the axial direction is determined
by the interface propagation velocity, as well as by the sampling
frequency. The instrument is capable of sampling at frequencies
up to 10 kHz for the whole cross section; in all the cases the maximum sampling frequency was used. The wire-mesh sensor thus
provides high-frequency data on the temporal variation of the local
void fraction at each junction of the mesh. Further details about the
operation of the WMS can be found in Prasser et al. [22].
The current studies were performed using the Erosion/Corrosion Research Center (E/CRC) large scale tower-boom loop facility
located at The University of Tulsa. The loop main components are
two diesel compressors, electrical compressor, slurry tank, test section, diaphragm pumps, and separator. The test section in this large
scale facility can be modied to accommodate different pipe diameters and can be oriented between 0 (horizontal) to 90 (vertical).
For these experiments, the pipe diameter was 76.2 mm (3 in.) and
18 m long.
Experiments were performed horizontally and near atmospheric pressure conditions. The uids used were air and either
water or water mixed with Carboxy Methyl Cellulose (CMC) to
increase the liquid viscosity. In order to measure changes on the

75

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

surface tension of water and CMC mixtures, Rodriguez [23] used a

Du-Noy ring tensiometer. It was found that when the liquid viscosity increased from 1 to 10 cP, the surface tension decreased
from 73 dyn/cm to 70 dyn/cm. The air required for the testing
was supplied by diesel compressors. The outlet of the test section
is connected to a separator, where air exhausts from the top and
liquid is pumped back to a 250 gallon liquid tank for recirculation
due to the considerable amounts of liquid required to maintain
stratied and annular ow inside the test section. A schematic of
the experimental facility is shown in Fig. 3. More details about
the experimental facility can be found in Vieira et al. [24]. First,
the WMS was placed at a distance 17 m downstream from the inlet
of the test section (L/D = 233) and 0.9 m upstream of the bend.
Then, all experiments were repeated with the WMS installed
0.6 m downstream of the bend. A photograph of the WMS in the
downstream location of the elbow is shown in Fig. 4.
Experiments were performed at different supercial gas velocities, two supercial liquid velocities and two liquid viscosities.
Fourteen experiments were performed in the upstream location
of the elbow. Sixteen experiments were performed in the downstream location of the elbow for nearly the same ow conditions
as for the upstream cases. A reasonable estimation of the systematic uncertainty propagation was performed for liquid and gas
supercial velocities in this study. The estimation of the systematic
uncertainty propagation was obtained previously by Vieira et al.
[25]. Tables 1 and 2 present the percentage error (uncertainty) of
liquid and gas supercial velocities, respectively.
Fig. 5 presents the test matrix for the upstream section measurements for 1 cP liquid viscosity conditions. Two supercial
liquid velocities, VSL, were used: 0.03 m/s and 0.2 m/s. On the other
hand, the supercial gas velocities, VSG, varied from 9 m/s to 34 m/
s. The test matrix is plotted on ow pattern maps to observe the
transitions from one ow regime to another as predicted by Barneas model [26].
The average void fraction measurements were obtained from
600,000 samples over a specied time interval of sixty seconds at
a xed test condition. The precision limit of the result Pr can be calculated, assuming no correlated precision limits for the precision
errors [27], using the following equation:

measurement. The systematic uncertainty, Br, is obtained from the

water calibration le measured for each liquid viscosity for both
upstream and downstream locations. The precision limit of the data
and systematic uncertainty for each ow condition are calculated
and presented in Tables 2 and 3 along with the supercial gas
and liquid velocities, liquid viscosities, and cross-sectional averaged
void fraction measurements.
For 1 cP liquid viscosity, all experiments fall within the range of
stratied and annular ow patterns. As the liquid viscosity increased
from 1 cP to 10 cP, some highly aerated slug ows (hydraulic jumps)
were observed for low supercial gas velocities. For each experiment, data from the WMS was obtained with a frequency of
10 kHz for a period of 60 s (see Table 4).
The WMS data were analyzed applying signal processing techniques such as probability density function (PDF) and cross-correlation techniques implemented in MATLAB.
The probability density function is a function that describes the
relative likelihood for a random variable to take on a given value.
The PDF returns a probability density estimate for the sample in
the vector x. The estimate is based on a normal kernel function
and is evaluated at 100 equally spaced points, xi, that cover the
range of the data in x. The PDF was used to determine the twophase ow pattern using the WMS data from the rst plane of
the WMS.
A cross-correlation was performed between the time varying
void fraction data measured by the WMS-planes for each upstream
and downstream location. This allows the determination of the
time for frequencies and velocities of periodic structures between
the planes. The cross-correlation gives the degree of linear dependence between two time series data sets, x and y. It was calculated
as the average product of x  lx and y  ly. The averaged product
dened in Eq. (2) gives the covariance of x and y in the limit as the
sample size approaches innity.

Pr c  StdDevag

h
i

C xy E xt  lx fyt s  ly g
Z
i
1 Th
fxt  lx gfyt s  ly g dt
lim
T!1 T 0

where c is the coverage factor and StdDev(ag) is the standard deviation of the sample of N readings of the void fraction ag. For N P 10,
c = 2 is assumed. The value of the standard deviation is calculated
from N readings over the specied time interval of the void fraction

Dual Wire Mesh Sensor

Boom
Gas
and
Liquid

N
1X
xi  lx yi  ly
N!1 N
i1

rxy Ex  lx y  ly  lim

For any time delay s, the covariance function between x(t) and
y(t) is:

Rxy s  lx ly

Rxy s lim

Liquid
76.2 mm ID Test Section

where the cross-correlation function between x(t) and y(t), Rxy(s) is

dened as:

1
T!1 T

Test
17 mSections

xtyt sds

The correlation coefcient function is dened as:

Gas
V11

V12

V21

V22
Flow
meter
250 gallon tank

Gas
V1

Collecting
Tank

Stirrer
Pump 1.1)

Rxy s  lx ly
C xy s
q
qxy s p
C xx 0C yy 0
Rxx 0  l2x Ryy 0  l2y

Gas &
Liquid

Pump 1.2)

Diesel Compressors

Pump 2.1)

Pump 2.2)

Fig. 3. Schematic of large scale boom loop.

After calculating the cross-correlation between the two WMS

signals, the maximum of the cross-correlation function indicates
the point in time where the signals are best aligned, i.e. the time
delay between the two signals is determined by the argument of
the maximum of the cross-correlation. More details about the
cross-correlation techniques used in this work can be found in Bendat and Piersol [28] and Orfanidis [29]. Fig. 6 shows an example of
the application of cross-correlation to the output signals of the
dual WMS for a stratiedslug condition in the upstream and

76

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

Dual Wire
Mesh Sensor

Standard
Elbow

Flow
Direction

Fig. 4. Installed wire mesh sensor in the downstream section.

Table 1
Uncertainty of supercial liquid velocities.
VSL (m/s)

Error (%)

0.03
0.2

8.5
5

Table 2
Uncertainty of supercial gas velocities.
VSG (m/s)

Error (%)

9
18
27
30
32
34

1.8
1.9
2.0
2.1
2.2
2.2

Stratified
Wavy

10

VSL, (m/s)

SLUG

ANNULAR

0.1
STRATIFIED
SMOOTH

0.01

0.001
0.01

Test
(#)

VSG
(m/s)

VSL
(m/s)

lL

ag
()

Pr
()

Br
()

Flow pattern observed

(cP)

1
2
3
4
5
6
7
8
9
10
11
12
13
14

9
18
30
9
18
27
34
9
18
27
34
9
18
25

0.03
0.03
0.03
0.20
0.20
0.20
0.20
0.03
0.03
0.03
0.03
0.20
0.20
0.20

1
1
1
1
1
1
1
10
10
10
10
10
10
10

0.93
0.96
0.98
0.83
0.90
0.94
0.97
0.89
0.96
0.97
0.97
0.82
0.88
0.92

0.02
0.02
0.02
0.14
0.09
0.05
0.03
0.03
0.02
0.02
0.01
0.12
0.12
0.08

0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.04
0.04
0.04
0.04
0.04
0.04
0.04

Stratied-wavy
Stratied-wavy
Annular
Stratiedslug transition
Stratied-wavy
Annular
Annular
Stratied-wavy
Stratied-wavy
Annular
Annular
Stratiedslug Transition
Stratiedslug transition
Annular

Annular
Dispersed-Bubble

DISPERSED BUBBLE

Table 3
Experimental test matrix, cross-sectional averaged void fraction measurements and
precision limit of the data; horizontal airwater or airwater & CMC (upstream
section).

0.1

STRATIFIED
WAVY

10

100

VSG, (m/s)
Fig. 5. Investigated area of ow conditions for stratied and annular ow
conditions; 1 cP water viscosity.

downstream sections of the elbow. The maximum cross correlation

coefcient is obtained at time delay of s = 0.0144 s and s = 0.0115 s
for the upstream and downstream locations, respectively.

Table 4
Experimental test matrix, cross-sectional averaged void fraction measurements and
precision limit of the data; horizontal airwater or airwater & CMC (downstream
section).
Test
(#)

VSG
(m/s)

VSL
(m/s)

lL

ag
()

Pr
()

Br
()

Flow pattern observed

(cP)

15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

9
18
27
36
9
18
27
36
9
18
27
36
9
18
27
36

0.03
0.03
0.03
0.03
0.20
0.20
0.20
0.20
0.03
0.03
0.03
0.03
0.20
0.20
0.20
0.20

1
1
1
1
1
1
1
1
10
10
10
10
10
10
10
10

0.94
0.97
0.98
0.99
0.85
0.91
0.96
0.98
0.91
0.95
0.96
0.97
0.82
0.90
0.94
0.97

0.02
0.02
0.02
0.02
0.15
0.07
0.04
0.02
0.03
0.03
0.02
0.01
0.14
0.08
0.04
0.02

0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.02

Stratied-wavy
Stratied-wavy
Annular
Annular
Stratied-wavy
Annular
Annular
Annular
Stratiedslug Transition
Stratied-wavy
Annular
Annular
Stratiedslug Transition
Stratied-wavy
Annular
Annular

4. Results and discussion

4.1. Sequence of cross-section images
Figs. 7 and 9, and Figs. 8 and 10 show the sequence of instantaneous cross-sectional void fraction for stratiedslug transition,
stratied wavy and annular ows, respectively, obtained by the
WMS raw data upstream and downstream the elbow. The experimental conditions are shown in the ow pattern map previously
described (Fig. 5). Fig. 7 shows that in stratiedslug transition
ows, waves can grow very fast, and some of them can ll the tube
generating pseudo-slugs. Fig. 7d shows that the level of liquid in

the pipe can be higher than half the diameter of the pipe. The wave
instability is the most notable feature of the stratied to slug transition. Fig. 8 shows that in the downstream section more liquid is
in the perimeter of the cross-sectional area of the pipe and waves
are smaller as compared with the upstream conditions. As can be
observed for the stratied ow (Fig. 9) at higher gas ow rates,
the holdup is very small. Fig. 9 shows that the average liquid layer
is thinner than the average liquid layer observed in Fig. 7a and c.
The liquid lm thickness decreases with increasing supercial
gas velocities for a xed supercial liquid velocity of 0.2 m/s.
Fig. 10 shows the downstream conditions where the liquid layer

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

1.2
Upstream
Downstream

Rxy()

0.8
0.6
0.4
0.2
0
-0.2
-10

-5

10

Time, (s)
Fig. 6. Cross-correlation sample result between two planes of the dual wire mesh
sensor; stratiedslug transition (VSG = 9 m/s, VSL = 0.2 m/s and, lL = 1 cP).

in the bottom part of the pipe tends to climb the inner and outer
pipe walls in an oscillating pattern.
4.2. Cross-sectionally averaged void fraction
Figs. 1113 show plots of instantaneous cross-section void fraction averages obtained from the output of the WMS placed
upstream and downstream of the elbow for stratiedslug, stratied wavy and annular ows, respectively. The observed stratied
slug transition pattern in Fig. 11 is characterized by the intermittent appearance of pseudo-slugs or highly-aerated slugs which
bridge the entire pipe section. Previous investigations [10,30] also
have identied this phenomenon by the occurrence of signicant
pressure uctuations in the stratiedslug transition. Similar uctuations are observed in the upstream section of the pipe as
observed in the downstream section of the elbow.
Fig. 12 shows that the mean void fraction in the upstream location for stratied ow (ag = 0.90) is smaller than for annular ow
(ag = 0.97) which is shown in Fig. 13. The periodic structures
obtained from the cross-sectionally averaged void fraction time
series show similar structures in the upstream and downstream
locations for stratied and annular ows. In both cases, the

77

uctuations of the void fraction values show the waviness of these

ow patterns.
The mean void fraction across a pipe cross-section was determined by averaging the cross-sectional void fraction data over an
interval of 60 s at a data acquisition frequency of 10 kHz. The mean
void fraction indicates the general trends of the measurements
with the main ow variables (VSG and VSL). This allows a simple tool
to analyze and relate multiphase ow data to other complex phenomena such as sand transport, corrosion and erosion studies. The
effect of the gas supercial liquid and gas velocities on the mean
void fraction, before and after the horizontal 90 elbow for a liquid
viscosity of 1 cP, is presented in Fig. 14. The uncertainty bars on the
graph represent the precision limit of the results. The mean void
fraction increases monotonically with the gas supercial velocity.
However, the increase at the lower gas supercial velocity is much
sharper than that at the higher gas supercial velocity. Fig. 15
shows the mean void fraction as a function of the gas and liquid
supercial velocities for a liquid viscosity of 10 cP. As the liquid
viscosity is increased, more shear between the liquid phase and
the pipe wall exists and a slight decrease in mean void fraction
for all gas supercial velocities is observed. On the other hand, similar trends of supercial gas velocities and mean void fraction are
observed when the liquid viscosity was increased.
4.3. Contours of cross-sectionally time averaged local void fractions
Figs. 1621 show the contours of cross-sectional time averaged
local void fractions for stratiedslug transition, stratied wavy
and annular ow. Figs. 16 and 17 show that in the downstream location more liquid tends to climbs the inner wall of the pipe in comparison with the upstream location for stratiedslug and
stratied wavy conditions. In this mechanism, a secondary ow in
the gas phase is created as a result of the variation of wave height
on the liquid lm around the periphery. The differences in wave
height give differences in the apparent roughness of the interface
as seen by the gas phase and this creates a secondary ow with
the gas moving upwards near the interface and downwards in the
center of the pipe. This secondary ow can create a circumferential

(a) t = 10 s

(b) t = 20 s

(c) t = 30 s

(d) t = 40 s

(e) t = 50 s

(f) t = 60 s

Fig. 7. Sequence of frames at 10 s intervals showing the location of the gas and liquid (in blue) for stratiedslug transition in the upstream section of the elbow (VSG = 9 m/s,
VSL = 0.2 m/s and, lL = 1 cP). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

78

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

(a) t = 10 s

(b) t = 20 s

(c) t = 30 s

(d) t = 40 s

(e) t = 50 s

(f) t = 60 s

Fig. 8. Sequence of frames at 10 s intervals showing the location of the gas and liquid (in blue) for stratiedslug transition in the downstream section of the elbow
(VSG = 9 m/s, VSL = 0.2 m/s and, lL = 1 cP). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

(a) t = 10 s

(b) t = 20 s

(c) t = 30 s

(d) t = 40 s

(e) t = 50 s

(f) t = 60 s

Fig. 9. Sequence of frames at 10 s intervals showing the location of the gas and liquid (in blue) for stratied-wavy ow in the upstream section of the elbow (VSG = 18 m/s,
VSL = 0.2 m/s and, lL = 1 cP). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

interfacial shear stress which can oppose gravity and transport the
liquid upwards.
Fig. 18 shows that for annular ows, the presence of the liquid
lm in the upper part of the tube is not clearly detected by the sensor. Several mechanisms have been proposed to explain the presence of very thin lms in the upper part of the tube in horizontal
annular ows, the more important and credible among these are
secondary ow mechanism [31] entrainment and re-deposition
mechanism [32], wave spreading mechanism [33] and pumping
action mechanism [34]. Figs. 19 and 20 show that if the liquid

viscosity is increased, lower values of void fraction are obtained in

the bottom of the pipe. This means that fewer bubbles are detected
by the wires placed in the bottom layers of the pipes. Fig. 21 shows
that the local values of averaged void fraction for annular ow in the
upstream and downstream locations of the elbow are almost the
same and are not affected by an increase of liquid viscosity.
The contour plots of the standard deviation distribution of the
time average void fraction are used to identify 2-D structures in
the liquid lm related with the amplitude of the interface oscillations in all ow patterns studied in this work.

79

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

(a) t = 10 s

(b) t = 20 s

(c) t = 30 s

(d) t = 40 s

(e) t = 50 s

(f) t = 60 s

Fig. 10. Sequence of frames at 10 s intervals showing the location of the gas and liquid (in blue) for stratied-wavy ow in the downstream section of the elbow (VSG = 18 m/s,
VSL = 0.2 m/s and, lL = 1 cP). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

0.8

Void fraction

Void fraction

0.8

0.6

0.4

0.6
0.4
Upstream
Downstream

0.2
0.2

Upstream
Downstream

10

15

20

Time, (s)
0

10

15

20

Time, (s)
Fig. 11. Representative cross-sectionally averaged time series of void fraction for
stratiedslug transition in the upstream and downstream section of the elbow
(VSG = 9 m/s, VSL = 0.2 m/s and, lL = 1 cP).

Fig. 13. Representative cross-sectionally averaged time series of void fraction for
annular ow in the upstream (VSG = 34 m/s), and downstream (VSG = 36 m/s) section
of the elbow (VSL = 0.03 m/s and, lL = 1 cP).

1.00
0.8

Mean void fraction

Void fraction

0.90
0.6
0.4
Upstream
Downstream

0.2
0

10

15

20

Time, (s)
Fig. 12. Representative cross-sectionally averaged time series of void fraction for
stratied-wavy in the upstream and downstream section of the elbow (VSG = 18
m/s, VSL = 0.2 m/s and, lL = 1 cP).

0.80
0.70
Upstream; Vsl=0.03 m/s
Downstream; Vsl=0.03 m/s
Upstream; Vsl=0.2 m/s
Downstream; Vsl=0.2 m/s

0.60
0.50
0.40

10

15

20

25

30

35

Superficial gas velocity, (m/s)

Fig. 14. Average mean void fraction, lL = 1 cP.

40

80

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

always equal to one, as conrmed by the standard deviation prole. As the liquid viscosity is increased, the standard deviation in
the bottom layers decreases because the liquid lm is traveling
at lower velocities, with fewer bubbles entrained. The analysis of
the standard deviation highlights that for VSG = 9 m/s and
VSL = 0.67 m/s with 10 cP, the presence of highly unstable liquid
lm can be veried by an increase of the standard deviation in
the core region.

1.00

Mean void fraction

0.90
0.80
0.70
Upstream; Vsl=0.03 m/s
Downstream; Vsl=0.03 m/s
Upstream; Vsl=0.2 m/s
Downstream; Vsl=0.2 m/s

0.60
0.50

4.4. Effect of the elbow on phase distribution close to the upper part of
the inner pipe wall

0.40
10

15

20

25

30

35

40

Superficial gas velocity, (m/s)

Fig. 15. Average mean void fraction, lL = 10 cP.

Figs. 2227 show the contour plots of standard deviation in the

cross-sectional area of the WMS. Figs. 22 and 23 illustrate the stochastic nature of liquid and gas distributions before and after the
bend for the stratiedslug transition and stratied-wavy conditions. Due to oscillations of the interface, typical of the interaction
between the liquid and gas phases, the signals of stratied ows
show higher standard deviation values in the bottom of the pipe
in the upstream section. On the other hand, in the downstream
section, centrifugal and gravitational forces in the gasliquid
stream leads to a broad scattered distribution of void fraction in
the bottom layers of the downstream location. Similar results are
obtained for 10 cP (see Figs. 25 and 26).
Under annular ow conditions, as shown in Figs. 24 and 27, in
the central region of the pipe (core region) the void fraction is

A crucial issue in uid transportation systems is the liquid

transport to the top of the pipe. It is vital (from the point of view
of ow assurance risk) to understand the relationship between
the continuous liquid lm around the periphery and the thicker
liquid pool at the bottom of the pipe. It is no problem to distribute
corrosion inhibitors in the pipe at high ow rates. Injected inhibitor liquids are then dispersed as a spray, and the droplets are carried along the line with the gas producing a protective lm all over
the internal pipeline surface [35]. However, in case of stratied
ows this mechanism of inhibitor distribution is no longer possible. Injected inhibitor liquids will collect at the bottom of the line
and will be transported along the line only at the bottom of the
pipeline. Under such conditions, predominantly the upper part of
the inner pipeline surface is not sufciently supplied with corrosion inhibitor and is, therefore, subject to sweet or sour corrosion,
resulting in localized corrosion and also corrosion cracking [35].
This phenomenon is generally referred to as the top-of-the-line
(TOL) corrosion [36].

Voidm
g(%)

95

90
80

80

60

30

20
5

10

95
90
80
60
30
20
5

95

95

90
80
70
60
50
40
30
20
10
5

Voidm
g(%)

95

95

95

95

80

90

70
50

70

30

90
70

80
30

40

20
5

(a) Upstream Location

(b) Downstream Location

90
80
70
60
50
40
30
20
10
5

Fig. 16. Cross-section local average void fraction for stratiedslug transition (VSG = 9 m/s, VSL = 0.2 m/s and, lL = 1 cP).

95
60

95

95

40

70
20

80

70
40
20
10

80

(a) Upstream Location

Voidm
g(%)

Voidm
g(%)

90
80
70
60
50
40
30
20
10
5

90
80
70
60
50
40
30
20
10
5

95

95

90
80

90
60

95

80

40

50 4 0
10

(b) Downstream Location

Fig. 17. Cross-section local average void fraction for stratied wavy ow (VSG = 18 m/s, VSL = 0.2 m/s and, lL = 1 cP).

81

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

95

90

80

95
90
90
80
70 80
70
40 30

95

Voidm
g(%)

Voidm
g(%)

90
80
70
60
50
40
30
20
10
5

90
80
70
60
50
40
30
20
10
5

95
90
70

95

95

90
70

(b) Downstream Location; VSG = 36 m/s

(a) Upstream Location; VSG = 34 m/s

Fig. 18. Cross-section local average void fraction for annular ow conditions (VSL = 0.2 m/s and, lL = 1 cP).

90
70
50
20

95

95

95
90

80

90

50

90
70

50

20
5

60
30

(a) Upstream Location

Voidm
g(%)

95

95
70

95
80
60
20
5

95

90
80
70
60
50
40
30
20
10
5

95

95

90
50
20

Voidm
g(%)

95

80

60
30

90
80
70
60
50
40
30
20
10
5

(b) Downstream Location

Fig. 19. Cross-section local average void fraction for stratiedslug transition (VSG = 9 m/s, VSL = 0.2 m/s and, lL = 10 cP).

Voidm
g(%)

95

95

90

50

90
70

70

20

20
5

95

50

90
80
70
60
50
40
30
20
10
5

20

(a) Upstream Location

95

95

95
80

90

90

60
30

60

95
80
60

Voidm
g(%)
90
80
70
60
50
40
30
20
10
5

30

(b) Downstream Location

Fig. 20. Cross-section local average void fraction for stratied-wavy ow (VSG = 18 m/s, VSL = 0.2 m/s and, lL = 10 cP).

In this section, the effect of the elbow on phase distribution

close to the upper part of the inner pipe wall in the upstream
and downstream sections is studied experimentally. The time series of local void fraction values along the top of the pipe is used to
analyze the effect of the standard elbow on the liquid lm around
the gas core. The selected area for the void fraction is shown in
Fig. 28.
The time series of averaged void fraction along the top of the pipe
for stratiedslug ow before and after the 90 elbow for a liquid
viscosity of 1 cP, is presented in Fig. 29. It can be observed that for

a supercial liquid velocity of 0.2 m/s, the minimum values of the

void fraction in the downstream section, which represent a quick
replenishment of the liquid, are lower than the values obtained in
the upstream section of the pipe. Because the effect of centrifugal
forces, less liquid is deposited at the bottom of the pipe and is more
uniformly distributed around the upper section of the pipe around
the gas core region. In Fig. 30, it can be observed that the uctuations of void fraction for stratied-wavy conditions in the downstream section are also substantially higher with respect to the
variation in the upstream location for the same supercial velocities

82

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

95

95

Voidm
g(%)

90
80
70
60
50
40
30
20
10
5

90
80
70
60
50
40
30
20
10
5

95

95
80

Voidm
g(%)

95

95

80

95 70
30

30

30

95
70
30

(b) Downstream Location; VSG = 36 m/s

(a) Upstream Location; VSG = 34 m/s

Fig. 21. Cross-section local average void fraction for annular ow (VSL = 0.03 m/s and, lL = 10 cP).

15
20
30

25

30
20

15

15
20

25

10
5

30

15

20

StdDev
g(%)
Voidst

10
15

20

25

30

30
25
20
15
10
5
1

30

30

20

10

10

10

30
25
20
15
10
5
1

15

10

10

StdDev
g(%)
Voidst

25

10

30

10

(a) Upstream Location

20
10

30

(b) Downstream Location

Fig. 22. Standard deviation of local average void fraction for stratiedslug transition (VSG = 9 m/s, VSL = 0.2 m/s and, lL = 1 cP).

5
10
20
20
10

20

15

25

1
10

20
30

15
20

10
20

30
25
20
15
10
5
1

30

30

30
20
10 5

15

25

30

30
25
20
15
10
5
1

StdDev
g(%)
Voidst

15

StdDev
g(%)
Voidst

(a) Upstream Location

30
20

20

(b) Downstream Location

Fig. 23. Standard deviation of local average void fraction for stratied-wavy ow (VSG = 18 m/s, VSL = 0.2 m/s and, lL = 1 cP).

and liquid viscosity. It can be concluded from the void fraction contour plots and time series that in the downstream section the presence of continuous liquid lm along the periphery of the tube is
more frequent than the upstream section of the elbow.
4.5. Reconstructed 3D images of stratiedslug transition and
stratied wavy ows
The iso-surface plots of the air phase in the two-phase ow
structures are obtained by xing the threshold of void fraction to
the individual air and liquid phases. In the current study, the

threshold values for the gas and liquid phases were xed to 0.9
and 0.1. Fig. 31 shows the structure of the waves and the slugs
formed at liquid viscosities of 1 cP and 10 cP. It can be observed
that with an increase of liquid viscosity from 1 cP to 10 cP, there
is a signicant change in the wave characteristics in the upstream
section. Therefore, a pseudo slug structure can be clearly observed
at 10 cP liquid viscosities. Fig. 31d shows that the downstream section of the pseudo-slug structure is even more aerated. This pseudo
slug structure may be described as liquid slug structures that lack
sufcient quantities of liquid to form a traditional slug. In stratied
ows, when a wave keeps growing, its amplitude will eventually

83

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

1
10
20
20

20

5
15

30
25
20
15
10
5
1

30
25
20
15
10
5
1

StdDev
g(%)
Voidst

StdDev
g(%)
Voidst

10

10

20

15

25

5
15
30

(b) Downstream Location; VSG = 36 m/s

(a) Upstream Location; VSG = 34 m/s

Fig. 24. Standard deviation of local average void fraction for annular ow (VSG = 0.03 m/s and, lL = 1 cP).

5
10

10

10
15

15

25
30
20

15

20

30
25

20
25

10

StdDev
g(%)
Voidst

10

30
25
20
15
10
5
1

StdDev
g(%)
Voidst

20

15

15

10

25

15

20

20

30

30

25

30
25
20
15
10
5
1

30

20

15

30

25

25 30

15
5

(a) Upstream Location

(b) Downstream Location

Fig. 25. Standard deviation of local average void fraction for stratiedslug transition (VSG = 9 m/s, VSL = 0.2 m/s and, lL = 10 cP).

StdDev
g(%)
Voidst

10

20
30

10
20

15
25

30

15
25

10
15

20
30

15

15

(a) Upstream Location

5
10

20

30
25
20
15
10
5
1

30

30
15

StdDev
g(%)
Voidst

30
25
20
15
10
5
1

20

30
10

20

(b) Downstream Location

Fig. 26. Standard deviation of local average void fraction for stratied-wavy ow (VSG = 18 m/s, VSL = 0.2 m/s and, lL = 10 cP).

be so large that the top of the wave hits the top of the pipe. When
the system is operating at relatively low pressures and at relatively
low gas and liquid ow rates, very long slugs with sizes reaching
several pipe diameters may form. These experiments clearly show
that the high viscosity liquids tend to promote slug ow.
Fig. 32 shows the structure of stratied-wavy ow at liquid viscosities of 1 cP and 10 cP. It can be observed that with an increase
of liquid viscosity from 1 cP to 10 cP, there is no signicant change
in the stratied gasliquid structure in the upstream section.

However, from Fig. 32d, it can be observed that in the downstream

section of the elbow, there is more liquid in the upper part of the
pipe.
4.6. Flow pattern identication using probability density functions
(PDF)
The probability density function was determined using the ks
density command in MATLAB. Fig. 33 shows the PDFs for six

84

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

StdDev
g(%)
Voidst

StdDev
g(%)
Voidst
30
25
20
15
10
5
1
1

1
5

1
15 0

10

30
25
20
15
10
5
1
5

10

10

20

1
5

10

5
15

10

15 5
2

(b) Downstream Location; VSG = 36 m/s

(a) Upstream Location; VSG = 34 m/s

Fig. 27. Standard deviation of local average void fraction for annular ow (VSL = 0.03 m/s and, lL = 10 cP).

Void fraction

0.95
0.9
0.85
0.8
0.75
Upstream
Downstream

0.7
0.65
0

10

20

30

40

50

60

Time, (s)
Fig. 30. Time series of averaged void fraction around the top of the pipe for
stratied-wavy ow (VSG = 18 m/s, VSL = 0.2 m/s; lL = 1 cP).

Fig. 28. Selected area for averaged void fraction time series around the top of the
pipe (in blue). (For interpretation of the references to color in this gure legend, the
reader is referred to the web version of this article.)

Void fraction

0.8
0.6
0.4
0.2
0

Upstream
Downstream

10

20

30

40

50

60

Time, (s)
Fig. 29. Time series of averaged void fraction around the top of the pipe for
stratiedslug transition (VSG = 9 m/s, VSL = 0.2 m/s and, lL = 1 cP).

representative experimental conditions in the test matrix. The gures are organized horizontally when the supercial gas velocity is
constant and vertically when the liquid supercial velocity is constant. As can be seen in Fig. 33, there are typical shapes or signatures as the supercial velocities change. A narrow single peak
for void fractions between 0.9 and 0.95 is typical of annular ow.

For the second type, a single peak at high void fraction with a small
tail down to lower void fractions is recognized as the signature of
stratied-wavy ow. As the gas velocity is increased, the lower
void fraction peak moves to higher values and the peaks appear
at void fractions of 0.97, which represent annular ows. As the
liquid velocity is increased, all dominant void fraction peak frequencies diminish. For lower gas velocities, a double peak is usually expected for slug ow. The observed peaks in the data are
not bimodal, these slugs are highly aerated not yet fully developed
and do not present the characteristic frequency for slug ow. This
is not surprising as the amount of liquid present in the pipe is too
small for slug ow to exist and as a result, a characteristic single
peak signature of stratied ow is still observed.
One interesting observation is that, for a constant VSG, the
observed differences between the maximum PDF values upstream
and downstream vary for both liquid velocities of 0.03 and 0.2 m/s.
For the lower liquid velocity, as the gas velocity increases from
9 m/s to 27 m/s, the differences between the maximum values
decrease; on the other hand, for higher liquid velocity of 0.2 m/s,
a slight increase of this difference is observed. A following broad
tail present in the PDF of the upstream section does not exist in
the PDF results for data collected after the elbow for supercial
gas and liquid velocities of 27 m/s and 0.2 m/s, respectively. The
secondary peak disappears and a single peak is observed at a maximum void fraction value.
4.7. Structure velocity using cross-correlation techniques
The velocities of the periodic structures, such as highly unstable
liquid lms and disturbance waves can be obtained by computing

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

85

(a) Upstream; L = 1 cP

(b) Dowstream; L = 1 cP

(c) Upstream; L = 10 cP

(d) Dowstream; L = 10 cP
Fig. 31. Inuence of liquid viscosity on the structure of liquid waves and slugs (VSG = 9 m/s, VSL = 0.2 m/s; lL = 1 cP).

(a) Upstream; L = 1 cP

(b) Dowstream; L = 1 cP

(c) Upstream; L = 10 cP

(d) Dowstream; L = 10 cP
Fig. 32. Inuence of liquid viscosity on the structure of stratied wavy-ow (VSG = 18 m/s, VSL = 0.2 m/s).

the cross-correlation of average cross-sectional void fraction time

series from both meshes, selecting the time delay which corresponds to the best correlation of the cross-correlation function,
and dividing the physical distance between the meshes by the time
lag at the maximum correlation coefcient. The velocities are plotted against the mixture velocity, the sum of the gas and liquid

supercial velocities, as shown in Fig. 34. At a constant liquid

supercial velocity, the experimental values of interface structure
velocity show a positive slope for stratiedslug transition, stratied wavy and annular ow patterns. A clear increase in the structure velocity with the mixture velocity is observed at low
supercial liquid velocity and 1 cP liquid viscosity. As the liquid

86

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

100

100
Upstream
Downstream

60
40
20

40

0
0.6

0.7

0.8

0.9

0.6

0.7

0.8

0.9

Void fraction

Void fraction

(a) V SG = 9 m/s; VSL = 0.03 m/s

(b) VSG = 9 m/s; VSL = 0.2 m/s

100

100
Upstream
Downstream

Upstream
Downstream

80

PDF, %

80

PDF, %

60

20

60
40
20

60
40
20

0
0.6

0.7

0.8

0.9

0.6

0.7

0.8

0.9

Void fraction

Void fraction

(c) VSG = 18 m/s; VSL = 0.03 m/s

(d) VSG = 18 m/s; VSL = 0.2 m/s

100

100
Upstream
Downstream

Upstream

80

PDF, %

80

PDF, %

Upstream
Downstream

80

PDF, %

PDF, %

80

60
40
20

Downstream

60
40
20

0
0.6

0.7

0.8

0.9

0.6

0.7

0.8

0.9

Void fraction

Void fraction

(e) VSG = 27 m/s; VSL = 0.03 m/s

(f) VSG = 27 m/s; VSL = 0.2 m/s

Fig. 33. Probability density function (PDF) for lL = 1 cP.

compared with the structure velocity for 1 cP liquid for the mixture
velocities considered.

Structure velocity, (m/s)

4.0
Upstream; Vsl=0.03 m/s, 1 cP
Downstream; Vsl=0.03 m/s, 1 cP
Upstream; Vsl=0.03 m/s, 10 cP
Downstream; Vsl=0.03 m/s, 10 cP

3.0

5. Conclusions

2.0

1.0

0.0
5

10

15

20

25

30

35

40

Mixture velocity, (m/s)

Fig. 34. Structure velocities for VSL = 0.03 m/s and liquid viscosities of 1 cP and
10 cP.

viscosity is increased from 1 cP to 10 cP, the inuence of mixture

velocity on structure velocities is not so pronounced at the downstream location. On the other hand, in the upstream location, the
structure velocity for 10 cP liquid viscosity is signicantly lower

A dual wire-mesh sensor was used to detect the local instantaneous cross-sectional distribution of the phases in gasliquid stratied and annular ows in the upstream and downstream locations
of a standard elbow. Data were obtained for a wide range of ow
rates in the horizontal orientation. Detailed information of the void
fraction distribution and interfacial structures are obtained. The
signals of the sensor, that are proportional to the liquid conductivity, are processed to obtain void fraction time series, cross-section
average time series, local void fraction distributions and structure
velocities. The effects of the supercial liquid and gas velocities
and liquid viscosity on the average and local void fraction have
been experimentally investigated. The analysis included times series plots of void fraction, contour plots of void fraction distribution
and local standard deviation, 3D image reconstruction, PDF of void
fraction and average void fraction for various ow regimes considered. These results shed additional light on the hydrodynamics of
horizontal gasliquid ows, in particular, regarding the

R.E. Vieira et al. / Experimental Thermal and Fluid Science 59 (2014) 7287

distribution of void fraction in the upstream and downstream section of a standard elbow.
The experimental study highlights the details of gas and liquid
distributions for stratied and annular gasliquid conditions inside
the pipe:
 In general, it is found, within the considered operating conditions, that the effect of the 90 horizontal bend on the time
average void fraction was to slightly increase the cross-sectional time averaged void fraction value.
 The sequence of instantaneous cross-sectional void fraction
showed that wave instability is the most notable feature of
the stratied to slug transition in both upstream and downstream locations of the elbow.
 The contour plots of cross-sectional averaged void fraction
show that, irrespective of liquid viscosity, the cross-sectional
distribution of the void fraction changes after the elbow. The
analysis of the void fraction time series around the top of the
pipe highlights that for stratied ows the liquid tends to climb
the inner wall of the pipe for the measurements after the elbow.
These results help to gain a better understanding of how the
liquid phase wets the inner wall at the top of the pipe and
assure continuous wetting of corrosion inhibitors.
 Three-dimensional structures have been identied for liquid
viscosities of 1 cP and 10 cP in stratiedslug transition and
stratied-wavy ows. Results show that high viscosity liquids
tend to promote slug ow. The pseudo slugs structures found
in the downstream section of the elbow are more aerated than
the structures identied in the upstream section.
 Probability density function was used to discriminate non-stable
stratied ows from stratied wavy ow, as evident in Fig. 33.
The PDF for non-stable stratiedslug transition shows a at
region without a peak which is different from previous works
found in literature. Examination of the probability density function plots of the cross-sectional averaged time series of void fraction leads to the conclusion that these different trends are linked
to hydrodynamic changes in ow patterns and transition zones.
On the other hand, within the downstream pipe ow of horizontal 90 elbows, the same unstable stratied, stratied-wavy and
annular ow patterns are observed.
 The contours of the standard deviation of the void fraction signal reect the stochastic behavior of the interface and the
amplitude of the interface oscillations.
 The velocities of the periodic structures were obtained by computing the cross-correlation of average cross-sectional void
fraction time series from both meshes. The velocities were plotted against the mixture velocity. Positive slopes were obtained
for liquid viscosities of 1 cP and 10 cP in the downstream section. Almost a constant trend was obtained when the liquid viscosity was increased to 10 cP for the same ow conditions in the
upstream section.

References
[1] [1] M.I. Cohen, An investigation of pressure drop in a two-phase twocomponent ow in bends, M.Sc. Thesis, MIT, 1957.
[2] J.R. Castillo, Study of two-phase ow in pipe bends, M.Sc. Thesis MIT, 1957.
[3] S.A. Shirazi, J. Shadley, B. McLaury, E. Rybicki, A procedure to predict solid
particle erosion in elbows and tees, J. Pressure Vessel Technol. 117 (1995) 4552.
[4] B.S. McLaury, J. Wang, S.A. Shirazi, J. Shadley, E. Rybicki, Solid particle erosion
in long radius elbows and straight pipes, in: SPE Annual Technical Conference
and Exhibition, San Antonio, Texas. SPE38842 (1997) pp. 977986.

87

[5] B.S. McLaury, S. Shirazi, E. Rybicki, Sand erosion in multiphase ow for slug and
annular ow regimes, CORROSION/2010 San Antonio, Texas, paper No.10377
(2010).
[6] N. Kesana, S. Grubb, B.S. McLaury, S.A. Shirazi, Ultrasonic measurement of
multiphase ow erosion patterns in a standard elbow, ASME J. Energy Resour.
Technol. 1353 (2013) 032905.
[7] D.C. Silverman, Rotating cylinder electrode geometry relationships for
prediction of velocity-sensitive corrosion, Corrosion 44 (1) (1988) 4249.
[8] J. Postlethwaite, E. Tinker, M. Hawrylak, Erosioncorrosion in slurry pipelines,
Corrosion 30 (1974) 285290.
[9] J. Villarreal, D. Laverde, C. Fuentes, Carbonsteel corrosion in multiphase slug
ow and CO2, Corros. Sci. 48 (2006) 23632379.
[10] K. Usama, Long liquid slugs in stratied gas/liquid ow in horizontal and
slightly inclined pipes. Ph.D. Thesis, Delft University, Netherlands, 2009.
[11] G. Gardner, P. Neller, Phase distributions ow of an airwater mixture round
bends and past obstructions, Proc. Inst. Mech. Eng. 184 (1969) 93101.
[12] K. Usui, S. Aoki, A. Inoue, Flow behavior and pressure drop of two-phase ow
through C-shaped bend in a vertical plane-I: upward ow, J. Nucl. Sci. Technol.
17 (1980) 875887.
[13] K. Usui, S. Aoki, A. Inoue, Flow behavior and phase distributions in two-phase
ow around inverted U-bend, J. Nucl. Sci. Technol. 20 (1983) 915928.
[14] S. Das, M. Biswas, A. Mitra, Non-Newtonian liquid ow in bend, Chem. Eng. J.
45 (1991) 165171.
[15] T. Bandyopadhyay, T. Banerjee, S. Das, Gas-non-Newtonian liquid ow through
elbows, Chem. Eng. Commun. 182 (2000) 2133.
[16] A. Ribeiro, T. Bott, D. Jepson, The inuence of a bend on drop sizes in horizontal
annular two-phase ow, Int. J. Multiphase Flow 27 (2001) 721728.
[17] P. James, B. Azzopardi, D. Graham, C. Sudlow, The effect of a bend on droplet
size distribution in two-phase ow, in: 7th International Conference on
Multiphase Flow in Industrial Plants, Bologna, 1315 September 2000.
[18] A. Azzi, L. Friedel, R. Kibboua, B. Shannak, Reproductive accuracy of two-phase
ow pressure loss correlations for vertical 90 degree bends, Forsch.
Ingenieurwesen 67 (2002) 109116.
[19] P.L. Spedding, E. Benard, Gasliquid two phase ow through a vertical 90
elbow bend, Exp. Therm. Fluid Sci. 31 (2007) 761769.
[20] M. Abdulkadir, D. Zhao, S. Sharaf, L. Abdulkareem, I. Lowndes, B. Azzopardi,
Interrogating the effect of 90 degree bends on airsilicone oil ows using
advanced instrumentation, Chem. Eng. Sci. 66 (2011) 24532467.
[21] N. Kesana, Erosion in multiphase pseudo-slug ow with emphasis on sand
sampling and pseudo-slug characteristics, Ph.D. thesis, The University of Tulsa,
Oklahoma, USA, 2013.
[22] H.M. Prasser, A. Bttger, J. Zschau, A new electrode-mesh tomograph for gas/
liquid ows, Flow Meas. Instrum. 9 (1998) 111119.
[23] J.C. Rodriguez, Effects of liquid viscosity and sand size on erosion in slug
multiphase ow, MSc. thesis, The University of Tulsa, Oklahoma, USA, 2008.
[24] R. Vieira, N. Kesana, B.S. McLaury, S. Shirazi, Sand erosion in multiphase ow
for low-liquid loading and annular conditions, in: Proceedings of the
International Mechanical Engineering Congress-American Society of
Mechanical Engineers, 2012, Houston, Texas, 812 November.
[25] R. Vieira, N. Kesana, B. McLaury, S. Shirazi, E. Schleicher, U. Hampel,
Experimental investigation of horizontal gasliquid stratied and annular
ow using wire mesh sensor, in: Proceedings of the ASME Fluids Engineering
Division Summer Meeting, 2013, paper no. FEDSM2013-16117, Incline Village,
Nevada, July 711.
[26] D. Barnea, A unied model for predicting ow pattern transitions for the whole
range of pipe inclinations, Int. J. Multiphase Flow 13 (1987) 112.
[27] 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.
[28] J. Bendat, A. Piersol, Random Data: Analysis and Measurement Procedures,
fourth ed., Wiley, New Jersey, 2010.
[29] S. Orfanidis, Optimum Signal Processing: An Introduction, second ed.,
McGraw-Hill, New York, 2007.
[30] P.Y. Lin, T.J. Hanratty, Detection of slug ow from pressure measurements, Int.
J. Multiphase Flow 13 (1987) 1321.
[31] J.E. Laurinat, T.J. Hanratty, W.P. Jepson, Film thickness distribution for gas
liquid annular ow in a horizontal pipe, PhysicoChem. Hydrodynam. 6 (1985)
179195.
[32] T.W.F. Russell, D.E. Lamb, Flow mechanism of two-phase annular ow, Can. J.
Chem. Eng. 43 (1965) 237245.
[33] D. Butterworth, Airwater annular ow in a horizontal tube, Prog. Heat Mass
Transfer 6 (1972) 235251.
[34] T. Fukano, A. Ousaka, Prediction of circumferential distribution of lm
thickness in horizontal and near-horizontal gasliquid annular ow, Int. J.
Multiphase Flow 15 (1989) 403420.
[35] P. Attwood, K. Van Gelder, C. Charnley, CO2 Corrosion in wet gas systems.
CORROSION/1996 San Antonio, Texas, 32, 1996.
[36] R. Nyborg, A. Dugstad, Top of line corrosion and water condensation in wet gas
pipelines. CORROSION/2007 San Antonio, Texas, paper No. 07555, 2007.