International Journal of Heat and Mass Transfer 127 (2018) 555–567

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Experimental study of buoyancy effect and its criteria for heat transfer of
supercritical R134a in horizontal tubes
Ran Tian a,b, Yue Zhang a, Yuezheng Ma a, Hui Li a, Lin Shi a,⇑
a
Key Laboratory for Thermal Science and Power Engineering of Ministry of Education, Department of Energy and Power Engineering, Tsinghua University, Beijing 100084, China
b
School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, China

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

Article history: Supercritical organic Rankine cycles (ORCs) are a promising waste heat recovery technology because of
Received 2 February 2018 their high thermal efficiency and lower exergy loss. Heat transfer characteristics must be considered
Received in revised form 24 July 2018 for vapor generator design in ORCs. This paper focuses on the buoyancy effect on heat transfer character-
Accepted 18 August 2018
istics in horizontal tubes of supercritical R134a, a widely used ORC working fluid, whose supercritical
Available online 23 August 2018
heat transfer characteristics have been less widely studied. Experiments of supercritical R134a flowing
in horizontal tubes with different inner diameters of 10.3 mm and 16 mm were conducted to obtain basic
Keywords:
experimental data. The influences of heat flux, mass flux, and tube diameter on heat transfer were inves-
Experiment
Buoyancy effect
tigated with an emphasis on the buoyancy effect. This study attempts to extend the applicability of exist-
Supercritical heat transfer ing buoyancy criteria to organic fluids in horizontal tubes to quantitatively analyze the buoyancy effect.
R134a Existing buoyancy criteria, which were developed for water, were evaluated and the threshold values
Horizontal tubes determining the onset of the buoyancy effect in heated horizontal tubes were determined for the present
data. A simple parameter of qd 0.7/G1.2 not based on wall temperature was developed to be correlated
with the maximum temperature difference between the top and bottom surfaces for convenience in engi-
neering use. This parameter was validated with both the present data and supercritical water data from
the literature.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction et al. [7] have comprehensively reviewed the experimental studies

Supercritical organic Rankine cycles (ORCs) are receiving con-
siderable attention owing to their advantages in the utilization of
renewable energy and low-grade waste heat energy, such as higher
thermal efficiency, lower exergy loss, and compact components
[1,2]. In the vapor generator of a supercritical ORC unit, the work-
ing fluid is heated by the heat source at supercritical pressures,
where the thermophysical properties of the working fluid undergo
significant variations (as shown in Fig. 1), leading to unique and
complex heat transfer phenomena [3]. Therefore, investigation of
the supercritical heat transfer characteristics of organic fluids is
of considerable importance for the vapor generator design of ORC
systems.
Supercritical heat transfer has mostly been studied in the field
of supercritical water-cooled reactors and fuel-fired boilers; thus,
previous studies mainly focused on water and CO2 in vertical
heated tubes. Pioro and Duffey [4,5], Huang et al. [6], and Cabeza
⇑ Corresponding author.
E-mail address: rnxsl@mail.tsinghua.edu.cn (L. Shi).
on water and CO2, showing that the experimental parameters
cover wide ranges: pressures of 22.5–35 MPa, heat fluxes of
0.018–9.44 MW/m2, mass fluxes of 100–10,000 kg/m2s, and inner
diameters of 1.27–38.1 mm for water; and pressures of 7.4–21
MPa, heat fluxes of 3.5–2600 kW/m2, mass fluxes of 100–4170
kg/m2s and inner diameters of 0.099–29 mm for CO2. However,
there is a lack of heat transfer experimental data for organic fluids
at supercritical pressures compared with water and CO2. Limited
studies can be found in the literature for organic fluids, the
majority of which were conducted in vertical tubes, such as
R134a [8–10], R12 [11], and R22 [12–14] heated in vertical tubes.
The vapor generators in ORC units are commonly horizontally
oriented, but unfortunately supercritical heat transfer of organic
fluids in heated horizontal tubes has been seldom studied. R134a
[15], 410a [16], and 404a [17] have been studied in horizontal
tubes, but in a cooling process. Therefore, the development of a
supercritical ORC raises new requirements for the study of super-
critical heat transfer.
The buoyancy effect due to a density gradient is the most
important factor for interpreting heat transfer mechanisms at

https://doi.org/10.1016/j.ijheatmasstransfer.2018.08.072
0017-9310/Ó 2018 Elsevier Ltd. All rights reserved.
556 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567

Nomenclature

Bo Jackson buoyancy criterion for vertical flow, ¼ Grb =Re2:7

b Greek symbols
BuJ Jackson buoyancy criterion for horizontal flow, b volumetric expansion coefficient (1/K)

 
¼ Gr b Re2 qb x 2 q density (kg/m3) RT
b q d
w
q
average density, ð1=ðT w  T b ÞÞ T bw qdT (kg/m3)
cp specific heat (kJ/kgK) l dynamic viscosity (lPas)
d diameter (m) k thermal conductivity ðW  m1  K1 Þ
G mass flux ðkg  m2  s1 Þ m kinetic viscosity (m2/s)
Grb Grashof number based on bulk density,
3
¼ gd ðqb  qw Þ=qb m2b Subscripts
Gr b Grashof number based on average density b bulk
3
¼ gd ðqb  q
Þ=qb m2b c critical point
Gr

4
Grashof number based on heat flux ¼ gbb qw d =kb m2b cal calculated
h enthalpy (kJ/kg) exp experimental
Kv acceleration criterion, ¼ 4qw dbb =ðRe2 lb cp;b Þ top top surface
Nu Nusselt number bottom bottom surface
p pressure (MPa) pc pseudo-critical point
Pr Prandtl number fc forced convection
Pr average Prandtl number w Wall
q heat flux (kW/m2) in inner
Re Reynolds number
T temperature (K) Abbreviations
DT w wall temperature difference between the top and bot- DB Dittus–Boelter correlation
tom surface HTC heat transfer coefficient
x axial location of the tube ORC organic Rankine cycle

80
40 1200 Tpc 160
buoyancy force [29], which is a special phenomenon that differs
(mW/m-K) from that of vertical flows. The top surface wall temperature is
35 70 140
1000 always higher than that of the bottom surface in buoyancy-aided
30
60
120 flow. Heat transfer on the top surface will be impaired at high
25 800 100 q/G conditions, leading to a local wall temperature peak, which is
viscosity
20 50 similar to that in vertical flows [30].
600 ( Pa-s) 80
15
For the purpose of quantitative analysis, various criteria for
3
400
desnity(kg/m ) 40 60 determining the onset of buoyancy effect in vertical heated tubes
10
cp(kJ/kg-K) 40 were proposed, such as Grb/Re2 [29,31], Gr
=ðRe3:425 Pr 0:8 Þ [19],
5 200 30
0:5

20 lw
0
Gr b =Re2:7 [19,32], and Gr b qqb 0:4 2:625
l =ðPr w Reb Þ [33]. Li [34]
w b
0 20 0
and Huang et al. [6] have reviewed the buoyancy parameters.
40 60 80 100 120 140 160
o Among these buoyancy parameters, Bo ¼ Gr b =Re2:7 (or
T( C)

3:425 0:8
Bo ¼ Gr =ðRe Pr Þ, which is equivalent to Bo) obtained from
Fig. 1. Thermophysical properties variations of R134a at 4.2 MPa and data is taken the theory of Hall and Jackson [24] appears to be the most appro-
from NIST Refprop database [44]. priate parameter and has been widely tested. Jackson noted that
the effect of buoyancy on heat transfer would be >5% when
Gr b =Re2:7 > 105 . Li [34] demonstrated that the mixed convection
supercritical pressures in both vertical and horizontal tubes data of air in vertical flows (both upwards and downwards) was
[18,19]. In vertical tubes, at a lower ratio of heat flux to mass flux well described by the buoyancy parameter of Bo. Licht et al. [35]
(q/G, kJ/kg), forced convection heat transfer is enhanced near the showed that Bo can distinguish forced convection from mixed con-
pseudo-critical point because much of the boundary layer is cov- vection for water in vertical flows. Jiang’s [36] results indicated
ered with a high-specific-heat fluid, which is called the integrated that Bo was consistent with CO2 experimental data for upward
effect of specific heat [20–22]. In mixed convection with a rela- flows. However, Liu [37] found that the threshold value of Bo⁄ for
tively high q/G, local heat transfer deterioration (HTD) [23] hap- n-decane was much smaller than that proposed by Jackson. Exper-
pens with a sharp wall temperature peak. Hall and Jackson [24] imental results of water in horizontal tube by Bazargan et al. [29]
proposed a two-layer density theory to explain HTD. The buoyancy showed that Bo was not applicable to horizontal flows. Deng et al.
force due to a radial density gradient accelerates the flow near the [38] found that Bo was inaccurate in predicting the buoyancy force
wall and further modifies the shear stress distribution, which con- of RP-3 kerosene. Therefore, further examination is required to see
sequently causes ‘‘relaminarization” in turbulent flow. As a result, whether these criteria are applicable for organic fluids in horizon-
the heat transfer ability is reduced. Computational results have tal flows.
proven this theory [25]. Experimental results have shown that Compared with vertical flows, studies on buoyancy criteria for
the heat flux, mass flux, and tube diameter affect the buoyancy horizontal flows are relatively limited. Petukhov and Polyakov
intensity [26–28]. In horizontal flows, significant temperature vari- [39] have proposed a buoyancy criterion for horizontal flows of
ation was observed in the circumferential direction caused by the Grq/Grth. When Grq/Grth > 1, the buoyancy effect is expected to
 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567 557

x2
q 
become significant. Jackson used BuJ ¼ Gr b Re2
b d
b
qw < 10 [18] 2. Experiment facilities and data reduction
to show the absence of buoyancy effects. Inconsistent results from
2.1. Experimental loop
the evaluation of these buoyancy criteria were reported. Bazargan
[29] demonstrated that Grq/Grth could show the buoyancy effect for
A schematic of the experimental loop, which consisted of the
water effectively in horizontal flows, even a small effect can be
main R134a circulating loop, cooling system, and pressure control
detected by the criterion of Petukhov and Polyakov. Yu et al. [40]
system, is shown in Fig. 2. Supercritical R134a was driven by the
found that both Grq/Grth and Grb/Re2 could predict buoyancy effects
circulating pump to flow through the loop. One part of the fluid
when HTD occurred on the top surface for supercritical water, but
was heated in the preheater to the desired inlet temperature. Then,
neither of the two criteria were adequate when heat transfer
the fluid flowed into the test tube where it was heated with a uni-
enhancement occurred on the top surface. Experimental results
form heat flux at a specific mass flux. The other part of the fluid
by Adebity and Hall [18] showed that BuJ and Grq/Grth were too
was not heated and flowed through the bypass directly to the plate
conservative in predicting the buoyancy effect for supercritical
heat exchanger inlet, where it mixed with the high-temperature
CO2 in a horizontal tube. Tanimizu and Sadr [41] concluded that
fluid from the test tube. The design of the bypass effectively
Grb/Re2 was more adequate for predicting the buoyancy effect than
reduced the cooling load on the heat exchanger and enlarged the
BuJ, however, results in [42] showed that BuJ gave improved predic-
adjustment range of the mass flow rate. After mixing, the fluid
tion compared to Grb/Re2. Thus, these criteria need to be re-
was cooled by water in the plate heat exchanger to reach a safe
examined for organic fluids.
temperature for the circulating pump and the flowmeter.
R134a is a widely used working fluid in ORC systems owing to
The main circulating pump was a key component in the loop
its high thermodynamic performance [43], whose property varia-
and should be chosen carefully. A shielded centrifugal pump
tions are shown in Fig. 1. Properties such as density and dynamic
designed for organic fluids was used because it provided a good
viscosity undergo a significant drop within a very narrow temper-
seal, a stable mass flow, and reduced pressure pulsations in the
ature range, while the specific heat and thermal conductivity have system. The preheater and test tube were heated electrically by a
a peak near the pseudocritical points. In this study, the buoyancy DC power supply with a large current and low voltage. The super-
effect on the heat transfer characteristics of supercritical R134a critical pressure was provided by a bladder-type pressurizer con-
in horizontal tubes was experimentally investigated. The influence nected to a nitrogen cylinder. The charging and discharging
of heat flux, mass flux, and tube inner diameter on heat transfer are processes of the nitrogen were completed by a set of electromag-
discussed in detail with an emphasis on the buoyancy effect. To netic valves, which could control the system pressure to the
quantitatively analyze the buoyancy effect, existing buoyancy cri- desired value. The pressurizer can also damp out pressure pulses
teria were evaluated. This study attempted to extend the applica- through the contraction and expansion of the bladder.
bility of these buoyancy criteria to organic fluids and determine A Coriolis mass flowmeter (Krohne Optimass 6000) was used to
their threshold values for the onset of the buoyancy effect in measure the mass flow rate. The pressure was measured by pres-
heated horizontal tubes. Finally, a simple parameter correlated sure transmitters (Yokogawa EJA510A). The temperatures in the
with the maximum temperature difference between the top and loop were measured with Pt100 sensors. The currents through
bottom surfaces was developed for convenience in engineering the test section and preheater were measured by current trans-
use. formers and the voltage was measured by an Agilent 34972A.

10.3 mm test section

P T T P
Cooling tower
16 mm test section

Water bypass
Bypass control
T

DC T
T

Water tanke power

T

plate
heat
P

Y-filter Electric preheater

exchanger N2
N2
supply
Water pump Pressurizer cylinders P
P T T P
Filter Coriolis
Water flow meter flow meter
Circulating pump

Low-temperature R134a Ball valve Pressure safety valve

Flow rate control valve P Pressure transmitter
High-temperature R134a
R134a charging valve
Cooling water T Temperature sensor (pt100)
Electromagnetic valve

Fig. 2. Schematic of the experimental system.

558 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567

The main technical parameters of the experimental system are
listed in Table 1.
2.2. Test sections
Two test sections of 316L stainless steel with different inner
diameters of 10.3 mm (1.2-mm-thick) and 16 mm (1.5-mm-
thick) were used in the experiments. The arrangement of the two
test sections was the same as shown in Fig. 3. Supercritical
R134a flowed horizontally through the test section. The heating
length was 2500 mm. A 500 mm non-heating section was set
before the heating electrodes at the inlet for the fully developed
flow and a 100 mm non-heating section after the heating elec-
trodes to avoid the influence of the elbows at the outlet. The inlet
and outlet temperatures of the fluid were measured with Pt100
sensors inserted in the tube after the mixing section. Differential
pressure transducers (Yokogawa EJA110A) were used to measure
the pressure drop. 118 T-type (Omega) thermocouples were
attached to the outer wall surface at 32 axial locations to measure
the wall temperature distribution. Types A and B were alternated
along the tube every 70 mm. There were two locations that used
type C to examine the temperature symmetry around the tube.
2.3. Experimental procedure
The experiments were conducted at various pressures, mass
fluxes, heat fluxes, and inlet temperatures. The first step was to
increase the pressure stepwise using the pressurizer. Then, the
cooling system and the circulating pump were started. The mass
flow rate was controlled by the pump frequency converter together
with the control valve. The required inlet temperature was
achieved by carefully adjusting the current and voltage of the pre-
heater power supply. The heat flux was controlled by adjusting the
Table 1
Technical parameters of the test loop.
Parameter Value
Design pressure 6 MPa
Design temperature 300 °C
Max. heating power for test section 20 kW
Max. heating power for preheater 40 kW
Mass flow rate 0–2 m3/h
Cooling capacity 50 kW
current and voltage of the test section power supply. When the
parameters became stable, the data was recorded by the data
acquisition system. Tests with the same pressure, mass flux, and
heat flux, but different inlet temperatures, were conducted to col-
lect experimental data covering a wide enthalpy range.
2.4. Data reduction and uncertainty analysis
In the present study, pressures were in the range 4.14–4.87 MPa
(1.02–1.2 pc), which is the optimal pressure range in an ORC sys-
tem [2]. The mass flux was from 400 to 1500 kg/m2s and the heat
flux varied from 10 to 110 kW/m2 with an inlet temperature of 70–
95 °C.
The local heat transfer coefficient (HTC) was calculated as:
q q
HTC ¼ ¼ ð1Þ
DT T w;in  T b
The heat flux q (kW/m2) in Eq. (1) was calculated as:
Q e UI  Q loss
q¼ ¼ ð2Þ
Ain pdL
where Qe is the heating power and Ain is the inside surface area of
the test tube. Qloss is the heat loss, which was measured before
the experiments by first evacuating the tube, and it was found to
be <1% of the power applied to the tube. U and I are the voltage
and current applied to the test tube. L is the heating length.
The local bulk temperature, Tb, in Eq. (1) was determined by the
local bulk enthalpy and the pressure using the database of NIST
Refprop, i.e. T = f (h, p). The bulk enthalpy was calculated from
the energy balance equation as:
qx
hb ¼ hin þ 4 ð3Þ
Gd
where the inlet enthalpy, hin, was determined from the inlet tem-
perature and pressure. x is the distance from the start of the heating
section and G is the mass flux (kg/m2 s).
The inner wall temperature, Tw,in, in Eq. (1) was calculated from
the outer wall temperature and the heat flux as:
2 2
! 2  
q do d q d do
T w;in ¼ T w;out þ v   v o ln ð4Þ
4k 4 4 2k 4 d
where qv is the volumetric heat generation rate, do is the outer
diameter, and k is the thermal conductivity of the solid wall.
The uncertainties were then calculated using Eq. (5):

P
DC power

PT100 100 2500 500 PT100

T T
Heated length Inlet

thermocouples 70 70 70 185
Isolation Isolation
B A B A
electrode electrode

top

bottom
A B C

Fig. 3. Test section.

 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567 559

"  2 #1=2
dy Xn
@lnf 6
¼ dxi ð5Þ
y i¼1
@xi
5
The uncertainty of the HTC was calculated as:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

HTCcal (kW/m K)
 2  2 4
d DT

2
dðHTCÞ dq
¼ þ ð6Þ
HTC q DT
3
According to Eqs. (2)–(5), the uncertainties of the heat flux and
the temperature difference are calculated by:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2
 2  2  2  2  2 Gungor & Winterton
dq dd dL dU dI dQ loss
¼ þ þ þ þ ð7Þ Kandlikar
q d L U I Q loss 1
1 2 3 4 5 6
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2
HTCexp (kW/m K)
dDT ðdT w;in Þ2 þ ðdT b Þ2
¼ ð8Þ
DT T w;in  T b Fig. 4. Comparison of measured HTC with predictions of correlations [45,46].

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
" !
u
u 2 2  #2
t 2 1 do din 1 2
dT w;in ¼ ðdT w;o Þ þ   do ln
4k 4 4 8k
The uncertainty of the bulk fluid temperature was estimated
from Eq. (3). In this research, the relevant uncertainties of the mea-
sured temperature, pressure and mass flux are given in Table 2. The
temperature difference between the bulk fluid and the inner wall is
>5 K. Then, the maximum uncertainty of the HTC is 11.03%.
2.5. Validation
As no experimental data for supercritical R134a heated in hor-
izontal tubes is found in the literature, two-phase boiling experi-
ments for R134a, which have been extensively studied in the
field of air-conditioners and heat pumps, were conducted at vari-
ous conditions to validate the present experimental system. Some
well-validated heat transfer correlations have been developed and
the experimental HTC was compared with the calculated data by
using the Gungor and Winterton correlation [45] and the Kandlikar
correlation [46]. The parameters of the boiling experiment were
p = 0.6–1.2 MPa, q = 2–12 kW/m2, G = 300–1000 kg/m2s, and vapor
qualities of 0.1–0.4, which are well within the validated ranges of
the two correlations. The validation results are shown in Fig. 4 with
the average error within 10%. This indicates that the test system
used in this study is reliable and convincing.
3. Heat transfer characteristics
In this section, the effects of heat flux, mass flux, and tube diam-
eter size on the heat transfer characteristics in horizontal tubes are
discussed with an emphasis on the buoyancy effect. The experi-
mental Nusselt number (Nuexp) was normalized by the Dittus–
Boelter correlation (NuDB) [47] to determine whether the heat
transfer is enhanced or deteriorated. Nuexp/NuDB > 1 means that
heat transfer is enhanced, otherwise it is deteriorated.
Table 2
Uncertainties of various parameters in the experiments.
Parameter
Pressure, p
Tin, Tout
Mass flow rate, G
Outer wall temperature, Tw,out
Current, I
Voltage, U
HTC
do 3.1. Heat flux effect
ðdqv Þ2 ð9Þ
din
To analyze the heat flux effect on heat transfer, wall tempera-
tures and HTCs at various heat fluxes with constant mass flux are
shown in Fig. 5. For the top surfaces, as shown in Fig. 5a with
increasing heat flux, the wall temperature increases and the nor-
malized Nusselt number decreases. At a relatively low heat flux
of q = 10 kW/m2, the wall temperature profile is flat, while at
higher heat fluxes of 40 and 60 kW/m2 there exist wall tempera-
ture peaks on the top surface. Correspondingly, Nuexp/NuDB is much
smaller than 1, indicating that the heat transfer deteriorated for
these conditions.
For the bottom surfaces, as shown in Fig. 5b, the wall tempera-
ture also increases with increasing heat flux, but unlike the top sur-
face, the bottom wall temperature increases slowly with the bulk
enthalpy without a temperature peak. This is because the heat
transfer is enhanced on the bottom surface as shown by
Nuexp/NuDB, which is >1. The heat transfer enhancement is reduced
with an increasing heat flux. When the heat flux increases to
60 kW/m2, Nuexp/NuDB is near 1, which means the heat transfer
on the bottom becomes normal and is no longer enhanced owing
to the high heat flux. Even though Nuexp/NuDB fell below 1.0 near
the pseudo-critical point (hpc), the heat transfer is not considered
to be deteriorated here on the bottom surface. Owing to the sharp
property variations, the DB correlation usually over-predicts the
HTC near the pseudo-critical point. Therefore, the values of
Nuexp/NuDB for all conditions are smaller than 1 near the pseudo-
critical enthalpy (hpc).
By comparing the heat transfer characteristics on the top and
bottom surfaces, it can be seen that the top wall temperature is
always higher than that on the bottom surface, and the tempera-
ture difference between the top and bottom surfaces increases
with increasing heat flux. The heat transfer ability on the bottom
surface is stronger than that on the top surface. The difference in
heat transfer ability between the top and bottom surfaces is caused
by the buoyancy effect, as reported in the literature [28–31]. If
there is no buoyancy force, the temperature on the top and bottom
surfaces should be consistent. For supercritical pressure fluids, the
significant density changes lead to a strong buoyancy force. The
Uncertainty
heated fluid of high temperature and low density near the tube
wall flows up to the top region of the tube owing to the buoyancy
0.2%
0.1 °C
force; the low temperature fluid with high density moves down to
0.2% the bottom part of the tube, leading to a secondary flow in the
±0.5 °C cross section. Chu [48] demonstrated that this buoyancy-driven
0.5% secondary flow restrained the turbulent energy in the top part,
0.5%
while the turbulence is enhanced in the bottom part, raising the
11.03%
non-uniform wall temperature profile in the circumferential
560 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567

300 320 340 360 380 400 420 440 460 300 320 340 360 380 400 420 440 460
520 520
Top surface Bottom surface
480 480
hpc

Tw,in(K)

Tw,in(K)
440 440

400 400

360 360
2
2.5
2
2.5 G=400 kg/m s
G=400 kg/m s hpc 2
2
q=10 kW/m
2.0 q=10 kW/m 2.0 q=20 kW/m
2
2
q=20 kW/m q=40 kW/m
2
2
1.5 q=40 kW/m 1.5 q=60 kW/m
2

Nuexp/NuDB
2
Nuexp/NuDB

q=60 kW/m
1.0 1.0

0.5 0.5

0.0 0.0
300 320 340 360 380 400 420 440 460 300 320 340 360 380 400 420 440 460
hb(kJ/kg) hb(kJ/kg)
(a) (b)
Fig. 5. Wall temperature and normalized Nusselt number at various heat fluxes with din = 10.3 mm, p = 4.26 MPa, and G = 400 kg/m2s for (a) the top surfaces and (b) the
bottom surfaces.

direction and leading to HTD at the top at high heat fluxes. From
the experimental results, it can be seen that heat flux is an impor-
tant parameter that influences the buoyancy intensity. Higher heat
flux leads to larger temperature deviation between the top and
bottom surfaces, indicating a stronger buoyancy effect.
3.2. Mass flux effect
Mass flux effects are shown in Fig. 6. An increase in mass flux has
a similar effect on heat transfer as that of a decrease in heat flux. For
the top surfaces, with an increasing mass flux, the wall temperature
decreases and the normalized Nusselt number increases from
below unity to larger than unity. At G = 1500 kg/m2 s, Nuexp/NuDB
is >1, indicating that the heat transfer on the top surface can also
be enhanced. For the bottom surfaces, the heat transfer is signifi-
cantly improved at mass fluxes of 1000 and 1500 kg/m2s with
Nuexp/NuDB much >1. The mechanisms of the heat transfer enhance-
ment can be ascribed to the combination of the buoyancy effect and
the integrated effect of the specific heat, as indicated in [49]. These
experimental results reveal that mass flux is also an important fac-
tor affecting the heat transfer, with a higher mass flux leading to a
weaker buoyancy effect. With very low q/G, i.e. the buoyancy effect
can be neglected, the circumferential temperature will be uniform
and heat transfer can be enhanced on both the top and bottom
surfaces.
various experimental conditions. The experimental results show
that the effect of tube diameter on the heat transfer differs at various
q/G values. In the case of G600-q20, which has a relatively low q/G
value of 0.03 kJ/kg, there is no obvious difference in the wall temper-
ature with an almost overlapped HTC of the two tubes. In the case of
G600-q30, which has an increased q/G of 0.05 kJ/kg, obvious wall
temperature differences occur on the top surfaces of the two tubes
with Tw of the 16-mm tube higher than that of the 10.3-mm tube.
However, the wall temperature on the bottom surface of the two
tubes are close to each other. With q/G increasing to 0.1 kJ/kg, as in
the case of G600-q60, the Tw difference on the top surfaces of the
two tube is further enlarged to 40 K at the Tw peaks. HTD on the
top surface is more severe with larger tube diameters. Even though
a large Tw difference exists between the top surfaces, the bottom
wall temperatures are still almost consistent in the two tubes.
It can be seen that the tube diameter mainly affects the wall
temperature on the top surfaces, and it has little influence on the
bottom surfaces. Heat transfer of supercritical flow was prone to
deteriorate in the larger tube. The tube diameter effect on heat
transfer can be explained by the tube diameter effect on the buoy-
ancy effect. In tubes with larger diameter the buoyancy has more
space to develop, leading to a stronger buoyancy force and greater
temperature nonuniformity in the circumferential direction. Simi-
lar results have also been reported in vertical flows [20] where heat
transfer deteriorates more easily in larger tubes.

3.3. Tube diameter size effect

Besides the effects of heat flux and mass flux, tube diameter is
another important parameter influencing the heat transfer charac-
teristics and buoyancy intensity in supercritical horizontal flows.
Fig. 7 shows the wall temperature and HTC of two horizontal
tubes with different inner diameters of 10.3 mm and 16 mm at
4. Buoyancy effect criterion
4.1. Evaluation of the buoyancy criteria and the onset values
To quantitatively evaluate the buoyancy effect on heat transfer,
a proper buoyancy criterion is needed to alert the investigator as to
 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567 561

300 320 340 360 380 400 420 440 300 320 340 360 380 400 420 440
480 480
Top surfaces Bottom surfaces
460 460

440 440

Tw,in(K)
2
q=40 kW/m

Tw,in(K)
420 420

400 400

380 380

360 360
2
2.5 2.5 G=400 kg/m s
hpc
2
q=40 kW/m 2
G=600 kg/m s
hpc
2
G=400 kg/m s 2
2.0 2 2.0 G=1000 kg/m s
G=600 kg/m s 2
2 G=1500 kg/m s
G=1000 kg/m s
1.5 2
G=1500 kg/m s 1.5
Nuexp/NuDB

Nuexp/NuDB
1.0 1.0

0.5 0.5

0.0 0.0
300 320 340 360 380 400 420 440 300 320 340 360 380 400 420 440
hb(kJ/kg) hb(kJ/kg)
(a) (b)
Fig. 6. Wall temperature and normalized Nusselt number at various mass fluxes with din = 10.3 mm, p = 4.26 MPa, and q = 40 kW/m2 for (a) the top surfaces and (b) the
bottom surfaces.

whether buoyancy may have a significant influence on heat trans- data are scattered abundantly. Therefore, BuJ is discussed no further
fer and, if so, how much it may affect heat transfer. This study in this section.
chooses the three most widely used criteria to evaluate their appli- Petukhov and Polyakov [39] proposed using Grq/Grth to evaluate
cability for the present data of supercritical R134a in horizontal the buoyancy effect on heat transfer for supercritical heat transfer
tubes. Before analyzing the buoyancy effect, the thermal accelera- in horizontal tubes. The buoyancy effect is expected to be signifi-
tion effect is evaluated by the Kv number [23]. Kv in the present cant when Grq/Grth > 1.
experiments ranges from 0.13 to 4.08  108, two orders of magni- 4
tude smaller than the threshold value of 3  106, which indicates Grq ¼ gbqw d =kb m2b and ð13Þ
that the thermal acceleration effect is quite weak and can be !
neglected in the present experiments. Pr 2=3  1
Grth ¼ 3  105 Pr0:5 Re2:75
b 1 þ 2:4 ð14Þ
Grb/Re2 represents the ratio of the buoyancy force to the inertial Re1=8
b
force and is used as a criterion to determine the free convection
1 qb qw e
effect [29,31]. When Grb/Re2 > 0.01, the effect of free convection where b ¼ ; q at T ¼ T w þT b
; Pr ¼ Thww h b lb
.The ranges of the
e
q T w T b 2 T b kb
cannot be ignored. Jackson proposed a different threshold value three buoyancy criteria in the present experiments are listed in
for horizontal flows; he stated that when Grb/Re2 < 103 [50], buoy- Table 3.
ancy has no effect on heat transfer in horizontal flows. The thresh- To evaluate the three criteria in Eqs. (10), (11), (13), and (14),
old value of this criterion will be evaluated in this section. the profiles of the three criteria together with the wall tempera-
Grb =Re2 < 0:01; or 103 ; ð10Þ tures are shown in Fig. 8. Fig. 8a shows the case with HTD. The
Grq/Grth criterion successfully predicts the weak buoyancy effect
3
where Gr b ¼ gd ðqb  qw Þ=q m Re ¼ Gd=lb .Jackson gave the
2
b b,
at the inlet with a lower value; then it exhibits a maximum value
following criteria through theoretical analysis [19] for negligible of 250 before the Tw peak. Afterwards, Grq/Grth decreases and
buoyancy effect in vertical and horizontal flows: reaches the minimum value of 10 at a higher enthalpy of
440 kJ/kg. The Grb/Re2 and Bo criteria have similar trends by which
5
Vertical : Bo ¼ Grb =Re2:7
b < 10 and ð11Þ they monotonously decrease with the bulk enthalpy but failed to
  predict the weak buoyancy at the inlet. Values of all the three
qb
x  2 criteria are much larger than their threshold values, indicating a
Horizontal : BuJ ¼ Grb Re2
b < 10 ð12Þ
qw d strong buoyancy effect, which is consistent with the HTD phe-
R Tw nomenon in this case.

Þ=qb m2b , q ¼ T 1T
where Grb ¼ gd ðqb  q
3
Tb
qdT. Even though BuJ is In the case of G600-q20 in Fig. 8b, there is no HTD, but there is
w b

proposed for horizontal tubes, the validation results in this paper an obvious temperature difference between the top and bottom
showed that the criterion of Bo showed better prediction perfor- surfaces. The values of the three parameters declined more com-
mance than BuJ. When BuJ is correlated with Nuexp or with the tem- pared to that in Fig. 8a, which means the buoyancy effect intensity
perature difference between the top and bottom surface (DTw), the is reduced. The peak of Grq/Grth is approximately at a position
562 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567

395
7000 p = 1.05 pc
390 2
6000 G=600 kg/m s
2
385 q = 20 kW/m
5000

HTC (W/m K)
380 din = 16 mm

2
Tw,in (K)

4000 din = 10.3 mm

375 p = 1.05 pc
2
370 G=600 kg/m s 3000
2
q = 20 kW/m
365 2000
din = 16 mm
360 din = 10.3 mm 1000

300 320 340 360 380 400 300 320 340 360 380 400
hb(kJ/kg) hb(kJ/kg)
(a) (b)
430 4500
p = 1.05 pc p = 1.05 pc
420 2
4000 2
G=600 kg/m s G=600 kg/m s
2 2
410 q = 30 kW/m 3500 q = 30 kW/m
3000 16 mm
HTC (W/m K)
Tw,in(K)

400 2 10.3 mm
2500
390
2000
380
1500
370
16 mm
10.3 mm 1000

360 500
300 320 340 360 380 400 420 300 320 340 360 380 400 420
hb(kJ/kg) hb(kJ/kg)

(c) (d)

500
p = 1.05 pc 3000 p = 1.05 pc
2
480 G=600 kg/m s
2 G=600 kg/m s
2
q = 60 kW/m
2
2500 q = 60 kW/m
460 16 mm
HTC(W/m -K)

2000 10.3 mm
440
2
Tw,in(K)

1500
420

400 1000

16 mm
380 10.3 mm 500

300 320 340 360 380 400 420 440 460 300 320 340 360 380 400 420 440 460
hb(kJ/kg) hb(kJ/kg)
(e) (f)
Fig. 7. Effects of the tube diameter on heat transfer for supercritical R134a in horizontal tubes (p = 1.05 pc). Filled symbols refer to the top surface and open symbols refer to
the bottom surface. (a) wall temperature with G = 600 kg/m2s, q = 20 kW/m2, (b) heat transfer coefficient with G = 600 kg/m2s, q = 20 kW/m2, (c) wall temperature with
G = 600 kg/m2s, q = 30 kW/m2, (d) heat transfer coefficient with G = 600 kg/m2s, q = 30 kW/m2, (e) wall temperature with G = 600 kg/m2s, q = 60 kW/m2, (f) heat transfer
coefficient with G = 600 kg/m2s, q = 60 kW/m2.

Table 3 To closely observe values near the threshold of these criteria for
Ranges of the buoyancy criteria in the present experiments.
better assessment, a case with a further reduced buoyancy effect is
Buoyancy criterion Grb/Re2 Bo Grq/Grth examined in Fig. 8c. The temperature differences between the top
Range in the present 0.001–0.79 2  107–2  104 1.5–882 and bottom surfaces are <3.5 K. In this case, the values of Grq/Grth
experiments and Grb/Re2 are still beyond the corresponding thresholds, which
means the buoyancy effect on heat transfer should be considered.
However, Bo is already 10 times smaller than 105, indicating a
where the DTw is at its maximum. However, the Bo and Grb/Re2 cri- negligible buoyancy effect. The three criteria give inconsistent pre-
teria reach their maximum values earlier at a lower bulk enthalpy, diction results in this case. Therefore, the threshold values of these
which is not as reasonable as that of Grq/Grth. parameters need to be reexamined.
 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567 563

2
G=600 kg/m s, q=60 kW/m
2 developed model for ‘‘Nufc” at supercritical pressures. In horizontal
-5
0.25 300 7×10 480 tubes, the difference in the HTC or the wall temperature between
Tw Bo
2 6 the top and bottom surfaces is the most representative parameter
0.20
250 Grb/Re Grq/Grth 460 to show the buoyancy effect. Therefore, Nutop/Nubottom was used as
5
200 the indicator and the onset of buoyancy effect is defined as
0.15
150
4 440
Nutop =Nubottom < 0:9; ð15Þ
0.10 top 3
420 where Nutop and Nubottom are the Nusselt number on the top and
100
2 bottom, respectively.
0.05
50 The Nutop/Nubottom is correlated with the buoyancy criteria in
1 400
Figs. 9–11. To intuitively show the wall temperature difference
0.00 0 bottom 0 influenced by the buoyancy, distributions of DTw versus the
380
300 320 340 360 380 400 420 440 460 buoyancy criteria were also given. As shown in these figures,
hb(kJ/kg) Nutop/Nubottom decreases and DTw increases with the increasing
buoyancy criteria, indicating a stronger buoyancy effect. The data
(a) is scattered more significantly at higher buoyancy criteria values.
2 2
This may be caused by the instability of heat transfer with strong
0.16 70 G=600 kg/m s, q=20 kW/m ×10
-5
420 buoyancy. When Nutop/Nubottom equals 0.9, the threshold
Tw 2.2
values of the buoyancy criteria for the onset of buoyancy can be
0.14 Bo
determined as follows: Grb/Re2 = 0.009, Bo = 1.2  106 and
60 top 2.0
2
Grb/Re 400
0.12 1.8 Grq/Grth = 5. The wall temperature difference, DTw, corresponding
50 Grq/Grth 1.6 to the onset of buoyancy is consistently to be 2–3 K. It can be seen
0.10
1.4 380 that the critical values for Grb/Re2 and Grq/Grth are close to the val-
0.08 40 ues given earlier. However, the critical value for Bo determined
1.2
bottom here is one order of magnitude smaller than the value given by
1.0
0.06 30
0.8
360 Jackson for vertical flow. This is caused by the difference in flow
0.04 direction. Buoyancy affects heat transfer at a lower Bo value in
20 0.6
horizontal tubes.
0.02 0.4 340
300 320 340 360 380 400
4.2. Simple criterion for engineering use: qd0.7/G1.2
hb(kJ/kg)
(b) The buoyancy criteria discussed above are based on the wall
-5 temperature and require iterative calculations to determine the
2 2 ×10
0.018 9 G=1500 kg/m s, q=40 kW/m 0.12 420 wall temperature. Therefore, the prediction accuracy is dependent
Tw 410
on the accuracy of heat transfer correlations [52]. Therefore, it is
0.016 8
Bo not convenient for engineering use. If a criterion that uses the
0.10 400
0.014
7 Grb/Re
2
operating parameters directly can be defined, it will be much
0.012 Grq/Grth 390 easier to use. As discussed in Sections 3.1 and 3.2, the mass flux
6
0.08 380
and heat flux have considerable effects on heat transfer, so the
0.010 top ratio of q/G is a widely used parameter to roughly predict heat
5 370
0.008 transfer characteristics. Some researchers have used q/G to predict
bottom 0.06 360 the onset of HTD in vertical flow [53]. However, the form of q/G is
4
0.006
350
too simple to provide adequate prediction performance. As shown
0.004 3 in Fig. 12, even though the four sets of experimental cases have the
0.04 340
same value of q/G (0.05 kJ/kg), the heat transfer phenomena differ
300 320 340 360 380 400
considerably. The increase in heat flux increases Tw, while the
hb(kJ/kg)
increase in mass flux decreases Tw. From G1500 to G600, q and G
(c) increase by the same multiples to keep q/G constant, but Tw on
the top surfaces increases obviously with HTD occurring at
Fig. 8. Variations of buoyancy criteria and wall temperature under various G = 600 kg/m2s. The HTC increases with increasing mass flux for
conditions, din = 10.3 mm and p = 4.62 MPa: (a) G = 600 kg/m2s, q = 60 kW/m2, (b) both the top and bottom surfaces. These results reveal that the
G = 600 kg/m2s, q = 20 kW/m2, and (c) G = 1500 kg/m2s, q = 40 kW/m2.
mass flux has a stronger influence on heat transfer than the heat
flux. Therefore, if the form qm/Gn is used, then m < n instead of
The above analyses show the qualitative applicability of the m = n must be used.
three criteria for buoyancy effect prediction. However, the thresh- According to Section 3.3, apart from the mass flux and heat flux,
old value proposed mainly based on water data appears to not be the tube diameter is also an important parameter affecting super-
applicable for the present data because of the differences in fluid critical heat transfer, and it should be considered as well. As shown
properties and experimental conditions. Then, this study attempts in Fig. 13, Grq/Grth in a 16-mm-diameter tube is always larger than
to extend the applicability of these criteria to organic fluids in hor- that in a 10.3-mm-diameter tube, indicating that a larger tube
izontal tubes and to determine their threshold values. diameter leads to stronger buoyancy. Schatte et al. [53] and Li
Before attempting to determine the threshold value, a definition et al. [54] studied the influence of the tube diameter in vertical
for the onset of buoyancy effect should be given. In vertical flow, flow. Bazargan [55] took the diameter effect into account for forced
Nu/Nufc < 0.95 [33] was set as the critical point where buoyancy convection in horizontal flow by using a parameter of qd0.2/G0.8.
has a significant effect on heat transfer. Nufc is the Nu of forced ver- However, it is quite ambitious to correlate this parameter with
tical flow without the buoyancy effect, calculated using the Jackson the wall temperature when stating that cases having the same
correlation [51]. However, in horizontal flows, there is no well- value of qd0.2/G0.8 should also have the same wall temperature.
564 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567

1.1 20

1.0
(0.009,0.9) 15
Nutop/Nubottom 0.9

0.8 10

Tw(K)
0.7
5
0.6

0.5 0 (0.009, 3)

0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20
2
Grb/Re Grb/Re
2

(a) (b)
Fig. 9. Effect of Grb/Re on heat transfer and its threshold value for the onset of buoyancy. (a) Nutop/Nubottom versus Grb/Re2, (b) DTw versus Grb/Re2.
2

1.1 20

1.0
-6
(1.2×10 , 0.9) 15
0.9
Nutop/Nubottom

0.8 10
Tw(K)

0.7
5
0.6
-6
0.5
0 (1.2×10 , 2)

0.0 1.0x10
-5
2.0x10
-5
3.0x10
-5 0.0 1.0x10
-5
2.0x10
-5
3.0x10
-5

Bo Bo
(a) (b)
Fig. 10. Effect of Bo on heat transfer and its threshold value for the onset of buoyancy. (a) Nutop/Nubottom versus Bo, (b) DTw versus Bo.

1.1 20

1.0
(5, 0.9) 15
0.9
Nutop/Nubottom

10
Tw(K)

0.8

0.7
5
0.6

0.5
0 (5, 2)

0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
Grq/Grth Grq/Grth
(a) (b)
Fig. 11. Effect of Grq/Grth on heat transfer and its threshold value for the onset of buoyancy. (a) Nutop/Nubottom versus Grq/Grth, (b) DTw versus Grq/Grth.

However, the experimental results of [49] together with the results
in this study showed that this parameter failed to describe the
effect of q, G, and d on heat transfer. As shown in Fig. 14, for the
two cases that have the same value of qd0.2/G0.8, there exist signif-
icant differences in the wall temperature. This is partly because
qd0.2/G0.8 was obtained based on the forced flow correlation, which
is not suitable for buoyancy-influenced flow. Moreover, the index
of G is smaller than that of q, which is not consistent with the
heat transfer phenomena. Based on these analyses, the parameter
qmdl/G n can be used with the following considerations:
 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567 565

10.0
G600-q30 bottom
410 q/G = 0.05 kJ/kg G800-q40 8.0

HTC(kW/m )
2
G1000-q50
G1500-q70 6.0
400 4.0
2.0
390
Tw,in (K)

0.0
6.0 top
q/G=0.05 kJ/kg G600-q30

HTC(kW/m )
2
380 bottom G800-q40
4.0
G1000-q50
G600-q30
G1500-q70
370 top G800-q40
G1000-q50 2.0
G1500-q70
360 0.0
300 320 340 360 380 400 420 300 320 340 360 380 400 420
hb(kJ/kg) hb(kJ/kg)
(a) (b)
Fig. 12. Heat transfer characteristics on conditions with the same q/G: (a) wall temperature and (b) HTC.

500 10.3 mm-G1000-q100

500 0.2 0.8
qd /G = 0.1594
2
G=600 kg/m s
2
20 kW/m
400 10.3mm 16 mm-G600-q60
480 0.2 0.8
qd /G = 0.1572
16mm
2
30 kW/m
300
10.3mm 460
16mm
Grq/Grth

Tw,in (K)

2
200 60 kW/m 440 top
10.3mm
16mm bottom
100 420

0 400

0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 300 320 340 360 380 400 420 440
h/hpc hb(kJ/kg)
0.2
Fig. 13. Buoyancy effect in different diameter tubes (corresponds to the cases in Fig. 14. Wall temperature in cases with the same value of qd /G0.8 [55]. (q is in
Fig. 7). kW/m2, d is in mm and G is kg/m2s).

(1) The indexes of q and G should satisfy the following condi-
tion: m < n.
(2) The diameter d should be the numerator. The value of l rep-
resents the effect of diameter on the heat transfer, with a lar-
ger l indicating a stronger influence.
(3) The target of qmdl/G n is to provide a simple and easy- to-use
tool to predict the heat transfer characteristics quickly, and
not to predict the wall temperature.
As mentioned above, the temperature difference between the
top and bottom surfaces, DT w , is a typical characteristic of super-
critical heat transfer in horizontal tubes. Herein, the maximum
DTw at a given operating condition is taken as the target variable
to predict. The final form is determined to be qd0.7/G1.2 through
data fitting and empirical analysis. Fig. 15 shows the relationship
between qd0.7/G1.2 and the max(DTw). Fig. 15a uses all the data of
R134a from the present experiments, which contains data for
two different diameters and various pressures. The max(DTw) is
clearly correlated with the proposed parameter of qd0.7/G1.2. The
data was fitted by an empirical equation of a quadratic function
with zero intercept. When the buoyancy is negligible, qd0.7/G1.2
equals to zero, corresponding to the zero intercept. With an
increasing qd0.7/G1.2, DT w increases, indicating that a larger
qd0.7/G1.2 corresponds to stronger buoyancy. At relatively small
qd0.7/G1.2, max(DTw) increases slowly with increasing qd0.7/G1.2.
At relatively large qd0.7/G1.2, the max(DTw) increases fast, then, a
small increase in heat flux or decrease in mass flux may result in
significant deterioration and a considerable rise of DTw. According
to the results in Section 4.1, the wall temperature difference corre-
sponding to the onset of buoyancy is 2–3 K. When max(DTw)
equals to 3 K, the qd0.7/G1.2 > 0.03 can be used as the critical value.
To further validate the proposed qd0.7/G1.2 parameter, experi-
mental data for supercritical water taken from the literature were
used in Fig. 15b. The parameter ranges of this water data are listed
in Table 4. qd0.7/G1.2 can also provide acceptable prediction results
for water. It should be noted that the range of the y-coordinates is
different because the parameter is not nondimensional. As the
number of data points taken from literature is limited, the critical
value for water cannot be obtained here. In the future, when more
data from various fluids are available, a look-up chart of
l
maxðDT w Þ  qm d =Gn can be developed for engineering use. In addi-
l
tion, the form of qm d =Gn needs further validation with more data,
especially with multiple fluids.
566 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567

300
100
din=10mm din=6.3 mm
250
1.02Pc R134a din=7.5 mm H2O
80
1.05Pc 200 din=26 mm
1.2Pc
din=43 mm

max( Tw)
60
max( Tw)

150
din=16mm
40 1.05Pc 100
1.2Pc
20 50
2
0 y=1696x +90x 0

0.01 0.1 1 2 3 4
0.7 1.2 0.7 1.2
qd /G qd /G
(a) (b)
Fig. 15. Relationship between qd0.7/G1.2 and the maximum wall temperature difference between the top and bottom surfaces: (a) R134a in present data, (b) water data from
the literature (see Table 4). (q is in kW/m2, d is in mm and G is kg/m2s).

Table 4
Parameter ranges of experimental data of supercritical water used in Fig. 15b.

din (mm) p (MPa) G (kg/m2 s) q (kW/m2)

Bazargan et al. [29] 6.3 24.2, 24.4 340, 432 300
Yamagata et al. [20] 7.5 24.5 1260 698, 930
Yu et al. [40,49] 26 23–25 100–1000 50–400
Lei et al. [28,30] 26, 43 23–25 100–300 50–180

5. Conclusions 4. A simple criterion of qd0.7/G1.2 was developed to provide a sim-

The buoyancy effects on the heat transfer characteristics of
supercritical R134a in horizontal tubes were experimentally inves-
tigated. The effects of heat flux, mass flux, and tube diameter were
discussed in detail. This study attempts to extend the applicability
of the buoyancy criteria to organic fluids in horizontal tubes and to
determine their threshold values. Finally, a simple criterion was
developed for engineering use.
1. The buoyancy effect was intensified by the increase in heat flux
and decrease in mass flux for horizontal flows. This heat
transfer enhancement can occur on both the top and bottom
surfaces. HTD was observed on the top surface at higher q/G.
There is no HTD on the bottom surface in the present
experiments.
2. A larger tube diameter leads to a stronger buoyancy force. The
effect of tube diameter on heat transfer differs at various q/G
values. At lower q/G with weak buoyancy, the wall temperature
and HTC are consistent in the two tubes. At higher q/G with
strong buoyancy, heat transfer on the top surface was prone
to deteriorate in the larger tube.
3. The buoyancy criteria of Grb/Re2, Bo, and Grq/Grth can predict the
buoyancy effect qualitatively. The Grq/Grth criterion successfully
predicted the weak buoyancy at the inlet and its peak location
shows good agreement with the wall temperature profiles.
Nutop/Nubottom < 0.9 was used to determine the onset value of
the criteria for the buoyancy effect for the present data in hor-
izontal tubes. The critical values for Grb/Re2 and Grq/Grth are
0.009 and 5, respectively, which are close to previously given
values. However, for the Bo criterion, it was 1.2  106, an order
of magnitude smaller than 105 for vertical flow, which means
that buoyancy influences heat transfer at a lower Bo value in
horizontal tubes.
ple and easy-to-use tool to predict heat transfer characteristics
quickly as the previous buoyancy criteria are based on the wall
temperature. The present experimental data and supercritical
water data from the literature were used to validate this param-
eter. The maximum temperature difference between the top
and bottom surfaces is clearly correlated with the proposed
parameter of qd0.7/G1.2.
Conflict of interest
The authors declare that there is no conflict of interest.
Acknowledgements
This work was supported by the National Key Research and
Development Program of China under (Grant No.
2016YFB0901405), the State Key Program of the National Natural
Science Foundation of China (Grant No. 51236004) and the Science
Fund for Creative Research Groups (No. 51621062).
References
[1] S. Lecompte, H. Huisseune, M.V.D. Broek, B. Vanslambrouck, M.D. Paepe,
Review of organic Rankine cycle (ORC) architectures for waste heat recovery,
Renew. Sustain. Energy Rev. 47 (2015) 448–461.
[2] H.X. Zhai, Q.S. An, L. Shi, V. Lemort, S. Quoilin, Categorization and analysis of
heat sources for organic Rankine cycle systems, Renew. Sustain. Energy Rev. 64
(2016) 790–805.
[3] I.L. Pioro, S. Mokry, S. Draper, Specifics of thermophysical properties and
forced-convective heat transfer at critical and supercritical pressures, Rev.
Chem. Eng. 27 (3–4) (2011) 191–214.
[4] I.L. Pioro, R.B. Duffey, Experimental heat transfer in supercritical water flowing
inside channels (survey), Nucl. Eng. Des. 235 (22) (2005) 2407–2430.
[5] I.L. Pioro, R.B. Duffey, Heat Transfer and Hydraulic Resistance at Supercritical
Pressures in Power-Engineering Applications, ACME, New York, 2006.
 R. Tian et al. / International Journal of Heat and Mass Transfer 127 (2018) 555–567
[6] D. Huang, Z. Wu, B. Sunden, W. Li, A brief review on convection heat transfer of
fluids at supercritical pressures in tubes and the recent progress, Appl. Energy
162 (2016) 494–505.
[7] L.F. Cabeza, A.D. Gracia, A.I. Fernández, M.M. Farid, Supercritical CO2 as heat
transfer fluid: a review, Appl. Therm. Eng. 125 (2017) 799–810.
[8] S.X. Zhang, H.Y. Gu, X. Cheng, Z.Q. Xiong, Experimental study on heat transfer
of supercritical Freon flowing upward in a circular tube, Nucl. Eng. Des. 280
(2014) 305–315.
[9] Y.L. Cui, H.X. Wang, Experimental study on convection heat transfer of R134a
at supercritical pressures in a vertical tube for upward and downward flows,
Appl. Therm. Eng. 129 (2018) 1414–1425.
[10] F. Feuerstein, S. Coelho, D. Klingel, X. Cheng, Large-scale heat transfer
experiments with supercritical R134a flowing upward in a circular tube,
ATW-Int. J. Nucl. Power 62 (2) (2017) 121–125.
[11] G. Richards, G.D. Harvel, I.L. Pioro, A.S. Shelegov, P.L. Kirillov, Heat transfer
profiles of a vertical, bare, 7-element bundle cooled with supercritical Freon
R12, Nucl. Eng. Des. 264 (2013) 246–256.
[12] P.X. Jiang, C.R. Zhao, B. Liu, Flow and heat transfer characteristics of r22 and
ethanol at supercritical pressures, J. Supercrit. Fluids 70 (2012) 75–89.
[13] H. Mori, T. Kaida, M. Ohno, S. Yoshida, Y. Hamamoto, Heat transfer to a
supercritical pressure fluid flowing in sub-bundle channels, J. Nucl. Sci. Tech.
49 (4) (2012) 373–383.
[14] S.K. Dubey, R.P. Vedula, K.N. Iyer, A.J. Gaikwad, Local heat transfer coefficient
measurements using thermal camera for upward flow of Freon 22 in a vertical
tube at supercritical conditions and development of correlations, Nucl. Eng.
Des. 328 (2018) 80–94.
[15] C.R. Zhao, P.X. Jiang, Experimental study of in-tube cooling heat transfer and
pressure drop characteristics of R134a at supercritical pressures, Exp. Therm.
Fluid Sci. 35 (7) (2011) 1293–1303.
[16] U.C. Andresen, Supercritical gas cooling and near-critical-pressure
condensation of refrigerant blends in microchannels, Georgia Institute of
Technology (2006).
[17] Y.R. Jiang, Heat transfer and pressure drop of refrigerant R404A at near-critical
and supercritical pressures PhD thesis, Iowa State University, 2004.
[18] G.A. Adebiyi, W.B. Hall, Experimental investigation of heat transfer to
supercritical pressure carbon dioxide in a horizontal pipe, Int. J. Heat Mass
Transf. 19 (7) (1976) 715–720.
[19] J.D. Jackson, W.B. Hall, Influences of buoyancy on heat transfer to fluids
flowing in vertical tubes under turbulent conditions, in: Turbulent forced
convection in channels and bundles, Hemisphere, Washington vol. 2(1), 1979,
pp. 613–640.
[20] K. Yamagata, K. Nishikawa, S. Hasegawa, T. Fujii, S. Yoshida, Forced convective
heat transfer to supercritical water flowing in tubes, Int. J. Heat Mass Transf. 15
(12) (1972) 2575–2593.
[21] J.R. Licht, Heat transfer and fluid flow characteristics in supercritical water, Ph.
D. Thesis, University of Wisconsin Madison, 2008.
[22] Z.H. Li, Y.X. Wu, G.L. Tang, D.L. Zhang, J.F. Lu, Comparison between heat
transfer to supercritical water in a smooth tube and in an internally ribbed
tube, Int. J. Heat Mass Transf. 84 (2015) 529–541.
[23] D.M. McEligot, J.D. Jackson, ‘‘Deterioration” criteria for convective heat transfer
in gas flow through non-circular ducts, Nucl. Eng. Des. 232 (3) (2004) 327–333.
[24] W.B. Hall, J.D. Jackson, Laminarization of a turbulent pipe flow by buoyancy
forces. In: Mechanical Engineering. 345 E 47TH ST, New York, NY 10017:
ASME-Am. Soc. Mech. Eng., 1969, vol. 91(11), p. 66.
[25] S. He, K. He, M. Seddighi, Laminarisation of flow at low Reynolds number due
to streamwise body force, J. Fluid Mech. 809 (2016) 31–71.
[26] Y.Y. Bae, Mixed convection heat transfer to carbon dioxide flowing upward and
downward in a vertical tube and an annular channel, Nucl. Eng. Des. 241 (8)
(2011) 3164–3177.
[27] S. Mokry, I. Pioro, P. Kirillov, Y. Gospodinov, Supercritical-water heat transfer
in a vertical bare tube, Nucl. Eng. Des. 240 (3) (2010) 568–576.
[28] X.L. Lei, H.X. Li, N.T. Dinh, W.Q. Zhang, A study of heat transfer scaling of
supercritical pressure water in horizontal tubes, Int. J. Heat Mass Transf. 114
(2017) 923–933.
[29] M. Bazargan, D. Fraser, V. Chatoorgan, Effect of buoyancy on heat transfer in
supercritical water flow in a horizontal round tube, J. Heat Transf. 127 (8)
(2005) 897–902.
[30] X.L. Lei, H.X. Li, W.Q. Zhang, N.T. Dinh, Y.M. Guo, S.Q. Yu, Experimental study
on the difference of heat transfer characteristics between vertical and
horizontal flows of supercritical pressure water, Appl. Therm. Eng. 113
(2017) 609–620.
567
[31] M. Yang, Numerical study on the heat transfer of carbon dioxide in horizontal
straight tubes under supercritical pressure, PloS one. 11 (7) (2016) e0159602.
[32] J.D. Jackson, Fluid flow and convective heat transfer to fluids at supercritical
pressure, Nucl. Eng. Des. 264 (2013) 24–40.
[33] S.H. Liu, Y.P. Huang, G.X. Liu, J.F. Wang, L.K.H. Leung, Improvement of
buoyancy and acceleration parameters for forced and mixed convective heat
transfer to supercritical fluids flowing in vertical tubes, Int. J. Heat Mass Transf.
106 (2017) 1144–1156.
[34] J.K. Li, Studies of buoyancy-influenced convective heat transfer to air in a
vertical tube PhD Thesis, The, University of Manchester, 1994.
[35] J. Licht, M. Anderson, M. Corradini, Heat transfer and fluid flow characteristics
in supercritical pressure water, ASME J. Heat Transf. 131 (7) (2009) 072502.
[36] P.X. Jiang, Y. Zhang, C.R. Zhao, R.F. Shi, Convection heat transfer of CO2 at
supercritical pressures in a vertical mini tube at relatively low Reynolds
numbers, Exp. Therm. Fluid Sci. 32 (2008) 1628–1637.
[37] B. Liu, Research, on in-tube convection heat transfer and thermal cracking of
supercritical pressure fluids PhD Thesis, Tsinghua University, 2013.
[38] H.W. Deng, K. Zhu, G.Q. Xu, Z. Tao, Heat transfer characteristics of RP-3
kerosene at supercritical pressure in a vertical circular tube, J. Enhanc. Heat
Transf. 19 (2012) 409–421.
[39] B.S. Petukhov, A.F. Polyakov, Gravitational effects on heat transfer in a single-
phase fluid near the critical point, in: B.E. Launder (Ed.), Heat Transfer in
Turbulent Mixed Convection, 1988. pp. 196–201.
[40] S.Q. Yu, H.X. Li, X.L. Lei, Y.F. Feng, Y.F. Zhang, H. He, T. Wang, Influence of
buoyancy on heat transfer to water flowing in horizontal tubes under
supercritical pressure, Appl. Therm. Eng. 59 (1) (2013) 380–388.
[41] K. Tanimizu, R. Sadr, Experimental investigation of buoyancy effects on
convection heat transfer of supercritical CO2 flow in a horizontal tube, Heat
Mass Transf. 52 (4) (2016) 713–726.
[42] T.H. Kim, J.G. Kwon, M.H. Kim, H.S. Park, Experimental investigation on validity
of buoyancy parameters to heat transfer of CO2 at supercritical pressures in a
horizontal tube, Exp. Therm. Fluid Sci. 92 (2018) 222–230.
[43] C. Yu, J.L. Xu, Y.S. Sun, Transcritical pressure organic Rankine cycle (ORC)
analysis based on the integrated-average temperature difference in
evaporators, Appl. Therm. Eng. 88 (2015) 2–13.
[44] E.W. Lemmon, M.L. Huber, M.O. MeLinden, NIST reference fluid
thermodynamic and transport properties RFFPROP, version 9.1, Nat. Inst.
Standard Technol. (2013).
[45] K.E. Gungor, R.H.S. Winterton, A general correlation for flow boiling in tubes
and annuli, Int. J. Heat Mass Transf. 29 (3) (1986) 351–358.
[46] S.G. Kandlikar, A general correlation for saturated two-phase flow boiling heat
transfer inside horizontal and vertical tubes, ASME J. Heat Transf. 112 (1)
(1990) 219–228.
[47] F.W. Dittus, L.M.K. Boelter, Heat Transfer in Automobile Radiators of the
Tubular Type vol. 2 (1930) 443–461.
[48] X. Chu, E. Laurien, Flow stratification of supercritical CO2 in a heated horizontal
pipe, J. Supercrit. Fluids 116 (2016) 172–189.
[49] S.Q. Yu, H.X. Li, X.L. Lei, Y.C. Feng, Y.F. Zhang, H. He, T. Wang, Experimental
investigation on heat transfer characteristics of supercritical pressure water in
a horizontal tube, Exp. Therm. Fluid Sci. 50 (2013) 213–221.
[50] J.D. Jackson, W.B. Hall, Forced convection heat transfer to fluids at supercritical
pressure, Turb. Forced Convect. Channels Bundles 2 (1) (1979) 563–611.
[51] J.D. Jackson, Consideration of the heat transfer properties of supercritical
pressure water in connection with the cooling of advanced nuclear reactors,
in: Proceedings of the 13th Pacific Basin Nuclear Conference, Shenzhen City,
China, October 21–25, 2002.
[52] Z.H. Hu, Heat transfer characteristics of water at near-critical and supercritical
pressures in vertical and inclined tubes PhD Thesis, Xi’an Jiaotong University,
2001.
[53] G.A. Schatte, A. Kohlhepp, C. Wieland, H. Spliethoff, Development of a new
empirical correlation for the prediction of the onset of the deterioration of heat
transfer to supercritical water in vertical tubes, Int. J. Heat Mass Transf. 102
(2016) 133–141.
[54] Z.H. Li, D.L. Zhang, Y.X. Wu, J.F. Lu, Q. Liu, A new criterion for predicting
deterioration of heat transfer to supercritical water in smooth tubes, vol. 34,
in: Proceedings of the CSEE, 2014, pp. 6304–6309.
[55] M. Bazargan, Forced Convection Heat Transfer to Turbulent Flow of
Supercritical Water in a Round Horizontal Tube PhD thesis, University of
British Columbia, 2001.