Condensation in Smooth

Horizontal Tubes
An experimental study of heat transfer and flow regimes during condensation of
refrigerants in horizontal tubes was conducted. Measurements were made in smooth,
M. K. Dobson round tubes with diameters ranging from 3.14 mm to 7.04 mm. The refrigerants
Exxon Production Research Co., tested were R-12, R-22, R-134a, and near-azeotropic blends of R-32/R-125 in 50
Houston, TX 77252 percent/50 percent and 60 percent/40 percent compositions. The study focused pri-
Mem. ASME. marily on measurement and prediction of condensing heat transfer coefficients and
the relationship between heat transfer coefficients and two-phase flow regimes. Flow
regimes were observed visually at the inlet and outlet of the test condenser as the
J. C. Chato heat transfer data were collected. Stratified, wavy, wavy annular, annular, annular
Professor Emeritus, mist, and slug flows were observed. True mist flow without a stable wall film was
Department of Mechanical and Industrial not observed during condensation tests. The experimental results were compared
Engineering, with existing flow regime maps and some corrections are suggested. The heat transfer
University of Illinois at Urbana-Champaign, behavior was controlled by the prevailing flow regime. For the purpose of analyzing
1206 W. Green Street, condensing heat transfer behavior, the various flow regimes were divided into two
Urbana, IL 61801
broad categories of gravity-dominated and shear-dominated flows. In the gravity
[email protected]
Life Mem. ASME.
dominated flow regime, the dominant heat transfer mode was laminar film condensa-
tion in the top of the tube. This regime was characterized by heat transfer coefficients
that depended on the wall-to-refrigerant temperature difference but were nearly
independent of mass flux. In the shear-dominated flow regime, forced-convective
condensation was the dominant heat transfer mechanism. This regime was character-
ized by heat transfer coefficients that were independent of temperature difference but
very dependent on mass flux and quality. Heat transfer correlations that were devel-
oped for each of these flow regimes successfully predicted data from the present
study and from several other sources.

Introduction able for determining the transition from one flow regime to the
next. The work described herein represents an extensive effort
Studies of condensation are numerous in the technical litera- to combine the topics of heat transfer and flow regimes in
ture. These studies include analytical efforts to model the phys- order to develop correlations that can be used in the design of
ics of condensation processes, experimental efforts to measure condensers.
the heat transfer behavior of certain fluids, and various combina-
tions of the two. Most investigators have collected data for a
limited number of fluids under a range of operating conditions
that was suitable for the applications of their interest. Efforts Background Review
to match their data with existing correlations or to develop A detailed discussion of the background and the literature
correlations to fit their own data have met with some success. was given by Dobson (1994) and Dobson et al. (1994a, 1994b).
Many correlations that are available come with no explicit range Other summaries were published by Breber (1988), Palen et
of parameters over which they can be expected to give accurate al. (1993), and Wang and Chato (1995). Here, only the most
results. Although there are handbooks and design manuals sup- significant points will be described.
plying reasonable recommendations for a design engineer who
is searching the technical literature, the overall literature pro- Flow Regimes in Horizontal Two-Phase Flow. Fundamen-
vides seemingly diverse reports about what correlation is tally, the two factors controlling the flow are gravity and vapor
"best." The differences between the lowest and highest pre- shear. At low vapor velocities gravity dominates and the con-
dicted heat transfer coefficients, in certain cases, can be as high
densate forms primarily on the top portion of the tube and flows
as a factor of two or three (Hinde et al., 1992).
downward into a liquid pool which is driven out axially partly
There is agreement in the literature that the mechanisms of by the vapor flow and partly by a gravitational head. In terms
heat transfer and pressure drop are intimately linked with the of void fraction, the flow regimes can be divided into the follow-
prevailing two-phase flow regime. This has led to many studies ing two groups: (1) those that occur at high void fractions; and
aimed at predicting what dimensionless parameters govern spe- ( 2) those that occur only at low void fractions. The first category
cific flow regime transitions, and at what values of these dimen- includes five flow regimes: stratified flow, wavy flow, wavy-
sionless parameters certain transitions are expected to occur. annular flow, annular flow, and annular-mist flow. The second
Although debates still exist in the literature concerning differ- category includes slug, plug, and bubbly flow. The five flow
ences in the flow regime predictions of various researchers, a regimes in the first category are arranged such that each succes-
basic understanding has been established of what the various sive flow regime corresponds to an increase in the vapor veloc-
flow regimes are and, in most cases, what parameters are suit- ity. Thus, the first two are gravity dominated, the third is influ-
enced about equally, and the last two are vapor shear dominated.
The three flow regimes in the second category are arranged
Contributed by the Heat Transfer Division for publication in the JOURNAL OF such that the transition from one flow regime to the next corre-
HEAT TRANSFER. Manuscript received by the Heat Transfer Division August 19,
1996; revision received August 4, 1997; Keywords: Condensation; Heat Ex- sponds to an increase in the liquid inventory (i.e., decrease
changers; Multiphase Flows. Associate Technical Editor: T. J. Rabas. ma).

Journal of Heat Transfer Copyright © 1998 by ASME FEBRUARY 1998, Vol. 120 / 193

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 At very low vapor velocities, the gravity controlled stratified the annular-mist flow regime because of the appearance of an
flow regime is observed. Because the vapor velocity is low, the annular film with a mixture of vapor and mist in the core flow.
liquid-vapor interface remains smooth (Chato, 1962). At low void fractions slug flow is formed when interfacial
As the vapor velocity is increased, the liquid-vapor interface waves grow sufficiently in amplitude to block the entire cross
becomes Helmholtz unstable, giving rise to surface waves section at some axial locations, leading to the visual appearance
(Carey, 1992) and wavy flow. The condensation process on the of slugs of liquid flowing down the tube. These slugs have been
top of the tube is similar to that in stratified flow. The condensa- shown by Hubbard and Dukler (1966) and Lin and Hanratty
tion process on the portion of the tube wall near the interface (1989) to create large pressure spikes due to rapid deceleration
between the liquid pool and the vapor is affected by the waves of the vapor flow. In other cases, a flow that would visually be
since it is alternately exposed to a thin condensate film flowing identified as slug flow does not create these large pressure
downward and the crest of a wave moving in the mean flow spikes. This regime has been designated by Lin and Hanratty as
direction. pseudo-slug flow. They proposed that, unlike true slugs, pseudo
As the vapor velocity is increased further, the wavy flow slugs either did not entirely block the tube or did so only mo-
becomes unstable and can result in two different transitions. mentarily. A regime with a similar qualitative description to
At high liquid fractions, the slug flow case that is described pseudo-slug flow was designated as proto-slug flow by Nichol-
subsequently occurs. At lower liquid fractions, the waves begin son et al. (1978), and wavy-annular flow by Barnea et al.
to wash up and around the tube wall leading to an annular (1980). This illustrates the considerable subjectivity in flow
flow. Before the annular flow pattern is reached, however, a regime classifications.
transitional flow pattern called wavy-annular flow is observed As condensation continues, the slugs coalesce into a predomi-
in which some liquid from the waves begins to wash up and nantly liquid flow with large bubbles within. This is referred to
around the circumference of the tube, but not enough to create as the plug flow regime. Turbulent fluctuations within the liquid
a symmetric annular film. This liquid moves primarily in the eventually break these plugs into smaller vapor bubbles that
mean flow direction rather than downward, creating the primary become dispersed throughout the liquid. This is called the bub-
difference from the wavy flow regime. bly flow regime. The slug, plug, and bubbly flow regimes occur
With still further increases in the vapor velocity, the liquid at the end of the condensation process when the liquid inventory
migration from the pool at the bottom of the tube to the top of is large (i.e., a is small). In combination, they occupy only 10
the tube continues until the film thickness becomes nearly uni- percent to 20 percent of the total quality range. The plug and
form. The visual appearance of this type of flow is one of an bubbly flow regimes are confined to the bottom of the quality
annular film of liquid on the wall and a high-speed vapor core range, and will not be discussed in detail because they were
in the center, hence the name annular flow. The liquid-vapor not encountered in our experiments.
interface in annular flow is nearly always characterized by sur-
face waves due to the high-speed vapor flowing over it. Flow Regime Mapping. Since the flow patterns strongly
At yet higher vapor velocities, the crests of the waves on the influence the heat and momentum transfer processes, it is im-
liquid film are sheared off by the vapor flow and entrained in portant for designers to predict what flow pattern is expected
the core in the form of liquid droplets. This is referred to as based on the flow rate, quality, fluid properties, and tube diame-

Nomenclature
Ag = dimensionless cross-sectional area L = test-section length X„ = turbulent-turbulent Lockhart Mar-
occupied by vapor, Ag/D2 Nu = Nusselt number (subscript indi- tinelli parameter, (pg/pi)°'5(pi/
01
Bd = Bond number, g(px - pg)D2la cates characteristic length, other Ms) ((l -x)lx)os
Bo = boiling number, q"IGilg thmD),hD/kx x = vapor quality
cvi = specific heat at constant pressure P = pressure z = axial location in tube
for liquid Pred = reduced pressure, PIPaaiaa
Pr, = liquid Prandtl number, /i,cp,,/&, Greek Letters
D = tube inner diameter
D+ = dimensionless diameter, q" = heat flux a = void fraction
D\lrwpi/fjLX R = radius of curvature aT = angle of inclination in Taitel-
F = two-phase multiplier defined in Reg = superficial vapor Reynolds num- Dukler criterion
Eq. (19) ber, GDxl pg P = fraction of the circumference
F2 = denominator in Eq. (27) Re, = superficial liquid Reynolds num- where filmwise condensation
Ftd = defined by Eq. (2) ber, GD(\ - x)lpx prevails, Eqs. (17a, b)
Fr, = Froude number, G2Ip\gD Re,D = liquid only Reynolds number, &i = angle subtended from the top of
Frso = Soliman's modified Froude num- GD/fii tube to the liquid level
ber, Eqs. (4a, b) Re„„ = vapor only Reynolds number, Hs — dynamic viscosity of vapor
g = acceleration due to gravity GD/fig refrigerant
G = mass flux Sug = Suratman number, pgDa/p,2, p, = dynamic viscosity of liquid
Ga = Galileo number, gpi(p\ — TSM = saturation temperature refrigerant
Pg)D\lp\
Ts — surface temperature of tube wall Vi = kinematic viscosity of liquid, ptlpt
h = heat transfer coefficient ug = vapor velocity/superficial vapor pa = density of air
h, = dimensionless liquid level height, velocity pg = density of vapor
hID Ui = liquid velocity/superficial liquid Pi = density of liquid
Z; = enthalpy of vaporization
: velocity a = surface tension of refrigerant
g

Weso = Soliman's modified Weber num- r„ = shear stress at the wall

i'„ - Rohsenow's (1956) modified en- 4>j = two-phase pressure drop
thalpy of vaporization, Eq. (10) ber, Eqs. (5a, b)
Ja, : liquid Jakob number, (c p ,,(r sat - x = location from leading edge or va- multiplier, (dPldz)l(dPldz),
Ts)liig por quality 4>u, = defined by Eq. (16)
X„ = laminar-turbulent Lockhart Mar- 4>2g = two-phase pressure drop
kr liquid thermal conductivity tinelli parameter
K,d -• defined by Eq. (1)
multiplier, (dP/dz)/(dP/dz)g
LJ0 = mass fraction of oil in refrigerant

194 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ter. One of the earliest attempts at a flow regime map was by growth on a stratified-wavy flow which led to one of two effects.
Baker (1954). The Baker map was based on observations of At low liquid fractions the wave washed liquid around the cir-
adiabatic gas-liquid flows in tubes ranging from 25.4 to 101.6 cumference of the tube leading to an annular flow. At high
mm in diameter. The data used included both air-water and oil- liquid fractions the wave bridged the tube cross section and led
water flows, providing a fairly wide range of fluid properties. to an intermittent (slug or plug) flow. Taitel and Dukler pre-
The horizontal and vertical coordinates on the Baker map are dicted this transition based on a modification of the Kelvin-
the superficial liquid and vapor mass fluxes, times scaling fac- Helmholtz stability analysis for an infinitesimal wave between
tors, that account for fluid property variations. Although subse- two horizontal plates. Taitel and Dukler extended this theory
quent flow regime maps have achieved improved accuracy, to predict the stability of finite waves inside a tube. In dimen-
Baker's work is historically the first widely recognized flow sionless form, their result can be expressed as follows:
regime map.
Mandhane et al. (1974) developed a flow regime map similar i a2A\ - (2h, i)2
l. (3)
to Baker's using a larger database of 5935 observations. The .(l - hY
abscissa and ordinate of the Mandhane map are the superficial
gas velocity and superficial liquid velocity, which makes it
All terms in the bracketed portion of Eq. (3) are exclusive
rather simple to use. The Mandhane map correctly predicted
functions of the Martinelli parameter, so the boundary is a func-
the flow regime for 68 percent of the observations in their
tion of only F,d and X„. As discussed below, the two transition
database, as opposed to 42 percent for the original Baker map.
criteria expressed by Eqs. (1) and (3) are the only ones relevant
The boundaries of the Mandhane map were constructed primar-
to cases studied here.
ily for air-water data in tubes larger than those used in the
present study. Comparisons with recent data for refrigerants Another approach to predicting flow regime transitions, spe-
revealed systematic problems with the Mandhane map (Wat- cifically for condensation, has been developed by Soliman
telet, 1994; Wattelet et al., 1994). These problems were caused (1982, 1983). Soliman distinguished between three flow re-
primarily by the much higher vapor densities of the refrigerants gimes that he deemed to be important for condensing flows:
in Wattelet's study than the air densities in the Mandhane data. wavy flow, annular flow, and mist flow. He developed two flow
This problem would be worse for the data from the present regime transition criteria, one for the wavy-to-annular transition
study, since the vapor density is considerably higher at conden- and one for the annular-to-mist transition. It is important to note
sation temperatures than at evaporation temperatures. that the wavy flow regime of Soliman includes the regimes
commonly called stratified, slug, and wavy flow. While these
Perhaps the most theoretically based flow regime map is that regimes have important differences from the standpoint of flow
of Taitel and Dukler (1976). They reasoned that each flow regime classification, particularly concerning the stability of
regime transition was based on a different set of competing the wavy interface, Soliman apparently concluded that these
forces and that a single parameter or set of coordinates should differences were less important than the significant stratification
not be expected to predict all flow regime transitions. The that they all had in common.
Taitel-Dukler map includes five flow regimes: stratified smooth,
stratified wavy, annular, intermittent (plug and slug), and dis- Soliman (1982) postulated that the wavy-to-annular transi-
persed bubble. The map was developed for adiabatic flows, but tion was based on a balance between inertial and gravitational
has been used with success by other investigators for diabatic forces on the liquid film. The Froude number, V2/gL, represents
flows (Barnhart, 1992; Wattelet, 1994; Wattelet et al., 1994). a balance between these forces. Soliman proposed that the ap-
Taitel and Dukler developed various approaches for predicting propriate velocity was the actual liquid velocity and the appro-
the transitions between the various flow regimes based on the priate length scale was the film thickness, S. These parameters
appropriate physical mechanisms. For the stratified to wavy were not known based solely on G, x, and the fluid properties.
flow transition, they hypothesizxd that waves would be formed Soliman obtained expressions for them based on relations for
when the pressure and shear forces acting on a wave were two-phase pressure drop in annular flow. Thus, his transition
sufficient to overcome viscous dissipation in the wave. Their criterion is opposite to that of Taitel and Dukler in that it as-
ideas were based on the work of Jeffreys (1925, 1926), and in sumes that an annular flow exists and seeks to determine when
dimensional form they deduced that waves would be expected gravitational forces will cause a transition to wavy flow. The
to form when a critical vapor velocity was exceeded. In dimen- resulting expression for the Froude number is given by
sionless form, the inequality can be expressed as
1 + 1.09X?,0
20 X« Ga°
K,d = Ftd^, (1)
for Re, s 1250, (4a)
The right hand side of Eq. (1) follows the nomenclature set 5
forth by Taitel and Dukler, where all quantities with tildes are 1 + 1.09XJ 1
Fr„ = 1.26 Re,"
dimensionless variables. In this scheme, actual phase velocities x„ Ga0'5
are scaled by superficial phase velocities and lengths are scaled
by the tube diameter. for Re, > 1250. (4b)
The first term on the right of the equal sign, referred to by
Taitel and Dukler as a modified Froude number, is defined as Based on comparisons with data in tubes of 4.8 mm to 25
follows: mm in diameter, and with fluids including water, refrigerants,
and acetone, Soliman (1982) concluded that wavy flow was
Gxlpg observed for FrSD < 7, and annular flow was observed for Frst)
F„,= (2) > 7. Dobson (1994) and Dobson et al. (1994a, 1994b) reported
Pi ~ Ps V^7 that Fr,„ = 7 served as a good indicator of the transition from
The second term in Eq. (2) is a superficial vapor Froude wavy to wavy-annular flow, although a symmetric annular flow
number (ratio of inertial to gravity forces in the vapor), while was not observed until around Frm = 1 8 .
the first is a modifier depending on the density ratio. Soliman (1983, 1986) also developed a parameter for pre-
The transition from a stratified-wavy flow to an intermittent dicting the transition from annular to mist flow. He postulated
or annular flow was observed experimentally by Butterworth that the primary forces tending to prevent entrainment were
(1972) and Dukler and Hubbard (1975). They described wave liquid viscous forces and surface tension forces, and that the

Journal of Heat Transfer FEBRUARY 1998, Vol. 120 / 195

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 primary force which promoted mist formation was vapor inertia. lis - i,g(l + 0.68 Ja,). (10)
He formulated a modified Weber number that represented a
balance between these forces: This correction shows that the assumption of a linear tempera-
ture profile in the original analysis is quite acceptable for Ja
Re° much less than unity.
We„ 2.45 for Re, < 1250, (5a) During condensation inside horizontal, smooth tubes at low
Suf(l + 1.09Xr9)0
vapor velocities, gravitational forces, which tend to pull conden-
Re 0.79 v 0.157 sate down the tube wall, are much stronger than vapor shear
We,„ = 0.85 forces, which tend to pull the condensate in the direction of the
Su° (l + 1.09X?,039)0-4
3
mean flow. Thus, a condensate film forms on the top of the tube
and grows in thickness as it flows around the circumference. The
for Re, > 1250. (5b) bottom portion of the tube is filled with a liquid pool that
transports the condensed liquid along the tube in the direction
Based on visual observations from various studies, Soliman of the mean flow. This heat transfer mechanism is very similar
(1986) concluded that annular flow was always observed for to that in external falling-film condensation for which extensive
WeM < 20, and pure mist flow with no stable liquid film on theory has been developed.
the wall was always observed for We.„ > 30. The region of Chato (1962) concentrated on stratified flows with low vapor
Wevo between 20 and 30 was reported to be a mix of annular velocities. He developed a similarity solution for the condensate
and mist flow, called annular-mist flow. film which was patterned after Chen's (1961) analysis of falling
film condensation outside of a horizontal cylinder. He applied
Gravity-Driven Condensation. The gravity-driven flow
this solution to the upper portion of the tube where falling-film
regimes as discussed in this report include the stratified, wavy,
condensation existed (i.e., to the liquid pool on the bottom).
and slug flow regions. These regimes are lumped together pri-
To predict the depth of the liquid pool, he developed a separate
marily because the dominant heat transfer mechanism in each
model for it based on open channel hydraulics. Both his analyti-
regime is conduction across the film at the top of the tube. This
cal model and experimental results for R-113 showed that the
type of condensation is commonly referred to as film condensa-
depth of the liquid level was relatively constant. This allowed
tion. Analytical studies of condensation began with Nusselt's
his heat transfer data to be approximated quite well by the
pioneering analysis (1916) of gravity-driven condensation of a
following Nusselt type correlation:
pure component on a vertical plate. He derived the local and
mean Nusselt numbers. The mean Nusselt number at x = L is
Ps)gi'igD2
given by Nu = — = 0.555 Piipi ~ (11)
k,n,{TsM - Ts)
hi 3
g(Pi ~ pg)iisL
= — = 0.943 (6)
NUi ktv,(Tm - Ts) The constant 0.555 is 76 percent of the value of 0.728 for
k, external condensation on a cylinder. This decrease in heat trans-
fer is due to the thickness of the liquid pool on the bottom of
The bracketed term in Eq. (6) can be expressed in dimensionless the tube which reduces the heat transfer to negligible amounts.
form as Jaster and Kosky (1976) proposed a correlation similar to
Ga*Pr, Chato's for stratified flow condensation. To account for the
Nu, = 0.943 (7) variation of the liquid pool depth in a manner consistent with
Ja, pressure-driven flow, they replaced the constant in the Chato
correlation with a function of the void fraction, a. This resulted
Dhir and Lienhard (1971) devised a simple way to extend
the analysis for the vertical wall to arbitrary axisymmetric bod-
Pg)gi'igD3
ies. They showed that the local Nusselt number can be predicted Nu = — = 0.728a 3M Piipi ~ (12)
by replacing "g" in Eq. (6) with an effective acceleration of kiHi(TSM - Ts)
gravity:
They recommend using Zivi's (1964) correlation for the void
x{gR)A fraction:
Uff • (8)
P gU3R4,: dx 1 - X (Ps
Jo 1 + (13)
Pi
In Eq. (8), R{x) is the local radius of curvature and g(x) is
the local gravity component in the x-direction. For the horizontal
cylinder, the effective gravity can be evaluated numerically and Jaster and Kosky's correlation overpredicts the Chato correla-
averaged over the circumference of the tube to show tion for all qualities greater than about 0.2. It had a mean devia-
tion of 37 percent with their own data, which it appeared to
GaD*Pr, consistently overpredict.
Nu = 0.729 (9) The correlations of Chato, and Jaster and Kosky, both neglect
Ja,
the heat transfer that occurs in the liquid pool at the bottom of
the tube. Chato showed that this heat transfer was normally
The subscript in GaD indicates that the diameter is the length negligible compared to that through the upper part of the tube
scale. based on conduction only. This assumption is reasonable for
Based on integral analyses, Bromley (1952) and Rohsenow his low speed stratified flows, but might not be for higher mass
(1956) corrected for the assumption of a linear temperature flux, low quality situations where wavy or stratified flow could
profile. This is equivalent to including the convective terms prevail in the presence of substantial convective heat transfer
which were originally neglected in the energy equation. Once in the bottom of the tube. Rosson and Myers (1965) collected
Bromley included cross-flow terms which were neglected in his experimental data in what they called the intermittent flow re-
original analysis, he and Rohsenow agreed that this effect could gime, which included stratified, wavy, and slug flows. They
be corrected for by replacing the latent heat in Eq. (6) by a measured the variation of heat transfer coefficient with angle
modified latent heat given by around the tube and found that the heat transfer coefficient

196 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 continuously decreased from the top to the bottom of the tube. Two-Phase Multiplier Correlations. The simplest method
Their data suggested that filmwise condensation occurred in the of heat transfer prediction in the annular flow regime is the
top of the tube with superimposed effects of vapor shear. They two-phase multiplier approach. This approach was pioneered for
proposed replacing the constant in the Nusselt's solution with predicting convective evaporation data by Denglor and Addoms
an empirically determined function of vapor Reynolds number: (1956), and was adapted for condensation by Shah (1979).
The theoretical hypothesis is that the heat transfer process in
PiiPi ~ Ps)gi'isD3 annular two-phase flow is similar to that in single-phase flow
Nutop = 0.31 Reg1 (14) of the liquid (through which all of the heat is transferred), and
thus their ratio may be characterized by a two-phase multiplier.
This reasoning is in fact very similar to that of Lockhart and
In the bottom of the tube, they postulated forced-convective Martinelli (1947), who pioneered the two-phase multiplier ap-
heat transfer. Using a heat and momentum transfer analogy, proach for predicting two-phase pressure drop. The single-phase
they recommended the following correlation: heat transfer coefficients are typically predicted by modifica-
0/,/,v8 Re, tions of the Dittus and Boelter (1930) correlation, which results
NubM = —F (15) in the following form:
| \ , In (1 + 5 P r , ) l
5
1/ ' Pr, Nu = 0.023 Re? (19)
Ps M?
where
where m i s a constant between 0.3 and 0.4, and F is the two-
1 + 11 (16) phase multiplier. While the two-phase multiplier can depend on
Xu more dimensionless groups than those indicated in Eq. (19),
XV
Rosson and Myers defined a parameter, 0, that represented the groups shown are the most prevalent. The type of single-
the fraction of the tube perimeter over which filmwise condensa- phase correlation shown is valid for turbulent flow, and is based
tion occurred. They recommended predicting the value of j3 as primarily on an analogy between heat and momentum transfer.
follows: It then follows that the same assumptions which are made in
its development are implicitly included in the form given by
Reg 6 Re?" Eq. (19).
P = Reg1 if < 6.4 X 10" (17a)
Ga The most widely cited correlation of the two-phase multiplier
type is that of Shah (1979). It was developed from his observa-
1.74 x 1(T5 Ga Reg 6 Re? tion that the mechanisms of condensation and evaporation were
0= if > 6.4 X 10" 5 . {lib)
VRe„ Re, Ga very similar in the absence of nucleate boiling. With this idea,
he set out to modify the convective component of his flow
Then, the circumferentially averaged Nusselt number was given boiling correlation for use during condensation. The form of
by his correlation is
Nu = /3Nu top + (1 - £ ) N u b o t . (18) 3.8
Nu - 0.023 Re? 8 Pr? 1 + (20)
Rosson and Myers compared their predicted values to their own
experimental data for acetone and methanol, and the agreement
was reasonable. A large amount of scatter was inherent due to
The bracketed term is the two-phase multiplier. It properly ap-
inaccuracies in their experimental technique, so it is difficult to
proaches unity as x approaches 0, indicating that it predicts the
discern whether the deviations were due to theoretical deficien-
single-phase liquid heat transfer coefficient when all liquid is
cies or experimental scatter.
present. As the reduced pressure is increased, the properties of
Tien et al. (1988) presented an analysis for gravity-driven the liquid and vapor become more alike and the two-phase
condensation that they proposed to be valid for stratified, wavy, multiplier decreases. Thus, the qualitative behavior is as ex-
and slug flow. Their analysis was similar to that of Rosson and pected. Shah compared his correlation with data from the open
Myers, although more deeply rooted in conservation equations literature for refrigerants, water, methanol, benzene, toluene,
than empirically determined expressions. This analysis ap- trichloroethylene, and ethanol. The mean deviation between his
proaches the correct values in the asymptotic limits. That is, correlation and the experimental data (474 points) was 17 per-
for a quality of zero it predicts a single-phase liquid Nusselt cent. Although he classified his correlation as a ' 'generalized''
number, and for situations where stratified flow exists rather correlation which was independent of flow regime, it is catego-
than slug flow, it reduces to the form of Rosson and Myers. To rized herein as an annular flow correlation since it will later be
use the Tien model, six simultaneous nonlinear equations must shown to be appropriate only for this flow regime.
be solved. Although novel and well structured, the technique is
probably too involved for a practical design correlation. Cavallini and Zecchin (1974) used the results of a theoretical
annular flow analysis to deduce the dimensionless groups that
Annular Flow Condensation. The annular flow regime should be present in an annular flow correlation. They then used
represents the situation where the interfacial shear stresses dom- regression analysis to justify neglecting many of the groups that
inate the gravitational forces and result in a nearly symmetric did not appear in their empirically developed correlation, which
annular film with a high speed vapor core. As might be ex- can be shown to be of the two-phase multiplier form by writing
pected, the heat transfer characteristics in the annular flow re- it in the following way:
gime differ significantly from those in the gravity-driven re-
gime.
A variety of approaches for predicting heat transfer during
annular flow condensation have been developed. While these
Nu = 0.023 Re?-' 2.64 1 +
)"te)]J
approaches can be divided into many different categories, they (21)
will be reduced to the following three for the purposes of this
review: (1) two-phase multiplier approaches; (2) shear-based Here, the bracketed term represents the two-phase multiplier.
approaches; and (3) boundary-layer approaches. These three Cavallini and Zecchin compared their correlation with experi-
categories of correlations will be discussed separately. mental data from six different studies with R-113, R-12, and

Journal of Heat Transfer FEBRUARY 1998, Vol. 120 / 197

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 R-22. The standard deviation between their correlation and the ox
experimental data sets varied from 8 percent to 47 percent. The Nu = 0.018 (El\ (ih
\f V>i
mean deviation of all the data combined was 30 percent, if each
point was given equal weight, and 22 percent, if each set was
given equal weight. XRe?2[Re,„-Re,]a7Pr?'' (23)
Chen et al. (1987) made no comparison with horizontal flow
Shear-Based Correlations. The use of shear-based correla- data for the annular flow regime.
tions for annular flow condensation dates back to the early
work of Carpenter and Colburn (1951). They argued that the
resistance to heat transfer in the turbulent liquid flow was en- Experimental Facility
tirely inside the laminar sublayer and that the wall shear stress Refrigerant and Water Circuits. Detailed descriptions of
was composed of additive components due to friction, accelera- the facility were given by Dobson (1994) and Dobson et al.
tion, and gravity. Although it was later pointed out by Soliman (1994a, 1994b). The refrigerant flow in the loop is driven by
et al. (1968) that their equation for the accelerational shear a variable-speed gear pump. Because the gear pump requires no
component was incorrect, the framework that they established lubrication, the loop can be operated in an oil-free environment.
at a relatively early point in the history of forced-convective Following the pump, the refrigerant flow rate is measured by
condensation remains useful. one of two flow meters. A Coriolis-type mass flow meter is
Soliman et al. utilized the framework established by Carpen- used for all flows that fall inside of its 0.90 kg/min range. A
ter and Colburn to develop their own semi-empirical heat trans- positive-displacement flow meter with a 0-3.8 1/min range is
fer correlation for annular flow. They accepted the basic validity used at high flow rates. A serpentine preheater is used to heat
of the form presented by Carpenter and Colburn, but refined it the subcooled liquid refrigerant to the desired state at the inlet
in several ways. First, they corrected the equation for predicting of the test condenser. Immediately after the preheater is a hori-
the wall shear stress due to phase change. Their corrected form zontal, adiabatic section. This adiabatic section provides fully
showed that this component was normally negligible in compar- developed flow to the test section. In this section, heat is trans-
ison to the frictional component except at low qualities. Second, ferred from the two-phase refrigerant to water in a counterflow
they implemented an improved correlation for the frictional heat exchanger. The flow regime is observed visually at the
pressure drop. Third, they used data from other researchers to inlet and outlet of the test condenser in sight glasses of inner
determine new values of the constants. Neglecting the gravita- diameters that match the test condenser. The ball valves in the
tional term, which is appropriate for horizontal flow, the Soli- loop create no disturbance in the flow in the open position.
man correlation can be written as After exiting the test condenser, the refrigerant is returned to a
subcooled liquid state by extracting heat from it in a water-
Pi cooled after-condenser and in a second water-cooled heat ex-
Nu = 0.036 Re to Pr?-65 ^ changer. A receiver between the after condenser and the heat
exchanger contains a two-phase mixture of the refrigerant. The
temperature of the receiver and, consequently, the pressure of
the entire system is maintained at the required level by a con-
*./M« + Bo£^)"^, trolled-temperature water bath around the receiver. In earlier
experiments, the pressure was maintained by an externally pres-
surized bladder. The refrigerant then passes through a filter and
where a, = * ( 2 - y) - 1 is recirculated through the pump.
a2 = 2(1 - x) The test section is cooled by water from the building supply
flowing counter currently through a concentric annular tube.
a, = 2(y- l ) ( x - 1), The temperature of the cooling water is set by a 1.5 kW, variac-
controlled heating element. At various axial and circumferential
locations (dependent on tube size) along the outer surface of
a4 = 3 + 2x
the test section, thermocouples were soldered into machined
x
grooves and mounted so that they did not protrude into the water

a5 - y
H (226-/)
flow. This arrangement provided an accurate measurement of
the tube wall temperature with negligible fin effects. Static pres-
sure taps at either end of the test section and at other, strategic
points in the loop provide direct or differential pressure read-
interface velocity ings.
7 = 1.25 for turbulent liquid. (22g)
mean film velocity Three tube sizes were chosen. The parametric ranges used
are summarized in Tables 1-3.
Soliman et al. (1968) compared the predictions of their corre-
lation to data for steam, R-113, R-22, ethanol, methanol, tolu- Instrumentation and Data Acquisition Systems. The
ene, and trichloroethylene. The agreement was correct in trend,
measurement devices used in this study included thermocou-
although even on log-log axes the deviations appeared quite
large. No statistical information regarding deviations was given.
ples, absolute pressure transducers, differential pressure trans-
Chen et al. (1987) developed a generalized correlation for
vertical flow condensation which included several effects com- Table 1 Condensation test envelope for the 4.S7 mm i.d. tube
bined with an asymptotic model. They stated, as did Carey Quantity Range
(1992), that their correlation for the shear-dominated regime Mass flux 75-650 kg/m^-s
was also appropriate for horizontal flow. Their correlation used Average quality 10%-90%
the general form of Soliman et al. (1968) with two modifica- Heat flux 5-15 kW/m 2
tions. First, the acceleration terms were neglected since they Tsat -Tg 1.8-4.0 °C
were normally negligible at the relatively high quality condi- Saturation temperature 35 - 60 °C
tions at which annular flow was expected to exist. Second, the Refrigerant R-134a, R-12, R-22, R-32/R-125
pressure drop model was replaced by one from Dukler (1960). Reduced pressure 0.21-0.40
The final result can be written as Flow regimes Wavy, annular, annular-mist

198 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 2 Condensation test envelope for the 3.14 mm i.d. tube of the test section. For the two-phase experiments, the heat
transfer coefficient was corrected for entrance effects, as dis-
Quantity Range
cussed by Dobson (1994) and Dobson et al. (1994a, b). The
Mass flux 75-800 kg/m 2 -s
significant components that contributed to uncertainty in the
Average quality 10%-90%
Heat flux 5-15 kW/m 2
two-phase experiments are the wall temperature, the saturation
1.8-4.0 °C
temperature, the inlet and outlet water temperatures, the heat
Tsat ~Ts
Saturation temperature 35-45°C
loss to the environment, the surface area of the test section, and
Refrigerant R-134a, R-22, R-32/R-125 the water flow rate. The fractional uncertainties of the heat
Reduced pressure 0.21-0.57 transfer coefficients were well under 0.15 for all but about a
Flow regimes Wavy, annular, annular-mist dozen tests, the highest value being 0.19. The maximum uncer-
tainty in the value of the quality was 0.1, with the vast majority
being at or under 0.03.
ducers, watt-hour transducers, and flow meters. The uncertain-
ties of the refrigerant temperature measurements with calibrated
type-T copper-constantan thermocouples were ±0.2°C, except Flow Regime Observations
at the heater inlet where they were ±0.3°C. The uncertainties This section presents the experimental flow regime observa-
of the water temperature measurements were ± O.TC. tions from this study and compares them to various flow regime
Absolute and differential pressures were measured through- predictors from the literature. As described before, these obser-
out the refrigerant loop with calibrated strain-gage-type pressure vations were gathered in tubes of 3.14, 4.57, and 7.04 mm inner
transducers. The uncertainties of the absolute pressure measure- diameters. The refrigerants used were R-134a, R-22, and 60/
ments were ±7 kPa at the test section inlet and ±35 kPa at 40 and 50/50 blends of R-32 and R-125. The properties of the
the other locations. The pressure drop across each of the test- 60/40 and 50/50 blends of R-32 and R-125 were so similar to
condensers was measured with differential pressure transducers. each other and so different from the other fluids that they will
Different combinations of transducers were used for experi- both be referred to only as R-32/R-125 for the remainder of
ments in the different tubes. Three kinds of differential pressure this paper.
transducers were used, the uncertainties of which were ±0.2
kPa for the 0 - 7 kPa unit, ±0.5 kPa for the 0-35 kPa unit, and Effect Of Mass Flux and Quality. The dominant factors
±1 kPa for the 0-105 kPa unit. affecting the flow regimes were the mass flux and the quality.
The power input into the refrigerant heater was measured At the lowest mass flux in the study, 25 kg/m 2 -s, smooth-
using two watt-hour transducers. One transducer measured the stratified flow was observed across the entire range of quality.
power input into the nine heaters that were controlled by the As the mass flux was increased to 75 kg/m 2 -s, interfacial waves
variac, while the other transducer measured the power input developed and wavy flow was observed for the entire range of
into the remaining eleven heaters. Both transducers produced a quality. For these two mass fluxes, the flow regime was not
pulse output for each 100 mJ of energy input. These pulses affected by changes in diameter or refrigerant.
were counted over the monitored period and divided by the At mass fluxes of 150 and 300 kg/m 2 -s, several different
elapsed time to yield a power measurement. The uncertainty of flow regimes were observed as the quality was changed. Slug
these devices was factory estimated at 0.2 percent of the reading. flow was observed at low qualities, followed sequentially by
They were indirectly verified by energy balance checks across wavy, wavy-annular, and annular flow. At these mass fluxes,
the refrigerant heater. the tube diameter and fluid properties influenced the range of
quality over which each of the flow regimes occurred. These
Two flow meters were used to monitor the refrigerant flow
effects are discussed later.
rate. For flow rates below 0.015 kg/s, a Coriolis-type mass flow
meter was used with an uncertainty of ±0.1 percent of the At the highest mass fluxes in this study, 500, 650, and 800
reading. For higher flow rates, a calibrated positive-displace- kg/m 2 -s, the flow regimes included slug flow at low quality,
ment flow meter was used to measure the volumetric flow rate followed by wavy-annular, annular, and annular-mist flow as
with an uncertainty of 0.75 percent of its full-scale reading. The the quality was increased. Although the quality range over
range of this flow meter was 0-3.8 1/min, which corresponds which each of these regimes occurred depended somewhat on
to a mass flow range of approximately 0-0.07 kg/s with R- fluid properties and diameter, annular or annular-mist flow nor-
134a at 35°C. The lowest flow rate at which this flow meter mally occurred for 70 percent to 80 percent of the condensation
was used, which was over 20 percent of its full scale range, path. These were the only mass fluxes where significant entrain-
corresponded with the upper limit of the Coriolis-type flow ment occurred. Detecting any details of the mist was difficult
meter. because it normally flowed inside a wall film, but it was clearly
visible at high qualities. The amount of mist showed significant
The flow rate of the water was determined by weighing a temporal variations, with periods of no entrainment being fol-
time sample using a graduated cylinder and stopwatch. The lowed by periods with a rather dense mist in the core flow. This
estimated uncertainty of these mass measurements is ± 1 gm. is consistent with the findings reported by Barnhart (1992),
All data were logged into a computer using data acquisition who found that droplets formed at slug initiation propagated far
hardware and software. The data acquisition program displayed downstream. Perhaps the most significant observation at these
all measured quantities on a graphical display in real time. In
addition, secondary quantities such as refrigerant quality, heat
transfer coefficient, and various dimensionless groups were pre-
sented on the screen. This allowed the user to closely monitor Table 3 Condensation test envelope for the 7.04 mm i.d. tube
the system parameters.
Quantity Range
Uncertainties in the experimental data were calculated using
Mass flux 25-800 kg/m2-s
the familiar methods described by Moffat (1988). The data
Average quality 10%-90%
from the present study were classified as single sample. The Heat flux 5-15 kW/m 2
uncertainty of each measured quantity consisted of the uncer-
Tsat -Ts 1.8-4.0 °C
tainty of the measurement device, the uncertainty of the data- Saturation temperature 35-45°C
acquisition system, interactions between the sensor and the ex- Refrigerant R-134a, R-22, R-32/R-125
perimental system, and conceptual errors. Two uncertainties Reduced pressure 0.21-0.57
were of interest in the present study, (1) the measured heat Flow regimes Stratified, wavy, annular, annular-
transfer coefficient, and (2) the quality at the inlet and outlet mist

Journal of Heat Transfer FEBRUARY 1998, Vol. 120 / 199

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 4 Property comparison of R-134a, R-22, and 50/50 R-32/R-125 1
Xslag
Property R-134a R-22 R-32/R-125 ~ 1 + 3.27P re S' 547 ' (24)
35*C/45*C 35 'C/45 "C 35X/45-C
PI (kg/m3) 1167/1126 1148/1073' 1613/956 This equation predicts that the most slug flow will occur for
p q (kg/m3) 43/56 58/77 86/114 R-32/R-125, followed by R-22 and R-134a. For R-32/R-125
PI /pq 27.2/19.9 19.9/13.9 11.8/8.4 at 45°C; for instance, Eq. (24) predicts that slug flow will be
W (ftPa-s) 179/163 190/179 135/126 observed for all qualities below 19 percent. For R-134a at 35°C,
|ij, (uPa-s) 12/13 14/15 14/15 it predicts a slug flow region confined to qualities below 11
w /ua 14.4/12.8 13.6/12.2 9.5/7.9 percent. This lengthening of the slug flow region with increases
(W /(iq) al (Pq / P l ) a 5 0.250/0.289 0.291/0.344 0.363/O.425 in reduced pressure was observed experimentally.
a (mN-m) 6.82/5.<S0 6.76/5.38 5.93/4.26
The tube diameter also affected the flow regime transitions.
Pred 0.218/0.285 0.268/0.343 0.430/0.549
As the tube diameter was reduced at a fixed mass flux, the
transition from wavy flow to wavy-annular flow, and the transi-
tion from wavy-annular flow to annular flow shifted to lower
high mass fluxes was that pure mist flow (without a stable wall qualities. This trend is summarized in Table 5, which shows
film) was observed only at qualities over 90 percent, and even the estimated quality at which these transitions occurred in each
then never at the outlet sight glass. Thus, what was mist flow of the tubes for R-134a at 35°C. The quality where the wavy
in the adiabatic section appeared to develop a stable liquid film to wavy-annular transition occurred shifted the most at a mass
as soon as condensation began. Over the range of parameters flux of 150 kg/m 2 -s, where it decreased from 37 percent in the
tested in this study, then, Soliman's (1986) idea of pure mist 7.04 mm i.d. tube to 23 percent in the 3.14 mm i.d. tube. The
flow as a distinct condensation regime was not verified. quality at which the wavy-annular to annular transition occurred
decreased from 65 percent to 45 percent at this same mass flux.
Effects of Fluid Properties and Tube Diameter. Al- At higher mass fluxes most of the quality range was associated
though mass flux and quality were the primary factors affecting with annular flow even in the 7.04 mm i.d. tube. Thus, the
the flow regimes, the thermophysical properties of the fluids diameter effects were less pronounced.
and the tube diameter played a secondary role. Their influence Pure stratified flow was observed only in the 7.04 mm i.d.
was most evident at the intermediate mass fluxes of 150 and tube, primarily because this was the only tube where a mass
300 kg/m 2 -s, where the mass velocities were not so high or flux of 25 kg/m 2 -s could be stabilized. The cause of the instabili-
low as to allow one flow regime to clearly dominate over an- ties could not be pinpointed, but it was probably due to interac-
other. The primary fluid properties that affect the flow regimes tions among the components of the entire loop. Thus, no obser-
are the vapor and liquid densities and viscosities, the ratios vations on the effect of diameter could be made with respect
between these quantities, and the surface tension. Values of to this transition. Entrainment was only observed at mass fluxes
these properties for the three fluids used in this study are listed of 500 kg/m 2 -s and above in each of the tubes. Although visu-
in Table 4 for saturation temperatures of both 35°C and 45CC. ally estimating the amount of entrainment was difficult, it ap-
Much of the variation in the fluid properties can be explained peared that entrainment was less pronounced in the smaller
by the reduced pressure. At high values of reduced pressure, tubes.
the liquid and vapor phases are more similar. This dictates
higher values of vapor density and viscosity and lower values
of liquid density and viscosity. The surface tension, which has Comparison With Flow Regime Maps
the interpretation of the work required to increase the interfacial A major goal of this study was to eliminate the need for
area (Carey, 1992), also decreases as the reduced pressure in- testing with every refrigerant that might be considered as a
creases and the phases become more similar. At a given temper- candidate in the future. Since the flow regimes clearly affect
ature, the reduced pressure is lowest for R-134a, followed by the heat transfer processes, it was necessary to evaluate the
R-22 and R-32/R-125. The values of the other properties follow potential of existing flow regime prediction techniques for con-
accordingly. densation conditions. Evaluations will be made herein for the
The most noticeable effect of the property differences on the maps of Mandhane et al. (1974), Taitel and Dukler (1976),
flow regimes was the extent to which annular flow prevailed and Soliman (1982, 1983). The potential role of surface tension
over wavy or wavy-annular flow. At a given mass flux where will also be briefly explored for the data in the 3.14 mm i.d.
wavy, wavy-annular, and annular flows all occurred, the size tube. The baseline case that will be used to evaluate each of
of the quality range occupied by annular flow was greater at these maps will be R-134a in the 7.04 mm i.d. tube at 35°C.
lower reduced pressures. Thus, the most annular flow was ob- The validity of the maps for other diameters, fluids, and temper-
served for R-134a at 35°C, and the least annular flow was ob- atures will be addressed by examining how their predictions
served for R-32/R-125 at 45°C. The differences in the observed change with diameter and reduced pressure.
flow regimes for these two cases were significant. At a mass
Mandhane Map. The predictions of the original Mandhane
flux of 300 kg/m 2 -s, the transition to annular flow that occurred
map are grossly different from the experimental observations,
at 25 percent quality for the low-temperature R-134a case was
but this map was developed primarily from air-water data. At
delayed until 50 percent for the high-temperature R-32/R-125
atmospheric pressure and temperature, the density of air is about •
case. Even at the highest mass flux of 650 kg/m 2 -s, significant
1.15 kg/m 3 . This is between 40 and 100 times lower than the
stratification was observed across the entire quality range for
vapor densities from this study. Hanratty (1994) pointed out
the high-temperature R-32/R-125 case. For the low-temperature
that the Mandhane map could be brought into better agreement
R-134a case at the same mass flux, a symmetric annular film
was established at 15 percent quality. Smaller changes in re-
duced pressure, such as changing the temperature of the same
fluid, produced similar but smaller changes. Table 5 Effect of tube diameter on R-134a flow regime transitions at
35°C
Another item that is affected by the reduced pressure is the
length of the slug flow region. Taitel and Dukler's (1976) crite- lube Wavy to wavy-annular Wavy-annular to annular
rion for slug flow is that X„ > 1.6. Combining this with the i.d. (mm) xforG=150 xforG=300 xforG=150 xforG=300
kg/m2-s kg/m2-s kg/m2-s kg/m^-s
fact that (pjp^dj,,/^)01 s 0.551 P°J>2 (Wattelet, 1994;
Wattelet et al., 1994) yields an algebraic expression for the
3.14 6.i3 Tus1 \4$ 6.29
4.57 0.35 0.25 6.56 038
quality below which slug flow is expected to occur: 7.04 0.37 0.25 0.65 0.41

200 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1
DISPERSED ' '1 y "I DlSFEKSEB
•=• 10
T BUBBLE, 1 •=• 10 BUBBLE,
: ELONGATED 1 „...„ * ELONGATED
SLUG
; BUBBLE 1 BUBBLE

i> r V D—{>•,
ykMiST:
1 1
I ^yt^^H
1 "
•3

a o.i r STRATIFIED \V
WEI I]
IT 3 0.1
; u

I 0.0 •
^>OCC*t\
' ' ' ' •"! 1 1 1 I 1 11 ID
\
\\ i Jul
L_ 1 l_LL.LI.li
<_
8.
s
C/i 0.0 r -
0.1 1 10 100, 1000 0.1 1 10 100 1000
Superficial vapor velocity , j [ft/s]
Superficial vapor velocity , j [ft/s]
O— G=25 kg/m2-s — • — G=150 kg/m 2 -s — A
— G=500 kg/m 2 -s
• G=25 kg/m-s • G=150 kg/m-s " • G=500 k g / m - s
2 D 2
~ * — G=75 kg/m -s — — G=300 kg/m -s 1 G=650 kg/m 2 -s
• G=75 kg/m 2 -s • G=300 kg/m 2 -s" " G=650 kg/m 2 -s

Fig. 1 Predicted flow patterns for R-134a on the Mandhane et al. (1074)
Fig. 2 Predicted flow patterns for R-32/R-125 at 45 C on the Mandhane
map using corrected superficial velocities
et al. (1974) map using corrected superficial velocities

with refrigerant data by accounting for the large differences in creased amount of annular flow from the 3.14 mm i.d. tube
vapor density. To effect this change, the superficial vapor veloc- would not be predicted using the Mandhane map.
ity of the refrigerant was multiplied by the square root of the The effect of reduced pressure is addressed in Fig. 2, which
ratio of the vapor density to the density of atmospheric air. The plots the condensation paths for R-32/R-125 at 45°C on the
resulting quantity is referred to herein as a modified superficial corrected Mandhane map. This combination of refrigerant and
vapor velocity: temperature was selected for comparison because it corre-
(25) sponded to the highest reduced pressure tested in this study. The
JpJPah- trends predicted by the Mandhane map regarding the change in
From a physical standpoint, this modified superficial velocity fluid properties are consistent with the experimental observa-
is proportional to the square root of the vapor kinetic energy. tions. Compared to the predictions of the low reduced-pressure
This simple change was made in both the vapor and liquid case (Fig. 1), the lines for a given mass flux were shifted left
superficial velocities and the results are plotted on the Mandhane for the high reduced-pressure case (Fig. 2). The mass flux of
coordinates in Fig. 1. The condensation paths are for mass fluxes 150 kg/m 2 -s was correctly predicted to lie entirely in the wavy
ranging from 25 kg/m 2 -s to 650 kg/m 2 -s of R-134a. The points flow regime. The quality at which annular flow was expected
on the map cover a range of quality from 5 percent to 95 percent to begin for the high mass flux cases was correctly increased
in 5 percent increments. Decreasing the quality corresponds to over the R-134a case, although a portion of the slug flow region
moving right to left on the map. was again occupied by either wavy or wavy-annular flow.
Making this simple correction to the Mandhane map pro-
duced much better agreement with the experimental data. The Taitel-Dukler Map. Figure 3 traces condensation paths for
25 kg/m 2 -s data were still correctly predicted to be in the stra- the baseline R-134a case on the Taitel-Dukler map. The points
tified flow regime. The data at 75 and 150 kg/m 2 -s were moved again range from 5 percent to 95 percent quality at 5 percent
from the being exclusively in the stratified flow regime to being increments, with increasing quality corresponding to moving
predominantly wavy flow, consistent with experimental obser- left on the map. Since the Taitel-Dukler map uses different
vations. Furthermore, the highest quality data at 150 kg/m 2 -s coordinates for different transitions, in general three different
were correctly predicted to be on the line between wavy and lines would be needed for each mass flux. However, for the
annular flow. The data at the high mass fluxes of 300, 500, and range of parameters in this study, no dispersed bubble flow was
650 kg/m 2 -s were all predominantly in the annular or annular-
mist flow regime. The only obvious deficiency was an overpre-
diction in the length of the slug flow region, particularly at 300 ANNULAR DISPERSED
kg/m 2 -s. Barnea et al. (1980), however, also noted that their i 11mii i iiiinil i inun
wavy-annular flow data occupied the slug flow region on the 10*
-Fw,G=650DISpERSED
Mandhane map. They suggested this was due to different inter-
pretations of what constituted slug flow, and that data observed H IOOO
to be in the wavy-annular flow regime should be classified as
slug flow for comparisons with the Mandhane map (or con-
versely that points lying near the right edge of Mandhane's slug 100 E*
flow region might be expected to correspond to wavy-annular
flow). Incorporating this suggestion results in excellent agree-
ment with the Mandhane map. H 10
Having established the basic validity of the corrected Mand-
hane map for the case of R-134a in the 7.04 mm i.d. tube,
questions remain about how it incorporates the effects of diame- 0.001 1
ter and reduced pressure. Neither coordinate on the Mandhane 0.001 0.01 0.1 10 100 1000
map includes a diameter term, and investigators as early as
Taitel and Dukler (1976) have noted systematic deviations in Fig. 3 Flow regime predictions of the Taitel-Dukler (1976) map for
the Mandhane map because of this deficiency. Thus, the in- R-134a at 35 C in the 7.04 mm i.d. tube

Journal of Heat Transfer FEBRUARY 1998, Vol. 120 / 201

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 800 is properly labeled as intermittent or annular flow in Taitel-
Dukler terminology. From a heat transfer standpoint, however,
700 the instability of the wavy flow near this boundary is less im-
600 portant than the significant stratification due to gravity. At
higher mass fluxes, the range of quality occupied by this hybrid
e 500 flow pattern becomes so small that proper classification is unim-
SLUG ANNULAR portant.
X 400
9 Of the boundaries from the Taitel-Dukler map that were used
E 300 in this study, only that between wavy and intermittent or annular
flow depends on the diameter. The parameter used for predicting
2 200 this transition, F,d, is proportional to D~os at a fixed mass flux
loo r "^fflHPTO^SW
and quality. Thus, decreasing the diameter increases the Froude
r. [STRATIFIED SK
number and decreases the mass flux at which the annular transi-
tion is expected to occur. This is consistent with the observed
0 0.2 0.4 0.6 0.8 1 trend of more annular flow in the smaller tubes. While the trend
Quality is physically correct, the predictions themselves were incorrect
when applied to the data in the 3.14 mm i.d. tube. The Taitel-
Fig. 4 Taitel-Dukler (1976) predictions on G-x coordinates for R-134a Dukler method predicts annular flow across nearly the entire
at 35°C in the 7.04 mm i.d. tube
range of quality at a mass flux of 75 kg/m 2 -s, while wavy flow
was observed exclusively at this mass flux. The predicted trend
observed or predicted. Thus, the only two parameters that were of the Taitel-Dukler map to an increase in the reduced pressure
needed were Kld and F,d, used, respectively, for predicting the was consistent with the experimental observations. For example,
stratified to wavy and wavy to annular, or intermittent, transi- the slug flow region was wider, as predicted by Eq. (24).
tions. At higher reduced pressures one would also expect the stra-
Plots of K,d versus X„ are shown for the two lowest mass tified-to-wavy and wavy-to-annular transitions to be shifted to
fluxes only, since these were the only mass fluxes at which any higher mass fluxes due to the lower vapor velocity at a given
stratified flow was predicted. The map predicted stratified flow mass flux and quality. Although this expected trend did occur,
up to 80 percent for the 25 kg/m 2 -s flux case, where stratified the observed shift was very modest. At a quality of 50 percent,
flow was observed across the entire range of quality. For a mass for instance, the wavy-to-annular transition shifted from a mass
flux of 75 kg/m 2 -s, the Taitel-Dukler map correctly predicted flux of 75 kg/m 2 -s for the R-134a case to 83 kg/m 2 -s for the
nearly all wavy flow. R-32/R-125 case. This small shift was due to two opposing
Plots of F,d versus X„, used for predicting the transition from trends brought about by the changes in fluid properties. At
wavy to annular or intermittent flow, are shown for all the mass constant mass flux and quality, the value of F,d for the high-
fluxes in the study. It is interesting to note that the transition temperature R-32/R-125 refrigerant was only 71 percent of
boundary is nearly parallel to the condensation path lines. This that for the low-temperature R-134a case, moving the curve
implies that a particular mass flux, to a close approximation, downward relative to the transition boundary on the Taitel-
will be either wavy or annular across the entire range of quality. Dukler map. However, the value of X„ for R-32/R-125 was 70
This causes no problems for the 25 and 75 kg/m 2 -s mass fluxes, percent greater than that for the R-134a case at constant quality,
where the map correctly predicts that no annular flow will be moving the curve to the right on the Taitel-Dukler map and
observed. For mass fluxes slightly above the transition boundary therefore closer to the boundary. At 50 percent quality, for an
(in the annular regime), the slug flow regime at low qualities example, these two effects result in the low-temperature R-
was always followed by regions of wavy, wavy-annular, then 134a case being only 11 percent higher relative to the transition
annular flow. The quality range over which the wavy and wavy- boundary than the high-temperature R-32/R-125 case. From a
annular flow regimes occurred decreased as mass flux was in- practical standpoint, this predicted shift in the transition bound-
creased. At a mass flux of 150 kg/m 2 -s, the wavy regime per- ary is insignificant. The apparent discrepancy between the pre-
sisted up to 25 percent quality, followed by wavy-annular flow dictions and the observations at mass fluxes slightly above the
up to 65 percent quality, and annular flow at higher qualities. annular flow boundary was also present with all the refrigerants
The predictions of the Taitel-Dukler map are translated onto used in this study. For example, at a mass flux of 150 kg/m 2 -
mass flux-quality (G-x) coordinates in Fig. 4. At low mass s wavy flow persisted at qualities up to 50 percent, while the
fluxes, stratified flow is predicted across the entire range of Taitel-Dukler map predicted annular flow above 20 percent
quality. At slightly higher mass fluxes, wavy flow is predicted quality.
across most of the quality range with a small amount of stratified
flow at low quality. At mass fluxes above 140 kg/m 2 -s, slug Soliman Transitions. Soliman (1982, 1986) developed
flow is predicted for qualities below 11.8 percent and annular criteria for two flow-regime transitions for condensation: (1)
flow is predicted for all higher qualities. It is close to this wavy or slug flow to annular flow, and (2) annular flow to mist
boundary that the observed flow regimes deviated most signifi- flow. His transition criteria are displayed on G-x coordinates
cantly from the Taitel-Dukler predictions. The length of the in Fig. 5. Several interesting observations can be made from
slug flow region was underpredicted, and this was consistently comparing the predictions of Soliman to those of Taitel and
followed by some wavy or wavy-annular flow that was not Dukler. First, at high qualities, Soliman's prediction of the
predicted by the Taitel-Dukler map. wavy-to-annular transition agrees fairly well with that of Taitel
The apparent discrepancy between the observed and predicted and Dukler. Unlike the Taitel-Dukler map, however, Soliman
flow regimes at mass fluxes slightly above the annular boundary predicts a wavy region at low qualities over the entire mass
of the Taitel-Dukler map is largely due to differences in termi- flux range of this study. This occurs partially because Soliman
nology. In an early experimental verification of the Taitel- lumps the wavy and slug flow regions together. At high mass
Dukler map, Barnea et al. (1980) used the term wavy-annular fluxes, the region predicted to be wavy flow by Soliman corre-
flow to refer to a hybrid pattern observed at the lowest gas rates sponds almost exactly with the slug flow region on the Taitel-
where the slug-to-annular transition occurred. A similar regime Dukler map. At lower mass fluxes however, the region predicted
has been termed "proto-slug" flow by Nicholson et al. (1978), to be occupied by wavy flow extends to higher qualities than
and "pseudo-slug" flow by Lin and Hanratty (1989). Since the slug flow boundary on the Taitel-Dukler map. This is consis-
this pattern occurs after the wavy flow has become unstable, it tent with the experimental data in both magnitude and trend if

202 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 [| U| I M , | I I I I M I I M I explored in Fig. 6. This figure compares the predicted wavy to
annular transition lines for R-134a at 35°C and R-32/R-125 at
•» MIST FLOW (3.14 nun) | 45°C (low and high reduced pressures). The predicted trends
are consistent with the experimental observations. The mass
flux below which no annular flow is predicted was increased
*> JvlISTFLOW(7.04mm)'
by approximately 30 kg/m 2 -s. Because testing was not per-
formed at mass fluxes between 75 and 150 kg/m 2 -s, this predic-
tion could not be verified. At the 150 kg/m 2 -s mass flux, Fig.
6 predicts a transition from wavy to wavy-annular flow at 41
percent quality for R-134a and 57 percent quality for R-32/
R-125. These predictions agree well with the experimentally
observed values, which were estimated at 37 percent and 50
percent, respectively. At mass fluxes near 1000 kg/m 2 -s, the
I I I I I I I I I I I I I I I predicted quality range for wavy flow for R-32/R-125 is very
0.2 0.4 0.6 0.8 1 near the 19 percent range for the slug flow regime predicted by
Quality Taitel and Dukler.
• WAVY TO ANN, D=7.04 mm ANN TO MIST, D=7.04 mm Because reliable surface tension data are not yet available
• WAVY TO ANN, D=3.14 mm ™ « « « ANN TO MIST, D«3.14 mm for this mixture, and drawing any conclusions concerning mist
flow from the visualization was difficult, only the lines for the
Fig. 5 Soliman (1982,1983) predicted flow regimes on G-x coordinates wavy to annular transition were included in Fig. 6. As another
for R-134a at 35°C way of assessing the effect of increasing reduced pressure, the
magnitude of the Weber number was examined as the tempera-
ture was increased for both R-134a and R-22. The Weber num-
the predicted transition line is considered to be that of wavy ber increased as temperature increased due to the reduced liquid
flow to wavy-annular flow. It was shown by Dobson (1994) viscosity and surface tension, but only slightly (less than 10
and Dobson et al. (1994a, 1994b) that the transition from wavy- percent as the temperature of both fluids was raised from 35 to
annular flow to annular flow was well predicted by a value of 55°C). This small change indicates that the decreased surface
Fr.,0 = 18, as opposed to Frs„ = 7 for the wavy to wavy-annular tension and liquid viscosity are nearly balanced by correspond-
transition. ing decreases in the density ratio. Based on these trends, one
Soliman's map also includes a distinct mist flow region, un- would expect slightly more entrainment to occur for R-32/
like the maps of Mandhane and Taitel-Dukler. According to R-125 than the other refrigerants at identical mass flux and
Soliman, mist or spray flow is a regime with all the liquid quality.
flowing as entrained droplets in the core flow and no stable film
on the wall. Annular mist flow would refer to a regime with a Potential Role of Surface Tension. In any situation where
stable liquid film on the wall and significant entrainment in the a free surface exists, the potential role of surface tension should
core flow. According to the observations made in the present not be overlooked. The analytical approach in the original
study, most of the region labeled as mist flow by Soliman's Taitel-Dukler maps clearly does ignore surface tension forces.
map would more properly be called annular-mist flow. Although The consensus of subsequent investigators is that surface ten-
the amount of entrainment was very significant, a stable liquid sion forces become increasingly important as the diameter is
film was always observed on the wall at qualities below 90 decreased, and may dominate for sufficiently small tube sizes.
percent. Even when the flow entered the first sight glass (after Of primary interest in this study was the effect on the wavy to
the adiabatic section) as mist flow, the outlet sight glass always annular transition, particularly in the 3.14 mm i.d. tube.
had annular-mist flow. This suggests that the net mass flux Galbiatti and Andreini (1992) started with Taitel and
toward the wall during condensation always results in a stable Dukler's analysis of a Kelvin-Helmholtz instability for the wavy
liquid film, no matter what the observations might indicate in to annular flow transition. In the original analysis, the stabilizing
an adiabatic section. This finding is important for interpreting effect of gravity is opposed by the effect of an accelerating
the annular-mist flow heat transfer data. vapor flowing over a wave. Galbiatti and Andreini included
surface tension forces that supplement the stabilizing gravity
If Soliman's mist flow region is interpreted as annular-mist
flow, the predictions seem quite reasonable. No mist flow is
predicted at mass fluxes below 300 kg/m 2 -s, which is consistent
with the observations. At a mass flux of 650 kg/m 2 -s, mist flow 800 i 11 i • i i i
is predicted for qualities greater than about 20 percent. This
again seems consistent with the experimental data, although a 700
great amount of subjectivity was present in categorizing a re-
gime as annular or annular mist. «r 600
The diameter effects predicted by Soliman's transition criteria t 500 ^
are also shown in Fig. 5. The lower mass flux limit at which
annular flow is predicted is relatively insensitive to the diameter % 400
change, much like the predictions of Taitel and Dukler. At E 300
mass fluxes slightly above this, however, the wavy to annular
transition line was shifted to lower qualities with decreasing
diameter. This was consistent with the experimental observa-
I 200
: WAVY FLOW
100
tions in both direction and magnitude. The predicted effect on
the mist flow regime was much more dramatic, with a significant 0 I I I I I I I I—I I I I I H — I — I -
stabilizing effect on the liquid film being predicted as the tube 0 0.2 0.4 0.6 0.8 1
diameter was decreased. This was consistent with the trend of Quality
the observations, although the transition was very difficult to 7.04 mm 1 i R-32/R-125, 45'C, 7.04 mm
detect visually.
The effect that increasing the reduced pressure has on the Fig. 6 Effect of reduced pressure on Soliman's (1982) wavy to annular
wavy to annular flow regime transition predicted by Soliman is flow regime transition

Journal of Heat Transfer FEBRUARY 1998, Vol. 120 / 203

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 1 1 1 1 1 1 •[•• i • i— i | ' r-i—i—[• i' T •• T
identical to the lower mass flux cases. As the quality was in-
A G=75 - creased to around 30 percent, the Nusselt number began to
. • G=150 . display a much more pronounced effect of quality. At mass
- A G=300 + - fluxes above 300 kg/m 2 -s, the dependence of the Nusselt num-
- D G=500 + o - ber on quality remained similar. Even at low qualities, the Nus-
" O G=650 + " selt numbers were substantially higher than those for the low
- + G=800 - mass flux cases. If the same data were plotted as Nusselt number
O D "
o D versus mass flux, the heat transfer coefficient remained rela-

+
o n " tively constant at low mass fluxes. At a sufficiently high mass
o+ •
D
° A - flux, the slope of the heat transfer versus mass flux curve exhib-
+ A ited a distinct change in slope which remained relatively con-
• o a A
m "J stant. The mass flux at which this change in slope occurred
increased as the quality was decreased. For the 25 percent qual-
i i ' ity points, this shift was delayed to a mass flux of 300 kg/
0 0.2 0.4 0.6 0.8 1 m 2 -s.
Average Quality The change in heat transfer behavior exhibited in Fig. 7 is
closely linked to changes in the two-phase flow regime. For the
Fig. 7 Variation of Nusselt number with quality for 60 percent/40 per- two lowest mass fluxes, 75 and 150 kg/m 2 -s, wavy or wavy-
cent R-32/R-125 mixture at 35°C in 3.14 mm i.d. test-section
annular flow prevailed over much of the quality range. In the
wavy flow regime, heat is transferred primarily by conduction
across the thin condensate film at the top of the tube. This
forces. Their final result for the transition criterion can be ob- condensate flows down the tube under the action of gravity,
tained by plotting the following modified form of F,d on the and is then transported out of the tube in the liquid pool at the
Taitel-Dukler map: bottom of the tube. The modest increase in heat transfer coeffi-
cient with quality in this flow regime occurs because of the
, F d
' = • F
'" • (26) corresponding increase in interfacial area. The primary item
/ 1.305g / 1.305 affecting the heat transfer coefficient in this flow regime is the
+
-\f gD2(pl-pg) V +
Bd film thickness. Since the film thickness is insensitive to mass
flux in the wavy flow regime, wavy flow heat transfer coeffi-
For values of the Bond number, Bd, that are much greater cients are also relatively insensitive to mass flux.
than unity, surface tension plays no significant stabilizing role.
At the highest mass fluxes of 500, 650, and 800 kg/m 2 -s,
The lowest value of the Bond number for this study, that for
annular flow prevailed over most of the quality range. In the
R-134a at 35°C in the 3.14 mm i.d. tube, was about 16. This
annular flow regime, correlations such as those of Soliman et
Bond number results in a value of the denominator of Eq. (26)
al. (1968) and Traviss et al. (1973) clearly illustrate the interde-
of only 1.04, indicating gravity dominance. This is surprising
because Galbiatti and Andreini stated that the surface tension pendence between pressure drop and heat transfer (h <*
and gravity forces were of equal magnitude for D = 3 mm, V A P / A Z ) . Since the pressure drop increases sharply as the
but the value of surface tension for their air-water system was quality is increased, the heat transfer coefficients in the annular
significantly larger than that for the high-temperature refriger- flow regime show significant quality dependence.
ants in this study. The value of diameter at which surface tension Perhaps the most interesting heat transfer trends were dis-
forces become important is quite sensitive to what wavelength played at a mass flux of 300 kg/m 2 -s. At this mass flux, the
is considered most unstable (1 = 5.5D was used by Galbiatti flat Nusselt number versus quality behavior that is characteristic
and Andreini), so this issue might warrant further attention if of wavy flow was observed at low qualities while annular flow
tubes in the 3 mm range are implemented commercially. While behavior was observed at higher qualities. The change in slope
the experimental observations from this study did not reveal of the Nusselt number versus quality graph occurred between
large differences that could be attributed to surface tension 30 percent and 40 percent quality, which corresponded closely
forces, the addition of oil that is present in real condensers with an observed change from the wavy-annular to the annular
could alter this finding. flow regime. At this mass flux, it would be inappropriate to use
a single heat transfer model over the entire quality range.
Experimental Heat Transfer Data
Tube Diameter Effects. The relationships between h and
Only the data from the 3.14 mm and 7.04 mm i.d. tubes were D that were predicted by annular and wavy flow heat transfer
used to develop the heat transfer coefficient models because of correlations agreed well with the experimental data. Although
increased accuracy in these tubes. Overall, 261 heat transfer this was expected for the 7.04 mm i.d. tube, a commonly used
points were collected in the 3.14 mm i.d. tube and 387 heat and tested size, some doubt existed about whether the heat
transfer points were collected in the 7.04 mm i.d. tube. These transfer behavior in the 3.14 mm i.d. tube would correspond
points included data in the smooth-stratified, wavy-stratified,
with that predicted by "large tube" correlations.
wavy-annular, annular, and annular-mist flow regimes.
The most noticeable effect of the tube diameter had to do
Overall Trends—Effect of Mass Flux and Quality. Fig- with the point at which the heat transfer mechanism changed
ure 7 presents heat transfer data for R-32/R-125 (60/40 per- from filmwise (wavy) to forced convective (annular). The pri-
cent) in the 3.14 mm i.d. tube at a saturation temperature of mary difference in the heat transfer behavior in the two tubes
35°C. These data illustrate the general effects of mass flux and was observed at a mass flux of 300 kg/m 2 -s. In the 3.14 mm
quality rather clearly. At the lowest mass flux of 75 kg/m 2 -s, i.d. tube, the heat transfer behavior showed a change in slope
the Nusselt number increased very modestly as the quality was between 30 percent and 40 percent quality as the flow regime
increased. A similar quality dependence was exhibited as the changed from wavy annular to annular. In the larger 7.04 mm
mass flux was doubled to 150 kg/m 2 -s. Perhaps more interest- i.d. tube, this transition was only observed for the highest quality
ingly, the Nusselt numbers remained nearly identical as the point (89 percent). Again, it corresponded closely with an ob-
mass flux was doubled. As the mass flux was doubled once served transition to annular flow at around 80 percent quality.
again to 300 kg/m 2 -s, a different trend began to emerge. At The heat transfer characteristics at the other mass fluxes were
low qualities, the heat transfer coefficients remained nearly similar in both tubes.

204 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 .5WU i
1
i ' ' ' ' bQ ' ' 5000 1 1 | 1 1 1 | T 1 1
1 ' ' ' \A '
t
J

-K]
^ . . A . \
A D
2500 7
<vL 5, O -J * 4000 ""
; o D •
£•g : /Zl^r-*
£ 2000
•H 3000
oefficie
0 * U3
1500 L. -
D 8
u •

: ^ 2000 '- -
sfer

1000 7 - n>
H
H
P
500 •_
A
O
D
R-134a
R-22
R-32/R-125
s 1» 1000 —A— R-134a
—O—R-22
-0-R-32/R-125
"
S ' •
:
SI S3 .
0 •t , , I , , . 1 , , i , , , i , .
0 i • .

() 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8

Average Quality Average Quality

Fig. 8 Wavy flow heat transfer comparison for R-134a, R-22, and 50 Fig. 10 Mixed flow regime heat transfer comparison for R-134a, R-22,
percent/50 percent R-32/R-125 in the 7.04 mm i.d. tube at 35°C, G = 75 and 50 percent/50 percent R-32/R-125 in the 7.04 mm i.d. tube at 35°C,
kg/m 2 -s, A T « 3°C G = 300 kg/m 2 -s

Effect of Refrigerants. A comparison of wavy flow heat was highest for R-32/R-125, followed by nearly identical values
transfer for R-134a, R-22, and 50 percent/50 percent R-32/ for R-134a and R-22. This resulted in R-32/R-125 having
R-125 is made in Fig. 8. This figure shows the heat transfer slightly better annular flow heat transfer than the other fluids
coefficients at a mass flux of 75 kg/m 2 -s and a saturation tem- at low qualities, where the heat transfer coefficient was close
perature of 35°C in the 7.04 mm i.d. tube. Since the wavy flow to the single-phase liquid value. As the quality was increased,
heat transfer coefficients are dependent on the refrigerant-to- the two-phase multiplier increased in value and became increas-
wall temperature difference, it was maintained near 3°C (2.87°C ingly important. The two-phase multiplier was highest for R-
to 3.11°C) for all the tests. Perhaps the primary conclusion from 134a, the lowest reduced pressure fluid. This resulted in R-134a
these data is that little difference exists in the wavy flow heat having higher heat transfer coefficients than the other fluids at
transfer coefficients of the three fluids, with a maximum differ- high qualities.
ence of around 10 percent. It appeared that the highest heat
transfer coefficients were obtained for R-22, followed by R- Figures 8 and 9 compared the heat transfer coefficients of
134a and R-32/R-125. Based on the wavy flow property index the fluids at mass fluxes where one flow regime prevailed over
from the Chato (1962) correlation, [p,(p, - pg)k]ilgl p,]a25, R- the entire quality range. At intermediate mass fluxes, a transition
32/R-125 would be expected to have the highest wavy flow quality was reached where the flow regime and heat transfer
heat transfer coefficients (4.7 percent above R-134a), followed characteristics changed from wavy flow to annular flow. This
by R-22 (3.5 percent above R-134a) and R-134a. These pre- point was also affected by the fluid. Based on the flow regime
dicted differences are very small in light of the considerable results, one would expect annular flow to occur at lower quali-
uncertainty in transport properties. The index itself also neglects ties for R-134a than for R-32/R-125. This is demonstrated in
the effect of void fraction variations between the fluids, which Fig. 10, which compares the heat transfer coefficients of R-
would tend to reverse the predicted differences. 134a, R-22, and R-32/R-125 at a mass flux of 300 kg/m 2 -s. At
this mass flux, the quality where the wavy-annular to annular
Figure 9 compares the annular flow heat transfer coefficients transition occurred was 40 percent for R-134a, 50 percent for
of R-134a, R-22, and 50 percent/50 percent R-32/R-125. These R-22, and 72 percent for R-32/R-125. The heat transfer points
data were collected at a mass flux of 650 kg/m 2 -s and a satura- were connected with lines in this figure to make changes in
tion temperature of 35CC in the 7.04 mm i.d. tube. The best slope clearer. The data showed that, as expected, the change in
fluid for annular flow heat transfer was R-134a, while the heat slope occurred slightly earlier for R-134a than for the other
transfer coefficients of R-22 and R-32/R-125 were nearly iden- fluids. At the highest qualities, where annular flow was observed
tical. The single-phase, liquid heat transfer index, kf0.6 c°f
„ 0 . 4 / , , 0.4
/p\ for all fluids, R-134a had the highest heat transfer coefficients
of the three fluids. Across the full range of quality, the heat
transfer coefficients of the three fluids were within 10 percent.
12000 1 • ' • 1 ' i i ,• i i

A
Effect of Temperature Difference. The refrigerant-to-wall
"a loooo temperature difference has an impact on the heat transfer coef-
A ficients in the wavy flow regime. This dependence occurs be-
S 8000
A
oiP -_ cause, for a falling film, a larger driving temperature difference
A results in a thicker film at a given location (hence lower heat
CO transfer coefficients). In the annular flow regime, a significant
| 6000
u amount of experimental and analytical evidence suggests a neg-
ligible impact of temperature difference or heat flux. Studying
| 4000
the impact of temperature difference on the heat transfer coeffi-
: D A R-134a
: cients then provides a nonvisual method of assessing the extent
t 2000 o R-22
of filmwise and/or forced-convective condensation.
D R-32/R-125
. , 1 1 1 1 . 1 1 • 1 I i i l i i
Precisely controlling the temperature difference during inter-
nal condensation experiments is very difficult. For this reason,
0.2 0.4 0.6 0.8 few internal condensation data are available for which the tem-
Average Quality perature difference was deliberately controlled. To gain an
Fig. 9 Annular flow heat transfer comparison for R-134a, R-22, and 50
added understanding of AT effects, the temperature difference
percent/50 percent R-32/R-125 in the 7.04 mm i.d. tube at 35°C, G = was controlled at two different levels for all mass fluxes during
650 kg/m 2 -s testing with R-22 and R-32/R-125 in the 7.04 mm i.d. tube.

Journal of Heat Transfer FEBRUARY 1998, Vol. 120 / 205

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 300 1 ' ' i | i i i
1 ' ' ' 1 ' ' ' . 500 I ' ' 1
1 ' 1 ' ' 1
1
:
A
250 400 - 0* 1
A

200 '- o ° j d*
A o 300 - J
150 7 A O -_ s
Z
J
O
^
100 1 -I
200 6 A AT==2C '-_
- A AT=2 C .~ • A
O
50 " - AT==3C _
~ O AT=3 C '- 100 • AT=4C :
.
0 1 i i I i i i 1 i > < 1 i i i i i ' 1

0 0.2 0.4 0.6 0.8 0

Average Quality
0 0.2 0.4 0.6 0.8 1
Fig. 11 Effect of A T on Nu for R-32/R-125 in the 7.04 mm test-section, Average Quality
G = 75 kg/m 2 -s, T,„ = 35°C Fig. 13 Effect of A T on Nu for R-32/R-125 in 7.04 mm test-section,
G = 300 kg/m 2 -s, Taat = 35°C

Figure 11 shows the variation of Nusselt number with quality dence on AT less than the -0.25 power dependence from Nus-
for R-32/R-125 at a saturation temperature of 35°C and a mass selt theory. Stated another way, the heat transfer in the bottom
flux of 75 kg/m 2 -s. The two different sets of points correspond of the liquid pool was no longer negligible. Figure 15 shows
to temperature differences of approximately 2°C (1.88°C to that at a mass flux of 650 kg/m 2 -s, the unmodified Nusselt
2.12°C) and 3°C (2.87°C to 3.11°C). As predicted by Nusselt numbers exhibited no dependence on temperature difference,
theory, the Nusselt numbers were lower for the higher tempera- as expected for annular flow.
ture difference data across the full range of quality. These same
data were corrected for varying temperature differences by com-
puting the quantity Nu/[Ga*Pr/Ja/] 0 2 5 based on liquid proper- Comparisons with Existing Heat Transfer Correla-
ties. If all of the heat transfer occurred by filmwise condensation tions
and none occurred in the bottom of the tube, this correction Several correlations from the literature were selected for com-
should collapse the two different sets of data onto a single line. parison with the experimental data. For the wavy flow data, the
Figure 12 shows that, to a large extent, making this correction correlations of Chato (1962), Jaster and Kosky (1976), and
does just that. As the quality approaches unity and the liquid Rosson and Myers (1965) were selected. The correlations of
pool vanishes, moreover, the value of Nu/[Ga*Pr/Ja;] 0 2 5 ap- Shah (1979), Cavallini and Zecchin (1974), Traviss et al.
pears to properly approach the accepted value of 0.728 for (1973), and Chen et al. (1987) were selected for comparison
condensation outside of a horizontal cylinder. with the annular flow data.
Figures 13 through 15 present similar plots for R-32/R-125
in the 7.04 mm tube as the mass flux was increased to 300 kg/ Gravity-Dominated Correlations. Chato's correlation
m2-s and 650 kg/m 2 -s. The trends are very interesting. At 300 was developed for stratified flow and recommended for use at
kg/m 2 -s, the unmodified Nusselt numbers increased with de- vapor Reynolds numbers of less than 35,000. It was compared
creasing temperature difference up to a quality of 75 percent. with 210 experimental data points that met this criteria, and had
At that point, they appeared to converge as would be expected a mean deviation of 12.8 percent. The primary deficiency in
for annular flow. In Fig. 14, the quantity Nu/[Ga*Pr/Ja ; ] a 2 5 is the Chato correlation was that it predicted no variation of the
plotted for the same set of data. The trends on this plot are Nusselt number with quality. Comparisons with the wavy flow
"flipped" in comparison to the data of Fig. 13, with the highest data show that such variations did exist.
values of Nu/[Ga*Pr/Ja,] 0 2 5 occurring for the highest values The range of applicability of the Jaster and Kosky correlation
of AT. This indicates that at this mass flux, additional heat is specified by an upper limit of a dimensionless wall shear.
transfer by forced-convective condensation resulted in a depen- The mean deviation between the 213 experimental points which

1.0 — p —1—r—
1 ' ' 1 1 1 1 1 1 ' ' 1
1 ' ' 1.2 1 ' 1
1 ' , 1 . 1 , 1 1 1
- O
1.0 A -
0.8 % -
A
A 0.8 - **k -

3 0.6 o* o -
O A

a. A
0.6 - *A
* - c* o •<1 -
O 0.4 c* S" 0.4 ~A A AT=2 C -
T3
55
A AT=2C z • O AT=3 C -
o AT=3C 0.2 • AT=4C -
. 1 . ,
- . 1 , .
0.2 1 , , , 1 , , , , i , ,
0.0 i 1 > i , , , i , , "
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
Average Quality Average Quality
0.0
Fig. 12 Effect of A T on Nu/[Ga*Pr/Ja]° 2 6 for R-32/R-125 in the 7.04 Fig. 14 Effect of A T on Nu/[Ga*Pr/Ja] 0 2 6 for R-32/R-125 in the 7.04
mm i.d. test-section, G = 75 kg/m 2 -s, Tm = 35°C mm i.d. test-section, G = 300 kg/m 2 -s, T„, = 35°C

206 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ;—r- i i | i i 1 | 1 T 1 1 ' ' 1 1 1 1
deviations occurred for some low Nusselt number data that were
in the wavy-annular flow regime and for some very high mass
700 -; flux, high quality data. In general, the Shah correlation seemed
600 -. to underpredict the experimental data.
tP
The mean deviation of the Cavallini and Zecchin correlation
500 tP -_ with the experimental data was 11.6 percent, slightly higher than
the Shah correlation. Despite the slightly higher mean deviation,
5 400

i i l i i i i l n nil
T
though, the predictions of the Cavallini and Zecchin correlation
300 a were more correct in trend than those of the Shah correlation.
When the Cavallini and Zecchin correlation was in error, it
200 o AT=3C tended to overpredict the experimental data. The largest errors
100 • AT=4C occurred at low qualities because this correlation approaches a
value of 2.18 times the single-phase Nusselt number at a quality
n l l l l l 1 , , of zero.
0 0.2 0.4 0.6 0.8 1 The mean deviation of the Traviss correlation was 11.8 per-
Average Quality cent, slightly higher than the Cavallini and Zecchin correlation.
The Traviss correlation tended to overpredict the experimental
Fig. 15 Effect of A T on Nu for R-32/R-125 in the 7.04 mm tube, G data, particularly at high qualities, where their empirical correc-
650 kg/m 2 -s, TM = 35°C
tion was used. If this correction was omitted, their correlation
would have underpredicted the high quality data.
met this criterion and the predicted values was 14.5 percent, The Chen correlation was the worst predictor of the annular
slightly higher than the simpler Chato correlation. Although the flow data with a mean deviation of 23.3 percent. This correlation
deviations between their correlation and the present data were significantly underpredicted nearly all of the data. The correla-
sometimes large, the mean deviation of 14.5 percent was sub- tion of Soliman et al. (1968) generally predicts lower Nusselt
stantially better than the 37 percent standard deviation with their numbers than the Chen correlation. Thus, it would have per-
own data. formed even worse against the present data.
The correlations of Chato, and Jaster and Kosky, were both One problem with the annular flow correlations that is not
able to predict most of the experimental data for the wavy flow apparent in a plot of experimental versus predicted Nusselt
regime within a range of ±25 percent. Neither predicted the numbers concerns their range of applicability. Figure 16 com-
variation with quality successfully, and neither accounted for pares experimental and predicted Nusselt numbers versus qual-
heat transfer in the bottom of the liquid pool. For very low ity for R-22 in the 7.04 mm i.d. tube at a mass flux of 225 kg/
mass fluxes, this assumption is reasonable. m2-s and a saturation temperature of 35°C. According to the
Taitel-Dukler maps, this mass flux should be annular at all
Because no guidelines were given the use of the Rosson and qualities above 12 percent. The velocity, with vapor only, would
Myers correlation, it was compared against the full database of be 3.9 m/s, 30 percent above the lower limit of applicability
points that were later used to develop the wavy flow correlation specified by Shah. Thus, the best information available in the
for this study. Although Rosson and Myers attempted to account literature would suggest that annular flow correlations should
for forced-convective condensation in the liquid pool at the be applicable over nearly the entire quality range. Still, Fig. 16
bottom of the tube, their correlation was actually a poor pre- shows that the Nusselt numbers were well above the annular
dictor of the experimental data. The correlation had a mean flow predictions at low qualities. As the quality reached about
deviation of 21.3 percent with the experimental data, almost 10 70 percent and the flow pattern became fully annular, the predic-
percent worse than the simpler Chato and Jaster and Kosky tions of the Cavallini and Zecchin correlation agreed very well
correlations. The most problematic part of the correlation with the experimental data. These data suggest the need for
seemed to be prediction of the parameter /3, which represents further development of heat transfer models in the wavy or
the fraction of the tube circumference occupied by filmwise wavy-annular regions at higher mass fluxes.
condensation. At low mass fluxes, this parameter should clearly
be related to the void fraction and approach unity as the quality
approaches unity. However, the empirical. expressions devel- Development of the Heat Transfer Correlations
oped by Rosson and Myers do not behave in this manner. The In this section, the development of the annular flow correla-
trends were very erratic, particularly for mass fluxes over 25 tion for forced-convective condensation is first described. This
kg/m 2 -s, where the relationship was not even monotonic.

Annular Flow Correlations. The annular flow correlations

that were selected for comparison with the experimental data 400
encompass at least one member of each of the three broad 350
classes—two-phase multiplier correlations (Shah, Cavallini,
and Zecchin), shear-based correlations (Chen et al.), and 300
boundary layer analyses (Traviss et al.).
250
Of the four correlations, only Shah's came with specific
guidelines for a lower limit of applicability. His correlation 200
should not be used at mass fluxes where the vapor velocity with
x = 1 was less than 3 m/s. For R-134a at 35°C, this corresponds 150
Exp.
to a lower mass flux limit of 130 kg/m 2 -s. This value is higher 100 Shah
for the other refrigerants, which have higher vapor densities. In Cav. Zech.
all cases, this represents a vapor velocity well above the wavy 50 t Traviss
to annular transition line on the Taitel-Dukler map. This crite-
rion was selected for each of the correlations so that they would
0
be compared on an equal basis. 0 0.2 0.4 0.6 0.8
Average Quality
The predictions of the Shah correlation agree fairly well with
the data, with a mean deviation of 9.1 percent. Nearly all of the Fig. 16 Comparison of annular flow correlations with heat transfer data
data were predicted within ±25 percent. The most significant for R-22 in the 7.04 mm tube, T,M = 35°C, G = 225 kg/m 2 -s

Journal of Heat Transfer FEBRUARY 1998, Vol. 120 / 207

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 discussion establishes the strong similarities between the most F2 is seldom necessary. With F2 confined to values of Re, >
theoretical approach for annular flow condensation, the bound- 1125, a power law function of Re, and Pr, was used to curve
ary-layer analyses, and the simpler two-phase multiplier ap- fit the function F2. For a range of Re, from 1125 to 105 and
proaches. Then, an annular flow correlation is developed using Pr, from 1 to 10. F2 was well approximated by the following
the two-phase multiplier approach. Next, the wavy flow data function:
from this study are used to develop a wavy flow correlation.
This correlation uses a similar approach to that tried unsuccess- F2 s 10.25 Re? 06 Pr?': (29)
fully by Rosson and Myers (1965), but with more attention
paid to carefully weighting the forced-convective and filmwise If this approximation is used in Eq. (28), and D + is evaluated
components. Finally, the data from the wavy-annular flow re- with a pressure drop correlation using Lockhart and Martinelli's
gime are compared with predictions of the wavy and annular two-phase liquid multiplier approach, the following equation is
flow correlations. These comparisons are used to develop limits obtained for the Nusselt number:
of applicability for the two correlations.
V0.079/2 Re°- 8 >/(X„) Pr,
Annular Flow Correlation. The most theoretical correla- Nu (30)
tion that has been widely used is the Traviss correlation. Under 10.25 Re 0 0 6 Pr 059
its rather stringent assumptions, this method provides an analyti- = 0.0194 Re?'81 Pr?' 4 >,(Z„).
cal prediction of the Nusselt number. Before a pressure drop
model is assumed, the correlation can be written as follows: This is identical in form, and close in value, to the commonly
used two-phase multiplier correlations. A two-phase multiplier
D+ Pr, correlation using the Dittus-Boelter correlation for the single-
Nu (27)
F 2 (Re„ Pr,) phase heat transfer coefficient, such as the one of Shah (1979),
takes the following form:
The term D+ is the tube diameter scaled by the turbulent
length scale, /i,/^lrwpi. A simple force balance indicates the 1 - x
proportionality between the wall shear and the pressure drop, Nu = 0.023 Re?'8 Pr? 4 F
establishing the fact that the annular flow Nusselt number is (31)
proportional to the square root of the pressure drop per unit
length. two-phase multiplier
The denominator of Eq. (27), F2, can be thought of as a
There is little difference in the form of Eq. (30) and (31),
dimensionless heat transfer resistance. Physically, this resis-
except that Shah used a slight modification of the Lockhart-
tance increases as the dimensionless film thickness increases,
Martinelli parameter for correlating his two-phase multiplier.
as would be expected from conduction arguments. It also in-
If the correlation of Traviss et al. (1973) proved to be the
creases with increasing Prandtl numbers, since high Prandtl
most accurate predictor of experimental data, modifying it
numbers reflect more effective transfer of momentum than heat.
would make little sense. In fact, even Traviss found that their
This function is dependent upon the liquid Reynolds number
correlation underpredicted their own data at high qualities. For
because, under the assumptions of a symmetric annular film and
this reason, they arbitrarily replaced their two-phase multiplier
no entrainment, specifying the liquid Reynolds number uniquely
(which they call F rather than <j>t) with F 1 1 5 at high qualities
specifies the dimensionless film thickness. A plot of F2 versus
(X„ < 0.155). Rather than modifying this function, which was
Re, for various values of Pr, shows that as Re, increases from
an empirical fit to two-phase pressure drop data to begin with,
0 to 1125 (the value where the fully turbulent region begins),
it appears equally reasonable to determine the function directly
F2 increases very rapidly. As Re, is increased further, F2 in-
from condensing heat transfer data. This is exactly the approach
creases much more slowly. Physically, this occurs because the
of the two-phase multiplier correlations.
primary resistance to heat transfer is contained in the laminar
sublayer and buffer regions. One additional benefit of dissecting the Traviss correlation
is that it demonstrates the robustness of the heat transfer correla-
Although the Traviss analysis was performed after the advent
tion for annular flow with respect to the rather serious assump-
of the simple shear-based correlations, it provides a useful
tions used in their derivation. These assumptions are (1) extrap-
method for understanding them. For relatively small changes
olating the universal velocity profile from single-phase pipe
in Re, when Re, > 1125, one could reasonably assume a con-
flow, (2) symmetric annular film, and (3) no entrainment of
stant value of F2 at a fixed Prandtl number. If the Prandtl number
liquid in the vapor core.
dependence could be expressed as a power law function, the
Nusselt number could then be expressed as With regard to the first assumption, the presence of waves at
the free surface would be expected to cause deviations from the
Nu = aD+ Pr? (28) universal velocity profile. However, the impact of these waves
is likely to be small in the region closest to the wall where most
This is exactly the form of the original shear-based correla- of the resistance to heat transfer occurs.
tion of Carpenter and Colburn (1951). Thus, these correlations With respect to the second assumption (the symmetric annu-
are justified for a narrow range of conditions by the more theo- lar film), gravity definitely creates a thinner film at the top of
retically sound analysis of Traviss. the tube than at the bottom. Since the heat transfer is relatively
Only a few more manipulations are required to show the insensitive to variations in film thickness provided that the film
equivalence between the Traviss analysis and the two-phase- remains thicker than the buffer layer, these effects would not
multiplier approaches. The first important observation is that be expected to significantly alter the heat transfer results.
annular flow is seldom encountered for liquid Reynolds num- The impact of violating the third serious assumption (that of
bers less than 1125. Using the criterion for annular flow that no entrainment), can also be assessed from Traviss' analysis.
Frso = 18 (Eqs. (4a) and (4b)), the corresponding equation This assumption is necessary when invoking the conservation
was solved for the quality above which annular flow could exist of liquid statement, which in turn determines the dimensionless
with Re, = 1125. For R-134a at 35°C, a value of over 99.5 film thickness. Since F2 only depends on Re, to the 0.06 power,
percent was obtained in the 7.04 mm i.d. tube and a value of though, it appears that the role of entrainment on heat transfer
over 97.5 percent was obtained in the 3.14 mm i.d. tube. This should be small, provided that the entrainment rate does not
indicates that the liquid film is seldom so thin that the fully become high enough to remove liquid from the laminar sublayer
turbulent region is not reached, thus the piecewise definition of or buffer layer.

208 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Based on the work above, the two-phase multiplier approach to determine the relative sizes of the entry and fully developed
was selected for correlating the annular flow heat transfer data regions. Low shear cases correspond to short entry regions, and
from this study. To make sure that the correlation was not biased the converse is true for high shear cases. However, only rela-
by data outside of the annular flow regime, only data with Fr„ tively low shear cases are amenable to film condensation analy-
> 20 were used to develop the correlation. This value was sis because high interfacial shear cases correspond to turbulent
reported by Dobson (1994) to provide a good indicator of the annular flows. The primary result of the analysis was to demon-
transition from wavy-annular to annular flow, and agreed well strate the validity of filmwise condensation solutions that as-
with the data from this study. The correlation developed was sume no axial velocity component for cases where a relatively
small axial velocity does exist.
2.22 The development of the correlation was guided by a combina-
Nu = 0.023 Re?" Pr? V 0.89
(32)
tion of careful data analysis and physical guidance from analyti-
cal solutions. The details are described by Dobson (1994) and
This form utilizes the single-phase heat transfer correlation Dobson et al. (1994a). The final correlation separates the heat
of Dittus and Boelter (1930) with a Prandtl exponent of 0.4. transfer by film condensation in the upper part of the horizontal
This value would normally be used for heating rather than cool- tube from the forced-convective heat transfer in the bottom
ing, but was selected for the following three reasons: (1) it was pool:
nearly identical to the value obtained by curve-fitting the Traviss
correlation; (2) it provided a better fit to the more accurate 0.23 Re°0' GaPr,
Nu = + (1 -6»,/7r)Nu forced, (33)
single-phase correlations in the range of Re and Pr from this 1 + LUX! Ja,
study; and (3) it has been used with some success by Shah for
fluids other than refrigerants. At a quality of zero, the Lockhart-
Martinelli parameter approaches infinity and Eq. (32) predicts where 9, = angle subtended from the top of tube to the liquid
the single-phase liquid Nusselt number. Values of the constants level and
in Eq. (32) were determined by regression analysis of the entire
set of data. The results show that Eq. (32) is a better predictor Nuforcecl = 0.0195 Re? 8 Pr?->,(X„) (34)
of the experimental data than any of the correlations from the
literature.
4>,{X„) = Jl.376 + ^jr . (35)
Wavy Flow Correlation. During wavy flow condensation
at sufficiently low mass fluxes, heat is transferred in the upper For 0 < Fr, s 0.7,
0.7
portion of the tube by filmwise condensation and in the pool at
the bottom of the tube by forced-convective condensation. At = 4.172 + 5.48 Fr, - 1.564 Fr? (36a)
very low vapor velocities, the heat transfer in the bottom of the
tube is much smaller than that in the top and can be readily c2 = 1.773 - 0.169 Fr, (36b)
neglected. Under these same conditions, the vapor flow has
little effect on the liquid film at the top of the tube and Nusselt's For Fr, > 0.7,
solution for external condensation on a cylinder can be used.
This approach was adopted by Chato (1962) and Jaster and d = 7.242 (37a)
Kosky (1976).
As the vapor velocity is increased, several deviations from c2 = 1.655 (37&)
ideal stratified flow conditions begin to emerge. First, the heat
transfer in the bottom of the tube may become significant Fr, is the liquid Froude number. Due to the 1.376 inside the
enough to warrant inclusion. This effect was observed experi- radical of Eq. (35), the correlation above matches the Dittus-
mentally at mass fluxes above 75 kg/m 2 -s, where the Nu/ Boelter single-phase correlation when x = 0.
[Ga Pr/Ja] 0 2 5 curves for different A7"s no longer converged 9, is geometrically related to the void fraction by the follow-
onto a single line. The second deviation from ideal stratified ing formula if the area occupied by the thin condensate film is
flow has to do with the interfacial shear exerted by the vapor neglected:
on the condensate film. For conditions approaching stagnant
vapor, little shear is exerted on the condensate. Under these 9, sin (20,)
a = (38)
conditions, the condensate falls almost vertically down the tube n 2-7T
and each axial location is independent of those around it. In
this situation, applying solutions from external condensation is If a void fraction model is assumed, this transcendental equa-
certainly justified. In the opposite extreme, that of very high tion must be solved to obtain the desired quantity, $,. Jaster and
vapor velocity and interfacial shear, gravitational forces become Kosky (1976) deduced an approximate relationship which is
negligible and annular flow is observed. Now the film thickness much easier to use, albeit for a slightly different purpose. In
is nearly independent of angle and filmwise condensation solu- the context of the present problem, their simplification can be
tions are inappropriate. In the intermediate region, the vapor stated as
velocity is high enough to exert a significant shear on the con-
densate film but not great enough to overwhelm gravity forces. 9, arccos (2a - 1)
Here, the condensate flows in a path with both axial and circum- (39)
ferential components.
Based on the data from this study, it appeared that some film The simplicity achieved by this assumption is well worth the
condensation was occurring in this intermediate regime. The modest errors associated with it. These errors are themselves
first question was whether the solutions from external filmwise mitigated by the fact that the forced-convective Nusselt number,
condensation would remain valid in this region. This question by which the quantity in Eq. (39) is multiplied, is normally
was addressed by solving the conservation equations for the considerably smaller than the filmwise Nusselt number. A void
film thickness in the entry and in the fully developed region. fraction correlation must be assumed to solve this equation. The
The solutions indicated that downstream of the entry region, calculations presented hereafter were made using the approxi-
the film thickness and the heat transfer coefficient is identical mation of Eq. (39) and the void fraction correlation of Zivi
to the zero shear case. The primary role of the vapor shear is (1964), Eq. (13).

Journal of Heat Transfer FEBRUARY 1998, Vol. 120 / 209

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 dence implies that a stable wall film is formed as soon as con-
densation begins. Stated another way, the nature of the conden-
sation process itself tends to preclude a dry wall. Although a
substantial amount of mist was entrained in the core flow at
high mass fluxes, the excellent agreement of the annular flow
correlation with experimental data containing significant en-
trainment indicates that the entrainment did not have a signifi-
cant effect on the heat transfer. These findings do not agree
with those of Soliman (1986), and are discussed in detail by
Dobson (1994) and Dobson et al. (1994a).
Design Recommendations.
For G > 500 kg/m 2 -s, use Eq. (32) for all qualities. For G
< 500 kg/m 2 -s, use Eq. (33) if Fr„ < 20, and use Eq. (32) if
0 50 100 150 200 250 300 350 400 Frso > 20.
Nu Experimental
Comparison With Experimental Data. The recom-
Fig. 17 Comparison of the wavy flow correlation, Eq. (33), with the mended correlations were applied to data from this study and
experimental data of the present study from the literature. The comparisons are summarized in Table
6 and discussed in detail by Dobson (1994).
Effect of Oil. Oil in the refrigerant decreases the heat trans-
The predictions of Eq. (33) are compared to the wavy flow fer and increases the pressure drop. Gaibel et al. (1994) dis-
data in Fig. 17. Overall, the agreement is very good. The mean cussed these effects. He found that for oil mass fractions UJ„ <
deviation between the predicted and experimental values was 0.05, the correction factor developed by Schlager et al. (1990)
6.6 percent, considerably better than the correlations from the gave acceptable values for the heat transfer when applied to the
literature. The worst deviations were found for the data at a Dobson correlations (Eqs. (32) and (33)).
saturation temperature of 45°C and some of the data in the 3.14
mm i.d. tube. For both of these situations, the uncertainties in Nuoil = Nu Dobson *exp(-3.2u; 0 ). (40)
both the heat transfer coefficients and the qualities were rather
large because the heat losses to the environment became a sig- For the pressure drop, the Souza et al. (1992) correction
nificant fraction of the overall heat transfer. For the lowest mass factor was found acceptable in the same oil concentration range.
flux tests in the 3.14 mm i.d. tube, for instance, the heat transfer APoil = AP pure (l + 12.4u>„ - 110.8w2). (41)
rates in the test section were on the order of only 50 W.
Insufficient data were developed to investigate the effects of
Limits of Application—Wavy-Annular Flow Regime. oil on the flow-regime transitions.
The correlations developed above were for the annular and
wavy flow regimes, respectively. The annular flow correlation Condensation of Zeotropes. Sweeney (1996) and
was developed using a set of data restricted to values of Frso > Sweeney and Chato (1996) correlated the data of Kenney et
20, which excluded wavy-annular flow. The wavy flow correla- al. (1994) obtained with Refrigerant 407c, a zeotropic mixture
tion was developed primarily with data for which FrM < 10, of R-32, R-125, and R-134a (23, 25, and 52 percent by mass),
including primarily wavy flow. A region exists in between these in a smooth tube. They found that the heat transfer rates could
limits where the physics of the condensation process is not be predicted by the following simple modification of the Dobson
so clear cut. The applicability of the wavy and annular flow correlations (Eqs. (32) and (33)):
correlations were assessed for data in this transition region. As
a first step, the wavy and annular flow correlations were used / G \03
to predict the Nusselt numbers for the entire set of data, regard- Nuwovy?2eollolx, = I —— I NuEq.(33) (42)
less of the flow regime. At values of Frso < 10, the Nusselt
numbers were much higher than predicted by the annular flow
correlation. In the worst case, this deviation was over a factor Nua„nlllariZ!,otr0pe — 0.71 ——- I NuEq.(32)> (43)
of 4. At values of Frso > 20, the data were entirely in the annular
flow regime, and the predictions were very good. At values of 2
Fr,0 between 10 and 20, some points were well predicted by where the mass flux, G, is in kg/m -s. Figure 18 shows the
the annular flow correlations, and others were predicted poorly. agreement between these correlations and a typical set of experi-
With the wavy flow correlation, at values of Frs„ < 10, all of mental data covering both wavy and annular flow regimes.
the data were predicted within a 20 percent range. At values of These correlations cannot be assumed to be generally applica-
Frso > 20, the measured Nusselt numbers were substantially ble to zeotropes without additional data on other zeotropes, but
larger than the predicted values. For the range of FrJO between they do indicate that the equations represent the underlying
10 and 20, the error in the prediction of the wavy flow correla- physical phenomena well.
tion was much less than the error produced by the annular
flow correlation. At mass fluxes at or above 500 kg/m 2 -s, all
experimental Nusselt numbers were substantially higher than Table 6 Comparison between the heat transfer correlations and experi-
those predicted by the wavy flow correlation, but were quite mental data
well predicted by the annular flow correlation.
Source Fluid Tube Diameter Mass Mux Mean Deviation
mm kg/m^-s %
Mist Flow Effects. Pure mist flow with no stable wall film Dobson (1994) R-134a 3.14 & 7.04 25-800 4.4
was never observed at qualities below 95 percent. Moreover, Dobson (1994) R-22 3.14&7.04 25-800 4.9
when it was observed at the highest qualities, it was always Dobson (1994) 50/50 R-32/R-125 3.14&7.04 25-800 5.9
after the adiabatic section and before the test-section inlet. At Dobson (1994) 60/40 R-32/R-125 3.14 & 7.04 25-800 6.1
the outlet sight glass, after condensation had begun in the test Altaian (1960) R-22 8.7 300-860 7.0
section, a stable wall film was always observed. This was true Bae(1970) R-22 12.5 320-660 13.5
for test-section outlet qualities as high as 90 percent. This evi- Traviss(1973) R-12 8.0 162- 1540 13.7

210 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 350 . • i • i i ' i > i ' i ' i 1
i ' i ' .
than those from which the map was developed. A simple correc-
tion with the ratio of vapor and air densities of the superficial
300 r o 9 J velocities of the vapor and liquid phases overcomes this prob-
T lem, and gives each axis the physical interpretation of square
250 8 - root of kinetic energy. With this correction applied, the Mand-
T '.
0 hane map was an excellent predictor of the experimental data
200 o -_ and the easiest of the three methods to use. Based on the data
3 A from this study and earlier suggestions from the literature, the
Z
150 8 -_ flow regime described as wavy annular in this study lies in the
; 6 right portion of the slug flow region on the Mandhane map.
100 0 Experimental Data - The region labeled as annular flow corresponds closely with
• Eq. (42) ". the establishment of a nearly axisymmetric annular film. The
50 T Eq. (43) -_ absence of a diameter dependence could prove problematic for
0 - , i , i 1 , 1 , , 1 , 1 , - very small tubes, but was not a significant problem for the range
() 0.2 0.4 0.6 0.8 1 of this study.
Average Quality, x The Taitel-Dukler map successfully distinguished the stra-
tified and wavy flow regimes at 25 and 75 kg/m 2 -s for each
Fig. 18 Comparison between smooth tube data and predictions of Eqs. refrigerant. Also, the increase in the length of the slug flow
(42) and (43), G = 300 kg/m 2 -s (220 klb m /ft 2 -hr) region as reduced pressure was increased was correctly pre-
dicted. The main difference between the predictions of the
Taitel-Dukler map and the observed flow regimes was found at
Instabilities. Probably because we were studying only local mass fluxes slightly above the wavy flow transition line. For
condensation, as opposed to complete condensation, we have these, the Taitel-Dukler map predicts that slug flow at low quali-
not encountered any instabilities, such as those reported by ties is followed immediately by annular flow for the remainder
Rabas and Minard (1987). Instabilities, however, could occur of the quality range. The experimental observations in this re-
in complete systems but would depend very strongly on the gion indicated that the slug flow regime was consistently fol-
interactions among the various components. lowed by wavy, wavy annular, and annular flows, as the quality
was increased. Observations of wavy-annular flow in the region
predicted to be annular flow are not due to a deficiency in the
Summary and Recommendations Taitel-Dukler technique, since wavy-annular flow represents an
Local (i.e., for small changes in quality) heat transfer and early stage of annular flow in their terminology. Observations
fluid flow data were collected and analyzed for four refrigerants of wavy flow where annular flow was predicted do represent
in two smooth tubes. Flow regimes were observed visually in an error, but are possibly due to subjectivity in classification.
three smooth tubes with diameters ranging from 3.14 to 7.04 Perhaps the most important conclusion concerning this map is
mm inner diameter. These observations were made over the that significant stratification of the liquid can be expected in
entire range of quality and a range of mass flux that comprises the lower part of the range labeled as annular flow. This stratifi-
the fluxes used from refrigerators to automotive air conditioners. cation is most significant at low qualities, or high values of the
Based on these observations and their comparisons with three Lockhart-Martinelli parameter. For pipe diameters smaller than
commonly used flow regime predictors, several conclusions can those used in this study, each of the flow regime transitions
be made. predicted by the Taitel-Dukler map should be examined for
The primary flow patterns that can be expected in refrigera- surface tension influences.
tion condensers are wavy, wavy annular, annular, and annular The Soliman flow regime predictors, although less theoreti-
mist. At low qualities, slug or pseudo-slug flow will occur. cally justified than those of Taitel and Dukler, matched the
From this study, it appears that mist flow without a stable liquid experimental observations very well. However, the definitions
film does not exist during condensation. At low vapor velocities, of the different regimes must be clearly understood. The wavy
the flow is gravity dominated, and the liquid tends to collect at flow region of Soliman includes stratified flow, wavy flow, and
the bottom of the tube. At high velocities, vapor shear controls slug flow. With this in mind, the annular flow transition pre-
and the liquid is distributed around the circumference in annular dicted by Soliman corresponded very well with the wavy to
fashion. Stratified flow without any interfacial waves was ob- wavy-annular transition from this study. It has been shown by
tained at the lowest mass flux in this study, but it is unlikely that Dobson (1994) and Dobson et al. (1994a, b) that the transition
such a low mass flux will be encountered in typical applications. from wavy-annular to annular flow is well predicted by a value
Three techniques for flow regime prediction were compared of Frto = 18 (as opposed to 7 for the wavy to wavy-annular
to the experimental observations. Comparing these flow regime transition). The region labeled as mist flow by Soliman was
predictors to each other presents several problems which would found to correspond well with the annular-mist flow regime in
reduce the usefulness of such an exercise. First, the subjectivity this study. However, pure mist flow without a stable wall film
involved in visually determining the flow regimes is consider- was confined to qualities greater than 90 percent, and even then
able. Differentiating between similar flow patterns is quite dif- appeared to become annular mist once condensation began.
ficult. Second, the number of categories included in the different The heat transfer behavior depended significantly on the flow
maps also presents a problem. While Soliman includes stratified, regime. At low mass fluxes, gravity-dominated wavy flow pre-
wavy, and slug flows in one regime, the Mandhane and Taitel- vailed, and film condensation was the dominant heat transfer
Dukler maps treat these as three separate regimes. Similarly, mode. In this flow regime, the heat transfer coefficients in-
the Mandhane and Taitel-Dukler maps do not have a separate creased slightly with increasing quality and mass flux. At the
mist flow regime, while Soliman's map does. The possibility highest mass fluxes, vapor shear dominated and annular or annu-
for an incorrect placement clearly increases with the number of lar-mist flow prevailed over nearly the entire quality range with
distinct flow regimes. Given these difficulties, a discussion of forced-convective condensation as the prevailing heat transfer
the merits and drawbacks of each technique seems more suit- mode. This regime was characterized by heat transfer coeffi-
able. cients that increased substantially with increases in quality and
The original map of Mandhane was an extremely poor pre- mass flux. At intermediate mass fluxes, wavy flow heat transfer
dictor of the experimental data from this study. This was be- behavior was observed at low qualities but became dominated
cause the vapor densities from this study were much higher by annular flow at higher mass fluxes. The quality at which this

Journal of Heat Transfer FEBRUARY 1998, Vol. 1 2 0 / 2 1 1

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 transition to forced-convective condensation occurred increased tional Heat Transfer Conference, Japan Society of Mechanical Engineers, Vol.
3, pp. 309-313.
with increases in tube diameter and reduced pressure of the Chato, J. C , 1962, "Laminar Condensation Inside Horizontal and Inclined
refrigerant. The effect of temperature difference on heat transfer Tubes," ASHRAE Journal, Vol. 4, pp. 52-60.
followed the expected -0.25 power dependence at low mass Chen, M. M., 1961, "An Analytical Study of Laminar Film Condensation. Part
fluxes in the wavyflowregime. As the massfluxwas increased, II: Single and Multiple Horizontal Tubes," ASME JOURNAL OF HEAT TRANSFER,
heat transfer contributions from the forced-convective conden- Vol. 83, pp. 55-60.
Chen, S. L„ Gerner, F. M., and Tien, C. L., 1987, "General Film Condensation
sateflowin the liquid pool reduced the dependence on tempera- Correlations," Experimental Heat Transfer, Vol. 1, pp. 93-107.
ture difference. Denglor, C. E. and Addoms, J. N„ 1956, "Heat Transfer Mechanism for Vapor-
The experimental data were compared to several correlations ization of Water in a Vertical Tube," Chemical Engineering Progress Symposium
Series, Vol. 52, No. 18, pp. 95-103.
from the literature and significant discrepancies were found. Dhir, V., and Lienhard, J., 1971, "Laminar Film Condensation on Plane and
The main problem with these correlations had to do with their Axisymmetric Bodies in Nonuniform Gravity," ASME JOURNAL OF HEAT TRANS-
range of applicability, i.e., they tended to yield good results FER, Vol. 93, pp. 97-100.
only in specific flow regimes. New correlations were developed Dittus, F. W., and Boelter, L. M. K., 1930, "Heat Transfer in Automobile Radi-
for the annular and wavyflowregimes. The annularflowcorre- ators of the Tubular Type," University of California Publications on Engineering,
Vol. 2, No. 13, Berkeley, CA, p. 443.
lation was developed using a two-phase multiplier approach. It Dobson, M. K., 1994, "Heat Transfer and Flow Regimes During Condensation
agreed with the data from this and other studies very well. in Horizontal Tubes," Ph.D. thesis, Dept. of Mechanical and Industrial Engi-
Analysis of the more theoretical approaches to annular flow neering, University of Illinois at Urbana-Champaign.
condensation revealed the similarity between these approaches Dobson, M. K., Chato, J. C , Wattelet, J. P., Gaibel, J. A., Ponchner, M„ Ken-
ney, P. J., Shimon, R. L., Villaneuva, T. C , Rhines, N. L„ Sweeney, K. A., Allen,
and that of the two-phase multiplier. A wavy flow correlation D. G., and Hershberger, T. T., 1994a, "Heat Transfer and Flow Regimes During
was developed that accounted for both film condensation at the Condensation in Horizontal Tubes," ACRC Technical Report 57, University of
top of the tube and forced-convective condensation in the liquid Illinois at Urbana-Champaign.
pool at the bottom of the tube. This correlation approaches the Dobson, M. K„ Chato, J. C , Hinde, D. K., and Wang, S. P., 1994b, "Experi-
correct limits in extreme cases—a single-phase liquid correla- mental Evaluation of Internal Condensation of Refrigerants R-12 and R-134a,"
ASHRAE Transactions, Vol. 100, No. 1, pp. 744-754.
tion at a quality of zero, and an external condensation for a Dukler, A. E., 1960, "Fluid Mechanics and Heat Transfer in Vertical Falling-
quality of 1 at low massfluxes.Based on a systematic analysis, Film Systems," Chemical Engineering Progress Symposium Series, Vol. 56, No.
a limit for switching from the wavy to the annular correlation 30, pp. 1-10.
was determined using Soliman's Froude number. Above a given Dukler, A. E., and Hubbard, M. G., 1975, "A Model for Gas-Liquid Slug Flow
in Horizontal and Near Horizontal Tubes," Ind. Eng. Chem. Fundamentals, Vol.
mass flux limit, the annular flow correlation could be applied 14, No. 4, pp. 337-347.
regardless of the Froude number. Gaibel, J. A., Chato, J. C , Dobson, M. K., et al., 1994, "Condensation of a
50/50 Blend of R-32/R-125 in Horizontal Tubes with and without Oil," ACRC
Although one of the new correlations could be used with a Technical Report 56, University of Illinois at Urbana-Champaign.
zeotropic refrigerant with he application of a simple multiplier, Hanratty, T. J., 1994, personal communication, University of Illinois at Urbana-
considerably more analyses and experimental data are needed Champaign, IL.
for the development of satisfactory understanding and reliable Hinde, D. K., Dobson, M. K., Chato, J. C , Mainland, M. E., and Rhines, N„
correlations for zeotropes. In addition, more reliable data are 1992,' 'Condensation of R- 134a With and Without Oils," ACRC Technical Report
26, University of Illinois at Urbana-Champaign.
needed to ascertain the void fraction under various flow condi- Hubbard, M. G., and Dukler, A. E., 1966, "The Characterization of Flow Re-
tions. gimes for Horizontal Two-Phase Flow: Statistical Analysis of Wall Pressure Fluc-
tuations," Proceedings of the 1966 Heat Transfer and Fluid Mechanics Institute,
Stanford University Press, Stanford, CA, pp. 100-121.
Acknowledgments Jaster, H. and Kosky, P. G„ 1976, "Condensation in a Mixed Flow Regime,"
International Journal of Heat and Mass Transfer, Vol. 19, pp. 95-99.
This work was supported in part by the Air Conditioning and Jeffreys, H., 1925, "On the Formation of Water Waves by Wind," Proceedings
Refrigeration Center of the University of Illinois, a consortium of the Royal Society, Vol. A107, p. 189.
of the University, eighteen industrial sponsors, and the National Jeffreys, H., 1926, "On the Formation of Water Waves by Wind," Proceedings
of the Royal Society, Vol. A110, p. 241.
Science Foundation. Kenney, P. J., Chato, J. C , Dobson, M. K., Wattelet, J. P., et al., 1994, "Con-
densation of a Zeotropic Refrigerant R-32/R-125/R-134a (23 percent/25 percent/
52 percent) in a Horizontal Tube," ACRC Technical Report 62, University of
References Illinois at Urbana-Champaign.
Altaian, M., Staub, F. W., and Norris, R. H., 1960, "Local Heat Transfer and Lin, P. Y. and Hanratty, T. J., 1989, "Detection of Slug Flow from Pressure
Pressure Drop for Refrigerant 22 Condensing in Horizontal Tubes," Chemical Measurements," International Journal of Multiphase Flow, Vol. 15, No. 2, pp.
Engineering Progress Symposium Series, Vol. 56, No. 30, pp. 151-159. 209-226,
Bae, S., Maulbetsch, J. S„ and Rohsenow, W. M., 1970, "Refrigerant Forced- Lockhart, R. W„ and Martinelli, R. C , 1947, "Proposed Correlation of Data for
Convection Condensation Inside Horizontal Tubes," Report No. DSR 72591-71, Isothermal, Two-Phase, Two-Component Flow in Pipes," Chemical Engineering
Massachusetts Institute of Technology, Cambridge, MA. Progress, Vol. 45, No. 1, pp. 39-48.
Baker, O., 1954, "Simultaneous Flow of Oil and Gas," Oil and Gas Journal, Mandhane, J. M., Gregory, G. A., and Aziz, K„ 1974, "A Flow Pattern Map
Vol. 53, pp. 185-195. for Gas-Liquid Flow in Horizontal Pipes," International Journal of Multiphase
Barnea, D„ Shoham, O., Taitel, Y., and Dukler, A. E., 1980, "Flow Pattern Flow, Vol. 1, pp. 537-553.
Transition for Gas-Liquid Flow in Horizontal and Inclined Pipes: Comparisons Moffat, R. J., 1988, "Describing the Uncertainties in Experimental Results,"
of Experimental Data with Theory," International Journal of Multiphase Flow, Experimental Thermal and Fluid Science, Vol. 1, pp. 3-17.
Vol. 6, pp. 217-225. Nicholson, M. K., Aziz, K., and Gregory, G. A., 1978, ' 'Intermittent Two Phase
Barnhart, J., 1992, "An Experimental Investigation of Flow Patterns and Liquid Flow: Predictive Models," 27th Canadian Chemical Engineering Conference,
Entrainment in a Horizontal-Tube Evaporator," Ph.D. thesis, Dept. of Mechanical Calgary, Alberta.
and Industrial Engineering, University of Illinois at Urbana-Champaign. Nusselt, W., 1916, "Die Oberflachenkondensation des Wasserdampfes," Z.
Breber, G., 1988, "Intube Condensation," Heat Transfer Equipment Design, Verein Deutscher Ingeniuere, Vol. 60, pp. 541 -575.
R. K. Shah, E. C. Subbarao, and R. A. Mashelkar, eds., Hemisphere Publishing Palen, J. W„ Kistler, R. S., and Yang, Z. F., 1993, "What We Still Don't Know
Corp., New York, pp. 1-23. about Condensation in Tubes," ASME Symposium, Condensation and Condenser
Bromley, L. A., 1952, "Effect of Heat Capacity on Condensation," Industrial Design, ASME, NY, pp. 19-53.
and Engineering Chemistry, Vol. 44, pp. 2966-2969. Rabas, T. J. and Minard, P. G„ 1987, "Two Types of Flow Instabilities Oc-
Butterworth, D., 1972, "A Visual Study of Mechanisms in Horizontal, Air- curring Inside Horizontal Tubes with Complete Condensation," Heat Transfer
Water Flow," AERE Report M2556, Harwell, England. Engineering, Vol. 8, No. 1, pp. 40-49.
Carey, V. P., 1992, Liquid-Vapor Phase Change Phenomena: An Introduction Rohsenow, W. M„ 1956, "Heat Transfer and Temperature Distribution in Lam-
to the Thermophysics of Vaporization and Condensation Processes in Heat Trans- inar Film Condensation," Transactions ASME, Vol. 78, pp. 1645-1648.
fer Equipment, Hemisphere, New York. Rosson, H. F„ and Meyers, J. A., 1965, "Point Values of Condensing Film
Carpenter, E. F. and Colburn, A. P., 1951, "The Effect of Vapor Velocity on Coefficients Inside a Horizontal Tube," Chemical Engineering Progress Sympo-
Condensation Inside Tubes," Proceedings of General Discussion on Heat Trans- sium Series, Vol. 61, No. 59, pp. 190-199.
fer, Inst. Mech. Eng. and ASME, pp. 20-26. Schlager, L. M„ Pate, M. B., and Bergles, A. E., 1990, "Performance Predic-
Cavallini, A., and Zecchin, R., 1974, "A Dimensionless Correlation for Heat tions of Refrigerant-Oil Mixtures in Smooth and Internally Finned Tubes,'' ASH-
Transfer in Forced-Convective Condensation," Proceedings of the Fifth Interna- RAE Transactions, Vol. 96, Part 1, pp. 161-182.

212 / Vol. 120, FEBRUARY 1998 Transactions of the ASME

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Shah, M. M., 1979, "A General Correlation for Heat Transfer During Film Sweeney, K. A., Chato, J. C , Ponchner, M., and Rhines, N. L., 1995, "The
Condensation Inside Pipes," International Journal of Heat and Mass Transfer, Effect of Oil on Condensation in a Microflnned Tube," ACRC Technical Report
Vol. 22, pp. 547-556. 87, University of Illinois at Urbana-Champaign.
Soliman, H. M„ 1982, "On the Annular-to-Wavy Flow Pattern Transition Dur- Taitel, Y., and Dukler, A. E., 1976, "A Model for Predicting Flow Regime
ing Condensation Inside Horizontal Tubes," The Canadian Journal of Chemical Transitions in Horizontal and Near Horizontal Gas-Liquid Flow," American Insti-
Engineering, Vol. 60, pp. 475-481. tute of Chemical Engineering Journal, Vol. 22, No. 1, pp. 47-55.
Soliman, H. M., 1983, "Correlation of Mist-to-Annular Transition During Conden- Tien, C. L., Chen, S. L., and Peterson, P. F., 1988, "Condensation Inside
sation," The Canadian Journal of Chemical Engineering, Vol. 61, pp. 178-182. Tubes," EPRI Project 1160-3 Final Report, EPRI, Palo Alto, CA.
Soliman, H. M„ 1986,' 'The Mist-Annular Transition During Condensation and Traviss, D. P., Rohsenow, W. M., and Baron, A. B., 1973, "Forced-Convective
its Influence on the Heat Transfer Mechanism," International Journal of Multi- Condensation in Tubes: A Heat Transfer Correlation for Condenser Design,"
phase Flow, Vol. 12, No. 2, pp. 277-288. ASHRAE Transactions, Vol. 79, No. 1, pp. 157-165.
Soliman, H. M., Schuster, J. R., and Berenson, P. J., 1968, "A General Heat Wang, S.-P., and Chato, J. C , 1995, "Review of Recent Research on Heat
Transfer Correlation for Annular Flow Condensation," ASME JOURNAL OP HEAT Transfer with Mixtures—Part 1: Condensation," ASHRAE Trans., Vol. 101, No.
l , p p . 1376-1386.
TRANSFER, Vol. 90, pp. 267-276.
Wattelet, J. P., 1994, "Heat Transfer Flow Regimes of Refrigerants in a Hori-
Souza, A. M„ Chato, J. C , and Wattelet, J. P., 1992, "Pressure Drop During zontal-Tube Evaporator," Ph.D. thesis, Dept. of Mechanical and Industrial Engi-
Two-Phase Flow of Refrigerants in Horizontal Smooth Tubes," ACRC Technical neering, University of Illinois at Urbana-Champaign.
Report 25, University of Illinois at Urbana-Champaign. Wattelet, J. P., Chato, J. C , Christoffersen, B. R„ Gaibel, J. A., Ponchner, M„
Sweeney, K. A., 1996, "The Heat Transfer and Pressure Drop Behavior of a Kenney, P. J., Shimon, R. L., Villaneuva, T. C , Rhines, N. L., Sweeney, K. A.,
Zeotiopic Refrigerant Mixture in a Microflnned Tube," M.S. thesis, Dept. of Allen, D. G., and Hershberger, T. T., 1994, "Heat Transfer Flow Regimes of
Mechanical and Industrial Engineering, University of Illinois at Urbana-Cham- Refrigerants in a Horizontal-Tube Evaporator," ACRC Technical Report 55, Uni-
paign. versity of Illinois at Urbana-Champaign.
Sweeney, K. A. and Chato, J. C , 1996, "The Heat Transfer and Pressure Drop Zivi, S. M„ 1964, "Estimation of Steady-State Steam Void-Fraction by Means
Behavior of a Zeotropic Refrigerant Mixture in a Microflnned Tube," ACRC of the Principle of Minimum Entropy Production," ASME JOURNAL OF HEAT
Technical Report 95, University of Illinois at Urbana-Champaign. TRANSFER, Vol. 86, pp. 247-252.

Journal of Heat Transfer FEBRUARY 1998, Vol. 120 / 213

Downloaded 21 Mar 2010 to 128.189.96.13. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm