International Journal of Heat and Mass Transfer 53 (2010) 453–465

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Correlations of two-phase frictional pressure drop and void fraction in mini-channel

W. Zhang a,1, T. Hibiki b, K. Mishima c,*
a
Graduate School of Energy Science, Kyoto University, Kyoto 606-8317, Japan
b
School of Nuclear Engineering, Purdue University West Lafayette, IN 47907-2017, USA
c
Research Reactor Institute, Kyoto University, Kumatori, Sennan, Osaka 590-0494, Japan

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

Article history: Alternative correlations of two-phase friction pressure drop and void fraction are explored for mini-chan-
Received 12 March 2008 nels based on the separated flow model and drift-flux model. By applying the artificial neural network,
Accepted 15 June 2008 dominant parameters to correlate the two-phase friction multiplier and void fraction are picked out. It
is found that in mini-channels the non-dimensional Laplace constant is a main parameter to correlate
the Chisholm parameter as well as the distribution parameter. Both previous correlations and the newly
Keywords: developed correlations are extensively evaluated with a variety of data sets collected from the literature.
Frictional pressure drop
Ó 2009 Elsevier Ltd. All rights reserved.
Two phase
Void fraction
Flow boiling
Mini-channel
Small diameter

1. Introduction air/water two-phase flow pressure drop in rectangular and circular

In relation to many cutting-edge electronic chips, avionics, com-
pact heat exchangers and bioengineering devices, mini-channel
cooling technologies have attracted considerable attention in re-
cent years. In comparison with single-phase flow, flow boiling is
deemed as an optimum option to be applied in mini-channels in
view of its extremely high heat transfer efficiency at the cost of
small wall temperature rises. However, a penalty of flow boiling
is the increased pressure drop and pressure fluctuation, which lim-
it the applicable range of flow boiling in such devices. Therefore, a
comprehensive understanding of pressure drop and void fraction
during two-phase flow in mini-channel is of considerable practical
importance for the design and performance evaluation of such
cooling devices.
Starting from studies on adiabatic two-phase flow, extensive
experimental and analytical efforts have been accumulated on
characteristics of two-phase flow and/or flow boiling pressure drop
in mini-channel. However, in regard to the applicability of existing
correlations to mini-channel, there exist some discrepancies in the
literature. Mishima and co-workers [1,2] extensively studied
* Corresponding author. Present address: Institute of Nuclear Technology Insti-
tute of Nuclear Safety System, Incorporated, 64 Sata, Mihama-cho, Mikata-gun,
Fukui 919-1205, Japan. Tel.: +81 770 37 9100; fax: +81 770 37 2008.
E-mail addresses: zhangweizhong@snerdi.com.cn (W. Zhang), hibiki@ecn.pur-
due.edu (T. Hibiki), mishima.kaichiro@inss.co.jp (K. Mishima).
1
Present address: Department of General Technology, Shanghai Nuclear Engineer-
ing Research and Design Institute, 29 Hong Cao Road, Shanghai 200233, China.
mini-channels with diameters ranging from 1 to 5 mm, and found
that the separated flow model could well predict their data and the
Chisholm’s parameter C [3] was successfully correlated by the
hydraulic diameter of channel. However, Triplett et al. [4] reported
that for bubbly and slug flows at high Reynolds numbers the exper-
imental two-phase frictional pressure drop data agreed reasonably
well with the predictions of a homogeneous model with a mixture
viscosity, whereas at low Reynolds numbers or for annular flow,
both the homogeneous mixture model and Friedel’s correlation
[5] predicted the data poorly. In addition, Tran [6] measured
two-phase flow pressure drop during a phase-change heat transfer
process with three refrigerants (R-134a, R-12, and R-113) under
pressures ranging from 138 to 856 kPa, and in two round tubes
(i.d.: 2.46, 2.92 mm) as well as a rectangular channel (i.d.: 4.06
mm). They reported that correlations for conventional channels
failed to predict their experimental data. In contrast, Kawahara
et al. [7] investigated nitrogen/water two-phase flow in a quartz
capillary with the inner diameter of 100 lm. They showed that the
two-phase friction multiplier data were in good agreement with
existing correlations for conventional channels.
To date, studies on void fraction in mini-channels are still lim-
ited. Experimental investigations on void fraction in mini-channel
with diameters in the order of 1 mm, or smaller than that, were ad-
dressed in the literature [1,2,4,8–11]. Among them, Kariyasaki et al.
[8] correlated their data in terms of the gas volumetric quality (or
homogeneous void fraction), b. Moriyama et al. [9] measured void
fractions during N2–R113 adiabatic gas–liquid two-phase flow in
extremely narrow channels with a clearance of 5–100 lm between

0017-9310/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2009.09.011
454 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465
Nomenclature
C Chisholm parameter
C0 distribution parameter
Dh hydraulic equivalent diameter of flow channel
dp/dz friction pressure gradient along channel axis
G mass flux
j superficial velocity
jT mixture volumetric flux, jg + jf
Lo* non-dimensional Laplace constant, {r/[g(qf  qg)]}0.5/Dh
p pressure
pcr critical pressure
Re Reynolds number, GDh/lf
Ref liquid Reynolds number, G(1  xeq)Dh/lf
Refo all-liquid Reynolds number, GDh/lf
Reg gas Reynolds number, GxeqDh/lg
v velocity
Vgj drift velocity
Wefo Weber number, G2Dh/(rqf)
X Martinelli parameter
xeq thermodynamic equilibrium quality
horizontal parallel plates, and proposed a drift-flux type correla-
tion to correlate their data. Later, Mishima and Hibiki [2], and Haz-
uku et al. [11] also reported that the measured void fractions could
be successfully reproduced by the drift-flux type correlations. In
contrast to this, Triplett et al. [4] claimed that homogeneous mod-
els provided the best predictions for their data in bubbly as well as
slug flow regimes, and existing correlations significantly over-pre-
dicted void fractions in annular flow regimes.
Therefore, it is evident that although existing experimental
works have revealed some unique phenomena in mini-channels,
there is still no general theory or correlation available, and some
discrepancies existing among the experimental results are not clar-
ified yet. In view of this, as an extension of efforts by Mishima and
co-workers [1,2] and a part of study dedicating to the development
of a series of correlations for flow boiling in mini-channel [12,13],
the purpose of this study is to summarize the previous studies on
two-phase frictional pressure drop and void fraction in mini-chan-
nels, evaluate their applicability, and then propose alternative
correlations for mini-channels.
2. Correlation development
2.1. Application of neural network
The artificial neural network (ANN) is an advance information
processing techniques. The ANN is composed of elements analo-
gous to the elementary functions of biological neurons [14,15].
One of the most important characteristics of the ANN is its capabil-
ity to learn from trained data and to predict for new data. The back-
propagation neural network (BPN) is one of the simple but
powerful ANNs. Owing to its objectivity of judgment, the BPN is re-
garded as a powerful alternative to current techniques for the pre-
diction of CHF [16] and the classification of flow regimes [17]. Since
any functional relationship can be approximated by a BPN if the
sigmoid layer has enough neurons [14], an architecture-fixed
BPN (layer and neuron numbers fixed) can be utilized to carry
out the input sensitivity (trial and error) analysis in order to select
a set of non-dimensional numbers which could well correlate the
two-phase frictional multiplier or void fraction. The BPN was
trained under MATLAB environment. The TRAINCGP algorithm in
the NEURAL NETWORK TOOLBOX of MATLAB was employed to
train the network in this study.
Greek symbols
a void fraction
q density
r surface tension
/2 two-phase friction multiplier
Subscripts
cal calculational value
exp experimental value
F friction
f saturated liquid or liquid
fo all flow taken as liquid
g saturated vapor or gas
tp two-phase
Mathematical symbol
exp exponential function
2.2. Frictional pressure drop
In order to correlate the data for two-phase friction pressure
drop, it is necessary to find out dominant experimental parameters
related to two-phase friction pressure drop. The hydraulic diame-
ter of channel, Dh, mass flux, G, pressure, p, and thermal equilib-
rium quality, xeq (or their alternatives) are often considered as
the related correlating parameters of local two-phase friction pres-
sure drop (or its equivalence, the two-phase frictional multiplier).
By combining physical properties with these experimental param-
eters, two-phase friction pressure gradient can be expressed in
many non-dimensional forms. For instance, the correlations of
Lockhart and Martinelli [18], Friedel et al. [5], and Zhang and Webb
[19] employed different non-dimensional numbers. However, as
demonstrated by numerous studies on two-phase friction pressure
drop since the pioneering work by Lockhart and Martinelli in
1940s, the separated flow model is the most commonly used meth-
od [20]. The success of this model to correlate the existing data has
demonstrated the usefulness of the Martinelli parameter, X, which
is a combination of the inertial and viscous forces of both phases.
Therefore, it may be deemed as one of the most dominant param-
eters to correlate two-phase friction pressure gradient for mini-
channels. A widely used correlation to calculate the two-phase
frictional multiplier is that proposed by Chisholm and Laird [21],
ðdp=dzÞtp C 1
/2f  ¼1þ þ 2; ð1Þ
ðdp=dzÞf X X
where Chisholm parameter C ranges in value from 5 to 20 for con-
ventional channels, depending on whether the liquid and gas flows
are laminar or turbulent. This successful performance of Eq. (1) has
been shown in the literature [1,7,22,23]. However, the functional
form of Chisholm parameter C needs to be clarified for mini/mi-
cro-channels. The dependence of the Chisholm parameter C on
experimental parameters was made clear by the application of the
BNP to the collected database. The output of the BNP is set to be
the two-phase frictional multiplier, /2f . The Martinelli parameter
is set as one of inputs to the BNP. The mean deviation is used as a
measure of predictive accuracy, defined as
1 X
Mean deviation ¼ jð/exp  /cal Þ=/exp j  100%; ð2Þ
N
where N is the data number. Based on the analysis of input sensitiv-
ity (trial and error method), it was found that the introduction of
 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465
the hydraulic diameter of channel as an input of the BNP can signif-
icantly improve the prediction accuracy by the BNP. The mass flux,
G, quality, xeq, and pressure, p, have minor effects on the results.
These findings are accordant with what were reported in previous
studies [1,2,7,24]. Further analysis by the BNP disclosed that the
non-dimensional Laplace constant, Lo*, defined as
" #0:5 ,
r 15
Lo  Dh ;
gðqf  qg Þ
if introduced into inputs, can work as well as the hydraulic diameter
of channel, and Weber number, Wefo, all-liquid Reynolds number,
Refo, the reduced pressure, p/pcr, and quality, xeq, have the secondary
effects on the predictive accuracy of the BNP (within ±5%) for col-
lected databases.
Based on the above discussion, this study attempts to modify
the correlation of Mishima and Hibiki [2] for the Chisholm param-
eter C in mini-channel:
C ¼ 21½1  expð0:319Dh Þ;
where Dh in mm, is the hydraulic diameter of channel. Since this
correlation is dimensional and may be difficult to scale physical
phenomena, a non-dimensional one would be desirable. Therefore
the hydraulic diameter of channel is replaced by the non-dimen-
sional Laplace constant (or confinement number) Lo* in the
Mishima–Hibiki correlation for the Chisholm parameter, C. The rea-
sons for choosing the Laplace constant as the best choice may be
stated as follows. First, this results from the application of the
ANN in this study. Second, theoretically the Laplace constant scales
the wave length of the Rayleigh–Taylor instability. When the bub-
bles are squeezed in the mini-channel, the formation of bubbles
and the bubble movement are limited by interfacial stability. The
Rayleigh–Taylor instability is the interfacial instability between
two fluids of different densities that are stratified in the gravity field
or accelerated normal to the interface. It is commonly observed that
the boundary between two stratified fluid layers at rest is not stable
if the upper fluid density is larger than the lower-fluid density.
Since the Rayleigh–Taylor instability can lead to the destruction of
the single common interface, it is important in the formation of
bubbles or droplets. In particular, the critical wavelength predicted
by the related stability analysis is one of the most significant length
scales for two-phase flow. Note that this instability is not limited to
the gravitational field. Any interface, and fluids that are accelerated
normal to the interface, can exhibit the same instability. In addition,
the Laplace constant is a nominal bubble size related to the capillary
parameter. Table 8 lists the average values of the non-dimensional
Laplace, Lo*, and the Chisholm parameter, C for each data set. Fig. 1
illustrates the relationship between Lo* and C. According to Sad-
atomi et al. [25] and Moriyama et al. [9], the friction pressure can
be correlated by the equation proposed by Chisholm and Laird
[21], Eq. (1), with C = 21 for conventional channels, and C = 0 for ex-
tremely narrow gaps. By noting these asymptotic values, the Chis-
holm parameter may be correlated by the following equation:
C ¼ 21½1  expð0:358=Lo Þ:
The applicable ranges of this correlation are as follows: 0.014 6
Dh 6 6.25 mm, Ref 6 2000, Reg 6 2000.
2.3. Void fraction
Void fractions can be successfully correlated by the drift-flux
model for conventional channels [26]. According to Moriyama
et al. [9], Mishima and co-workers [1,2], Takamasa et al. [27], Hazuku
et al. [11], and Hibiki and Ishii [28], the drift-flux model is applicable
to correlate the data of void fraction in mini-channels. Therefore,
based on it, a correlation will be proposed in this study.
455
25
-0.358/ Lo
20 C =21(1-e )
Chisholm Parameter, C [-]
ð3Þ
-0.674/Lo
C =21(1-e )
10 C =21(1-e
-0.142/Lo
)
5
0
-2 -1 0 1 2 3
10 10 10 10 10 10
ð4Þ Non-D. Laplace Constant, Lo [-]
Fig. 1. Chisholm parameter C as a function of non-dimensional Laplace constant.
According to the drift-flux model, the relationship between the
gas velocity, vg, and the mixture volumetric flux, jT, can be ex-
pressed by the following equation:
v g  jg =a ¼ C 0 jT þ V gj ; ð6Þ
where C0 is the distribution parameter, and Vgj is the drift velocity.
In a conventional round tube, the distribution parameter and the
drift velocity are given by equations corresponding to flow regimes
[26]. Since the flow regime maps for mini-channels are still under
development [2,29], for simplicity, this study tends to develop a sin-
gle equation for either the distribution parameter or the drift veloc-
ity. According to the existing studies [2,4,8], the behavior of void
fraction for the annular flow regime may be different with those
for the bubbly and churn flows. Therefore, the discussion here is
limited to bubbly and churn flows. Following our previous study
[2], the drift velocity is assumed to be zero in mini-channels. Then,
Eq. (6) can be simplified as
jg = a ¼ C 0 jT : ð7Þ
And then the development of a correlation for void fraction turns to
the development of an equation for the distribution parameter.
In order to develop an equation for the distribution parameter,
dominant non-dimensional numbers should be determined first.
The liquid superficial velocity, jf, gas superficial velocity, jg, hydrau-
lic diameter of channel, Dh, and pressure, p (or their alternatives)
are often considered the related correlating parameters. Many
non-dimensional numbers can be made if the above experimental
parameters are combined with physical properties of gas or liquid,
and thus the distribution parameter can be expressed in many
ð5Þ non-dimensional forms. For instance, Ishii [26] assumed the distri-
bution parameter depends on the density ratio, qg/qf, and the Rey-
nolds number, Ref, defined by jfqfDh/lf. In consideration of various
void distributions such as wall void peak in bubbly flow systems
and the bubble size being one of key parameters to govern the void
distribution, Hibiki and Ishii [30] introduced into the correlation of
the distribution parameter the bubble Sauter mean diameter,
which can be predicted from a correlation for the non-dimension-
alized bubble Sauter mean diameter, the non-dimensional Laplace
length, and the Reynolds number. The relationships between the
distribution parameter and non-dimensional numbers were made
clear as well by the application of the ANN to the collected data-
base, in the way as shown in the pressure drop part. It is shown
456 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465

2.0 tion parameter at small non-dimensional Laplace constants (i.e. for

conventional channels) may be approximated to be 1.2 according to
-1.39/Lo Ishii [26]. Therefore, the following equation for the distribution
C0 =1.20+0.380 e
parameter could be obtained by using the least square method:
Distribution Parameter, C0 [-]

1.6
C 0 ¼ 1:20 þ 0:380 expð1:39=Lo Þ: ð9Þ

The applicable conditions of this correlation are as follows:

0.200 6 Dh 6 4.90 mm, Ref 6 2000, Reg 6 1000, air–water two-
1.2 phase flow, atmospheric pressure.
Data Source
Kariyasaki et al. 3. Results and discussion
Moriyama et al.
0.8
Mishima et al. 3.1. Pressure drop
Mishima and Hibiki
Triplett et al. 3.1.1. Collected database
Hazuku et al.
0.4 Available 13 data sets for two-phase friction pressure drop in
0.2 1 10 40 mini/micro-channels are tabulated in Table 1. These data sets can
Non-D. Laplace Constant, Lo [-] be classified into three groups, that is, adiabatic liquid–gas flow,
adiabatic liquid–vapor flow, and flow boiling groups. Adiabatic li-
Fig. 2. Distribution parameter as a function of non-dimensional Laplace constant. quid–gas flow involves two fluids, between which there are no
mass and heat transfers. It is the simplest two-phase flow, and
has been extensively investigated by many researchers. Adiabatic
that only using the non-dimensional Laplace constant, Lo*, defined liquid–vapor flow involves only one type of fluid with two states,
in Eq. (3), can predict the distribution parameter with the smallest i.e. liquid and vapor. Although the experimental condition is kept
mean deviation of 9.56%. Other non-dimensional numbers have the to be adiabatic, mass and heat transfers between two states may
secondary effects on the predictive accuracy of the BNP. occur due to variation of pressure drop during flow. The pressure
Therefore, as an extension of Mishima and Hibiki’s work in drop gradient in mini-channels is a little bit higher than in conven-
1996, this study attempts to modify the correlation of the distribu- tional channels. Taking this point into account, existing correla-
tion parameter proposed by Mishima and Hibiki [2]: tions may be unsuitable for mini-channels. Experimental studies
C 0 ¼ 1:2 þ 0:510 expð0:691Dh Þ; ð8Þ by Ungar and Cornwell [31], Zhang and Webb [19], and Cavallini
et al. [32] deal with adiabatic liquid–vapor flow. Tran [6] and Yu
where the unit of Dh is in mm. Since this correlation is dimensional, et al. [33] conducted experiments on flow boiling in mini-channels,
it is also desirable to develop a non-dimensional correlation. There- which is the most complex two-phase flow, involving thermal and
fore the non-dimensional Laplace constant is determined to replace hydraulic coupling. Our collected database includes all the three
the hydraulic diameter in the Mishima–Hibiki correlation. The rela- groups of two-phase flow, containing several working fluids. The
tionship between the average values of the non-dimensional La- covered hydraulic diameters of channel range from 0.07 to
place, Lo*, and those of the distribution parameter, C0 for each 6.25 mm. It should be noted that although the data bases of
data set is illustrated in Fig. 2. The asymptotic value of the distribu- Moriyama et al. [9] and Kawahara et al. [7] are deemed to be for

Table 1
Databases for two-phase frictional pressure drop in mini-channels.

Symbols Reference Adiabatic Geometry Diameter or Working fluids Flow Channel Data
or diabatic gap  width (mm) direction material number
s Moriyama et al. [9] Adiabatic Rectangular duct (0.007, 0.025, R113–N2 Horizontal Nickel + Pyrex 104
0.052, 0.098)  30 glass plates
4 Mishima et al. [1] Rectangular duct (1.07, 2.45, 5.00)  40 Water–air Vertical upward Acrylic resin 306
~ Mishima and Round tube 1.05, 2.05, 3.12, 4.08 Water–air Vertical upward Pyrex glass 299
Hibiki [2]
O Triplett et al. [4] Round tube 1.10, 1.45, 1.09, 1.49 Water–air Horizontal Pyrex, acrylic, 192
semi-triangular poly-carbonate
. Lee and Lee [23] Rectangular duct (0.4, 1.0, 2.0, 4.0)  20 Water–air Horizontal Acrylic 42
} Kawahara et al. [7] Round duct 0.10 Water–N2 Horizontal Fused silica 64
/ Chung and Round duct 0.0495, 0.0996, Water–N2 Horizontal Fused silica 0
Kawaji [24] 0.250, 0.526
} Liu et al. [35] Round tubes, 0.91, 2, 3.02, 0.99 (Water, ethanol, Vertical Pyrex glass 205
square ducts  0.99, 2.89  2.89 oil)–air,
g Ungar and Adiabatic Round tube 1.46, 1.78, 2.58, 3.15 Ammonia–vapor Horizontal Unknown 133
Cornwell [31]
Zhang and Webb [19] Round tube 2.13, 3.25, 6.20 R134a, R22, Horizontal Aluminum, copper 51
Round multi-port R404a–vapor
 Cavallini et al. [32] Rectangular 1.4 (R134a, R236ea, Horizontal Aluminum 38
multi-port R410A)–vapor
Tran [6] Diabatic Round tube 2.46 (R134a, R12)–vapor Horizontal Brass, stainless steel 440
f

Yu et al. [33] Round tube 2.98 Water–vapor Horizontal Stainless steel 327
Total (13 databases) Adiabatic, Round tube, 0.007–6.25 (Water, R12, R113, Vertical, Nickel, Pyrex, acrylic, 2201
diabatic Rectangular duct R22, R134a, R404a, horizontal poly-carbonate, silica,
ammonia)– aluminum, copper,
(air, N2, vapor) brass, stainless steel
 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465 457

6 drop, the homogeneous model was shown to be capable to

10
correlate experimental data within an acceptable margin of error,
5 TV TT although the observed flow patterns were much less homogeneous
10
Liquid Reynolds Number, Ref [-]

in mini/micro-channels [4,7]. Here an extensive evaluation of six

4
exiting correlations based on the homogeneous model was made.
10 The mean deviation is used as a measure of predictive accuracy.
For the convenience of comparison, the two-phase friction pres-
3
10 sure gradient predicted by the homogeneous model is also ex-
pressed as a product of the single-phase pressure gradient for the
2 liquid phase being considered to flow alone, and the two-phase
10
friction multiplier. Table 2 shows the comparison of six correla-
1 tions with the seven databases for adiabatic gas–liquid two-phase
10 flow. As reported by Kawahara et al. [7], the correlation of Dukler
0 et al. [36] presented reasonably good predictions for seven dat-
10 abases for adiabatic gas–liquid two-phase flow. Seventeen data
VV VT sets among 26 in all were predicted within the mean deviation
-1
10 -1 of 20%. For all data in Table 2, the correlation of Dukler et al. gave
0 1 2 3 4 5 6
10 10 10 10 10 10 10 10 the smallest total mean deviation of 18.0%, followed subsequently
Gas Reynolds Number, Reg [-] by the Beattie–Whalery correlation [37] with 22.9%. Table 3
showed the evaluation with databases for liquid–vapor two-phase
Fig. 3. Distribution of all adiabatic data in plot of Ref versus Reg. flow. As reported by Ungar and Cornwell [31], and Triplett et al. [4],
all of six correlations based on the homogeneous model give satis-
factory predictions for three databases. The best performance is
presented by the correlation of Ackers [38] with the smallest total
micro-channel according to the channel definitions of Kandlikar
mean deviation of 9.47%. Table 4 lists the comparison of the six
[34], they are included in this collected database since their corre-
correlations with the data for flow boiling. Here the total two-
sponding flow regimes are suitable for evaluation of the existing
phase frictional pressure drops predicted by the correlations were
models. Moreover, the flow patterns observed in the experiment
compared to the experimental values. The agreement is relatively
of Kawahara et al. [7] were liquid film flows along the channel wall
poor. Fig. 4(a) shows the behavior of the correlation of Dukler
with gas flowing in the channel core, and thus the separated flow
[36] with the data obtained by Mishima et al. [1] and Mishima
model is deemed to hold for prediction of their data. No data were
and Hibiki [2] for adiabatic air–water flow in rectangular and circu-
collected from Chung and Kawaji [24] for lack of enough parame-
lar channels, respectively. There exist some deviations of predic-
ters. It should be mentioned that Liu et al. [35] recently presented
tion at the low experimental multiplier. Further examination of
an extensive data base for two-phase friction pressure drop in ver-
Table 2 found that the Dukler correlation tends to behave poor
tical capillaries of circular and square cross-sections using air as
for channels with relatively large hydraulic diameter. Fig. 4(b)
the gas phase and water, ethanol, or an oil mixture as the liquid
demonstrated that the correlation works well for the data of Ungar
phase. A negative friction pressure drop was observed for the bulk
and Cornwell [31]. However, it systematically under-predicts the
of experimental data at very low liquid velocities, which means
data of Tran [6], as shown in Fig. 4(c). The similar performance
that the total pressure drop is less than the hydrostatic pressure
can be observed for the Beattie–Whalery correlation [37].
drop and local down flows of liquid film may occur. All negative
friction pressure drop data were excluded here. Data of Tran [6]
3.1.2.2. Existing correlations based on separated flow model. It was
using R113 as a test fluid were also excluded since these data were
reported that the Lockhart–Martinelli type correlation can repre-
not verified by us with their own correlation. Since some proper-
sent the experimental data of two-phase friction pressure drop
ties of fluid are not available in hand, a part of data by Zhang
reasonably well for many flow conditions, see Mishima and co-
and Webb [19] and Cavallini et al. [32] were not included in this
workers [1,2] as well as Zhao and Bi [22]. An extensive evaluation
database. All collected data are obtained from the open literature
of nine existing correlations is presented here. Among these corre-
and verified by the correlations developed by their authors.
lations, seven (correlations of Mishima and Hibiki [2], Tran [6], Lee
Fig. 3 shows the distribution of all adiabatic data in the plot of
and Lee [23], Yu et al. [33], Zhang and Webb [19], Qu and Mudawar
Ref versus Reg. Most of collected data fall into the flow conditions of
[39], Lee and Mudawar [40]) are developed for mini/micro-chan-
liquid and/or gas being laminar, and few data in the turbulent–tur-
nels. Since the Lockhart and Martinelli’s classical correlation [18]
bulent (TT) condition, if flow conditions are divided in the way of
is the foundation of most of recently developed correlations, and
Lockhart and Martinelli [18]. Therefore, low Reynolds number flow
that of Friedel [5] is one of the most accurate correlations for con-
conditions (VV, TV, and VT) are often encountered in mini-channel,
ventional channels [41], thus both correlations are included here as
and flow at such conditions may have features of laminar flow. This
well. The evaluation results of each correlation with databases for
is a new characteristic of two-phase flow in mini-channel. From
adiabatic liquid–gas flow, adiabatic liquid–vapor and flow boiling
the definition of Reynolds number it is evident that low Reynolds
are tabulated in Tables 5–7, respectively. Since the correlations of
number results from the limitation on flow rate due to high pres-
Yu et al., Qu and Mudawar, and Lee and Mudawar were developed
sure drop gradient in the experiment, and the inherent small
just for one or two flow regimes (laminar or turbulent flow of gas
dimension of mini-channel. However, most of existing correlations
or liquid), therefore only a part of data falling into their respective
were developed for turbulent flow, so they may be unsuitable in
applicable ranges were used to evaluate the behaviors of these cor-
principle for such flow conditions in mini-channel.
relations. From Table 5, correlations of Mishima–Hibiki, and Lee–
Lee generally predict the seven databases satisfactorily. Fourteen
3.1.2. Evaluation of correlations among 26 data sets are predicted within the mean deviation of
3.1.2.1. Existing correlations based on homogeneous model. Accord- 20% by Mishima–Hibiki correlation, 16 by that of Lee and Lee. In
ing to some previous studies by Ungar and Cornwell [31], Triplett all, Mishima–Hibiki correlation presents the smallest total mean
et al. [4], and Kawahara et al. [7] on two-phase frictional pressure deviation of 16.6%, followed subsequently by that of Lee and Lee
458 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465

Table 2
Evaluation of correlations of frictional pressure drop based on homogeneous model with data for liquid–gas flow.

Reference Diameter or gap  width Working Geometry Mean deviationa

(mm) fluids
McAdams Ackers Cicchitti Dukler Beattie–Whalery Lin et al.
[45] [38] [46] [36] [37] [47]
Moriyama et al. [9] 0.025  30 R113–N2 Rectangular duct 153 158  5.64 38.7 216
0.098  30 108 114  26.4 157
11.8
Mishima et al. [1] 1.07  40 Water–air Rectangular duct 13.3 15.4 21.7 10.5 8.98 17.7
2.45  40 14.9 15.3 18.3 13.5 16.7
11.7
5.00  40 15.7 15.6 15.6 24.2 15.5
15.6
Mishima and Hibiki 1.05 Water–air Round tube 14.9 29.1 63.0 12.5 7.50 39.7
[2] 2.05 8.25 10.1 19.7 11.4 8.33 14.0
3.12 13.5 14.7 20.7 16.8 14.5
11.0
4.08 20.2 42.4 28.7 21.6
19.8 20.8
Triplett et al. [4] 1.10 Water–air Round tube, round 18.9 25.9 56.7 16.6 11.8 28.2
1.45 tube 13.9 20.6 56.7 22.1 23.8
11.8
1.09 Semi-triangular 29.3 44.8 115 14.9 50.1
10.9
Lee and Lee [23] 0.4  20 Water–air Rectangular duct 106 130 167 21.5 68.2 148
4.0  20 17.6 18.2 38.9 22.7 17.8 19.4
Kawahara et al. [7] 0.10 Water–N2 Round duct 66.9 127  15.7 60.1 110
Liu et al. [35] 0.91 Water–air, Round tube, 15.4 15.0 14.9 22.2 24.3 14.7
2.00 Water–air, Round tube, 15.3 15.8 16.4 13.0 22.0 16.3
3.02 Water–air, Round tube, 25.9 26.9 28.6 32.3 21.2 27.9
2.89  2.89 Water–air, Square duct, 18.3 17.4 17.6 32.7 17.1 17.2
0.91 Ethanol–air, Round tube, 85.8 99.2 109 17.9 87.5 106
2.00 Ethanol–air, Round tube, 28.2 30.0 31.2 17.9 37.8 30.9
3.02 Ethanol–air, Round tube, 7.33 7.35 7.36 7.19 7.38 7.36
0.99  0.99 Ethanol–air, Square duct, 54.4 60.9 66.0 12.8 63.1 64.2
2.89  2.89 Ethanol–air, Square duct, 18.7 20.6 22.0 13.3 23.2 21.7
3.02 Oil–air, Round tube, 21.3 116 134 24.5 83.9 76.9
2.89  2.89 Oil–air, Square duct 11.0 62.8 69.9 12.2 56.7 47.5
Total 30.4 41.9 71.7 18.0 22.9 46.5

The symbol  denoting that the mean deviation is larger than 200%.
a P
Mean deviation defined as (1/N) |(/f,exp  /f,cal)//f,exp|  100%, an underlined value denoting the smallest of mean deviations by existing correlations for each data set.

with 20.2%. The underlined value in the Tables 2–7 denotes the
smallest mean deviation among those by existing correlations for
each data set. The smallest mean deviations predicted by Lee and
Lee’s correlation are the greatest in number. From Table 5, the cor-
relations of Qu–Mudawar and Lee–Mudawar predict the data well
within their applicable ranges. Table 6 lists the evaluation results
of nine existing correlations with three databases for liquid–vapor
flow. Lee and Mudawar’s correlation works well for their applica-
ble ranges. For all flow conditions, the Mishima–Hibiki correlation
behaves best in total. Table 7 shows the evaluation with data for
flow boiling. The classical Lockhart–Martinelli correlation has the
smallest total mean deviation of 35.1%. The second is the Mishi-
ma–Hibiki correlation. The correlations of Tran, and Zhang and
Webb work poorly for databases obtained by other authors. The
correlations proposed by Mishima and Hibiki, and Lee and Lee have
similar performance for data of liquid–gas flow. However the for-
mer correlation works a little bit better for both liquid–vapor flow
and flow boiling. Fig. 5 further illustrates the behavior of prediction
by Mishima–Hibiki correlation. However, Mishima–Hibiki correla-
tion over-predicted the data of Ungar and Cornwell by 30%, as
shown in Fig. 5(b). The similar performance of prediction by the
Lee–Lee correlation can be observed.
3.1.2.3. Newly developed correlation. This section shows the evalua-
tion of the newly developed correlation, Eq. (5). The results are tab-
ulated in Tables 5–7. The bold value in the tables denotes the

Table 3
Evaluation of correlations of frictional pressure drop based on homogeneous model with data for liquid–vapor flow.

Reference Diameter or Working fluids Geometry Mean deviation

gap  width (mm)
McAdams Ackers Cicchitti Dukler Beattie– Lin et al.
[45] [38] [46] [36] Whalery [37] [47]
Ungar and 1.46 Ammonia- Round tube 8.87 10.9 36.4 8.88 8.88 11.1
Cornwell [31] 1.78 Vapor 14.7 11.4 16.7 19.1 16.1 14.3
2.58 6.63 7.18 24.2 8.12 6.23 7.38
3.15 11.0 25.5 7.61 7.75 9.15
7.38
Zhang and Webb 2.13 R134a–vapor, Multi-port, round 13.8 4.95 2.31 15.6 9.50 11.4
[19] 3.25 R134a–vapor, tube, 18.5 9.56 19.8 14.1 16.5
8.29
6.20 R134a–vapor, Round tube, 15.9 7.12 18.1 11.7 13.4
5.84
3.25 R22–vapor Round tube, 17.2 10.1 19.0 14.2 15.4
8.11
Cavallini et al. [32] 1.4 R134a–vapor, Rectangular 18.3 9.08 6.08 20.4 14.3 15.9
R236ea–vapor multi-port 22.4 10.4 25.0 20.2 20.1
5.20
Total 12.6 9.47 18.4 14.3 11.4 12.4
 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465 459

Table 4
Evaluation of correlations of frictional pressure drop based on homogeneous model with data for flow boiling.

Reference Diameter or gap  width Working Geometry Mean deviationa

(mm) fluids
McAdams Ackers[38] Cicchitti Dukler Beattie–Whalery Lin et al.
[45] [46] [36] [37] [47]
Tran et al. 2.46 R12, Round 49.0 40.3 35.4 53.1 44.2 45.0
[6] 2.46 R134a tube 59.0 50.6 62.4 55.7 55.9
42.9
Yu et al. 2.98 Water Round 76.8 64.1 162 40.8 45.5 91.6
[33] tube
Total 64.8 54.5 92.5 51.5 49.3 69.2

a P
Mean deviation defined as (1/N) |(DpF,tp,exp  DpF,tp,cal)/DpF,tp,exp|  100%.

smallest mean deviation for each data set among those by all cor-
relations including the new one. From Table 5, 13 data sets among
26 in all were predicted within 20%. The total mean deviation for
(a) 102 all data of liquid–gas flow was 17.9%, a little bit higher than that
Dukler Correlation predicted by Mishima–Hibiki correlation. The table shows for all
[-]

flow conditions the newly developed correlation works satisfacto-

f,cal

1
rily for liquid–vapor two-phase flow with the smallest total mean
10 deviation of 21.7%, about 6% less than that of Mishima–Hibiki cor-
Predicted Multiplier, φ

relation. For flow boiling data, the newly developed correlation

works best among the 10 correlations, as shown in Table 7. Fig. 6
further illustrates its performance.
0
10
Data Source 3.1.3. Recommendation
+30%
Mishima et al. It should be noted that the application of the newly developed
-1 -30% Mishima and Hibiki correlation out of the verified range is not recommended. Since
10 -1 0 1 2 most of data in this collected database fall into the flow condition
10 10 10 10 of either liquid flow or gas flow being laminar, it can be expected
Experimental Multiplier, φ f,exp [-] that the newly developed correlation would be able to predict
2 the friction pressure drop in such flow conditions. For other flow
(b) 10 conditions such as both liquid and gas flow being turbulent, the
Dukler Correlation Reynolds number, Ref, may be an important non-dimensional num-
Predicted Multiplier, φf,cal [-]

ber to correlate the Chisholm parameter. Furthermore, although

the newly developed correlation could predict all the data for the
flows of three groups, i.e. adiabatic liquid–gas flow, adiabatic li-
1
quid–vapor flow and flow boiling, within an acceptable margin of
10 error, it was found from Fig. 1 that most of the averages of the Chis-
holm parameter, C, for liquid–gas flow are above the curve de-
picted by the newly developed equation, however, those for
+30% liquid–vapor flow are below the curve. For liquid–gas two-phase
Data Source flow, Eq. (5) may work better if the constant of 0.358 is replaced
Ungar and Cornwell with 0.674. For liquid–vapor flow, however, the constant of
0 -30%
10 0 1 2
0.142 would be better.
10 10 10 For adiabatic liquid–gas two-phase flow:
Experimental Multiplier, φ f,exp [-]
C ¼ 21½1  expð0:674=Lo Þ: ð10Þ
(c) 10 6
For adiabatic liquid–vapor two-phase flow:
Dukler Correlation
Predicted Pressure Drop, Δ pF,cal [-]

5 C ¼ 21½1  expð0:142=Lo Þ: ð11Þ

10
On this point, future experimental investigation is recommended to
10
4 accumulate more data sets for clarifying the reasons why the differ-
ence occurs between adiabatic liquid–gas flow and adiabatic liquid–
3
vapor flow.
10
3.2. Void fraction
2
10 +30% Data Source
Tran 3.2.1. Collected database
1 -30% Six available datasets in the literature for void fraction of two-
10 1
10 10
2
10
3 4
10
5
10 10
6 phase flow in mini-channels were collected in this study and tab-
Experimental Pressure Drop, Δ pF,exp [-] ulated in Table 9. The hydraulic diameters range from 0.025 mm to
4.90 mm, and channel geometry are circular tube or rectangular
Fig. 4. Evaluation of Dukler correlation (1964) with data for (a) liquid–gas flow, (b) duct. The working fluids contain water and R113 as liquid phases,
liquid–vapor flow, and (c) flow boiling. with air and nitrogen as gas phases. The flow orientation is vertical
 460
Table 5
Evaluation of correlations of frictional pressure drop based on separated flow model with data for liquid–gas flow.

Reference Diameter or Working Geometry Mean deviationa

gap  width fluids
Lockhart– Friedel [5] Mishima– Tran [6] Lee–Lee [23] Yu et al. [33] Zhang and Qu and Lee and This
(mm)
Martinelli [18] Hibiki [2] Webb [19] Mudawar [39] Mudawar [40] study
Moriyama et al. [9] 0.025  30 R113–N2 Rectangular duct 65.9  3.31  5.62 /  3.25 3.66 3.38
0.098  30 47.0  8.77  11.6 /  9.59 9.17

W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465
7.89
Mishima et al. [1] 1.07  40 Water–air Rectangular duct 9.70 77.5 11.4 156 19.0 51.0  46.9 15.6 7.73
2.45  40 13.4 90.1 11.4 108 18.1 /  27.3 42.1 10.3
5.00  40 18.0 65.7 6.95 29.2 11.4 /  / / 14.5
Mishima and Hibiki [2] 1.05 Water–air Round tube 15.9 183 14.4  36.4 50.7  16.4 11.3 27.1
2.05 13.1 108 8.64 192 22.3 57.7  12.7 21.5 22.4
3.12 17.7 98.0 8.24 145 21.0 64.7  4.82 28.2 23.5
4.08 25.6 147 123 30.9 54.5  18.5 26.8 20.9
16.8
Triplett et al. [4] 1.10 Water–air Round tube, round tube 13.9 70.9 14.3 81.5 22.2 35.1  23.9 20.1 16.4
1.45 Semi-triangular 16.3 69.9 15.2 71.8 28.9 43.3  16.8 23.7 21.5
1.09 21.8 133 22.4 127 17.2 38.0  14.1 13.3 14.7
Lee and Lee [23] 0.4  20 Water–air Rectangular duct 59.3  55.6  12.5 /  110 71.9 27.9
4.0  20 19.3 48.5 23.7 31.8 56.5  / 22.9 25.8
15.6
Kawahara et al. [7] 0.10 Water–N2 Round duct 75.9  17.4  10.7 34.1  38.3 47.2 11.7
Liu et al. [35] 0.91 Water–air, Round tube, 20.3 52.4 20.6 37.5 22.1 /  25.5 23.4 17.9
2.00 Water–air, Round tube, 15.0 49.5 27.0 31.6 12.5 /  25.0 11.4 14.6
3.02 Water–air, Round tube, 23.8 28.4 19.5 12.9 33.6 /  21.1 28.7 22.8
2.89  2.89 Water–air, Square duct, 21.3 18.9 11.6 11.8 32.2 /  17.0 25.7 18.7
0.91 Ethanol– Round tube, 60.8 199 63.0 67.3 18.0 /  25.8 30.0 50.6
air,
2.00 Ethanol– Round tube, 26.0 68.6 41.2 28.5 17.8 / 139 36.0 31.5 32.6
air,
3.02 Ethanol– Round tube, 6.36 11.0 12.7 6.17 7.11 / 32.6 11.5 8.72 8.77
air,
0.99  0.99 Ethanol– Square duct, 41.2 132 44.7 55.3 12.9 /  30.7 26.3 35.5
air,
2.89  2.89 Ethanol– Square duct, 15.2 51.2 31.5 15.7 13.1 / 114 21.5 15.3 23.7
air,
3.02 Oil–air, Round tube, 35.7 197 54.8 97.1 24.5 /  30.4 24.5 25.3
2.89  2.89 Oil–air, Square duct 19.1 126 31.3 45.2 12.2 / 170 14.7 12.2 12.5
Total 25.7 138 16.6 181 20.2 49.3  24.3 23.9 17.9

The symbol  denoting that the mean deviation is larger than 200%.
The symbol / meaning that data are out of applicable parametric ranges of a correlation.
a P P
Mean deviation is defined as (1/N) |(/fo,exp  /fo,cal)//fo,exp|  100% for correlations of Friedel [5], Tran [6] and Zhang and Webb [19], however, it is defined as (1/N) |(/f,exp  /f,cal)//f,exp|  100% for other correlations.
 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465 461

Table 6
Evaluation of correlations of frictional pressure drop based on separated flow model with data for two-phase liquid–vapor flow.

Reference Diameter or Working Geometry Mean deviation

gap  width fluids
Lockhart– Friedel Mishima– Tran Lee– Yu Zhang and Qu and Lee and This
(mm)
Martinelli [5] Hibiki [2] [6] Lee et al. Webb [19] Mudawar Mudawar study
[18] [23] [33] [39] [40]
Ungar and 1.46 Ammonia- Round tube 42.0 102 30.3 134 19.0 36.4 111 6.33 13.3 15.1
Cornwell 1.78 Vapor 15.5 30.2 12.3 82.6 5.82 42.5 65.4 23.7 13.0 8.48
[31] 2.58 36.0 101 38.6 96.9 16.0 32.9 102 16.9 6.44 21.7
3.15 42.9 91.4 49.5 96.4 25.1 32.6 94.9 36.0 29.8
12.1
Zhang and 2.13 R134a–vapor, Multi-port, 52.6 14.8 20.5 45.4 73.5 29.0 4.25 / 45.9 35.8
Webb 3.25 R134a–vapor, round tube, 45.7 8.96 27.6 35.1 77.8 / / / 38.7
5.23
[19] 6.20 R134a–vapor, Round tube, 47.8 11.1 42.0 29.4 94.0 / / / 47.5
3.95
3.25 R22–vapor Round tube, 51.1 10.5 30.5 34.2 83.3 / / / 40.4
5.22
Cavallini 1.4 R134a–vapor, Rectangular 39.7 7.42 4.78 47.7 51.6 38.2 5.71 / 37.8 11.2
et al. [32] R236ea–vapor multi-port 6.81 14.2 12.2 46.3 7.96 42.4 27.5 / 15.8 10.1
Total 37.5 58.8 27.6 79.3 34.3 36.1 60.9 19.1 11.7 21.7

Table 7
Evaluation of correlations of frictional pressure drop based on separated flow model with data for flow boiling.

Reference Diameter or Working Geometry Mean deviation

gap  width fluids
Lockhart– Friedel Mishima– Tran Lee– Yu Zhang and Qu and Lee and This
(mm)
Martinelli [5] Hibiki [2] [6] Lee et al. Webb [19] Mudawar Mudawar study
[18] [23] [33] [39] [40]
Tran [6] 2.46 R12, Round 47.8 36.1 38.0 25.9 40.3 56.9 31.1 / 56.9 46.5
2.46 R134a tube 23.1 35.1 29.0 25.2 65.1 33.5 / 119 28.1
13.2
Yu et al. 2.98 Water Round 39.0 181 48.3  84.6 28.8  / 51.9 31.4
[33] tube
Total 35.1 97.4 39.2  55.0 40.6  / 92.9 32.8

Table 8
Averages of parameter C for each datasets in mini-channels.

Reference Adiabatic or diabatic Geometry Diameter or gap  width (mm) Working Flow Direction Lo* Parameter C
fluids
Moriyama et al. [9] Adiabatic Rectangular duct 0.007  30 R113–N2 Horizontal / /
0.025  30 21.50 0.22
0.052  30 / /
0.098  30 5.50 0.69
Mishima et al. [1] Adiabatic Rectangular duct 1.07  40 Water–air Vertical 1.30 6.66
2.45  40 upward 0.58 11.28
5.00  40 0.30 22.69
Mishima and Hibiki [2] Adiabatic Round tube 1.05 Water–air Vertical 2.61 7.57
2.05 upward 1.34 12.38
3.12 0.88 15.07
4.08 0.67 13.16
Triplett et al. [4] Adiabatic Round tube 1.10 Water–air Horizontal 2.49 4.95
Semi- 1.45 1.89 6.79
triangular 1.09 2.51 2.87
1.49 / /
Lee and Lee [23] Adiabatic Rectangular duct 0.4  20 Water–air Horizontal 3.49 2.03
1.0  20 1.43 7.56
2.0  20 0.75 11.59
4.0  20 0.41 13.04
Kawahara et al. [7] Adiabatic Round duct 0.10 Water–N2 Horizontal 27.38 0.24
Chung and Kawaji [24] Adiabatic Round duct 0.0495 Water–N2 Horizontal 54.89 0.15
0.0996 27.28 0.22
0.250 10.87 1.74
0.526 5.17 3.18
Ungar and Cornwell [31] Adiabatic Round tube 1.46 Ammonia- Horizontal 1.42 2.85
1.78 Vapor 1.16 6.74
2.58 0.80 3.49
3.15 0.66 2.67
Zhang and Webb [19] Adiabatic Round tube, 2.13 R134a–vapor Horizontal 0.28 6.89
Round multi-port 3.25 R134a–vapor 0.23 10.30
6.20 R134a–vapor 0.12 8.71
3.25 R22–vapor 0.26 9.01
/ R404a–vapor / /
Cavallini et al. [32] Adiabatic Rectangular multi- 1.4 R134a–vapor Horizontal 0.54 8.92
port R236ea–vapor 0.62 10.45
R410A–vapor / /
462 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465

(a) 10 2 (a) 10 2
Newly Developed Correlation
Mishima-Hibiki Correlation
Predicted Multiplier, φf,cal [-]

Predicted Multiplier, φ f,cal [-]

1 1
10 10

0 0
10 10
Data Source Data Source
+30% Mishima et al. +30%
Mishima et al.
Mishima and Hibiki -30% Mishima and Hibiki
-1 -30% -1
10 -1 0 1 2
10 -1 0 1 2
10 10 10 10 10 10 10 10
Experimental Multiplier, φf,exp [-] Experimental Multiplier, φf,exp [-]

(b) 10
2
(b) 10 2
Mishima-Hibiki Correlation Newly Developed Correlation
[-]

Predicted Multiplier, φ f,cal [-]

f,cal
Predicted Multiplier, φ

1 1
10 10

+30% Data Source +30% Data Source

Ungar and Cornwell Ungar and Cornwell
0 -30% 0
-30%
10 0 1 2 10 0 1 2
10 10 10 10 10 10
Experimental Multiplier, φ f,exp [-] Experimental Multiplier, φf,exp [-]

(c) 10 6 (c) 10
6
Predicted Pressure Drop, Δ pF,cal [-]

Mishima-Hibiki Correlation Newly Developed Correlation

Predicted Pressure Drop, Δ pF,cal [-]

5
10 10
5

4
10 10
4

3
10 10
3

2
10 +30% Data Source 10
2 +30% Data Source
Tran Tran
1
-30% -30%
10 1 2 3 4 5 6 10 1
1
10 10 10 10 10 10 10 10
2
10
10
3
10 10
4 5 6

Experimental Pressure Drop, Δ pF,exp [-]

Experimental Pressure Drop, Δ pF,exp [-]
Fig. 5. Evaluation of Mishima–Hibiki correlation (1996) with data for (a) liquid–gas
Fig. 6. Evaluation of newly developed correlation with data for (a) liquid–gas flow,
flow, (b) liquid–vapor flow, and (c) flow boiling.
(b) liquid–vapor flow, and (c) flow boiling.

upward or horizontal. All data were obtained under atmospheric

pressure. Hibiki–Ishii [28]. The corresponding relations between the symbols
Fig. 7 shows the distribution of collected data in the plot of Ref utilized in Figs. 7–10 and the datasets are listed in Table 9. The
versus Reg. It is evident that most of data fall into the flow condi- evaluation results are illustrated in Fig. 8. It indicates that the
tions of liquid and/or gas being laminar, and few data into the tur- Mishima–Hibiki correlation presents the best predictive accura-
bulent–turbulent condition, similar to that shown in Fig. 3 for cies, with the smallest mean deviation of 12.9%. Most of predic-
pressure drop. It can be concluded that low Reynolds number flow tions are well within the error band of ±30%. The second is 26.6%,
conditions are often encountered in mini-channel, and flow may given by the Hibiki–Ishii correlation. As pointed out by Hibiki
have features of laminar flow under such conditions. and Ishii [28], this correlation is applicable only to bubbly flow in
a mini-channel. It under-predicts the data at high void fraction
3.2.2. Evaluation of correlations as shown in Fig. 8(d). The correlation of Kariyasaki et al. over-pre-
3.2.2.1. Existing correlations. In what follows, an evaluation is pre- dicts the data about 30% for the data of low and intermediate void
sented of four existing correlations for mini-channels: correlations fractions. This correlation tends to predict the data well within the
of Kariyasaki et al. [8], Moriyama et al. [9], Mishima–Hibiki [2], and acceptable margins of errors at high void fractions over 60%. The
 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465 463

Table 9
Collected databases for void fraction of two-phase flow in mini-channels.

Symbols Reference Geometry Diameter or Working fluids Channel material Flow orientation
gap  width (mm)
s Kariyasaki et al. [8] Circular tube 1.00, Water–air Acrylic Vertical upward
2.40,
4.90
4 Moriyama et al. [9]) Rectangular duct 0.025  30, R113–N2 Nickel + Pyrex Horizontal
0.052  30, glass plates
0.096  30
. Mishima et al. [1] Rectangular duct 1.07  40, Water–air Aluminum Vertical upward
2.45  40
} Mishima and Hibiki [2] Circular tube 1.09, Water–air Aluminum Vertical upward
2.15,
3.08,
3.90
/ Triplett et al. [4] Circular tube 1.10 Water–air Pyrex Horizontal
} Hazuku et al. [11] Circular tube 1.02 Water–air Acrylic Vertical upward
Total (6 databases) Circular tube, 0.025–4.90 Water/R113- Acrylic, nickel, Vertical, horizontal
Rectangular duct Air/N2 Pyrex, aluminum

5 3.2.3. Recommendation
10
It should be mentioned that Serizawa et al. [42] calculated the
TV TT cross-sectional averaged void fraction for air–water two-phase
Liquid Reynolds Number, Re f [-]

4
10
3 flow and slug flow, respectively. Their data were correlated with
10
2
10
1 526 lm) as well as micro-channels (with diameters below
10
0 mand correlation was observed. The reasons for the contradictory
10
VV
-1
10 -1 0 1 2 3
10 10 10 10 10
Gas Reynolds Number, Re g [-]
Fig. 7. Distribution of collected data in plot of Ref versus Reg.
correlation of Moriyama et al. over-predicts the data about 40% for
low void fractions less than 30%. At high void fractions, this corre-
lation fails to correlate the data.
3.2.2.2. Newly developed correlation. Fig. 9 illustrates the total
behavior of the newly developed correlation, Eq. (9). The mean
deviation is 12.7%, slightly smaller than that of the Mishima–Hibiki
correlation. From the figure, most of data are well predicted within
the error band of ±30%. Fig. 10 shows the comparison of the newly
developed correlation and the existing correlations with the data
obtained by Hazuku et al. [11]. The values of gas velocity are plot-
ted versus those of mixture volumetric flux. The figure indicates
that the newly developed one works as well as the Mishima–Hibiki
correlation. The Hibiki–Ishii correlation [28] predicts the data sat-
isfactorily at the low values of mixture volumetric flux, however,
under-predicts the data at the high values. The predictions of cor-
relations proposed by Kariyasaki et al. and Moriyama et al. largely
deviate from the data. It should be noted that since Kariyasaki et al.
uses several equations to correlate the void fraction in terms of the
gas volumetric quality, b (= jg/(jf + jg)), instead of the drift-flux
model, and the Hibiki–Ishii correlation also consists of several
equations, thus, their predictions are not on straight lines.
flow in a 20 lm i.d. silica tube from high-speed video pictures by
assuming symmetrical shape of bubbles and gas slugs for bubbly
the Armand correlation [43]. In contrast, Kawahara and co-workers
[7,44] measured the void fractions by analyzing the recorded
images of the gas–liquid interface in the observation windows of
several mini-channels (with diameters ranging from 250 lm to
100 lm), and reported that for micro-channels a strong deviation
of the void fraction–volumetric quality relationship from the Ar-
conclusions obtained by Serizawa et al. and Kawahara et al. may be
VT due to the surface conditions of channel wall and the design of the
4 5
mixing chamber. It was reported that the two-phase flow struc-
10 10 tures in micro-channels (with hydraulic diameter of 20–100 lm)
is more seriously affected by the wettability between the tube
and the fluids, and a formation of dry area between gas slug and
the tube wall can be observed in the experiment when the flow
is very low [42]. The unique void fraction–volumetric quality rela-
tionship observed by Kawahara and co-workers may arise from the
design of the mixing chamber, which may be prone to produce
large bubbles resulting in the flow patterns of liquid film flowing
along the channel wall with gas flow in the core, and difficult to
yield fine bubbles for bubbly flow pattern. In this study it was as-
sumed that the channel wall of mini-channels would be all wetted
and no dry area would exist, and consequently the existing theo-
ries might be extendedly applied. For the above reasons, the col-
lected databases in this study do not contain the data sets of
Serizawa et al. and Kawahara and co-workers, and are tailored
for mini-channel according to the channel definitions of Kandlikar
[34], instead of micro-channel.
As mentioned in the friction pressure drop part, since most of
data in this collected data base fall into the flow condition of either
liquid or gas flows being laminar, it can be expected that the newly
developed correlation would be able to predict the void fraction in
this flow condition for mini-channels. For other flow conditions
such as both liquid and gas flow being turbulent, the Reynolds
number, Ref, may be an important non-dimensional number to cor-
relate the distribution parameter. In addition, the density ratio, qg/
qf may be also a significant non-dimensional number to correlate
the distribution parameter since the distribution parameter should
become unity as the density ratio approaches unity [26]. However,
since existing available data sets are for air–water two-phase flow
464 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465

(a) 100 (b) 100

Predicted Void Fraction, αcal [%]

Predicted Void Fraction, αcal [%]

Mean Dev.: Mean Dev.:
80 27.1% 80 63.8%

60 60

40 40
+30%
+30%
20 Correl. of 20
Correl. of
Kariyasaki et al. Moriyama et al.
-30% -30%
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Exp. Void Fraction, αexp [%] Exp. Void Fraction, αexp [%]

(c) 100 (d) 100

Predicted Void Fraction, αcal [%]

Correl. of Mishima-Hibiki Correl. of Hibiki-Ishii

Predicted Void Fraction, αcal [%]

80 80

60 60

40 40
+30%

20 20 +30%
Mean Dev.: 12.9% Mean Dev.: 26.6%
-30% -30%
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Exp. Void Fraction, αexp [%] Exp. Void Fraction, αexp [%]

Fig. 8. Evaluation of existing correlations with collected data.

100 10
Newly Developed Correlation Prediction of Correlations
Present Study
Predicted Void Fraction, α cal [%]

80 8 Kariyasaki et al.
Moriyama et al.
Gas Velocity, vg [m/s]

Mishima-Hibiki
60 6
Hibiki-Ishii

40 4
+30%

20 D =1.02 mm
2 h
Mean Dev.: 12.7 %
Data of Hazuku et al.
-30%
0
0 20 40 60 80 100 0
0 1 2 3 4 5 6 7
Exp. Void Fraction, α exp [%]
Mixture Volumetric Flux, j [m/s]
T
Fig. 9. Evaluation of newly developed correlation with collected data.
Fig. 10. Evaluation of correlations with data obtained by Hazuku et al. [11].

under atmospheric pressure, the effect of the density ratio on the 4. Conclusions
distribution parameter cannot be reflected in this newly developed
correlation. Therefore, for other flow conditions, pressures, and flu- Accurate prediction of two-phase frictional pressure drop and
ids, further study is needed. Moreover, the asymptotic value of the void fraction in mini-channels is essential to have a good under-
distribution parameter was assumed to be 1.2 for simplicity in the standing of two-phase flow as well as the successful modeling of
newly developed correlation. This may be valid for circular chan- two-phase flow in such channels. This study gave an extensive
nel. For rectangular channels, however, its value may be different. evaluation of existing correlations with a collected database cover-
 W. Zhang et al. / International Journal of Heat and Mass Transfer 53 (2010) 453–465
ing a wide range of running parameters, and proposed alternative
correlations for two-phase friction pressure drop and void fraction
in mini-channels. The detailed conclusions could be drawn as
follows:
(1) An extensive evaluation shows that the Dukler correlation
based on the homogeneous model works satisfactorily for
data sets of adiabatic liquid–gas flow as well as liquid–vapor
flow, however, systematically under-predicts the data of
flow boiling. All the tested correlations based on the homo-
geneous model predict well the data for liquid–vapor flow.
(2) The correlations proposed by Mishima and Hibiki, and Lee
and Lee, based on the separated flow model generally pre-
dict all of the collected data within an acceptable margin
of error. The good behaviors of both correlations for the pre-
diction of liquid–gas two-phase frictional pressure drop are
confirmed.
(3) By applying the ANN, the main non-dimensional number to
correlate the two-phase friction multiplier is picked out. A
correlation of the Chisholm parameter C for mini-channel
is newly developed using the non-dimensional Laplace
constant.
(4) An extensive evaluation of the existing void fraction correla-
tions indicates that the Mishima–Hibiki correlation also pre-
sents the best performance.
(5) By applying the ANN, the non-dimensional Laplace constant
number is found to successfully correlate the distribution
parameter. An alternative correlation of void fraction for
mini-channel is proposed using this non-dimensional
Laplace constant.
References
[1] K. Mishima, T. Hibiki, H. Nishihara, Some characteristics of gas–liquid flow in
narrow rectangular ducts, Int. J. Multiphase Flow 19 (1993) 115–124.
[2] K. Mishima, T. Hibiki, Some characteristics of air–water two-phase flows in
small diameter tubes, Int. J. Multiphase Flow 22 (1996) 703–712.
[3] D. Chisholm, A theoretical basis for the Lockhart–Martinelli correlation for
two-phase flow, Int. J. Heat Mass Transfer 10 (1967) 1767–1778.
[4] K.A. Triplett, S.M. Ghiaasiaan, S.I. Abdel-Khalik, A. LeMouel, B.N. McCord, Gas–
liquid two-phase flow in microchannels. Part II: void fraction and pressure
drop, Int. J. Multiphase Flow 25 (1999) 395–410.
[5] L. Friedel, Improved friction pressure drop correlations for horizontal and
vertical two-phase pipe flow, in: Proceedings of the European Two-Phase Flow
Group Meeting, Ispra, Italy, 1979.
[6] T.N. Tran, Pressure drop and heat transfer study of two-phase flow in small
channels, Ph.D. Dissertation, Texas Tech University, Lubbock, TX, 1998.
[7] A. Kawahara, P.M.-Y. Chung, M. Kawaji, Investigation of two-phase flow
pattern, void fraction and pressure drop in a microchannel, Int. J. Multiphase
Flow 28 (2002) 1411–1435.
[8] A. Kariyasaki, T. Fukano, A. Ousaka, M. Kagawa, Isothermal air–water two-
phase up- and downward flows in a vertical capillary tube (first report, flow
pattern and void fraction), Trans. JSME (Ser. B) 58 (1992) 2684–2690 (in
Japanese).
[9] K. Moriyama, A. Inoue, H. Ohira, The thermohydraulic characteristics of two-
phase flow in extremely narrow channels (the frictional pressure drop and
void fraction of adiabatic two-component two-phase flow), Trans. JSME (Ser. B)
58 (1992) 401–407.
[10] Z.Y. Bao, M.G. Bosnick, B.S. Haynes, Estimation of void fraction and pressure
drop for two-phase flow in fine passages, Trans. Inst. Chem. Eng. 72 (1994)
625–632.
[11] T. Hazuku, N. Tamura, N. Fukamachi, T. Takamasa, T. Hibiki, M. Ishii, Axial
development of vertical upward bubbly flow in a mini-pipe, in: Proceedings of
the 2005 ASME Summer Heat Transfer Conference, San Francisco, California,
USA, 2005.
[12] W. Zhang, T. Hibiki, K. Mishima, Correlation for flow boiling heat transfer in
mini-channels, Int. J. Heat Mass Transfer 47 (2004) 5749–5763.
[13] W. Zhang, T. Hibiki, K. Mishima, Y. Mi, Correlation of critical heat flux for flow
boiling of water in mini-channels, Int. J. Heat Mass Transfer 49 (2006) 1058–
1072.
[14] Y.H. Pao, Adaptive Pattern Recognition and Neural Networks, Addison-Wesley,
Reading, MA, 1989.
[15] P.D. Wasserman, Neural Computing: Theory and Practice, Van Nostrand-
Reinhold, New York, 1989.
465
[16] G.H. Su, K. Fukuda, D. Jia, K. Morita, Application of an artificial neural network
in reactor thermo-hydraulic problem: prediction of critical heat flux, J. Nucl.
Sci. Technol. 39 (2002) 564–571.
[17] Y. Mi, M. Ishii, L.H. Tsoukalas, Flow regime identification methodology with
neural networks and two-phase flow models, Nucl. Eng. Des. 204 (2001) 87–
100.
[18] R.W. Lockhart, R.C. Martinelli, Proposed correlation of data for isothermal two-
phase two-component flow in pipes, Chem. Eng. Prog. 45 (1949) 39–48.
[19] M. Zhang, R.L. Webb, Correlation of two-phase friction for refrigerants in
small-diameter tubes, Exp. Therm. Fluid Sci. 25 (2001) 131–139.
[20] G.F. Hewitt, N. Hall-Taylor, Annular Two-phase Flow, Pergamon Press, New
York, 1970.
[21] D. Chisholm, A.D.K. Laird, Two-phase flow in rough tubes, Trans. ASME 80
(1958) 276–286.
[22] T.S. Zhao, Q.C. Bi, Pressure drop characteristics of gas–liquid two-phase flow in
vertical miniature triangular channels, Int. J. Heat Mass Transfer 44 (2001)
2523–2534.
[23] H.J. Lee, S.Y. Lee, Pressure drop correlations for two-phase flow within
horizontal rectangular channels with small heights, Int. J. Multiphase Flow 27
(2001) 783–796.
[24] P.M.-Y. Chung, M. Kawaji, The effect of channel diameter on adiabatic two-
phase flow characteristics in microchannels, Int. J. Multiphase Flow 30 (2004)
735–761.
[25] Y. Sadatomi, T. Sato, S. Saruwatari, Two-phase flow in vertical noncircular
channels, Int. J. Multiphase Flow 8 (1982) 641–655.
[26] M. Ishii, One-dimensional drift-flux model and constitutive equations for
relative motion between phases in various two-phase flow regimes. ANL
Report ANL-77-47, 1977.
[27] T. Takamasa, T. Hazuku, N. Fukamachi, N. Tamura, T. Hibiki, M. Ishii,
Experimental study on interfacial area transport of bubbly flow in mini-
channels, in: Proceedings of the Fifth International Conference on Multiphase
Flow, Yokohama, Japan, Paper No. 490, 2004.
[28] T. Hibiki, M. Ishii, One-dimensional drift-flux model for various flow
conditions, in: Proceedings of the NURETH-11, Avignon, France, 2005, Paper
No. 014.
[29] K.A. Triplett, S.M. Ghiaasiaan, S.I. Abdel-Khalik, A. LeMouel, B.N. McCord, Gas–
liquid two-phase flow in microchannels. Part I: two-phase flow patterns, Int. J.
Multiphase Flow 25 (1999) 377–394.
[30] T. Hibiki, M. Ishii, Distribution parameter and drift velocity of drift-flux model
in bubbly flow, Int. J. Heat Mass Transfer 45 (2002) 707–721.
[31] E.K. Ungar, J.D. Cornwell, Two-phase pressure drop of ammonia in small
diameter horizontal tubes, in: Proceedings of the AIAA 17th Aerospace Ground
Testing Conference, Nashville, TN, 1992.
[32] A. Cavallini, D.D. Col, L. Doretti, M. Matkovic, L. Rossetto, C. Zilio, Two-phase
frictional pressure gradient of R236ea, R134a and R410A inside multi-port
mini-channels, Exp. Therm. Fluid Sci. 29 (2005) 861–870.
[33] W. Yu, D.M. France, M.W. Wambsganss, J.R. Hull, Two-phase pressure drop,
boiling heat transfer, and critical heat flux to water in a small-diameter
horizontal tube, Int. J. Multiphase Flow 28 (2002) 927–941.
[34] S.G. Kandlikar, Fundamental issues related to flow boiling in minichannels and
microchannels, Exp. Therm. Fluid Sci. 26 (2002) 389–407.
[35] H. Liu, C.O. Vandu, R. Krishna, Hydrodynamics of Taylor flow in vertical
capillaries: flow regimes, bubble rise velocity, liquid slug length, and pressure
drop, Ind. Eng. Chem. Res. 44 (2005) 4884–4897.
[36] A.E. Dukler, M. Wicks III, R.G. Cleveland, Pressure drop and hold-up in two-
phase flow, AIChE J. 10 (1964) 38–51.
[37] D.R.H. Beattie, P.B. Whalley, A simple two-phase flow frictional pressure drop
calculation method, Int. J. Multiphase Flow 8 (1982) 83–87.
[38] W.W. Akers, H.A. Deans, O.K. Crosser, Condensation heat transfer within
horizontal tubes, Chem. Eng. Prog. Symp. Ser. 55 (1959) 171–176.
[39] W. Qu, I. Mudawar, Measurement and prediction of pressure drop in two-
phase micro-channel heat sinks, Int. J. Heat Mass Transfer 46 (2003) 2737–
2753.
[40] J. Lee, I. Mudawar, Two-phase flow in high-heat-flux micro-channel heat sink
for refrigeration cooling applications. Part I: pressure drop characteristics, Int.
J. Heat Mass Transfer 48 (2005) 928–940.
[41] J.G. Collier, J.R. Thome, Convective Boiling and Condensation, third ed., Oxford
University Press, Oxford, 1994.
[42] A. Serizawa, Z. Feng, Z. Kawara, Two-phase flow in microchannels, Exp. Therm.
Fluid Sci. 26 (2002) 703–714.
[43] A.A. Armand, The resistance during the movement of a two-phase system in
horizontal pipes, Izv. Vses. Teplotekh. Inst. 1 (1946) 16–23 (AERE-Lib/Trans
828).
[44] A. Kawahara, M. Sadatomi, K. Okayama, M. Kawaji, P.M.-Y. Chung, Effects
of channel diameter and liquid properties on void fraction in adiabatic
two-phase flow through microchannels, Heat Transfer Eng. 26 (2005)
13–19.
[45] W.H. McAdams, Heat Transmission, second ed., McGraw-Hill, New York,
1942.
[46] A. Cicchitti, C. Lombardi, M. Silvestri, G. Solddaini, R. Zavalluilli, Two-phase
cooling experiments – pressure drop, heat transfer and burnout measurement,
Energ. Nucl. 7 (1960) 407–425.
[47] S. Lin, C.C.K. Kwok, R.Y. Li, Z.H. Chen, Z.Y. Chen, Local frictional pressure drop
during vaporization for R-12 through capillary tubes, Int. J. Multiphase Flow 17
(1991) 95–102.