You are on page 1of 13

Prediction of two-phase pressure gradients of

refrigerants in horizontal tubes


M.B. Ould Didi, N. Kattan, J.R. Thome*
Laboratory of Heat and Mass Transfer (LTCM), Department of Mechanical Engineering,
Swiss Federal Institute of Technology Lausanne (EPFL), CH-1015 Lausanne, Switzerland
Received 19 March 2001; received in revised form 17 September 2001; accepted 26 October 2001
Abstract
Two-phase pressure drop data were obtained for evaporation in two horizontal test sections of 10.92 and 12.00 mm
diameter for ve refrigerants (R-134a, R-123, R-402A, R-404A and R-502) over mass velocities from 100 to 500 kg/m
2
s and vapor qualities from 0.04 to 1.0. These data have then been compared against seven two-phase frictional pressure
drop prediction methods. Overall, the method by Mu ller-Steinhagen and Heck (Mu ller-Steinhagen H, Heck K. A
simple friction pressure drop correlation for two-phase ow in pipes. Chem. Eng. Process 1986;20:297308) and that by
Gro nnerud (Gro nnerud R. Investigation of liquid hold-up, ow-resistance and heat transfer in circulation type eva-
porators, part IV: two-phase ow resistance in boiling refrigerants. Annexe 1972-1, Bull. de lInst. du Froid, 1979) were
found to provide the most accurate predictions while the widely quoted method of Friedel (Friedel L. Improved friction
drop correlations for horizontal and vertical two-phase pipe ow. European Two-phase Flow Group Meeting, paper E2;
June 1979; Ispra, Italy) gave the third best results. The data were also classied by two-phase ow pattern using the
Kattan-Thome-Favrat (Kattan N, Thome JR, Favrat D. Flow boiling in horizontal tubes. Part 1: development of a
diabatic two-phase ow pattern map. J. Heat Transfer 1998;120:1407; Kattan N, Thome JR, Favrat D. Flow boiling in
horizontal tubes. Part 2; new heat transfer data for ve refrigerants. J Heat Transfer 1998;120:14855; Kattan N, Thome
JR, Favrat D. Flow boiling in horizontal tubes. Part 3: development of a new heat transfer model based on ow pat-
terns. J. Heat Transfer 1998;120:15665) ow pattern map. The best available method for annular ow was that of
Mu ller-Steinhagen and Heck. For intermittent ow and stratied-wavy ow, the best method in both cases was that of
Gro nnerud. It was observed that the peak in the two-phase frictional pressure gradient at high vapor qualities coin-
cided with the onset of dryout in the annular ow regime. #2002 Elsevier Science Ltd and IIR. All rights reserved.
Keywords: Refrigerant; Two-phase ow; Pressure drop; Horizontal tube; Calculation; R134a; R123; R402A; R404A; R502
Pre vision des gradients de pression des frigorige` nes en
e coulement diphasique dans des tubes horizontaux
Mots cles : Frigorige` ne ; E

coulement diphasique ; Chute de pression ; Tube horizontal ; Calcul ; R134a ; R123 ; R402A ; R404A ;
R502
0140-7007/02/$22.00 # 2002 Elsevier Science Ltd and IIR. All rights reserved.
PI I : S0140- 7007( 01) 00099- 8
International Journal of Refrigeration 25 (2002) 935947
www.elsevier.com/locate/ijrefrig
* Corresponding author. Tel.: +41-21-693-59-81; fax: 41-21-693-59-60.
E-mail address: john.thome@ep.ch (J.R. Thome).
1. Introduction
Prediction of two-phase pressure drops in direct-
expansion evaporators, condensers and two-phase
refrigerant transfer lines is important for accurate
design and optimization of refrigeration, air-condition-
ing and heat pump systems. Taking, for example, direct-
expansion evaporators, the optimal use of the two-phase
pressure drop to obtain the maximum ow boiling heat
transfer performance is one of the primary design goals.
In these evaporators, typically a two-phase pressure
drop equivalent to a loss of 1.4

C (2.5

F) in saturation
temperature from inlet to outlet is set as the design limit.
Yet, pressure drops predicted using leading methods
dier by up to 100%. Putting this into perspective, if an
evaporator is inaccurately designed with a two-phase
pressure drop only one-half the real value, then the sys-
tem eciency will suer accordingly from the larger
than expected fall in saturation temperature and pres-
sure through the evaporator. On the other hand, if the
predicted pressure drop is too large by a factor of two,
then fewer tubes of longer length could have been uti-
lized to obtain a more compact unit. Hence, accurate
prediction of two-phase pressure drops is a particularly
important aspect of rst law and second law optimiza-
tions of these systems.
In the present study, experimental test data available
from Kattan [1] have been compared to the following
seven widely quoted prediction methods for the fric-
tional pressure drop in two-phase ows: Lockhart and
Martinelli [2], Friedel [3], Gro nnerud [4], Chisholm [5],
Banko [6], Chawla [7] and Mu ller-Steinhagen and
Heck [8]. Of these methods, that of Friedel is often the
most recommended for use, i.e. refer to Whalley [9] and
Collier and Thome [10], based on its comparion to a
database of more than 40,000 points. In contrast, in a
recent study conducted on a large diversied database,
Tribbe and Mu ller-Steinhagen [11] found that the
method proposed by Mu ller-Steinhagen and Heck [8]
provided the best accuracy.
The two-phase pressure drop test data of Kattan cover
ve refrigerants: two pure uids (R-134a and R-123), one
azeotropic mixture (R-502) and two zeotropic mixtures
(R-402A and R-404A) over mass velocities from 100 to
500 kg/m
2
s for saturation pressures ranging from 0.112 to
0.890 MPa. While obtained during the heat transfer stud-
ies described in Kattan [1], these data were not reported
there and are hence presented here for the rst time.
Nomenclature
a parameter in Eq. (36) (Pa m
1
)
b parameter in Eq. (36) (Pa m
1
)
B parameter of Chisholm
C constant of Lockhart and Martinelli (m)
d
i
tube internal diameter
E parameter of Friedel
F parameter of Friedel
f friction factor
f
Fr
Froude friction factor
g acceleration due to gravity (m s
2
)
G factor in Eq. (36) (Pa m
1
)
H factor of Friedel
L tube length (m)
m
.
total
total mass velocity of liquid plus vapor
n exponent (kg m
2
s
1
)
p pressure (Pa)
dp/dz frictional pressure gradient (Pa m
1
)
p
total
total pressure drop (Pa)
p
static
static head pressure drop (Pa)
p
mom
two-phase momentum pressure drop (Pa)
p
frict
two-phase frictional pressure drop (Pa)
p
G
vapor-phase pressure drop (Pa)
p
L
liquid-phase pressure drop (Pa)
T
sat
saturation temperature (

C)
X
tt
Martinelli parameter
Y Chisholm parameter
Greek symbols
vapor quality
c vapor void fraction
, density (kg m
3
)

Lo
two-phase multiplier for liquid only

Go
two-phase multiplier for vapor only

Ltt
two-phase multiplier of Martinelli relative to
liquid

Gtt
two-phase multiplier of Martinelli relative to
vapor

gd
two-phase multiplier of Gro nnerud
j dynamic viscosity (N s m
2
)
o surface tension (N m
1
)
Dimensionless numbers
Fr Froude number
Re Reynolds number
We Weber number
Subscripts
G vapor or gas
Go vapor only (all ow as vapor)
h homogeneous
L liquid
Lo liquid only (all ow as liquid)
tp two-phase
936 M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947
2. Two-phase pressure drops
The two-phase pressure drops for ows inside tubes
are the sum of three contributions: the static pressure
drop p
static
, the momentum pressure drop p
mom
and
the frictional pressure drop p
frict
as:
p
total
p
static
p
mom
p
frict
1
For a horizontal tube, there is no change in static
head so p
static
= 0. The momentum pressure drop
reects the change in kinetic energy of the ow and is
for the present case given by:
p
mom

m
.
2
total
1
2
,
L
1 c


2
,
G
c
!
out

1
2
,
L
1 c


2
,
G
c
!
in
& '
2
where m
.
total
is the total mass velocity of liquid plus
vapor and is the vapor quality. In the present study,
the void fraction E is obtained from the Steiner [12] ver-
sion of the drift ux model of Rouhani and Axelsson
[13] for horizontal tubes:
c

,
G

1 0.12 1

,
G

1
,
L

1.18 1 go ,
L
,
G

0.25
m
.
2
total
,
0.5
L
!
1
3
Using the experimental values for the inlet and outlet
vapor quality, the momentum pressure drop is calcul-
able. Hence, the experimental two-phase frictional pres-
sure drop is obtainable from Eq. (1) by subtracting the
calculated momentum pressure drop from the measured
total pressure drop.
3. Frictional two-phase pressure drop correlations
The two-phase frictional pressure drop correlations of
the seven methods compared to the present experi-
mental data are described in this section.
3.1. Friedel [3] correlation
This method [3] is for vapor qualities from 0 4 - 1
and utilizes a two-phase multiplier as:
p
frict
p
L

2
Lo
4
where p
L
is calculated for the liquid-phase as
p
L
4f
L
L,d
i
m
.
2
total
1
2
1,2,
L
5
The liquid friction factor and liquid Reynolds number
are obtained from
f
0.079
Re
0.25
6
Re
m
.
total
d
i
j
7
using the liquid dynamic viscosity j
L
. His two-phase
multiplier is correlated as:

2
Lo
E
3.24FH
Fr
0.045
h
We
0.035
L
8
where Fr
h
, E, F and H are as follows:
Fr
h

m
.
2
total
gd
i
,
2
h
9
E 1
2

2
,
L
f
G
,
G
f
L
10
F
0.78
1
0.224
11
H
,
L
,
G

0.91
j
G
j
L

0.19
1
j
G
j
L

0.7
12
The liquid Weber We
L
is dened as:
We
L

m
.
2
total
d
i
o,
h
13
and the homogeneous density ,
h
is used:
,
h


,
G

1
,
L

1
14
Friedels method is typically that recommended when
the ratio of (j
L
/j
G
) is less than 1000.
3.2. Lockhart and Martinelli [2] correlation
This method [2] gives the two-phase frictional pres-
sure drop based on a two-phase multiplier for the
liquid-phase or vapor-phase, respectively, as:
p
frict

2
Ltt
p
L
15
p
frict

2
Gtt
p
G
16
where Eq. (5) is used for p
L
and p
G
is obtained from
M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947 937
p
G
4f
G
L,d
i
m
.
2
total

2
1,2,
G
17
The single-phase friction factors of the liquid and
vapor, f
L
and f
G
are calculated using Eq. (6) with their
respective physical properties. The corresponding two-
phase multipliers are

2
Ltt
1
C
X
tt

1
X
2
tt
. for Re
L
> 4000 18

2
Gtt
1 CX
tt
X
2
tt
. for Re
L
- 4000 19
where X
tt
is the Martinelli parameter for both phases in
the turbulent regimes dened as
X
tt

1


0.9
,
G
,
L

0.5
j
L
j
G

0.1
20
The value of C in Eqs. (18) and (19) depends on
the regimes of the liquid and vapor. For both uids
turbulent, C is equal to 20, as is always the case in the
present database. The correlation of Lockhart and
Martinelli [2] is applicable to the range of vapor qualities
from 0 - 41.
3.3. Gronnerud [4] correlation
This method [4] was developed specically for refrig-
erants and it is as follows:
p
frict

gd
p
L
21
and

gd
1
dp
dz

Fr
,
L
,
G

j
L
j
G

0.25
1
2
6
6
6
4
3
7
7
7
5
22
where Eq. (5) is used for p
L
and his two-phase multi-
plier is a function of
dp
dz

Fr
f
Fr
4
1.8

10
f
0.5
Fr

23
If the liquid Froude number Fr
L
is greater than or
equal to 1, then the friction factor f
Fr
is set to 1.0; if Fr
L
is less than 1, then:
f
Fr
Fr
0.3
L
0.0055 ln
1
Fr
L

2
24
where
Fr
L

m
.
2
total
gd
i
,
2
L
25
The correlation of Gro nnerud is applicable to vapor
qualities from 0 4 - 1.
3.4. Chisholm [5] correlation
Chisholm proposed a detailed empirical method [5]
for a wide range of operating conditions. His two-phase
frictional pressure drop gradient is given as
dp
dz

frict

dp
dz

Lo

2
Lo
26
The monophase frictional pressure gradients are
taken from the standard expressions for the liquid and
vapor phases:
dp
dz

Lo
f
L
2m
.
2
total
d
i
,
L
27
dp
dz

Go
f
G
2m
.
2
total
d
i
,
G
28
where the friction factors are obtained with Eq. (6)
using Eq. (7) and the respective dynamic viscosities of
the liquid and the vapor.
For laminar ows (Re-2000)
f
16
Re
29
and for turbulent ows (assumed to be at Re52000
to avoid an undened interval in his method) the
expression of Blasius [Eq. (6)] is used. These expressions
are applied using either the liquid phase or the vapor
phase Reynolds numbers. The parameter Y is obtained
from the ratio of the monophase frictional pressure
gradients:
Y
2

dp,dz
Go
dp,dz
Lo
30
His two-phase multiplier is then determined as:

2
Lo
1 Y
2
1 B
2n ,2
1 x
2n ,2

2n

31
where n is the exponent from the friction factor expres-
sion of Blasius (n=0.25). If 0-Y-9.5, then the para-
meter B is calculated as follows depending on the mass
velocity:
938 M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947
B
55
m
. 1,2
total
for m
.
total
51900 kg,m
2
s
B
2400
m
.
total
for 500 - m
.
total
- 1900 kg,m
2
s
B 4.8 for m
.
total
4500 kg,m
2
s 32
If 9.5-Y-28, then
B
520
Ym
. 1,2
total
for m
.
total
4600 kg,m
2
s
B
21
Y
for m
.
total
> 600 kg,m
2
s 33
If Y>28, then
B
15000
Y
2
m
. 1,2
total
34
The correlation of Chisholm is applicable to vapor
qualities from 0 441.
3.5. Banko [6] correlation
This method [6] is an extension of the homogeneous
model (not shown here).
3.6. Chawla [7] correlation
Chawla [7] suggested a correlation based on the velo-
city ratio between the vapor and liquid phases (not
shown here).
3.7. Muller-Steinhagen and Heck [8] correlation
This two-phase frictional pressure gradient correla-
tion is [8]:
dp
dz

frict
G 1
1,3
b
3
35
where the factor G is
G a 2 b a 36
where a and b are the frictional pressure gradients for all
the ow liquid (dp/dz)
Lo
and all the ow vapor (dp/
dz)
Go
, obtained respectively from Eqs. (27) and (28).
This model is essentially an empirical two-phase
extrapolation between all liquid ow and all vapor ow
and as such is applicable for 0 441. Recently,
Tribbe and Mu ller-Steinhagen [11] have shown that this
method gave the best results from a comparison of
competing methods against a database covering air-oil,
air-water, water-steam and several refrigerants.
4. Experimental test sections and measurement methods
New experimental data for two-phase pressure drops
were obtained for two 3.013 m long horizontal test sec-
tions of 10.92 and 12.00 mm diameter covering ve
refrigerants (R-134a, R-123, R-402A, R-404A and R-502)
over mass velocities from 100 to 500 kg/m
2
s and vapor
qualities from 0.04 to 1.0. R-134a is a new refrigerant
that has replaced R-12 and in part R-22. R-123 is a new
refrigerant that has replaced R-11. R-402A and R-404A
are new refrigerant mixtures that are near-azeotropes
that are replacing R-502. The test data were obtained
for evaporating conditions inside horizontal, copper
tube test sections of 3.013 m length that were heated by
counter-current ow of hot water. Two identical test
sections were connected in series in the test loop, one
located above the other (referred to later as the upper
and lower test sections). Two internal tube diameters
were tested: 10.92 and 12.00 mm.
The two-phase pressure drops were measured with a
choice of two dierent dierential pressure transducers,
the rst operating over the range from 010 kPa and the
second from 020 kPa. Each had an accuracy of 0.5%
FS and they were calibrated in the laboratory before
use. The ow rate of subcooled refrigerant before the
preheater was measured with a Coriolis meter, which
was calibrated in the laboratory and accurate to 0.2%
of the reading. The saturation pressures at the inlet and
outlet of the test sections were measured with absolute
pressure gauges that were accurate to 0.05% FS (1000
kPa) and the mean of these two pressures was used to
determine the saturation temperature and hence the
physical properties of the refrigerants. The physical
properties of the refrigerants were obtained using
REFPROP [14]. The test sections were set to the hor-
izontal using a special high sensitivity level.
The inlet vapor qualities of the refrigerant were
obtained from energy balances on the electrical pre-
heater. Inlet vapor qualities were 0.04 or greater. The
outlet vapor qualities were obtained from an energy
balance on the hot water, which from liquidliquid tests
were found to be accurate to 1% on average with a
maximum deviation of 2%. The inlet and outlet vapor
qualities were thus accurate to about 0.01 and 0.02,
respectively. A complete description of the experimental
test facility and the associated heat transfer data are
described in Kattan et al. [15,16,17].
The measured two-phase pressure drops are a combi-
nation of the frictional pressure drop and the momen-
tum pressure drop of the evaporating uid. Hence, the
momentum pressure drop was calculated using the inlet
and outlet vapor qualities together with Eqs. (2) and (3)
and subtracting its value from the measured pressure
drop to obtain the frictional pressure drop. The
momentum pressure drops were about 5 to 15% of the
measured pressure drops for the present test conditions.
M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947 939
5. Experimental results and discussion
The experimental frictional pressure drops have been
compared to all seven methods described earlier. The
calculated frictional pressure drops were obtained using
the mean vapor quality for each data set at the test
conditions. The in the test data from inlet to outlet
of the test section were on the order of 0.05 to 0.20. The
experimental frictional pressure drops were then con-
verted into frictional pressure gradients by dividing by
the test section length. The methods of Banko [6] and
Chawla [7] gave particularly poor predictions, in part
because the test conditions are distant from their range
of application. Hence, the results for these two methods
are not presented here. A selection of test data, com-
pared graphically to the ve remaining methods, are
shown below.
Fig. 1 depicts the R-134a data in the 10.92 mm tube
at a mass velocity of 300 kg/m
2
s at T
sat
=4

C. These
test data for the lower test section span nearly the entire
range of vapor quality. The experimental values go
through a maximum at a vapor quality of 0.85, which
corresponds to the transition from annular ow to
annular ow with partial dryout (i.e. annular ow to
stratied-wavy ow transition) predicted by the Kattan
et al. [15] ow pattern map (described in Section 7). The
Gro nnerud method [4] predicts these data best and
notably also predicts the peak in the data. The bottom
graph shows the experimental values normalized by the
predicted values for each method. As can be seen, the
experimental values range from as low as 50% and as
high as 250% of the predicted ones.
Fig. 2 depicts similar R-134a data obtained for the
same test conditions for the upper test section. Once
again the Gro nnerud method [4] gives the best predic-
tion and also closely follows the slope of the test data.
The peak in the pressure gradient at high vapor qualities
is again evident, although not as marked as in Fig. 1.
Fig. 3 shows R-134a test data for the 12.0 mm tube
obtained at 300 kg/m
2
s and T
sat
=2

C in the upper test
section at low vapor qualities ranging from 0.04 to
0.095. The variation in pressure gradient with vapor
quality is at with some scatter in the data, where the
ow pattern is predicted to be going through the transi-
tion from stratied-wavy ow into intermittent ow in
this range.
Fig. 4 presents R-134a test data for the 12.0 mm tube
obtained at 300 kg/m
2
s and T
sat
=2

C in the lower
evaporator test section at vapor qualities ranging from
0.11 to 0.37. As opposed to Fig. 3, these data (some data
points are superimposed by the predicted values in the
upper graph) now have a signicant upward slope. The
ow pattern is predicted to be going from the inter-
mittent regime at the lower qualities into the annular
regime at the high end of this range.
Fig. 5 presents R-134a test data for the 12.0 mm tube
obtained at the lowest mass velocity tested, namely 100
kg/m
2
s, at T
sat
=10

C in the lower evaporator test
section at qualities ranging from 0.22 to 0.77. The ow
Fig. 1. R-134a data for lower 10.92 mm tube at 300 kg/m
2
s and
T
sat
=4

C.
Fig. 1. R-134a data for upper 10.92 mm tube at 300 kg/m
2
s and
T
sat
=4

C.
Fig. 2. R-134a data for upper 10.92 mm tube at 300 kg/m
2
s
and T
sat
=4

C.
940 M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947
pattern is predicted to be stratied-wavy over the entire
range. Inevitably, there is more scatter in these data at
lower mass velocity as the pressure gradients are quite
small. For R-134a, limited sets of data were also
obtained at mass velocities of 200, 400 and 500 kg/m
2
s.
Fig. 6 shows the R-123 data in the 12 mm tube at a
mass velocity of 300 kg/m
2
s at T
sat
=30

C. These test
data for the lower test section cover only a narrow range
of vapor quality. The Gro nnerud [4] and Mu ller-Stein-
Fig. 4. R-134a data in 12 mm tube at 300 kg/m
2
s and
Fig. 3. Low vapor quality data for R-134a in 12 mm tube at
300 kg/m
2
s and T
sat
=2

C.
Fig. 5. R-134a data in 12 mm tube at 100 kg/m
2
s and
T
sat
=10

C.
Fig. 6. R-123 data in the 12 mm tube at 300 kg/m
2
s and
T
sat
=30

C.
M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947 941
hagen and Heck [8] methods best predict these data. The
bottom graph shows that the experimental values range
from as low as 20% and as high as 138% of the pre-
dicted ones. Hence, for example, the Chisholm method
over predicts these data by about three times.
Fig. 7 depicts the R-402A data in the 12 mm tube at
318 kg/m
2
s at T
sat
=2.4

C. These test data for the
lower test section span vapor qualities from 0.13 to 0.47.
The Friedel method [3] is most eective for these data,
although none of the methods capture the slope in the
experimental data.
Fig. 8 displays the R-404A data in the 12 mm tube at
318 kg/m
2
s at T
sat
=2.5

C. These test data for the
lower test section span vapor qualities from 0.12 to 0.59.
The Gro nnerud method [4] best predicts these data.
Fig. 9 presents the R-502 data in the 12 mm tube at
300 kg/m
2
s at T
sat
=2.5

C. These test data for the
lower test section span vapor qualities from 0.12 to 0.62.
Once again the Gro nnerud method [4] best predicts
these data followed by that of Friedel [3] and then
Mu ller-Steinhagen and Heck [8]. As for all these graphs,
the Chisholm method [5] typically tends to signicantly
over predict the experimental values. The Lockhart and
Martinelli method [2] also does not provide a satisfac-
tory description of the data in these graphs.
Fig. 7. R-402A data in the 12 mm tube at 318 kg/m
2
s and
T
sat
=2.4

C.
Fig. 7. R-402A data in the 12 mm tube at 318 kg/m
2
s and
T
sat
=2.4

C.
Fig. 8. R-404A data in the 12 mm tube at 318 kg/m
2
s and
T
sat
=2.5

C.
Fig. 9. R-502 data in the 12 mm tube at 300 kg/m
2
s and
T
sat
=2.5

C.
Fig. 9. R-502 data in the 12 mm tube at 300 kg/m
2
s and
T
sat
=2.5

C.
942 M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947
6. Statistical comparison to experimental results
Since the two-phase pressure drops at the lowest mass
velocity tested (100 kg/m
2
s) resulted in very small pres-
sure drops at low vapor qualities, some of these are not
accurately measured with the present transducers.
Hence, starting from an initial total of 868 measured
pressure drops, the statistical comparison of the stan-
dard, the mean and the average deviations was carried
out rst for all those test data greater than 0.5 kPa (788
points) and then for all test data greater than 1.0 kPa
(691 points).
Table 1 depicts the statistical comparion of the meth-
ods to the data greater than 0.5 kPa with the various
test conditions shown at the bottom. More than half of
the data are for R-134a for which numerous heat trans-
fer data were taken. The ranking of the methods are
shown in the right-most column. The Gro nnerud
method [4] gives the best overall results, with rankings
of 1/1/2 for the three types of deviation. The method of
Mu ller-Steinhagen and Heck [8] comes in a close second
with rankings of 2/2/1. In third place is the method of
Friedel [3] which always ranks third in all categories.
The Lockhart and Martinelli method [2] always ranks
fourth while that of Chisholm consistently ranks fth.
As a second test of the methods, Table 2 depicts the
statistical comparion of the methods to the data greater
than 1.0 kPa. The deviations are now smaller as the
scatter in experimental data at low pressure drops has
been eliminated by their exclusion. The ranking of the
methods are shown in the right-most column, which are
Fig. 10. Two-phase ow pattern map of Kattan et al. [1517]
for R-134a at conditions indicated (dashed lines show the pro-
gression in the ow pattern from a vapor quality of 0.01 to 0.99
for a xed mass velocity).
Table 1
Comparison to all data for pressure drops larger than 0.5 kPa
Correlation Deviation Refrigerants All Rank
(%) R134a R134a
a
R123 R402a R404a R502
Lockhart and Martinelli [2] Standard 311.24 73.58 58.97 100.33 121.75 97.29 121.85 4
Mean 123.64 45.12 45.44 76.00 88.89 71.14 89.94 4
Average 142.18 8.57 43.93 91.34 124.68 99.57 83.86 4
Friedel [3] Standard 201.53 47.77 50.14 55.21 66.23 53.64 74.55 3
Mean 83.06 28.79 40.07 40.31 48.67 40.40 52.38 3
Average 68.34 12.53 39.99 11.75 30.49 35.09 26.46 3
Gro nnerud [4] Standard 111.59 41.31 34.47 40.76 32.89 30.61 40.07 1
Mean 37.20 31.41 24.74 29.62 24.30 26.16 29.89 1
Average -0.14 2.57 1.41 24.96 10.47 7.39 8.35 2
Chisholm [5] Standard 387.40 97.81 119.73 91.85 115.61 94.58 144.01 5
Mean 156.35 58.75 92.83 68.07 85.88 67.14 102.74 5
Average 213.47 28.83 181.63 67.20 117.85 111.21 113.44 5
Mu ller-Steinhagen Standard 144.33 35.02 44.13 33.07 40.47 35.04 48.00 2
and Heck [8] Mean 52.25 19.91 32.55 24.69 28.54 26.54 32.14 2
Average 17.74 22.75 26.75 24.00 7.32 0.63 6.87 1
Number of experimental
points
249 178 25 150 130 56 788
T
sat
(

C) 1.3 to 10.3 4.44 30.7 1.3 to 10.3 1.3 to 10.3 2.5


(%) 4 to 80 18 to 100 7 to 70 6 to 74 7 to 84 7 to 84
Mass ux (kg/m
2
s) 100 to 500 100 to 400 100 to 300 100 to 318 100 to 318 100 to 300
P
sat
(bar) 2.8 to 4.2 3.42 1.12 6.3 to 8.9 5.9 to 8.4 6.19
P
crit
(bar) 41.35 41.35 36.07 41.35 37.32 40.75
s
L
(mN/m) 11.87 to 10.2 11.04 14.5 9.2 to 7.67 7.95 to 6.41 9.3
a
tube diameter is 10.92 mm.
M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947 943
similar to those in Table 1. Here the method of Mu ller-
Steinhagen and Heck [8] just outperforms the Gro n-
nerud method [4] with rankings of 1/1/3 versus 2/2/1 for
the three types of deviation. The method of Friedel [3]
comes in third with rankings of 3/3/2. The Lockhart and
Martinelli method [2] again always ranks fourth while
that of Chisholm [5] consistently ranks fth. Looking at
the best method, it has statistical deviations of 31.5%,
24 and 14.7%, respectively, which are quite poor from
a thermal design point of view.
7. Statistical comparison by ow pattern
Kattan et al. [17] showed that classifying ow boiling
heat transfer data by ow regime and then developing a
simplied ow structure and heat transfer model for each
regime, a very signicant increase in accuracy could be
obtained compared to methods oblivious to ow pat-
tern. Therefore, it makes sense to classify the two-phase
frictional pressure drop data by ow regime too,
although the above empirical methods do not model any
particular regime, in order to determine which is most
appropriate for a particular ow pattern.
In this light, Fig. 10 depicts their ow pattern map
prepared for the case of a 12 mm diameter tube for R-
134a at a saturation temperature of 10

C and a heat
ux of 10 kW/m
2
. Their map shows the calculated
transition curves plotted on an easy to read graph of
mass velocity vs. vapor quality, where I refers to Inter-
mittent ow, A refers to Annular ow, SW refers to
Stratied-Wavy ow, S refers to Stratied ow and MF
refers to Mist ow. In this study, such a map was cal-
culated for all the uids at their test conditions to
determine the local ow pattern at the inlet and at the
outlet of the test section for each data point. Then, only
data points for which the ow pattern at the inlet was
the same as at the outlet were selected. This resulted in a
database of two-phase frictional pressure drops for
annular ows, intermittent ows and stratied-wavy
ows but none in the stratied and mist ow regimes.
Table 3 depicts the statistical results for pressure drops
larger than 0.5 kPa where there are now 283 points total
with 100 for annular ow, 96 for intermittent ow and
87 for stratied-wavy ow. These results conrm the
following:
Annular ow. The best method is that of Mu ller-
Steinhagen and Heck [8] with rankings of 1/1/4 and for
which the deviations are 18.1, 11.5 and 32.5%. The
second best method is that of Friedel [3] with rankings
of 2/2/3. The method of Gro nnerud [4] is in this case in
a tie for the fourth and fth positions.
Intermittent ow. The best method is that of Gro n-
nerud [4] with very high rankings of 1/1/2 but with
deviations of 34.9, 27.0 and 32.5%. The second best
method is that of Mu ller-Steinhagen and Heck [8] with
Table 2
Comparison to all data for pressure drops larger than 1.0 kPa
Correlation Deviation
(%)
Refrigerants All Rank
R134a R134a
a
R123 R402a R404a R502
Lockhart and Martinelli [2] Standard 88.00 61.42 58.97 82.32 70.60 51.24 86.24 4
Mean 75.35 37.96 45.44 67.03 64.37 49.98 72.28 4
Average 97.32 -14.83 43.93 84.35 95.64 77.84 63.44 4
Friedel [3] Standard 56.29 40.01 50.14 46.53 36.53 24.21 49.41 3
Mean 48.38 24.25 40.07 37.20 32.37 21.88 39.05 3
Average 36.54 16.56 39.99 7.14 13.39 11.56 12.79 2
Gro nnerud [4] Standard 28.25 40.69 34.47 37.92 21.03 19.46 32.81 2
Mean 22.95 31.30 24.74 29.52 19.04 21.30 26.95 2
Average 11.03 2.68 1.41 23.86 13.90 15.95 10.25 1
Chisholm [5] Standard 109.63 81.55 119.73 81.28 69.87 49.35 101.48 5
Mean 94.28 49.15 92.83 65.66 63.22 47.76 83.05 5
Average 155.97 20.51 181.63 62.95 90.79 86.40 91.36 5
Mu ller-Steinhagen and Heck [8] Standard 33.64 31.26 44.13 27.35 22.24 15.29 31.48 1
Mean 27.67 17.92 32.55 22.54 19.44 13.31 23.97 1
Average 2.85 25.00 26.75 26.59 16.43 16.56 14.67 3
Number of experimental points 215 173 25 130 109 39 691
a
Tube diameter of 10.92 mm.
944 M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947
rankings of 2/2/1 and deviations just slightly worse than
Gro nnerud [4]. The method of Friedel [3] here comes in
third with rankings of 3/3/3.
Stratied-wavy ow. The best method is a tie between
that of Gro nnerud [4] with rankings of 1/1/4 and that of
Mu ller-Steinhagen and Heck [8] with rankings of 2/2/2.
The deviations for the Gro nnerud method [4] are 28.8,
23.5 and 25.5% while those of Mu ller-Steinhagen and
Heck [8] are 33.1, 26.9 and 13.1%. The method of
Friedel [3] here comes in third with rankings of 3/3/3.
Similarly, Table 4 depicts the statistical results for
pressure drops larger than 1.0 kPa where there are now
263 points total with 100 for annular ow, 95 for inter-
mittent ow and 68 for stratied-wavy ow. These
results demonstrate the following:
Annular ow. The results are exactly the same as for
Table 3.
Intermittent ow. The results are the same as for
Table 3 but based on one fewer data point.
Stratied-wavy ow. The best method is that of
Gro nnerud [4] with rankings of 1/1/4 and deviations of
21.6, 29.8 and 4.0%. The method of Mu ller-Steinha-
gen and Heck [8] with rankings of 2/2/3 is now tied with
the method of Friedel [3] whose rankings are now 3/3/1.
Overall, classifying the ow by local ow pattern and
then using the best two-phase frictional pressure method
for that particular ow pattern results in a signicant
improvement in accuracy as illustrated by the smaller
statistical deviations. For example, the top ranked
method for annular ow (Mu ller-Steinhagen and Heck
[8]) gives a standard deviation of 18.1% in Table 3
compared to its overall value of 48% for all data in
Table 1. Similarly, the top ranked method for Inter-
mittent ow and stratied-wavy ow (Gro nnerud [4])
gives standard deviations of 34.9 and 28.8% in Table 3,
respectively, compared to its overall value of 40.1% for
all data in Table 1.
Even so, the accuracies of these methods are still
rather poor for thermal design considering that the two-
phase pressure drop is only predicted to within 18.1 to
28.8% standard deviation. For the future, new tests are
getting underway to extend the present database to
other ow regimes and a larger range of tube diameters.
Then, new prediction methods specic to the particular
ow structure of each ow regime will be developed
similar to the Kattan et al. ow boiling heat transfer
model [1517].
8. Conclusions
The conclusions of the present study are the following:
An extensive new two-phase pressure drop
database for ve refrigerants originating from
the study of Kattan [1] is presented here for the
rst time.
Comparing seven of the most quoted methods in
the literature to this database irrespective of
ow pattern (788 data points), the method of
Mu ller-Steinhagen and Heck [8] and the method
of Gro nnerud [4] consistently gave the best pre-
dictions while that of Friedel [3] was only third
best.
Table 3
Comparison by ow pattern for pressure drops larger than 0.5 kPa
Correlation Deviation
(%)
Regimes All
A Rank I Rank SW Rank
Lockhart and Martinelli [2] Standard 29.13 3 85.63 4 67.41 4 57.14
Mean 21.47 3 65.75 4 46.31 4 74.71
Average 32.54 5 132.05 4 17.72 3 38.29
Friedel [3] Standard 20.98 2 57.02 3 45.43 3 31.54
Mean 13.08 2 44.48 3 34.44 3 38.39
Average 28.59 3 40.69 3 10.10 1 6.62
Gro nnerud [4] Standard 37.70 4 34.92 1 28.83 1 22.56
Mean 29.42 5 27.03 1 23.54 1 30.07
Average 11.70 2 1.96 2 25.48 4 4.42
Chisholm [5] Standard 37.95 5 114.52 5 82.36 5 64.49
Mean 27.22 4 86.75 5 62.89 5 82.82
Average 2.31 1 156.35 5 55.63 5 68.89
Mu ller-Steinhagen and Heck [8] Standard 18.11 1 37.69 2 33.06 2 20.33
Mean 11.54 1 27.59 2 26.85 2 23.74
Average -32.52 4 0.46 1 13.11 2 15.50
Number of experimental points 100 96 87 283
M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947 945
Comparing the seven methods to pressure drop
data classied by two-phase ow pattern, the
method of Mu ller-Steinhagen and Heck [8] was
best for annular ows while the method of
Gro nnerud [4] gave the best predictions for both
intermittent and stratied-wavy ows.
The statistical deviations of the best methods
still remain quite large with respect to the accu-
racy desired for reliable thermal design of eva-
porators and condensers.
The peak in the experimental two-phase fric-
tional pressure gradient was observed to coin-
cide with the onset of dryout in annular ows at
high vapor qualities, similar to the equivalent
peak in the ow boiling heat transfer coecient.
Acknowledgements
This project was supported by a grant from Swiss
National Science Foundation under contract 21/
57210.99.
References
[1] Kattan N. Contribution to the heat transfer analysis of
substitute refrigerants in evaporator tubes with smooth or
enhanced tube surfaces. PhD thesis No 1498, Swiss Fed-
eral Institute of Technology, Lausanne, Switzerland, 1996.
[2] Lockhart RW, Martinelli RC. Proposed correlation of
data for isothermal two-phase two-component ow in
pipes. Chem Eng Progr 1949;45:3945.
[3] Friedel L. Improved friction pressure drop correlations for
horizontal and vertical two-phase pipe ow. European
Two-Phase Flow Group Meeting, Paper E2; 1979 June;
Ispra, Italy.
[4] Gro nnerud R. Investigation of liquid hold-up, ow-resis-
tance and heat transfer in circulation type evaporators,
part IV: two-phase ow resistance in boiling refrigerants.
Annexe 1972-1, Bull. de lInst. du Froid, 1979.
[5] Chisholm D. Pressure gradients due to friction during the
ow of evaporating two-phase mixtures in smooth
tubes and channels. Int J Heat Mass Transfer 1973;16:
34758.
[6] Banko SG. A variable density single-uid model for two-
phase ow with particular reference to steam-water ow. J
Heat Transfer 1960;11:26572.
[7] Chawla JM. Wa rmeu bergang und Druckabfall in waager-
echten Rohren bei der Stro mung von verdampfenden
Ka ltemitteln. VDI-Forschungsh 1967;523 Lg1Lg2.
[8] Mu ller-Steinhagen H, Heck K. A simple friction pressure
drop correlation for two-phase ow in pipes. Chem Eng
Process 1986;20:297308.
[9] Whalley PB. Two-phase ow and heat transfer. Oxford:
Oxford University Press, 1996.
[10] Collier JG, Thome JR. Convective Boiling and
Condensation. 3rd ed. Oxford: Oxford University Press,
1994.
[11] Tribbe C, Mu ller-Steinhagen H. An evaluation of the
performance of phenomenological models for predicting
pressure gradient during gas-liquid ow in horizontal
pipelines. Int J Multiphase Flow 2000;26:101936.
Table 4
Comparison by ow pattern for pressure drops larger than 1.0 kPa
Correlation Deviation
(%)
Regimes All
A Rank I Rank SW Rank
Lockhart and Martinelli [2] Standard 29.13 3 79.13 4 47.73 4 53.23
Mean 21.47 3 62.21 4 33.65 4 73.23
Average 32.54 5 128.64 4 3.85 2 34.78
Friedel [3] Standard 20.98 2 54.54 3 30.04 3 28.46
Mean 13.08 2 42.63 3 25.74 3 35.50
Average 28.59 3 38.96 3 1.95 1 2.62
Gro nnerud [4] Standard 37.70 4 34.01 1 23.06 1 21.58
Mean 29.42 5 26.31 1 21.56 1 29.76
Average 11.70 2 2.79 2 28.78 4 4.00
Chisholm [5] Standard 37.95 5 114.24 5 57.98 5 61.76
Mean 27.22 4 87.04 5 52.08 5 82.33
Average 2.31 1 157.19 5 39.93 5 66.00
Mu ller-Steinhagen and Heck [8] Standard 18.11 1 36.04 2 25.61 2 18.52
Mean 11.54 1 26.36 2 23.29 2 21.80
Average 32.52 4 0.69 1 19.90 3 17.81
Number of experimental points 100 95 68 263
946 M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947
[12] Steiner D, VDI-Wa rmeatlas (VDI Heat Atlas), Verein
Deutscher Ingenieure, VDI-Gesellschaft Verfahren-
stechnik und Chemieingenieurwesen (GCV), Du sseldorf,
1993, Chapter Hbb.
[13] Rouhani Z, Axelsson E. Calculation of void volume frac-
tion in the subcooled and quality boiling regions. Int J
Heat Mass Transfer 1970;13:38393.
[14] REFPROP. NIST Refrigerant Properties Database 23,
Gaithersburg, MD, 1998, Version 6.01.
[15] Kattan N, Thome JR, Favrat D. Flow boiling in hor-
izontal tubes. Part 1: development of a diabatic two-phase
ow pattern map. J Heat Transfer 1998;120:1407.
[16] Kattan N, Thome JR, Favrat D. Flow boiling in hor-
izontal tubes. Part 2: new heat transfer data for ve
refrigerants. J Heat Transfer 1998;120:14855.
[17] Kattan N, Thome JR, Favrat D. Flow boiling in horizontal
tubes. Part 3: development of a new heat transfer model
based on ow patterns. J Heat Transfer 1998;120:15665.
M.B. Ould Didi et al. / International Journal of Refrigeration 25 (2002) 935947 947

You might also like