Nuclear Engineering and Design xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Nuclear Engineering and Design

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

Characterization of horizontal air–water two-phase flow

Ran Kong, Seungjin Kim ⇑
230 Reber Building, Department of Mechanical and Nuclear Engineering, The Pennsylvania State University, University Park, PA 16802, United States

h i g h l i g h t s

A visualization study is performed to develop flow regime map in horizontal flow.

Database in horizontal bubbly flow is extended using a local conductivity probe.

Frictional pressure drop analysis is performed in horizontal bubbly flow.

Drift flux analysis is performed in horizontal bubbly flow.

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

Article history: This paper presents experimental studies performed to characterize horizontal air–water two-phase flow
Received 22 December 2015 in a round pipe with an inner diameter of 3.81 cm. A detailed flow visualization study is performed using a
Accepted 9 June 2016 high-speed video camera in a wide range of two-phase flow conditions to verify previous flow regime
Available online xxxx
maps. Two-phase flows are classified into bubbly, plug, slug, stratified, stratified-wavy, and annular flow
regimes. While the transition boundaries identified in the present study compare well with the existing
ones (Mandhane et al., 1974) in general, some discrepancies are observed for bubbly-to-plug/slug, and
plug-to-slug transition boundaries. Based on the new transition boundaries, three additional test condi-
tions are determined in horizontal bubbly flow to extend the database by Talley et al. (2015a). Various
local two-phase flow parameters including void fraction, interfacial area concentration, bubble velocity,
and bubble Sauter mean diameter are obtained. The effects of increasing gas flow rate on void fraction,
bubble Sauter mean diameter, and bubble velocity are discussed. Bubbles begin to coalesce near the
gas–liquid layer instead of in the highly packed region when gas flow rate increases. Using all the current
experimental data, two-phase frictional pressure loss analysis is performed using the Lockhart–Martinelli
method. It is found that the coefficient C = 24 yields the best agreement with the data with the minimum
average difference. Moreover, drift flux analysis is performed to predict void-weighted area-averaged bub-
ble velocity and area-averaged void fraction. Based on the current database, functional relations for vg
vs. <j> and <a> vs. <jg>/<j> have been studied. It is found that vg – <j> method predicts the
void-weighted area-averaged bubble velocity and area-averaged void fraction better compared to
<a> – <jg>/<j> method.
Ó 2016 Elsevier B.V. All rights reserved.

1. Introduction developed for vertical two-phase flow are directly applied to hori-
zontal two-phase flow in nuclear system analysis codes (TRACE
Horizontal two-phase flows are common in many practical V5.0 Theory Manual, 2007; RELAP5-3D Code Manual, 2009). There-
engineering applications including nuclear reactors. Understand- fore, more work needs to be performed to characterize horizontal
ing horizontal two-phase flow is essential to correctly model the two-phase flow.
nuclear power system. However, there has been little investigation The common flow regime maps for horizontal two-phase flows
of horizontal flows compared with vertical flows. The highly asym- include that developed by Mandhane et al. (1974) through exper-
metric void distribution in horizontal flow, which is due to the imental work and that developed by Taitel and Dukler (1976)
effect of the buoyancy force, adds more difficulties in experimental through theoretical modeling. The experimental database used
studies. As such, previous work on vertical flow cannot be directly by Mandhane et al. (1974) covers a wide range of physical proper-
extended to horizontal flow. Despite this, the closure relations ties and flow rates. However, their map shows low accuracy in pre-
dicting the bubbly flow regime because little reliable experimental
⇑ Corresponding author. data exists for bubbly flow. Different from Mandhane et al.’s (1974)
E-mail address: [email protected] (S. Kim). map, Taitel and Dukler’s (1976) map indicates that the transition

http://dx.doi.org/10.1016/j.nucengdes.2016.06.016
0029-5493/Ó 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
2 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx

Nomenclature

ai interfacial area concentration [m1] q density [kg/m3]

C0 distribution parameter h azimuthal angle [°]
D pipe diameter [m] l dynamic viscosity [Pa * s]
Dsm bubble Sauter-mean diameter [m] /f two-phase frictional multiplier
f single-phase friction factor X Martinelli parameter
fb bubble frequency [s1]
j superficial velocity [m/s] Mathematical symbols
R pipe radius [m] <> area-average
Re Reynolds number  void-weighted area-average
vf liquid velocity [m/s]
vg bubble velocity [m/s]
Subscripts
Vgj drift velocity [m/s] f liquid phase
g gas phase
Greek symbols
a void fraction

from bubbly flow to intermittent flow (plug and slug flows) formed on horizontal flow (França and Lahey, 1992;
depends on gas flow rate. Moreover, their map shows that the tran- Kocamustafaogullari et al., 1994) due to limited database. Additional
sitions from stratified or stratified wavy flows to intermittent flow experiments should be performed to obtain a more comprehensive
require higher liquid flow rates than Mandhane et al.’s (1974) map. database to perform drift flux analysis in horizontal bubbly flow.
In order to verify these transition boundaries, additional study The current work seeks to perform experiments to characterize
needs to be performed using a high speed video camera. horizontal flow. First, a detailed flow visualization study is per-
In order to model horizontal two-phase flow phenomena, formed to verify the transition boundaries in horizontal flow. Based
experimental database has been established by previous research- on the new transition boundaries, additional test conditions are
ers. Kocamustafaogullari and Wang (1991), Huang (1993), determined. Using the more comprehensive experimental data-
Kocamustafaogullari and Huang (1994), Kocamustafaogullari base in horizontal bubby flow, the evolution of local two-phase
et al. (1994), Bottin et al. (2014), and Talley et al. (2015a) obtained flow parameters is investigated. The frictional pressure drop anal-
a local database in bubbly flow through experiments. ysis and the drift flux analysis are also performed using the more
Kocamustafaogullari et al. (1994) found that the local and area- comprehensive experimental database.
averaged void fraction and interfacial area concentration increase
with decreasing liquid flow and increasing gas flow in horizontal 2. Experimental facility
bubbly flow. The same trend was also observed by Iskandrani
and Kojasoy (2001) using hot-film anemometry in a 5.03 cm inner Experiments are performed in an adiabatic air–water test facil-
diameter transparent horizontal pipe and by Bottin et al. (2014) ity, which is made of 3.81 cm inner diameter, D, clear acrylic pipes,
using an optical probe in a 10.0 cm inner diameter horizontal pipe. as shown in Fig. 1. The total length of the test section downstream
Talley et al. (2015a) performed experiments in horizontal dis- of the two-phase injector is 9.5 m, yielding a development length
persed bubbly flow using a four-sensor conductivity probe (Kim of 248 diameters. The test facility is designed to investigate bubbly,
et al., 2000) in a 3.81 cm inner diameter horizontal pipe. However, plug, slug, stratified, stratified wavy, and annular flows.
the existing database is still limited either because local data was There are three instrumentation ports along the test section for
only taken at only one axial location or the database does not cover local multi-sensor conductivity probe measurement, the centers of
a wide range of flow rates in horizontal bubbly flow. Additional which are located at L/D = 44, 116, and 244 downstream of the
experiments need to be performed to obtain a more comprehen- two-phase injector. These ports can also be used for pressure mea-
sive database including local interfacial area concentration infor- surement and flow visualization study. To perform flow visualiza-
mation for bubble interaction modeling in horizontal bubbly flow. tion study, a high-speed video camera with a maximum resolution
Considering that two-phase pressure drop is an important of 512  512 pixels is employed. It is capable of 2000 frames per
parameter to determine the performance of the two-phase system, second (fps) at this resolution. To obtain local experimental data-
the two-phase pressure loss is investigated by experiment in this base, the state-of-the-art local four-sensor conductivity probe
paper. Lockhart and Martinelli (1949) developed a correlation to (Kim et al., 2000) is employed to obtain various time-averaged
predict the two-phase frictional pressure drop. Using this correla- two-phase flow parameters including void fraction a, interfacial
tion, Lee and Lee (2001) investigated air–water frictional pressure area concentration ai, bubble velocity vg, bubble frequency fb, and
drop in horizontal rectangular channels. The effects of 90° elbow bubble Sauter-mean diameter Dsm. A specially designed traversing
and 45° elbow on two-phase pressure loss in horizontal bubbly unit allows the four-sensor conductivity probe to move along the
flow were investigated by Kim et al. (2010). radial direction of the pipe cross-section and to be rotated around
In two-phase flow, void fraction and bubble velocity are two of the axis of the test section at every 22.5° in the azimuthal direction
the fundamental parameters. The void fraction describes the gas dis- without stopping the flow. Considering that the gas distribution in
tribution and is an important parameter for hydrodynamic and ther- horizontal two-phase flow is highly asymmetric, such a measuring
mal design in various two-phase systems, while the bubble velocity mechanism is indispensable to obtain reliable local data through-
determines the transport of the void fraction and interfacial area out the entire flow area. Thus, as shown in Fig. 2, a total of 129 local
concentration. Many studies have been performed using the drift data points are measured across the entire pipe cross-section.
flux approach given by Zuber and Findlay (1965) to investigate ver- More detailed descriptions for the test facility and considerations
tical flow (Ishii, 1977; Kataoka and Ishii, 1987; Hibiki and Ishii, 2002, on local four-sensor conductivity probe measurement can be found
2003; Goda et al., 2003), but relatively little research has been per- in previous studies (Talley et al., 2015a,b).

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx 3

Fig. 1. Simplified schematic diagram of the test facility (not to scale).

ter two-phase flow. This study is also used to determine additional

test conditions in horizontal bubbly flow. Videos are acquired at an
axial location of L/D = 244 from the inlet to ensure that the two-
phase flow is well developed. In the present study, videos are taken
in bubbly, plug, slug, stratified, stratified-wavy, and annular flows
(Mandhane et al., 1974) with superficial liquid (jf) and gas (jg)
velocities ranging from 0.02–6.00 m/s and 0.10–19.60 m/s,
respectively.
In the process of determining flow regimes, captured two-phase
flow videos are classified into horizontal two-phase flow regimes
which are defined as follows:

(1) In bubbly flow (shown in Fig. 3(a)), it is observed that liquid

Reynolds number (Ref) plays a major role in bubble distribu-
tion pattern for a fixed gas flow rate. Here, Ref is calculated
based on superficial liquid velocity. At relatively lower Rey-
nolds number, small near-spherical or distorted bubbles are
concentrated mostly in the upper half portion of the pipe
Fig. 2. Mesh for measurements of local two phase flow parameters (flow direction cross-section. As Ref becomes larger, bubbles start to get dis-
is out of the page and the non-dimensional radial distance along an azimuthal angle tributed across the entire pipe cross-section.
h is denoted by r/Rh.).
(2) In plug flow (shown in Fig. 3(b)), elongated gas plugs move
in the upper half portion of the pipe with an asymmetric
In this paper, both the flow regime identification and the local round nose. Unlike the slug flow, the plug flow is character-
conductivity probe measurements will be presented. ized by plug bubbles with thin elongated tails, followed by
liquid regions with few dispersed bubbles present.
(3) In slug flow (shown in Fig. 3(c)), gas slug bubbles move with
3. Results and discussion a higher velocity than the small bubbles in the lower half
portion of the pipe. It is characterized by gas slugs with a
The discussion of the experimental results is divided into four bullet-shaped nose and blunt tails followed by a distinct liq-
sections, namely: (1) verification of flow regime transition bound- uid slug containing many small bubbles. More importantly,
aries, (2) evolution of various local two-phase flow parameters, (3) bubble clusters are present on the order of the gas slug
frictional pressure loss analysis, and (4) drift-flux analysis. depth in the wake region as indicated in Fig. 3(c). The pres-
ence of these small bubbles, which can provide large inter-
3.1. Verification of flow regime transition boundaries faces for heat and mass transfer, is an important
characteristic to distinguish slug flow from plug flow.
In order to assess the current horizontal flow regime maps and (4) In stratified flow (shown in Fig. 3(d)), the gas phase is sepa-
to improve the accuracy of flow regime prediction in horizontal rated from the liquid phase and is flowing in the upper por-
pipes, a flow visualization study is performed in horizontal air–wa- tion of the pipe while liquid flows in the lower portion of the

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
4 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx

Flow Direction

Ref = 143,000 Ref = 162,000 Ref = 181,000

(a) bubbly flow (jg = 0.25m/s)

(b) plug flow

liquid slug bubble clusters gas slug

(c) slug flow

(d) stratified flow (e) stratified wavy flow (f) annular flow
Fig. 3. Characteristic images of various flow regimes in horizontal air-water two-phase flow.

pipe. A smooth interface without disturbances is present the flow regime map developed by Taitel and Dukler (1976), it is
between the two phases. The only difference between strat- found that the present study captures the trend in the transitions
ified flow and stratified-wavy flow (shown in Fig. 3(e)) is the from bubbly-to-intermittent (plug and slug) flows. Disagreement
appearance of disturbances at the interface between the two is noticed at the transitions from stratified-to-intermittent and
phases. stratified wavy-to-intermittent flows. It may be explained that in
(5) In annular flow (shown in Fig. 3(f)), a continuous gas core is the work by Taitel and Dukler (1976), the Kelvin–Helmholtz theory
present with a liquid film around the periphery of the pipe. underestimates the wave height, such that higher jf is needed for
The liquid film is rough and wavy, and is thicker at lower the transitions of stratified-to-intermittent and stratified wavy-
portion of the pipe. Dispersed liquid droplets entrained from to-intermittent to happen. Another shortcoming of Taitel and
the liquid film can also be observed in the gas core. Dukler’s (1976) map is that both the plug and slug flows are trea-
ted as intermittent flow, while the presence of small bubbles in
In total, videos of 263 flow conditions are taken and classified slug flow can have a large contribution to the interfacial area con-
into different flow regimes based on the definitions above. The centration. In the future work, the theoretical modeling of transi-
results are shown in Fig. 4. In order to mitigate subjectivity in tion from plug to slug flow should be included.
developing boundaries between flow regimes, these videos are
viewed and classified by multiple independent researchers. In 3.2. Evolution of various local two-phase flow parameters
Fig. 5, newly developed regime transition boundaries in the pre-
sent study are shown together with the flow regime boundaries Based on the new transition boundaries, the database estab-
suggested by Mandhane et al. (1974) and those by Taitel and lished by Talley et al. (2015a) is extended to relatively higher gas
Dukler (1976). It is found that the new transition boundaries gen- flow rate region in horizontal bubbly flow. Three additional test
erally agree well with Mandhane et al.’s (1974) map. Disagreement conditions are added. All twelve test conditions presented in the
occurs, however, at the transition from bubbly-to-plug and bubbly- current study are summarized in Table 1, nine of which were taken
to-slug flows. In Mandhane et al.’s (1974) map, transitions occur at by Talley et al. (2015a). The effects of gas and liquid flow rates on
a constant jf = 4.20 m/s regardless of jg. However, the present various local two-phase flow parameters such as void fraction,
results suggest that the transition is dependent on jg. At a lower interfacial area concentration, bubble velocity, and bubble Sauter
jg, transition occurs at a lower jf. Additionally, a discrepancy can mean diameter are investigated in this section.
also be found at the transition from plug flow to slug flow. The First, to gain confidence with the four-sensor conductivity
boundary observed in the present study suggests that it may probe measurements, superficial gas velocity at each port deter-
require more gas to make the transition from plug to slug at a jf less mined from the probe measurements (jg,probe) is benchmarked with
than 1 m/s, and require less gas with a jf above 1 m/s. This finding that based on gas rotameter and pressure measurements
is opposite to that by Mandhane et al. (1974) but more physical in (jg,flowmeter) for every test condition. As shown in Fig. 6, a good
that, the higher turbulence level at higher jf can break up the tails agreement is found between the two measurements yielding an
of larger bubbles into small bubbles, while the presence of these average percent difference of ±5.2% for all the twelve test
small bubbles is the characteristic of slug flow. Compared with conditions discussed in the present study.

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx 5

Fig. 4. Flow regime identification results on horizontal air-water two-phase flow.

Fig. 5. Comparison between present boundaries and the existing ones (Mandhane et al., 1974; Taitel and Dukler, 1976) in horizontal two-phase flow.

Table 1
Summary of the test conditions (Runs 1–9 from Talley et al., 2015a).

Run No. 1 2 3 4 5 6 7 8 9 10 11 12
jf [m/s] 4.00 4.00 5.00 5.00 5.00 6.00 6.00 6.00 6.00 5.00 6.00 5.00
jg [m/s] 0.11 0.19 0.11 0.17 0.33 0.09 0.15 0.30 0.58 0.58 1.00 1.00

In previous work, Talley et al. (2015a) found that increasing jg continue to be investigated. Fig. 7 shows the contour plots of a gen-
increases both the local peak and area-averaged values of a and erating using MATLAB for all the test conditions at jf = 5.00 m/s.
ai, while increasing jg has smaller effects on the Sauter mean diam- Only profiles of a are shown here because the profiles of a and ai
eter and bubble velocity. With the additional experimental data, the are similar in bubbly flow regime. In the figures, r/R denotes probe
effects of increasing jg on various two-phase flow parameters radial position with positive values at upper (or right) and negative

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
6 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx

A similar trend was reported by Kocamustafaogullari and Wang

(1991).
The effects of increasing jg on bubble Sauter mean diameter and
bubble velocity for a fixed jf are shown in Fig. 8. Only the data along
vertical axis (90°) are shown in these figures. It is observed in Fig. 8
(a) that as jg increases, the most significant bubble size increase
occurs near the center of the pipe instead of near the top wall. This
suggests that in transition from bubbly flow to plug or slug flow,
most of the bubble coalescence happens near the gas–liquid layer
instead of in the highly packed region, which is consistent with
previous flow visualization study (Talley et al., 2015b). Here, the
gas–liquid layer is defined as the region below where bubbles
are most densely packed together.
In Fig. 8(b), the liquid velocity is estimated using the same
approach employed by Kocamustafaogullari and Huang (1994) to
compare the bubble velocity and the liquid velocity, where:
 r 1=n
v f ðrÞ ¼ v f ;max 1 ð1Þ
R
with

ðn þ 1Þð2n þ 1Þ jf
v f ;max ¼ ð2Þ
2n2 1  hai
Fig. 6. Benchmark of four-sensor conductivity probe measurements for jg,probe
versus jg,flowmeter. where vf, vf,max, n are the estimated local liquid velocity, maximum
liquid velocity, and profile power, respectively. Here, n = 7 is used
based on previous study (Kocamustafaogullari and Huang, 1994).
values at lower (or left) region of the pipe cross section. It is found However, these expressions assume a symmetric and nearly uni-
that the effect of jg on a is consistent with previous study (Talley form void fraction profile in the pipe, which is more valid for high
et al., 2015a). In relatively high gas flow rate region, the peak values jf conditions with low local void fractions. Consistent to the finding
of a and ai can reach up to approximately 0.6 and 2000 [1/m], by Talley et al. (2015a), the bubbles have a smaller velocity than the
respectively, which is different from vertical upward bubbly flow. estimated liquid velocity for all test conditions and bubbles tend to

(a) jg = 0.11 [m/s], < > = 0.026 [-] (b) jg = 0.17 [m/s], < > = 0.042 [-] (c) jg = 0.33 [m/s], < > = 0.078 [-]

(d) jg = 0.58 [m/s] < > = 0.118 [-] (e) jg = 1.00 [m/s] < > = 0.193 [-]
Fig. 7. Effect of jg on a for jf = 5.00 m/s at L/D = 244.

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx 7

Increased system pressure at higher jf also causes bubble Sauter

mean diameter to decrease. Bubbles move faster at higher jf, which
confirms that gas phase follows liquid phase in horizontal bubbly
flow.

3.3. Frictional pressure loss analysis

In horizontal two-phase flow, there is no gravitational pressure

drop and the acceleration pressure drop is assumed to be negligi-
ble. Thus, only the frictional pressure drop remains in horizontal
air–water two-phase flow. In the present study, the frictional pres-
sure drop is analyzed using the conventional approach of Lockhart
and Martinelli (1949). This approach assumes that the two-phase
frictional pressure drop is the summation of pressure drop caused
by each phase and the interaction between two phases given by:
C 1
/2f ¼ 1 þ þ ð3Þ
X X2
where /f and X, are the two-phase frictional multiplier and Marti-
nelli parameter, respectively, defined by:
2/ f
ðdp=dzÞF ðdp=dzÞF
/2f  and X 2  g
f
ðdp=dzÞF ðdp=dzÞF

Here, the frictional loss for each phase is obtained by:

 k
dp 2f
¼ qk jk
2
ð4Þ
dz F D
where f, D, q, and j denote the single-phase friction factor, pipe
diameter, density, and superficial velocity, respectively. The sub-
script k denotes the liquid or gas phase index of f or g. The friction
factor for turbulent flow is obtained using Blasius formulation
(1913) given by:
 
q j D 0:25
f ¼ 0:079 k k ð5Þ
lk
The frictional pressure drop is predicted using Eq. (3) provided
that C is known. A value of C = 20 was suggested by Chisholm
(1967) for turbulent-gas turbulent-liquid two-phase flow through
straight pipes. In the present study, the absolute percent difference
from the experimental data is calculated. The best fit value of C
value is determined by minimizing the average absolute percent
difference. Based on the data acquired in the present study in
Fig. 8. Effects of jg on (a) Dsm and (b) vg for jf = 5.00 m/s at L/D = 244 of vertical axis bubbly flow regime, C = 24 is found to give the best fit to experi-
(90°), arrows denote trend with increasing jg.
mental data with an average percent difference of ±1.10%, while
C = 20 gives ±1.40% as shown in Fig. 10. Additional experiments
move at a higher velocity in the lower portion of the pipe. However, should be performed to obtain a more comprehensive database
in the current study it is found that increasing jg shows different in horizontal bubbly flow to obtain the best fitting value of C in
effects on bubble velocity in different radial positions. When (r/ the future.
R)90 > 0.7, increasing jg decreases bubble velocity while opposite
effect is observed when (r/R)90 < 0.7. This may because densely 3.4. Drift flux analysis
packed bubbles provide strong resistance to the motion of the liquid
as indicated by Talley et al. (2015b). At higher gas flow rates, more Considering that bubble velocity and void fraction are two
densely packed bubbles introduce more resistance to the liquid, important parameters to describe two-phase flow, the correlations
thus the liquid phase moves slower at (r/R)90 > 0.7. Correspondingly, to model vg and <a> in two-phase flow are investigated using
the liquid phase moves faster at (r/R)90 < 0.7 due to continuity. the drift flux approach in the present study. The one-dimensional
The effects of increasing jf on void fraction distribution, bubble relationship between vg and <j>, which can be obtained from
Sauter mean diameter, and bubble velocity for a fixed jg are shown local conductivity probe and flow rotameters, respectively,
in Fig. 9. Again, only the data along vertical axis (90°) are shown. provides a convenient way to predict vg. The conventional
Talley et al. (2015a) found that decreasing jf increases the degree vg – <j> formulation is given by (Zuber and Findlay, 1965):
of peaking in a and ai distributions, which is due to the decreased
hjg i
turbulence level. A consistent trend is observed for higher gas flow hhv g ii ¼ ¼ C 0 hji þ hhV gj ii ð6Þ
rate test conditions. Bubbles are more concentrated near the top hai
wall which is confirmed by the observation that no bubbles are Using Eq. (6), vg can be predicted provided that C0 and
detected when (r/R)90 < 0.2 for the lowest jf test condition. vgj are known. Here C0 is the distribution parameter and
Increasing jf promotes bubble breakup due to increased turbulence. vgj is the drift velocity defined as:

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
8 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx

Fig. 9. Effects of jf on (a) a, (b) Dsm, and (c) vg for jg = 0.20 m/s at L/D = 244 of vertical axis (90°), arrows denote trend with increasing jf.

haji be less than unity. Using all the experimental data, the values of
C0 ¼ ð7Þ
haihji the distribution parameter and drift velocity were determined to
be 0.99 and 0.09 m/s, respectively.
In the current study, the drift flux analysis is performed using
hav gj i
hhV gj ii ¼ ð8Þ the same approach to verify Kocamustafaogullari et al. (1994)’s
hai
conclusion. Data measured at all the axial locations are used to per-
form the analysis. With experimentally acquired vg, <jf>, and
where <j> is the total superficial velocity given by: <jg>, vg is plotted against <j> to determine the values for C0
and vgj as shown in Fig. 11. From the figure, it is found that
hji ¼ hjf i þ hjg i ð9Þ the effect of jf suggested by previous research is not observed using
Here, <> and  represent the area-averaged quantity over a the current experimental database. Most of the data with different
cross-sectional area and the void-weighted area-averaged quantity, jf agree well with one relationship. The values of C0 and vgj are
respectively. Among the previous studies to perform drift flux anal- determined to be 1.05 and 1.21 m/s from the plot, respectively.
ysis in horizontal flow, the most significant one for horizontal bub- The significantly negative drift velocity in the present study high-
bly flow is that of Kocamustafaogullari et al. (1994). The authors lights the difference between the vertical and horizontal bubbly
found a systematic effect of the liquid flow rate on C0 and vgj, flows. In vertical flow the bubbles move faster than the mixture
such that for relatively low values of <jf> and <j>, vgj < 0 and due to buoyancy effect, whereas it is opposite in the horizontal
C0 > 1.0. As <jf> or <j> increases, vgj can be positive and C0 can flow. The significantly different drift velocity from previous study

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx 9

Fig. 12. Relationship of <a> – <b> in horizontal bubbly flow.

Fig. 10. Predicted two-phase multiplier using Lockhart–Martinelli approach (1949).

and <b> is given by:

hjg i
hbi ¼ ð12Þ
hji
The linear relationship between <a> and <b>, as shown in
Fig. 12, suggests that C0 + vgj/<j> is a nearly constant value.
From the slope of the plot, this value is determined to be 0.85.
Moreover, C0 + vgj/<j> can also be calculated for each condition
in the present study using the results of C0 and vgj obtained
from vg – <j> plane. This value is found to be within ±10% of
the fit value, which is 0.85. This indicates that constant C0 +
vgj/<j> for horizontal bubbly flow is an acceptable hypothesis.
Noting that:

hjg i
hhv g ii ¼ ð13Þ
hai

and <jg> can be obtained from flow rotameter measurements, <a>

can be calculated once vg is predicted or vice versa. The com-
parison of measured and predicted values of vg and <a> using
vg – <j> plane is shown in Fig. 13. It is found that this approach
predicts the data very well, with an average disagreement of ±2.7%
for vg and ±2.7% for <a>. As a comparison, the area-averaged
Fig. 11. Relationship of vg – <j> in horizontal bubbly flow.
void fraction and void-weighted area-averaged bubble velocity are
also predicted from the <a> – <b> plot. Comparison of measured
(Kocamustafaogullari et al., 1994) might be explained for the fol- and predicted values of vg and <a> is shown in Fig. 14. This
lowing reason. Approximately one third of the test conditions in approach performs well with a disagreement of ±4.9% for vg
Kocamustafaogullari et al. (1994) would be classified as intermit- and ±4.2% for <a>. Obviously, vg – <j> plane predicts vg
tent flow based on the flow regime map of Taitel and Dukler and <a> with greater accuracy. This result demonstrates that the
(1976), whereas these conditions may contain elongated bubble vg – <j> plane provides closure relations for void-weighted
with a velocity different than that of a small bubble. area-averaged bubbly velocity and area-averaged void fraction
Apart from the relationship between vg and <j>, the <a> – modeling with relatively good prediction. However, additional
<b> plane (Zuber and Findlay, 1965) provides another approach experiments should be performed to obtain a more comprehensive
to analyze the experimental data. From Eq. (6) one can obtain: database in horizontal bubbly flow to compare these two methods
in the future.
hai ¼ Khbi ð10Þ
where K is given by:
4. Conclusions
1
K¼ hhV gj ii
ð11Þ
C0 þ Experiments are performed to characterize horizontal air–water
hji
two-phase flow. In order to verify the existing flow regime

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
10 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx

Fig. 14. Comparison of measured and predicted (a) vg and (b) <a> by <a> – <b>
Fig. 13. Comparison of measured and predicted (a) vg and (b) <a> by vg –
plane.
<j> plane.

concentration, bubble velocity, and bubble Sauter mean diameter

transition boundaries for horizontal two-phase flow, a detailed are obtained for additional test conditions using a four-sensor con-
flow visualization study is performed using a high-speed video ductivity probe at three axial locations. Unlike vertical upward
camera for a total of 263 flow conditions, which are classified into bubbly flow, the local void fraction and interfacial area concentra-
bubbly, plug, slug, stratified, stratified-wavy, and annular flows. As tion can reach 0.6 and 2000 [1/m], respectively, in horizontal bub-
a result, new transition boundaries are developed and compared bly flow. With increasing gas flow rate, most of the bubble
with the existing ones. Unlike Mandhane et al.’s (1974) map, the coalescences happen near the gas–liquid layer instead of in the
present results suggest that the transition from bubbly flow to most highly packed region. Bubbles are found to move faster in
plug/slug flow depends on jg. Additionally, it may require more the lower region of the pipe while opposite effect is observed in
gas to make the transition from plug to slug at a jf less than 1 m/ the higher region when gas flow rate increases.
s, and require less gas with a jf above 1 m/s. Unlike Taitel and The frictional pressure loss in horizontal bubbly flow is pre-
Dukler’s (1976) map, the present results suggest that it requires dicted using Lockhart–Martinelli method. For all the twelve condi-
less jf for the plug-to-stratified transition and the slug-to- tions studied in this paper, the coefficient C = 24 is found to give
stratified wavy transition. the best agreement with the data with the minimum average per-
Base on the new transition boundaries, three additional test cent difference of ±1.10%.
conditions at relatively higher gas flow rate region in horizontal Based on the extended database, the drift-flux analysis is per-
bubbly flow are determined to extend the experimental database formed using both the vg vs. <j> relationship and the <a> vs.
established by Talley et al. (2015a). Various local time-averaged <jg>/<j> relationship. In the first method, the distribution parame-
two-phase flow parameters such as void fraction, interfacial area ter C0 and the drift velocity vg are found to be 1.05 and

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016
 R. Kong, S. Kim / Nuclear Engineering and Design xxx (2016) xxx–xxx 11

1.21 m/s, respectively. The significantly negative drift velocity Hibiki, T., Ishii, M., 2003. One-dimensional drift-flux model for two-phase flow in a
large diameter pipe. Int. J. Heat Mass Transfer 46 (10), 1773–1790.
indicates that the gas phase moves slower than the mixture in hor-
Huang, W.D., 1993. Experimental studies for two-phase bubbly flow interfacial
izontal bubbly flow. In the second method, a constant value of 0.85 parameters in a horizontal channel (M.S. thesis). The University of Wisconsim-
is found for C0 + vgj/<j> from the slope of the plot. It is found Milwaukee.
that the vg – <j> plane can predict the vg and <a> with Ishii, M., 1977. One-Dimensional Drift-Flux Model and Constitutive Equations for
Relative Motion between Phases in Various Two-Phase Flow Regimes. Argonne
an average difference of ±2.7% and ±2.7%, respectively, while the National Lab., Ill (USA).
<a> – <b> plane can predict the vg and <a> with an average dif- Iskandrani, A., Kojasoy, G., 2001. Local void fraction and velocity field description in
ference of ±4.9% and ±4.2%, respectively. Additional experiments horizontal bubbly flow. Nucl. Eng. Des. 204, 117–128.
Kataoka, I., Ishii, M., 1987. Drift flux model for large diameter pipe and new
should be performed to compare these two methods in the future. correlation for pool void fraction. Int. J. Heat Mass Transfer 30 (9), 1927–1939.
Kim, S., Fu, X.Y., Wang, X., Ishii, M., 2000. Development of the miniaturized four-
sensor conductivity probe and the signal processing scheme. Int. J. Heat Mass
Acknowledgments Transfer 43 (22), 4101–4118.
Kim, S., Kojasoy, G., Guo, T., 2010. Two-phase minor loss in horizontal bubbly flow
This work was supported by the U.S. Nuclear Regulatory Com- with elbows: 45° and 90° elbows. Nucl. Eng. Des. 240 (2), 284–289.
Kocamustafaogullari, G., Huang, W.D., 1994. Internal structure and interfacial
mission, Office of Nuclear Regulatory Research. velocity development for bubbly two-phase flow. Nucl. Eng. Des. 151 (1), 79–
This report was prepared as an account of work sponsored by an 101.
agency of the U.S. Government. Neither the U.S. Government nor Kocamustafaogullari, G., Wang, Z., 1991. An experimental study on local interfacial
parameters in a horizontal bubbly two-phase flow. Int. J. Multiphase Flow 17
any agency thereof, nor any of their employees, makes any war- (1), 553–572.
ranty, expressed or implied, or assumes any legal liability or Kocamustafaogullari, G., Huang, W.D., Razi, J., 1994. Measurement and modeling of
responsibility for any third party’s use, or the results of such use, average void fraction, bubble size and interfacial area. Nucl. Eng. Des. 148 (2–3),
437–453.
of any information, apparatus, product, or process disclosed in this Lee, H.J., Lee, S.Y., 2001. Pressure drop correlations for two-phase flow within
report, or represents that its use by such third party would not horizontal rectangular channels with small heights. Int. J. Multiphase Flow 27
infringe privately owned rights. The views expressed in this paper (5), 783–796.
Lockhart, R.W., Martinelli, R.C., 1949. Proposed correlation of data for isothermal
are not necessarily those of the U.S. Nuclear Regulatory two-phase, two-component flow in pipes. Chem. Eng. Prog. 45, 39–48.
Commission. Mandhane, J.M., Gregory, G.A., Aziz, K., 1974. A flow pattern map for gas—liquid
flow in horizontal pipes. Int. J. Multiphase Flow 1 (4), 537–553.
RELAP5-3D Code Manual, 2009. Volume I: Code Structure, System Models and
References Solution Methods, INEEL-EXT-98-00834.
Taitel, Y., Dukler, A.E., 1976. A model for predicting flow regime transitions in
Bottin, M., Berlandis, J., Hervieu, E., Lance, M., Marchand, M., Öztürk, O., Serre, G., horizontal and near horizontal gas-liquid flow. AlChE J. 22 (1), 47–55.
2014. Experimental investigation of a developing two-phase bubbly flow in Talley, J.D., Worosz, T., Kim, S., 2015a. Characterization of horizontal air-water two-
horizontal pipe. Int. J. Multiphase Flow 60, 161–179. phase flow in a round pipe part II: measurement of local two-phase parameters
Chisholm, D., 1967. A theoretical basis for the Lockhart–Martinelli correlation for in bubbly flow. Int. J. Multiphase Flow 76, 223–236.
two-phase flow. Int. J. Heat Mass Transfer 10 (12), 1767–1778. Talley, J.D., Worosz, T., Kim, S., Buchanan, J.R., 2015b. Characterization of horizontal
França, F., Lahey, R., 1992. The use of drift-flux techniques for the analysis of air-water two-phase flow in a round pipe part I: flow visualization. Int. J.
horizontal two-phase flows. Int. J. Multiphase Flow 18 (6), 787–801. Multiphase Flow 76, 212–222.
Goda, H., Hibiki, T., Kim, S., Ishii, M., Uhle, J., 2003. Drift-flux model for downward TRACE V5.0 Theory Manual, 2007. Field Equations, Solution Methods and Physical
two-phase flow. Int. J. Heat Mass Transfer 46 (25), 4835–4844. Models. Division of Risk Assessment and Special Projects.
Hibiki, T., Ishii, M., 2002. Distribution parameter and drift velocity of drift-flux Zuber, N., Findlay, J., 1965. Average volumetric concentration in two-phase flow
model in bubbly flow. Int. J. Heat Mass Transfer 45 (4), 707–721. systems. J. Heat Transfer 87 (4), 453–468.

Please cite this article in press as: Kong, R., Kim, S. Characterization of horizontal air–water two-phase flow. Nucl. Eng. Des. (2016), http://dx.doi.org/
10.1016/j.nucengdes.2016.06.016