A STUDY OF FLOW PATTERNS FOR GAS/

LIQUID FLOW IN SMALL DIAMETER TUBES

N. K. Omebere-Iyari and B. J. Azzopardi

Multiphase Flow Research Group, Nottingham Fuel and Energy Centre, School of Chemical,
Environmental and Mining Engineering, The University of Nottingham, Nottingham, UK.

Abstract: Time resolved void fraction data and flow pattern information have been obtained for
two-phase air/water flows in a small diameter (5 mm) vertical pipe using conductance probes.
The time averaged void fractions are seen to agree with values measured using a completely
different approach.
Analysis of high speed videos reveals that the probability density function (PDF) technique is
inadequate for accurately delineating the transition between slug and churn flows but performs
better for the churn to annular flow transition. Instead a novel approach has been developed
for transitions between flow patterns using the velocity of structures. This gives good agreement
with the present experiments. Additionally, a modification to the bubble-to-slug flow transition of
Taitel et al. (1980) gives improved predictions in narrow passages.
A flow pattern specific method for void fraction prediction has been applied and its predictions
are in good agreement with the measured mean void fraction. The slug flow model gives better
predictions of void fraction in churn flow that the annular flow model.
The velocities of disturbance waves on the wall film in annular flow are well predicted using the
model of Pearce (1979). However, pipe diameter dependence of one of the constants is required.
Keywords: two-phase flow; flow pattern; void fraction; small diameter; vertical pipe; conduc-
tance probes; compact heat exchanger; nuclear fuel bundles.

INTRODUCTION the researchers to have worked with tubes of

Correspondence to:
Professor B.J. Azzopardi,
Multiphase Flow Research
Group, Nottingham Fuel and
Energy Centre, School of
Chemical, Environmental and
Mining Engineering, The
University of Nottingham,
University Park, Nottingham NG7
2RD, UK.
E-mail: barry.azzopardi@
nottingham.ac.uk
DOI: 10.1205/cherd05059
0263–8762/07/
$30.00 þ 0.00
Chemical Engineering
Research and Design
Trans IChemE,
Part A, February 2007
# 2007 Institution
of Chemical Engineers
Two-phase flow in one form or another exists
in at least 60% of all process industry heat
exchange equipment (Hewitt, 1981). How-
ever, most available data has been obtained
for tube diameters typical of shell and tube
exchangers. Two-phase flows in a small
diameter tube have become important for
compact heat exchanger applications and
for nuclear reactors with tight passages. Com-
pact heat exchangers are increasingly being
applied to the chemical and process indus-
tries because of the significant benefits they
offer. The national energy saving potential of
this technology has been estimated to be
worth 8 PJ year21 (ETSU, 1994). They are
used because of their greater effectiveness,
smaller volume, improved safety and power
savings when compared to most conventional
heat exchanger types.
Much of the published information on two-
phase flow characteristics have been confined
to conduits with diameters greater than
20 mm. However, Oya (1971), Barnea et al.
(1983), Wambsganss et al. (1992), Bao et al.
(1994), Holt et al. (1999), Cheng and Lin
(2001) and Okawa et al. (2004) are some of
smaller diameters. The recent review papers
by Cheng and Mewes (2006) and Thome
(2004) show that the challenges of exchanging
heat with smaller surface areas such as those
encountered in electronic devices are increas-
ing research interest in fluid flow in small diam-
eter channels. Barnea et al. (1983) found the
pipe diameter effect to be minimal for upward
air/water flow in vertical tubes with diameters
ranging from 4 to 12.3 mm. Holt et al. (1999)
used circular and non-circular channels with
hydraulic diameters ranging from 3.9 to
10 mm to collect two-phase pressure drop
and void fraction data in order to investigate
typical conditions found in compact evapor-
ators. The CISE (Premoli et al., 1970) and
Lockhart/Martinelli correlations were found to
accurately predict the circular void fraction
data only. Although the results obtained were
compared with a novel flow pattern specific
method, the flow patterns transitions were
not studied. Cheng and Mewes (2006) con-
clude that any future research should include
studies of two-phase flow pattern transitions
in small and mini channels.
Workers including Hibi et al. (2000), Okubo
et al. (2000) and Sakashita et al. (1999) report

180 Vol 85 (A2) 180–192

 A STUDY OF FLOW PATTERNS FOR GAS/LIQUID FLOW IN SMALL DIAMETER TUBES 181

that several nuclear installations expect to attain high conver-

sion ratios by reducing the moderation of neutrons. This
results in the decrease of the core water volume and corre-
sponds to a reduction in the cross-sectional area of flow chan-
nels. Okawa et al. (2004, 2005) working with a tube whose
inside diameter is 5 mm, obtained experimental data of the
deposition rate and entrainment fraction in annular flow to
investigate the onset of critical heat flux condition, which is
essential to the aforementioned nuclear power plants.
All heat exchanger design involves optimizing heat transfer
and pressure drop characteristics, both of which are affected Figure 1. Bubble, slug, churn and annular flow patterns (Taitel et al.,
by the various two-phase flow patterns. Although the 1980).
pressure drop characteristics are often overlooked, the
mechanical pumping power expended to overcome friction
and gravity is sometimes equal to that consumed in heat of slug flow which occurs at still higher gas flows. The
transfer when dealing with gases. It has been reported shape of the Taylor bubbles is distorted resulting in narrower
(Kays and London, 1998) that this mechanical energy is bubbles and the direction of the liquid phase in churn flow is
worth four to 10 times its heat equivalent in most thermal sys- oscillatory or alternating. Annular flow is characterized by
tems. In essence, mechanical energy can be transformed continuity of the gas phase along the pipe in the core, occur-
into heat with maximum efficiency (e.g., through friction) but ring at the highest gas flow rates. Liquid flows in the upward
when converting heat into mechanical energy (e.g., in a direction, both as a thin wavy film and as entrained droplets in
power station) some of the energy will be lost and 100% effi- the gas core.
ciency is never attainable. As a consequence, knowledge of
the void fraction is useful because it is required for determin- Flow Pattern Transitions
ing the acceleration and gravity components of pressure
The widely applied models by Taitel et al. (1980), Jayanti
drop, from which the frictional component can be inferred.
and Hewitt (1992) and Barnea (1986) for the bubble-
In the present study, void fraction data and flow pattern
to-slug, slug-to-churn and churn-to-annular flow transitions
information have been obtained from a small diameter
respectively are to be tested to determine their suitability for
(5 mm) pipe. High-speed video pictures, time-varying void
small channels of similar size to those found in compact
fraction data and probability density functions (PDFs) have
heat exchangers.
been used to discriminate between the various flow patterns.
Taitel et al. (1980) postulated that discrete bubbles com-
An empirical flow pattern map for the conditions of our exper-
bine into larger vapour spaces whose diameter approximates
iment has been developed and compared with the most
to the tube by agglomeration. This is assumed to occur as
commonly used correlations for flow pattern transitions in
the void fraction reaches 0.25 and results in a transition to
two-phase vertical upflow. The velocities of the various struc-
slug flow.
tures present in the flow were determined and have been
Jayanti and Hewitt (1992) concurred that the flooding
compared with the relationship proposed by Nicklin et al.
mechanism appears to be the most likely cause of the tran-
(1962) and Pearce (1979) for slug velocity and the velocities
sition from slug to churn flow in vertical tubes and suggested
of disturbance waves on wall film respectively. Some of our
an improvement to the modelling of the flooding mechanism
data has been taken in similar conditions to those of Holt
given by McQuillan and Whalley (1985) which was capable of
et al. (1999), who used the method of quick closing valves
predicting the transition over the full range of liquid rates.
for void fraction measurements. The results from the two
Watson and Hewitt (1999) found that the model of Jayanti
sets have been compared. The mean void fraction values
and Hewitt (1992) gives better predictions of the slug-to-
obtained have been tested against a flow pattern specific
churn flow transition than the methods by McQuillan and
method to assess its predictive capability.
Whalley (1985), Mishima and Ishii (1984) and Brauner and
Barnea (1986) in a 32 mm diameter tube.
The model for the transition from annular to intermittent
FLOW PATTERNS flows (slug or churn) developed by Barnea (1986) is pro-
Flow Pattern Classification posed to occur when the gas core is blocked at any location
by the liquid. There are two mechanisms by which this block-
The classification of flow patterns in co-current vertical age may occur. The first mechanism which predominates at
gas–liquid systems, by Hewitt and Hall-Taylor (1970), into low liquid flow rates is results from the instability of the annu-
four major groups: bubble, slug, churn and annular flow is lar configuration while the second mechanism is caused by
now widely accepted. These flow patterns are depicted in spontaneous blockage of the gas core as a result of axial
Figure 1. transfer of liquid in the film.
In bubble flow, the gas phase is approximately uniformly
distributed in a continuous liquid phase in the form of discrete
Flow Pattern Specific Model for Void Fraction
bubbles. Slug flow occurs at slightly higher gas flow rates. In
this flow pattern, most of the gas phase is situated in large Published models in the literature for predicting void fraction
bullet shaped bubbles referred to as ‘Taylor bubbles’. Suc- in slug and annular flows have been applied to the present
cessive Taylor bubbles are separated by liquid slugs that work. For churn flow, which is the least understood flow pat-
bridge the pipe and usually contain small gas bubbles. tern intermediate between slug and annular flows, both
Churn flow is a more chaotic, frothy and disordered version models have been applied to determine the most appropriate.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180–192
182 OMEBERE-IYARI et al.

Slug model
Slug flow is assumed to consist of a succession of identical
unit cells. A unit cell comprises a liquid slug and a cylindrical
Taylor bubble surrounded by a falling liquid film. A zero void
fraction is assumed in the liquid slugs based on the obser-
vation of de Cachard and Delhaye (1996) in their 10 mm
diameter vertical pipe. Ros (1961) has proposed that the con-
dition for a non-aerated liquid slug is given by
(rl  rg )gD2
Bo ¼ , 140
s
This corresponds to a diameter of less than 32 mm for the low
pressure systems being considered in this study. The
assumption therefore appears to be reasonable.
The gas mass balance gives
ugs
ug ¼
1
The widely used expression by Nicklin et al. (1962) for the
Taylor bubble velocity is
Assuming the liquid film is sufficiently thin and the droplet
velocity relative to the gas phase is sufficiently small. Hence,
rl Euls
_ d ¼ kd C  kd
m (14)
ugs
Whalley and Hewitt (1978) and Ishii and Mishima (1989)
show that increasing the interfacial shear force causes an
increase in the amount of droplets in equilibrium but increas-
ing the liquid surface tension has a reverse effect. The rate of
(1) droplet entrainment is thus assumed to be characterized by
the dimensionless number, pe which is defined as the ratio
of interfacial force to the surface tension force acting on the
liquid film surface.
fi rg u2gs
pe ¼ (15)
s=a

(2)
The mean liquid film thickness, ā is determined from the bal-
ance between the interfacial shear force and wall friction
force acting on the liquid film:

ug ¼ 1:2(uls þ ugs ) þ u0 (3) fi rg u2g  fw rl u2lf (16)

From equations (2) and (3), the average void fraction may be As the liquid film is thin, ug and ulf are approximated by ugs
expressed by and D=4au
fs respectively. The interfacial friction factor fi and
the wall friction factor fw are estimated using correlations by
ugs
1g ¼ (4) Wallis (1969):
1:2(uls þ ugs ) þ u0
 
a

u0 is determined by the correlation of White and Beardmore fi ¼ 0:005 1 þ 300 (17)
D
(1962)
16
fw ¼ or 0:005, whichever is greater (18)
u0 ¼ G(gD)1=2 (5) Relf

G ¼ 0:345(1  e½0:01Nv =0:345 )(1  e½3:37Bo=r  ) (6) Furthermore, the volume of the entrained droplets is
assumed to be proportional to the dimensionless term, pe.
r ¼ 10 when Nv . 250 (7)
_e
m
/ pe (19)
r ¼ 69(Nv )0:35 when 18 , Nv , 250 (8) rl

r ¼ 25 when Nv , 18 (9) From equations (11) and (14), the following expression is
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi obtained:
rL (rL  rG )gD3
Nv ¼ , 3:105 (10) _ e kd Euls
m
m2L ¼ (20)
rl ugs

Annular model Okawa et al. (2004) correlated annular flow data by

_ d equals
In equilibrium annular flow, the deposition rate, m
_ e.
the entrainment rate, m kd Euls
¼ ke pne (21)
ugs
_e ¼m
m _d (11)
where
m_ d is assumed to be proportional to the droplet concentration
in the gas core, C. ke ¼ 3:1  102 m s1 and n ¼ 2:3 for pe ,0:0675 (22)

_ d ¼ kd C
m (12) ke ¼ 1:6  103 m s1 and n ¼ 1:2 for 0:0675, pe ,0:295

where kd is the mass transfer coefficient of droplet deposition. (23)

_ d can be approximated by
m
ke ¼ 6:8  104 m s1 and n ¼ 2:3 for pe .0:295 (24)
r Euls
C l (13)
ugs The use of the Okawa et al. (2004) correlation is justified as it
was developed using experimental data obtained with the
where E is the entrained fraction. same fluid pairing as the present work (air/water) and in

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180– 192
 A STUDY OF FLOW PATTERNS FOR GAS/LIQUID FLOW IN SMALL DIAMETER TUBES 183

Figure 2. Schematic of the small diameter rig.

pipes whose range of diameters (5–42 mm) includes that
used in the present study.
The void fraction is determined from
 
D  2a
2
1g ¼ E (25)
D the two-phase flow travels for 0.67 m upwards, 1.07 m hori-
EXPERIMENTAL
The vertical two-phase flow experiments were carried out
on a specially designed facility in the School of Chemical,
Environmental and Mining Engineering (SChEME). The test
fluids used were air and mains tap water at ambient tempera-
ture. The gas and liquid superficial velocities ranged from
0.062 to 64.5 m s21 and 0.03 to 0.65 m s21 respectively at
pressures ranging from atmospheric to 2 bara. These flow
rates gave slug, churn and annular flow patterns. The flow
loop is shown schematically in Figure 2.
Air was drawn from the compressed air main and supplied
to the mixing unit where it combines with water drawn from a
storage tank by means of a centrifugal pump. Inflow of air and
water are controlled using separate banks of calibrated rota-
meters. The mixing unit consisted of an annular section
surrounding a porous wall.
Water enters the main pipe from the periphery to form a film
on the wall whilst the air passes along the middle. Down-
stream of the mixer, the two phases travel for 1.97 m up to
the test section along a 0.005 m diameter stainless steel
vertical pipe, where time varying, cross-section averaged
void fractions were measured using a cascade of eight iden-
tical conductance probes placed sequentially along the pipe.
Pressure measurements were taken with a pressure gauge
mounted close to the test section. Beyond the test section,
zontally and finally a further 3.12 m vertically downwards.
The outlet stream is collected in a cylindrical vessel of
100 mm diameter where the phases are separated. The
pressure in the test section was controlled using a gate
valve positioned at the separator inlet. After separation, the
liquid is collected from the bottom of the separator and air
passes out through a pipe at the top. The exit liquid could
either be recycled or taken off to be measured. During
measurements, water is diverted into a measuring cylinder
where the volumetric flow rate was measured for a sufficiently
long time to minimize uncertainty of measurement. Air is
metered at the top of the separator vessel using a calibrated
wet gas flow meter which allowed the flow rates to be
measured over a period of time. The liquid level in the
separator vessel was kept constant to ensure accuracy.
The pulsations associated with slug flow meant rotameters
would be inappropriate because of the large oscillations in
gas flow rate. The error of the liquid and gas flow rate
measurements are considered to be acceptable and range
from 0.71% to 1.64% and 0.59% to 5.05%, respectively.
Detailed calculations of the measurement errors were made
by Omebere-Iyari (2006).

Table 1. Physical properties of air and water at test conditions.

Pressure Liquid density Gas density Liquid viscosity Gas viscosity Surface tension
Fluids (bara) (kg m23) (kg m23) (Pa s21) (Pa s21) (N m21)

Air–water 1– 2 998 1.2–2.4 0.001 1.8  1025 0.073

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180–192
184 OMEBERE-IYARI et al.
Figure 3. Conductance probe set-up.
The physical properties of the fluid pairing are presented in
Table 1. The air –water mixture has been chosen because of
the minimal risk involved and its ease of handling whilst pro-
viding valuable flow pattern information.
The conductance technique has been chosen to measure
void fraction because water is an electrical conductor, albeit a
poor one, while air is essentially resistive. In this technique,
the cross-sectional averaged void fraction can be estimated
once the relationship between electrical impedance and
phase distribution has been established. Andreussi et al.
(1988) and Tsochatzidis et al. (1992) are some of the
researchers to have used this method successfully. The
probes consist of two stainless steel ring electrodes mounted
in acrylic resin housing (Figure 3). They were carefully
manufactured so that the electrodes had the same diameter,
D as the test section (5 mm). The distance between the
electrode plates, De, and width of each plate, s, are 1.7 and
0.5 mm respectively. The electronic circuitry employed for
the conductance probe is similar to that of Fossa (1998)
and Fossa and Guglielmini (1998).
The probes give a voltage output (Vout) which is pro-
portional to the resistance of the two-phase mixture. This
response is converted to dimensionless conductance (Ge
)
by referring to the value obtainable when the pipe is full of
liquid (Vfull). To account for the variation in water resistance,
Vfull was measured before every experimental run and
adjusted if a considerable drift was observed. The calibration
procedure involved artificially creating instantaneous liquid
fractions over the conductance probes using plastic plugs
with cylindrical rods of known diameters. Ge
is determined
from the voltage responses, Vout and Vfull and correlated
with the corresponding liquid/void fractions. A unique third
Figure 4. Conductance probe calibration curve compared with theory.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180– 192
Figure 5. Trends in the mean void fraction.
order polynomial is then obtained for each probe with Ge

as a function of void fraction (Figure 4).
The measurements from the probes are acquired using a
PC installed with a National Instrument (NI) DAQ card. A
modified version of the Labview program by Guglielmini
(2002) was used to sample and convert the voltage signals
to void fraction. Cross correlation of consecutive time-series
yielded vital structure velocity information. Sampling rates
of 1 and 5 kHz were used with data being taken for 60 and
12 s respectively.
Although flow pattern observation with the naked eye is the
simplest method, discrimination of flow features proves more
difficult at higher flow rate. A KODAK HS 4540 High Speed
Video System was positioned at the test section so that the
flow pattern and related structures in the two-phase flow
could be observed. Video footage for varying flow conditions
corresponding to slug, churn and annular flow patterns were
recorded for detailed study in slow motion. The video signals
were downloaded both as a sequence of digital still pictures
and in analogue VHS format. Proper illumination from a
halogen lamp was necessary in order to obtain good
quality pictures.
RESULTS
Void Fraction
The mean void fraction is plotted against the gas super-
ficial velocity for constant liquid superficial velocity in
Figure 5. As expected, the mean void fraction generally
increases with gas flow rate. At the liquid superficial velocities
of 0.1 m s21 and 0.4 m s21, we observe that an increase in
pressure from atmospheric to 2 bara results in a decrease
in mean void fraction for low gas superficial velocities but
Figure 6. Mean void fraction at low liquid superficial velocities.
 A STUDY OF FLOW PATTERNS FOR GAS/LIQUID FLOW IN SMALL DIAMETER TUBES
Figure 7. Regimes corresponding to changes in velocity structure
velocity.
has a negligible effect at higher gas superficial velocities.
Figure 6 shows that low liquid flow rates are characterised
by an increase, a reduction and further increase in the
mean void fraction with increasing gas flow rate.
Flow Structures
The motion of flow structures in the various two-phase flow
patterns, for example, void waves in bubble flow, liquid slugs
and Taylor bubbles in slug flow, huge waves in churn flow or
disturbance waves in annular flow are characterized by their
structure velocity. The time varying void fractions from the
two highest probes are cross correlated to yield a time
delay, from which structure velocity is determined as the dis-
tance of separation is known.
Figure 7 highlights that three distinct regions exist when
structure velocity is compared with gas superficial velocity.
These are a positive slope followed by a negative slope
and then a positive one. These changes might be linked to
the various flow pattern transitions. It can be seen from
Figure 8 that the experimental structure velocities are con-
sistent with the commonly used correlation for slug velocity
by Nicklin et al. (1962) at low mixture velocities. Comparisons
have been made with Nicklin et al. (1962) because recent
correlations are based on a similar format and the Nicklin
et al. (1962) correlation has been shown to be applicable to
various flow structures including slugs, void waves and
wisps (Azzopardi, 2004).
Figure 8. Correlating structure with equation of Nicklin et al. (1962).
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180–192
185
The effect of pressure is studied for several gas flow rates
at two liquid velocities, i.e., 0.1 and 0.4 m s21. An increase in
the pressure from 1.1 to 2 bara, results in a slight reduction in
the structure velocity whilst retaining the characteristic
change in slope.
The time varying void fraction data show the simultaneous
occurrence of huge waves and liquid slugs (Figure 9), con-
firming the findings made by Sekoguchi and Mori (1997)
using advanced measuring devices for longitudinal distri-
butions of the liquid hold-up along a tube. The void fraction
of a structure must equal zero to be classified as a liquid slug.
Visualization
The need for a detailed flow visualization scheme arose
when initial attempts to describe the flow by observations
with the naked eye proved inconclusive due to the two-
phase flow mixture moving too fast. Slug, churn and annular
flows were observed with high speed video pictures at 1.1
bara, a constant liquid superficial velocity of 0.1 m s21 for
gas superficial velocities of 0.12, 1.02, 1.43, 2.00, 4.07,
6.10 and 20.3 m s21. The observations were made at a
speed of 750 frames s21.
At the lowest gas flow rate, the flow is characterized by
Taylor bubbles (whose diameters approximate to the tube
diameter) followed by non-aerated liquid slugs. A thin liquid
film present on the pipe wall appears to travel counter-current
to the gas flow.
Increasing the gas flow rate to 1.02 and 1.43 m s21, the
flow behaviour is essentially the same albeit the liquid slugs
occur less frequently. When the gas superficial velocity
reaches 2 m s21, the liquid slug becomes aerated and
although the flow pattern is still slug flow, the rate of falling
liquid on the wall is very much increased. At about 4 m s21,
features of churn flow are present. The process of a falling
liquid film levitated only for limited height by the gas phase
is repeated until the gas core possess sufficient momentum
to transport the accumulated liquid further downstream. At
time intervals averaging 0.8 s, a fast moving chaotic two
phase mixture akin to a highly aerated liquid slug body is
observed passing through the test section. At a gas velocity
of 6.1 m s21, the churn flow is faster and the occasional
swift liquid body bears fewer semblances with a liquid slug.
The characteristic oscillatory motion is a major feature of
the flow. Increasing the gas velocity to 20.3 m s21 gives
fully developed annular flow. Most of the liquid phase travels
on the tube wall and occasional bridging of the pipe by the
liquid phase can be seen.
Probability Density Function
PDF is the rate of change of the probability that void frac-
tion values lie within a certain range versus void fraction. The
total area under the probability density function must equal
unity. The probability density function is determined using
the hist command in MATLAB.
The PDF of time varying void fraction has been used to
classify the flow patterns in the same manner as Costigan
and Whalley (1997). Typical examples for slug, churn and
annular flows are given in Figure 10. This is for a liquid super-
ficial velocity of 0.65 m s21 and gas superficial velocities of
0.15, 1.05, 5.22 and 23.9 m s21 respectively at 1.1 bara.
The slug flow regime is characterized by twin peaks
186 OMEBERE-IYARI et al.

Figure 9. Void fraction time-series for ULS ¼ 0.1 m s21, UGS ¼ 0.14 m s21, P ¼ 1.1 bara, acquisition frequency ¼ 5 kHz; sampling time ¼ 12 s.

corresponding to the liquid slug and the Taylor bubble. A
single peak existing at a high void fraction is a feature of
annular flow. In addition to a high void fraction peak, churn
flow possesses a tail extending to low void fractions.
COMPARISONS WITH PREVIOUS WORK
In the present experiments, some void fraction measure-
ments were taken at exactly the same conditions as those
of Holt et al. (1999), which were taken at 1.5 bara in a
5 mm diameter vertical tube for liquid superficial velocities
of 0.02, 0.08 and 0.12 m s21. The results from both sources
are given in Figures 11– 13. The conductance probe tech-
nique was employed in our experiments whilst Holt et al.
(1999) used the quick closing valves method. Most of the
data lies in the annular flow regime, where the liquid phase
can be transported both as a thin film on the pipe wall and
as entrained droplets in the liquid core. The droplets in the
core are insulated from the liquid film by the non-conducting
air phase, and are therefore not accounted for by the conduc-
tance technique. Hence, the experimental void fraction (or
more appropriately film fraction) obtained through conduc-
tance is expected to be greater than the actual void fraction
measured using quick closing valves for annular flow where
there is liquid entrainment.
At the lowest liquid superficial velocity of 0.02 m s21, the
flow pattern obtained is annular. Figure 11 shows that the
mean void fractions from both sets of results are indistin-
guishable. This means that the entrainment of the liquid
phase in the gas core is negligible. Confirmation is found in
the requisite critical liquid film flux predicted by the relation-
ship of Govan et al. (1988) for the onset entrainment. For
these experiments, entrainment is predicted to occur at
liquid superficial velocities of 0.084 m s21 and above. When
the liquid flow rate is increased to 0.08 m s21, churn and
annular flows are present. Figure 12 shows that at the
higher gas flow rates where annular flow exists, the mean
void fraction from Holt et al. (1999) is in good agreement
with present experiments. However, at gas velocities below
10 m s21, the values differ considerably. A possible reason
for this is that critical liquid film flux which predicts the
onset of entrainment is only applicable to annular flows.
Figure 13 gives the data obtained at a liquid velocity of
0.12 m s21. The correlation of Okawa et al. (2004) is used
to estimate the equilibrium entrainment fraction, which is sub-
tracted from the experimental void fraction to obtain an actual
corrected void fraction value. Figure 13 shows that the corre-
lation underpredicts the void fraction at gas velocities of up to
10 m s21 and gives good agreement between 10 and
40 m s21 whilst overpredicting the entrainment above

Figure 10. Typical PDF plots.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180– 192
 A STUDY OF FLOW PATTERNS FOR GAS/LIQUID FLOW IN SMALL DIAMETER TUBES 187

Figure 11. Mean void fractions at ULS of 0.02 m s21.

Figure 13. Mean void fraction at ULS of 0.12 m s21 (corrected for
entrainment).
40 m s21. The reason for this underprediction is because the
entrainment correlation has been developed for annular flows
but the flow pattern at low gas superficial velocities is either or maximum voidage for bubble flow and the bubble/pipe
slug or churn. diameter ratio (dimensionless bubble size) (Figure 15).
Assuming a constant bubble size, the relationship suggests
that critical voidage reduces as pipe diameter is decreased.
DISCUSSION Omebere-Iyari (2006) provides an explanation for this
Flow Pattern Transitions
The comparisons of flow patterns identified using PDFs
with observations from a high speed video camera for a con- Table 2. Flow pattern identification from PDFs and visual
stant liquid superficial velocity of 0.1 m s21 at 1.1 bara are observations.
given in Table 2. The PDF plots agree well with visual obser- Flow pattern
vation of flow pattern transitions except for cases of slug flow UGS (Costigan and Flow pattern
near the boundary with churn flow. The presence of different (m s21) PDF Whalley, 1997) observed
structures within a given flow pattern might be responsible. 0.1 Slug Slug
The empirical correlations by Barnea et al. (1983) for the
slug-to-churn and churn-to-annular flow transitions and the
bubble-to-slug flow transition of Taitel et al. (1980) are com-
pared with our results at 1.1 bara in Figure 14. The transitions
by Barnea et al. (1983) perform well for the slug-to-churn flow
transition at low liquid flow rates only. It is less reliable for the
transition to annular flow.
The correlation of Taitel et al. (1980) which assumes that
the transition from bubble to slug flow occurs at a critical 0.36 Churn Slung
void fraction of 0.25 is in poor agreement with our exper-
iments, as can be seen in Figure 14. Bubble coalescence
which is the mechanism for this transition increases sharply
beyond a certain critical void fraction.
Hence, an accurate estimate of the critical void fraction is
essential for the correct prediction of the bubble-to-slug flow
transition. Recent work by Song et al. (1995), Cheng et al.
(2002) and Guet et al. (2002) on the effect of bubble size 4.4 Churn Churn
show that an inverse relationship exists between the critical

10.7 Annular Annular

Figure 12. Mean void fractions at ULS of 0.08 m s21.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180–192
188 OMEBERE-IYARI et al.
Figure 14. Flow pattern transitions at 1.1 bara.
behaviour using phase distribution in bubble flow. In essence,
it has been shown through experimentation that there is an
increased wall peak distribution when the dimensionless
bubble size is large (as with smaller tubes). This accelerates
the rate of bubble coalescence and increases the chances of
creating large bubble at the wall before dispersion to the core.
If a sufficient number of large bubbles are generated in the
wall region some of them can reach the core without break-
up and form the Taylor bubbles associated with slug flow. It
appears therefore that larger dimensionless bubble sizes
are more favourable towards slug flow formation and explains
the reduction in the critical voidage for bubbles of this magni-
tude. Assuming the critical void fraction to be 0.05, the
relationship of Taitel et al. (1980) gives an improved predic-
tion for the transition to slug flow for our data as shown in
Figure 16.
The changes in the slope of structure velocity obtained
from cross correlation delineate the slug, churn and annular
flow patterns very well at 1.1 bara based on comparisons
with pictures from the high speed video camera. Visual
observations also show that for the churn/annular flow tran-
sition, the theoretical relationship of Barnea (1986) is less
reliable at the conditions examined.
At a liquid superficial velocity of 0.1 m s21 and gas superficial
velocity of 6.1 m s21, the transition of Barnea (1986) wrongly
predicts the existence of annular flow (Figure 16). The slug-
to-churn flow transition by Jayanti and Hewitt (1993) which
Figure 15. Critical void fraction for various bubble sizes.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180– 192
Figure 16. Improved flow pattern transitions at 1.1 bara.
gives flooding as the mechanism for the transition is in good
agreement with our transition using structure velocity.
The effect of pressure on flow pattern transitions is illus-
trated in Figure 17 using structure velocity. The data at 2
bara shows the same characteristic change in structure vel-
ocity as the 1.1 bara data (Figure 8) and can therefore be
used to delineate the flow pattern transitions at 2 bara. The
slug/churn flow transition moves to lower gas superficial vel-
ocities as the pressure is decreased from 2 to 1.1 bara. This
results in a larger slug flow and smaller churn flow region at
the higher pressure. The location of the churn/annular flow
transition is less dependent on system pressure.
Our use of structure velocity for flow pattern discrimination
is advantageous due to its objectivity. In addition, the reason-
able success of the aforementioned method has shown that
changes in the characteristics of flow structures are associ-
ated with flow pattern transitions and may be used for further
study of flow pattern transition mechanisms.
Void Fraction Prediction
The mean void fraction values were compared to the flow
pattern specific method described earlier (Figure 18). A stat-
istical analysis of its performance is presented in Table 3. The
terms, F and S, which are the correction factor and the trans-
formed standard deviation respectively, were shown to be
Figure 17. Effect of pressure on the experimental slug/churn and
churn/annular flow transitions.
 A STUDY OF FLOW PATTERNS FOR GAS/LIQUID FLOW IN SMALL DIAMETER TUBES 189

structures such as huge waves which we observed in fully

developed slug flow is not accounted for (Figure 9). It is
likely that the velocities of the huge waves differ significantly
from those of the liquid slugs. The model could therefore be
improved by using a robust method which takes the velocities
of the various structures in the flow into consideration.

Disturbance Wave Velocity

Figure 18. Comparisons of void fraction data with a flow pattern
specific method.
useful analytical tools by Govan (1988). F is defined as the
average factor by which the calculated value must be multi-
plied to give the experimental value, 1/exp(m); and S is
given by exp (s) 2 1.
Azzopardi (1997) in his extensive review of drops in annu-
lar flow indicates that in spite of the importance of waves,
there has been limited amount of theoretical work in compari-
son with the experimental aspect. Of the correlations dis-
cussed by Azzopardi (1997) for the film wave velocity in
annular flows, the work by Pearce (1979) based on simplified
equations of motion for the film and vapour core is tested
against our data. Pearce (1979) correlated the film wave vel-
ocity using
pffiffiffiffiffiffiffiffiffiffiffi
K u
lf þ ugs rg =rl
uw ¼ pffiffiffiffiffiffiffiffiffiffiffi (29)
K þ rg =rl

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi When tested against data from a horizontal tube of 31.8 mm

u n
u1 X diameter and using K ¼ 0.8, the calculated wave velocity
s¼t (e(i)  m)2 (26)
n i¼1 showed good agreement with the measured wave velocity
  obtained by cross correlation. The time-averaged film
P(i) velocity u
lf is determined from
e(i) ¼ log (27)
E(i)
1X n Dm _ lf
m¼ e(i) (28) u
lf ¼ (30)
n i¼1 4  rl  d

P(i) is the predicted or calculated void fraction and E(i) is the
experimental void fraction.
The slug flow model is applied to data identified as slug
and churn flow and the annular model to experiments ident-
ified as cases of churn and annular flow. Flow pattern identi-
fication is performed using the methods we found to be most
reliable. These methods are the modification to Taitel et al.
(1980) for the bubble/slug flow transition and the use of
structure velocity for the slug/churn and churn/annular flow
transitions.
The void fraction predictions are marginally better when the
slug flow model is employed for churn flow than with the
annular flow model (Table 3). This can be explained by
the fact that the huge waves in churn flow behave more
like liquid slugs while annular flow with smaller drop and
waves is significantly different. Overall, the flow pattern
specific method tends to underpredict the data especially
for the slug flow experiments. We attribute this to the fact
that the slug flow model employs a correlation for the slug
velocity only (Nicklin et al., 1962). The motion of other
_ lf is calculated using
The liquid film flow rate in annular flow, m
the method suggested by Hewitt and Govan (1990). Here, the
expression below containing the entrainment and deposition
correlations is integrated:
dm_ lf 4
_dm
¼ (m _ e) (31)
dz d
When applying the relationship of Pearce to our data, the
film thickness is computed by using the experimental void
fraction (assuming the liquid film is symmetric around the
wall). The best fit after optimization is obtained with
K ¼ 0.51 (Figure 19). Several data points have been ignored
due to very noisy cross correlation signals. The wave velocity
correlation is applied to other data obtained from different

Table 3. Statistical performance of the flow pattern specific method

for void fraction prediction.

Option 1 (slug model for slug Option 2 (slug model for

and churn flows and annular slug flow and annular model
model for annular flow) for churn and annular flows)

F 1.14 1.19
S 0.17 0.19

Ideally F ¼ 1 and S ¼ 0. Figure 19. Correlating wave velocities in annular flow.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180–192
190 OMEBERE-IYARI et al.

Figure 20. Wave velocity for various pipe diameters.

pipe diameters. Figures 20 and 21 highlight the effect of pipe

diameter on K. The 10 and 32 mm data are from Willets
(1987) and Wolf (1995) respectively, and the 25 mm and
42 mm data are sourced from Schadel (1988). The value
of K is about 0.9 for pipes with diameters in the range of
25–42 mm. However, K decreases to 0.61 and 0.51 as the
diameter is reduced to 10 mm and 5 mm, respectively.
Figure 22. Description of film thickness parameters in vertical annular
To explain the decrease of K with pipe diameter, an under- flow.
standing of the film thickness structure in vertical annular flow
and the physical definition of K is essential. The film thick-
ness wave in vertical annular flow is described in Figure 22.
exist on the liquid film circumference and translates to the
K and the mean film thickness are given by equations (32)
increased growth of the disturbance wave peaks, ap given
and (33).
that a smaller amount of sites are available for atomization.
a
From the work of Schadel (1988), the mean film thickness
K ¼ pffiffiffiffiffiffiffiffiffi (32) and intermittency for similar gas and liquid flow rates are
at ap
found to be independent of pipe diameter. This implies that
a
¼ at (1  I) þ ap jI (33) the increase in the disturbance wave peak, ap must corre-
spond to a decrease in the wave trough or base film, at.
where j is the factor that accounts for the non-uniformity of the Assuming that the wavelength is at least twice as large as
disturbance wave. I, the intermittency is given by as w/l. the disturbance wave width and the volume of liquid con-
Visualization studies by Hewitt and Lovegrove (1969) in a tained in the disturbance wave is smaller than that which
32 mm diameter tube show that disturbance waves appear forms the rest of the film, the reduction in at will be small in
as coherent bands. However, Azzopardi et al. (1983) comparison to the increase in ap. From the definition of K in
observed the waves to be circumferentially localized in a equation (32), the decrease of K with pipe diameter is there-
large tube having a diameter of 125 mm. Martin and fore reasonable. Support for our findings can be found in the
Azzopardi (1985) have indicated that the breakdown in autospectral density measurements of Martin and Azzopardi
coherence is systematic with pipe diameter. This increased
coherence in smaller pipes means that fewer imperfections

Figure 23. The effect of tube diameter on autospectral densities of

Figure 21. Variation of K with pipe diameter. film thickness measurements (Martin and Azzopardi, 1985).

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180– 192
 A STUDY OF FLOW PATTERNS FOR GAS/LIQUID FLOW IN SMALL DIAMETER TUBES
(1985) in 10 mm, 32 mm and 58 mm diameter pipes
(Figure 23). Here, the broadening with diameter is indicative
of an decrease in the degree of coherence. A distinct change
in the sharp-nature of the peak is observed between the pipe
diameters of 32 mm and 10 mm. This behaviour corresponds
to the sudden change in K when the diameter is reduced from
25 mm to 10 mm. We continue to investigate the reasons for
the aforementioned pipe diameter effects and our findings
may be the subject of a future publication.
CONCLUSIONS
Conductance probes have been successfully used to
acquire valuable time-varying void fraction data and flow pat-
tern information in a 5 mm diameter pipe. The use of high
speed video and PDF plots have enabled the study of flow
pattern transitions.
A novel approach has been developed for determining slug-
to-churn and churn-to-annular flow transitions in vertical two
phase flow based on changes in the structure velocity. The
PDF method of Costigan and Whalley (1997) fails to delineate
the exact location of the slug/churn transition accurately. The
Jayanti and Hewitt (1992) slug/churn flow transition model is
reliable at the conditions studied. The theoretical correlation of
Barnea (1986) performs remarkably better than the empirical
one of Barnea et al. (1983) for churn/annular flow transition. It
is shown that better predictions of the bubble-to-slug flow
transition in small passages is obtained by modifying the
relationship of Taitel et al. (1980) to account for the variation
of the critical void fraction with bubble size.
A flow pattern specific method modelled on slug and annu-
lar flow gives good predictions of void fraction. The slug flow
model performs better than the annular flow model for the
prediction of void fraction in churn flow. We believe that
huge waves in churn flow being more similar in behaviour
to liquid slugs than they are to the drops and waves in annu-
lar flow provides an explanation.
The relationship for the velocities of disturbance waves on
the wall film in annular flow by Pearce (1979) is applied to
both our data and other experiments performed in pipes
with diameters ranging from 5 mm to 42 mm. The correlating
constant, K increases sharply with pipe diameter before
attaining a steady value of 0.9.
NOMENCLATURE
ā mean film thickness, m
Bo Bond number [defined in equation (1)]
C droplet concentration in the gas core, kg m23
D tube diameter, m
E entrainment fraction
fi interfacial friction factor
fw wall friction factor
F correction factor
j non-uniformity factor [in equation (32)]
I intermittency
k mass transfer coefficient, m s21
K correlating parameter for wave velocity
m mass transfer rate or flux, kg m22s21
n power of density ratio
Nv dimensionless inverse viscosity [defined in equation (10)]
P pressure, bar
r function of Nv [in equation (16)]
S transformed standard deviation
u, U cross-sectional area-averaged velocity, m s21
ū time-averaged film velocity, m s21
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180–192
191
uo rise velocity of a Taylor bubble in stagnant fluid
z channel length, m
Greek symbols
1g void fraction
pe ratio of interfacial force to surface tension
r density, kg m23
s surface tension, N m21
Subscripts
d deposition
e entrainment
fs superficial liquid film
g gas phase
gs superficial gas phase
l liquid phase
lf liquid film
ls superficial liquid phase
p peak
s structure
t trough
w wave
REFERENCES
Andreussi, P., Di Donfrancesco, A. and Messia, M., 1988, An impe-
dance method for the measurement of liquid hold-up in two-
phase flow, International Journal of Multiphase Flow, 14; 777– 785.
Azzopardi, B.J., 1997, Drops in annular two-phase flow, International
Journal of Multiphase Flow, 23: 1– 53.
Azzopardi, B.J., 2004, Bubbles, drops and waves: differences or
underlying commonality, 42nd European Two-Phase Flow Group
Meeting, Genoa, 23–25 June.
Azzopardi, B.J., Taylor, S. and Gibbons, D.B., 1983, Annular two
phase flow in large diameter pipes, International Conference on
the Physical Modelling of Multiphase Flow, Coventry, UK.
Bao, Z.Y., Bosnich, M.G. and Haynes, B.S., 1994, Estimation of void
fraction and pressure drop for two-phase flow in fine passages,
Transactions of the Institution of Chemical Engineers, 72 (Part
A): 625 –632.
Barnea, D., 1986, Transition from annular flow and from dispersed
bubble flow—unified models for the whole range of pipe incli-
nations, International Journal of Multiphase Flow, 12 (5): 733– 744.
Barnea, D., Luninski, Y. and Taitel, Y., 1983, Flow pattern in horizontal
and vertical 2 phase flow in small diameter pipes, Canadian Jour-
nal of Chemical Engineering, 61(5): 617– 620.
Brauner, N. and Barnea, D., 1986, Slug/churn transition in upward
gas-liquid flow, Chem. Eng. Sci., 41(1): 159– 163.
Cheng, H., Hills, J.H., and Azzopardi, B.J., 2002, Effects of initial
bubble size on flow pattern transition in a 28.9 mm diameter
column, International Journal of Multiphase Flow, 28(6): 1047–
1062.
Cheng, L. and Mewes, D., 2006, Review of two-phase flow and flow
boiling of mixtures in small and mini channels, International Journal
of Multiphase Flow, 32: 183 –207.
Cheng, T.-W. and Lin, T.L., 2001, Characteristics of gas-liquid two-
phase flow in small diameter inclined tubes, Chem. Eng. Sci., 56:
6393–6398.
Costigan, G. and Whalley, P.B., 1997, Slug flow regime identification
from dynamic void fraction measurements in vertical air-water
flows, International Journal of Multiphase Flow, 23(2): 263– 282.
de Cachard, F. and Delhaye, J.M., 1996, A slug-churn flow model for
small-diameter airlift pumps, International Journal of Multiphase
Flow, 22(4): 627 –649.
ETSU, 1994, Guide to Compact Heat Exchangers, Department of the
Environment’s Energy Efficiency Best Practice Programme, Report
Good Practice Guide 89.
Fossa, M., 1998, Design and performance of a conductance probe
for measuring liquid fraction in two-phase gas-liquid flow, Flow
Measurement and Instrumentation, 9: 103– 109.
Fossa, M. and Guglielmini, G., 1998, Dynamic void fraction measure-
ments in horizontal ducts with sudden area contraction, Inter-
national Journal of Heat Transfer, 14: 3807–3815.
192 OMEBERE-IYARI et al.
Govan, A.H., 1988, A note on statistical methods for comparing
measured and calculated values, United Kingdom Atomic Energy
Authority Report AERE-M3621.
Govan, A.H., Hewitt, G.F., Owen, D.G. and Bott, T.R., 1988, An
improved CHF modelling code, 2nd UK National Heat Transfer
Conf., Glasgow, UK.
Guet, S., Ooms, G. and Oliemans, R.V.A., 2002, Influence of bubble
size on the transition from low-Re bubbly flow to slug flow in a ver-
tical pipe, Experimental Thermal and Fluid Science, 26: 635– 641.
Guglielmini, P., 2002, Caratterizzazione sperimentale del moto bifase
intermittente in presenza di singolarita mediante sonde ad
impedenza, Laurea Thesis, Universitá degli studi di Genova.
Hewitt, G.F., 1981, Introduction to two-phase flow problems in the
process industry, in Bergles, A.E. et al. (eds) Two-Phase Flow
and Heat Transfer in the Power and Process Industries, (Hemi-
sphere Publishing Corporation, Washington, DC, USA).
Hewitt, G.F. and Hall-Taylor, N.S., 1970, Annular Two-Phase Flow,
(Pergamon Press, Oxford).
Hewitt, G.F. and Govan, A.H., 1990, Phenomenological modelling of
non-equilibrium flows with phase change, International Journal of
Heat and Mass Transfer, 33: 229–242.
Hewitt, G.F. and Lovegrove P.C., 1969, Frequency and velocity
measurements of disturbance waves in annular two-phase flow,
UKAEA, Report AERE R4304.
Hibi, K., Kugo T., Tochihara, H., Shimada, S., Okubo, T., Iwamura, T.,
and Wada, S., 2000, Conceptual designing of reduced-moderation
water reactors (2) design for PWR type reactors, Proceedings of
the 8th International Conference on Nuclear Engineering, No. 8423.
Holt, A.J., Azzopardi, B.J., and Biddulph, M.W., 1999, Calculation of
two-phase pressure drop for vertical upflow in narrow passages by
means of a flow pattern specific model, Trans Institution of Chemi-
cal Engineers, 77(Part A): 7– 15.
Ishii, M. and Mishima, K., 1989, Droplet entrainment correlation in
annular two-phase flow, Int J Heat Mass Transfer, 32: 1835–1846.
Jayanti, S. and Hewitt, G.F., 1992, Prediction of the slug-to-churn flow
transition in vertical two-phase flow, International Journal of Multi-
phase Flow, 18(6): 847– 860.
Kays, W.M. and London, A.L., Compact heat exchangers (McGraw-
Hill, New York, USA).
Martin, C.J. and Azzopardi, B.J., 1985, Waves in vertical annular
flow, Physiochem. Hydrodyn., 6: 257–265.
McQuillan, K.W. and Whalley, P.B., 1985, Flow patterns in vertical
two-phase flow, International Journal of Multiphase Flow, 11(2):
161– 175.
Mishima, K. and Ishii, M., 1984, Flow regime transition criteria for
upward two-phase flow in vertical tubes, International Journal of
Heat and Mass Transfer, 27(5): 723–737.
Nicklin, D.J., Wilkes, J.O. and Davidson, J.F., 1962, Two-phase flow
in vertical tubes, Trans Inst Chem Engineers, 40: 61–68.
Okawa, T., Kotani, A. and Kataoka, I., 2004, Experiments for equilibrium
entrainment fraction in a small vertical tube, 5th International Confer-
ence on Multiphase Flow, Yokohama, Japan, 30 May–4 June.
Okawa, T. Kotani, A. and Kataoka, I., 2005, Experiments for liquid phase
mass transfer rate in annular regime for a small vertical tube,
International Journal of Heat and Mass Transfer, 48(3–4): 585–598.
Okubo, T. Shirakawa, T. Takeda, R. Yokoyama, T. Iwamura, T. and
Wada, S., 2000, Conceptual designing of reduced-moderation
water reactors (1) design for BWR type reactors, Proceedings of
the 8th International Conference on Nuclear Engineering, No.
8422.
Omebere-Iyari, N.K. 2006, The effect of pipe diameter and pressure
in vertical two-phase flow, PhD thesis, The University of
Nottingham, UK.
Oya, T., 1971, Upward liquid flow in small tube into which air streams
(1st report, experimental apparatus and flow patterns), Bulletin of
the JSME, 14(78): 1320– 1329.
Pearce, D.L., 1979, Film waves in horizontal annular flow: space time
correlator experiments, Central Electricity Research Laboratories,
Report RD/L/N/ 111/79.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A2): 180– 192
Premoli, A., Francesco, D and Prina, A., 1970, An empirical corre-
lation for evaluating two-phase mixture density under adiabatic
conditions, European Two-Phase Flow Group Meeting.
Ros, N.C.J., 1961, Simultaneous flow of gas and liquid as encoun-
tered in well tubing, J Pet Techn, 13: 1037– 1049.
Sakashita, Y., Yokoyama, T. and Tsuboi, Y., 1999, Core character-
istics of breeding BWR for BARS (BWR with advanced recycle
system), Proceedings of the 7th International Conference on
Nuclear Engineering, No. 7328.
Schadel, S.A., 1988, Atomization and deposition rates in
vertical annular two-phase flow, PhD thesis, University of Illinois,
USA.
Sekoguchi, K. and Mori, K., 1997, New development of experimental
study on interfacial structures in gas-liquid two-phase flow, Proc.
Experimental Heat Transfer, Fluid Mechanics and Thermodyn-
amics, Brussels.
Song, C.H., No, H.C., and Chung, M.K., 1995, Investigation of bubble
flow developments and its transition based on the instability of void
fraction waves, International Journal of Multiphase Flow, 21(3):
381– 404.
Taitel, Y. Barnea, D. and Dukler, A.E., 1980, Modelling flow pattern
transitions for steady upward gas-liquid flow in vertical tubes,
AIChE J, 26(3): 345– 354.
Thome, J.R., 2004, Boiling in microchannels: a review of experimen-
tal and theory, International Journal of Heat and Fluid Flow, 25:
128– 139.
Tsochatzidis, N.A., Karapantsios, T.D., Vostoglou, M.V. and Karabe-
las, A.J., 1992, A conductance probe for measuring liquid fraction
in pipes and packed beds, International Journal of Multiphase
Flow, 18: 653–667.
Wallis, G.B., 1969, One Dimensional Two-Phase Flow, (McGraw-Hill,
London, UK).
Wambsganss, M.W., Jendrzejczyk, J.A. and France, D.M., 1992,
Two-phase flow and pressure drop in flow passages of compact
heat exchangers, SAE Paper No. 920550 presented at Society
of Automotive Engineers International Congress and Exposition,
Detroit.
Watson, M.J. and Hewitt G.F., 1999, Pressure effects on the slug to
churn transition, International Journal of Multiphase Flow, 25:
1225–1241.
Whalley, P.B. and Hewitt, G.F., 1978, The correlation of liquid entrain-
ment fraction and entrainment rate in annular two-phase flow,
United Kingdom Atomic Energy Authority, Harwell, Report AERE -
R 9187. HTFS RS 237.
White, E.T. and Beardmore, R.H., 1962, The velocity of rise of single
cylindrical air bubbles through liquid contained in vertical tubes,
Chem Eng Sci, 17: 351 –361.
Willets, I.P., 1987, Non-aqueous annular two-phase flow, PhD thesis,
University of Oxford, UK.
Wolf, A., 1995, Film structure of vertical annular flow, PhD thesis,
Imperial College, University of London, UK.
ACKNOWLEDGEMENTS
This work has been undertaken within the Joint Project on Tran-
sient Multiphase Flows. The authors wish to acknowledge the con-
tributions made to this project by the Engineering and Physical
Sciences Research Council (EPSRC), the Department of Trade
and Industry and the following: Advantica; AspenTech; BP
Exploration; Chevron; ConocoPhillips; ENI; FEESA; Granherne/
Subsea 7; Institutt for Energiteknikk; Institut Français du Pétrole;
Norsk Hydro; Petrobras, Scandpower; Shell; SINTEF; Statoil and
TOTAL. The authors wish to express their sincere gratitude for
this support.
The manuscript was received 28 November 2005 and accepted for
publication after revision 23 October 2006.