Unified Model of Heat Transfer in

Gas/Liquid Pipe Flow

H.-Q. Zhang, SPE, Q. Wang, SPE, C. Sarica, SPE, and J.P. Brill, SPE, the U. of Tulsa

Summary (Zhang et al. 2003a, 2003b). The major advantage of this model
A unified model of multiphase heat transfer is developed for dif- compared with previous mechanistic models is that the predictions
ferent flow patterns of gas/liquid pipe flow at all inclinations –90° for both flow-pattern transition and flow behavior are incorporated
to +90° from horizontal. The required local flow parameters are into a single unified model based on slug dynamics. Multiphase
predicted by use of the unified hydrodynamic model for gas/liquid heat transfer depends on the hydrodynamic behavior of the flow.
pipe flow recently developed by Zhang et al. (2003a, 2003b). The The objective of this study is to develop a unified heat-transfer
model prediction of the pipe inside convective-heat-transfer coef- model for gas/liquid pipe flow that is consistent with the unified
ficients are compared with experimental measurements for a hydrodynamic model.
crude-oil/natural-gas system in horizontal and upward-vertical
flows, and good agreement is observed. Modeling
Basic Concepts. In this study, only the temperature changes
Introduction caused by heat transfer between the multiphase flowing fluids and
As oil and gas production moves to deep and ultradeep waters, their surroundings are considered. When fluids flow through a pipe
flow-assurance issues such as wax deposition, hydrate formation, and the surrounding temperature is colder than the fluids, heat is
and heavy-oil flow become very crucial in transportation of gas, lost from the fluids, resulting in a decline in temperature. Fig. 1
oil, and water to processing facilities. These flow-assurance prob- shows a pipe segment of length dl and an ID dI through which a
lems are strongly related to both the hydraulic and thermal behav- single fluid flows with density ␳ and specific heat cP. TB1 is the
iors (such as liquid holdups, local fluid velocities, pressure gradi- inlet temperature, and TB2 is the outlet temperature; v is the ve-
ent, slug characteristics, and convective-heat-transfer coefficients locity of the fluid, and q is the heat flux to the surroundings. The
corresponding to different phases and flow patterns) of the mul- heat lost from the fluid in the pipe segment must equal the heat
tiphase flow. Therefore, multiphase hydrodynamics and heat trans- transferred to the surroundings,
fer need to be modeled properly to optimize the design and op-
eration of the flow system. 共TB1 − TB2兲vA␳cP = q␲dIdl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
Compared to experimental and modeling studies of multiphase
hydrodynamics, very limited research results can be found in the Then,
open literature for multiphase heat transfer. Davis et al. (1979)
presented a method for predicting local Nusselt numbers for strati- ⭸TB q␲dI
=− . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
fied gas/liquid flow under turbulent-liquid/turbulent-gas condi- ⭸l vA␳cP
tions. A mathematical model based on the analogy between mo-
mentum transfer and heat transfer was developed and tested using If heat flux is defined as q⳱U(TB−TO) then,
heat-transfer and flow-characteristics data taken for air/water flow
⭸TB 4U共TB − TO兲
in a 63.5-mm-inside-diameter (ID) tube. =− . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)
Shoham et al. (1982) measured heat-transfer characteristics for ⭸l dIv␳cp
slug flow in a horizontal pipe. The time variations of temperature,
heat-transfer coefficients, and heat flux were reported for the where TB is the bulk temperature of the fluid, TO is the en-
different zones of slug flow. Substantial difference in heat- vironmental temperature outside the pipe, and l is pipe length in
transfer coefficient was found to exist between the bottom and top the flow direction.
of the slug. Eq. 3 is a differential equation that can be solved for TB as a
Most previous modeling studies were aimed at developing function of l if U is known. U is the overall-heat-transfer coeffi-
heat-transfer correlations for different flow patterns (Knott et al. cient. For the simple case shown in Fig. 1, U must include con-
1959; Kudirka et al. 1965; Aggour 1978; Shah 1981; Ravipudi and vective heat transfer from the bulk fluid to the pipe inside wall (h),
Godbold 1978; Rezkallah and Sims 1987). Kim et al. (1997) conductive heat transfer through the pipe wall, and convective heat
evaluated 20 heat-transfer correlations against experimental data transfer from the pipe outside wall to the surroundings. This can be
collected from the open literature and made recommendations for expressed as
different flow patterns and inclination angles. However, these rec- 1
ommended correlations did not give satisfactory predictions when U= , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
compared with experimental results by Matzain (1999). 1 dI dO dI
+ ln +
Manabe (2001) developed a comprehensive mechanistic model h 2kP dI hOdO
for heat transfer in gas/liquid pipe flow. The overall performance
was better than previous correlations in comparison with experi- where dO is the pipe outside diameter, h is the inside convective-
mental data; however, some inconsistencies in the hydrodynamic heat-transfer coefficient, kP is the pipe thermal conductivity, and
model and the heat-transfer formulations for stratified (annular) hO is the outside convective-heat-transfer coefficient.
and slug flows need to be improved. For multiphase flow, the above expressions are much
A unified hydrodynamic model has been developed for gas/ more complex, and the hydrodynamics of the flow system must
liquid pipe flow at the Tulsa U. Fluid Flow Projects (TUFFP) be considered.

Bubbly Flow and Dispersed-Bubble Flow. In multiphase heat-

transfer modeling, bubbly flow and dispersed-bubble flow can be
Copyright © 2006 Society of Petroleum Engineers
treated as pseudosingle-phase flow. The fluid physical properties
This paper (SPE 90459) was first presented at the 2004 SPE Annual Technical Conference are adjusted on the basis of liquid holdup. In this study, the
and Exhibition, Houston, 26–29 September, and revised for publication. Original manuscript
received for review 27 May 2004. Revised manuscript received 17 May 2005. Paper peer
Petukhov (1970) correlation is used for dispersed-bubble flow and
approved 26 May 2005. turbulent-bubbly flow,

114 February 2006 SPE Production & Operations

 Fig. 2—Control volume and temperatures in stratified flow.

Fig. 1—Heat transfer from single-phase fluid inside pipe to out- ⭸TM 4UM共TM − TO兲
=− , . . . . . . . . . . . . . . . . . . . . . . . (13)
side surroundings. ⭸l dI共␳LcPLvSL + ␳GcPGvSG兲
where TM is the bulk temperature of the mixture.

冉冊 fM M M

冉 冊
NReNPr
2 ␮L 0.25 Stratified/Annular Flow. As shown in Fig. 2, TF and TC are the

冑
Nu =
NM , . . . . . . . . . . (5) bulk temperatures of the liquid film and gas core, respectively, in
fM M2 Ⲑ 3 ␮LW a stratified or annular flow. As in Fig. 1, the inlet and outlet of the
1.07 + 12.7 共N − 1兲
2 Pr control volume are designated by Subscripts 1 and 2, respectively.
The heat transfer caused by circumferential temperature gradient
where NM M M
Nu, NRe, and NPr are the mixture Nusselt number, Reyn- and thermal conduction in the axial direction are neglected.
olds number, and Prandtl number, respectively, in which the liq- We assume that there is no phase change in the two-phase flow
uid viscosity (␮L), liquid thermal conductivity (kL), and mixture and neglect the energy changes caused by changes in pressure,
velocity (vM) are used: volume, velocity, and elevation. The energy equation for the liquid
film can be obtained from a balance between the energy change
␳MvMdI across the liquid-film control volume and the heat exchange with
M
NRe = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6)
␮L the gas core and the surroundings,

cPM␮L 共TF2 − TF1兲vFHLFA␳LcPL = −qFSFdl − qISIdl , . . . . . . . . . . . . . (14)

Pr =
NM , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7)
kL where TF1 and TF2 are the inlet and outlet temperatures of the
control volume for the liquid film, vF is the film velocity, HLF is
where the vM is the sum of the superficial liquid and gas velocities the holdup of the liquid film, SF and SI are the perimeters of the
(vM=vSL + vSG), ␳M is the mixture density, and cPM is the mixture- liquid film in contact with the pipe wall, and at the interface, dl is
specific heat. In bubbly flow, the gas bubbles tend to be concen- the length of the control volume, qF and qI are the heat fluxes from
trated at the center of the pipe; therefore, liquid density and spe- the liquid film to the outside pipe environment and to the gas core,
cific heat are used in these numbers rather than ␳M and cPM. For respectively, through the interface. From Eq. 14, we can obtain
dispersed-bubble flow, a mixture density and specific heat are used
and defined as ⭸TF qFSF + qISI
=− . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (15)
⭸l vFHLFA␳LcPL
␳M = 共1 − HL兲␳G + HL␳L , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)
Similarly, the energy equation for the gas core is
cPM = 共1 − HL兲cPG + HLcPL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
⭸TC qCSC − qISI
HL is liquid holdup; ␳L and ␳G are the densities of the liquid and =− , . . . . . . . . . . . . . . . . . . . . . . . . . . . (16)
⭸l vC共1 − HLF兲A␳CcPC
gas phases, respectively; cPL and cPG are the specific heats of the
liquid and gas, respectively. ␮L and ␮LW are liquid viscosities where TC is the gas-core temperature, vC is the gas-core velocity,
corresponding to the bulk temperature of the mixture and the tem- SC is the perimeter of the gas core in contact with the pipe wall, ␳C
perature at the pipe wall, respectively. fM in Eq. 5 is the mixture and cPC are the representative density and specific heat for the gas
friction factor at the pipe wall, which is a function of the mixture core, respectively, and qC is the heat flux from the gas core to the
Reynolds number (Zhang et al. 2003a). outside pipe environment.
If the bubbly flow is laminar and fully developed, the Nusselt Assuming the axial temperature gradient is the same for the
number is a constant. According to Shah and London (1978): liquid film and the gas core, the combined energy equation can be
obtained by combining Eqs. 15 and 16,
Nu = 3.657 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10)
NM
B共qFSF + qISI兲 = qCSC − qISI , . . . . . . . . . . . . . . . . . . . . . . . . . . (17)
Then, the convective-heat-transfer coefficient for bubbly or dis- where
persed-bubble flow is obtained from
vC共1 − HLF兲␳CcPC
NM B= . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18a)
MukL vFHLF␳LcPL
hM = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11)
dI
The heat fluxes are defined as
On the basis of the surface area of the pipe inside wall, the two- qF = UF共TF − TO兲 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18b)
phase overall heat-transfer coefficient for bubbly or dispersed-
bubble flow can be calculated from qC = UC共TC − TO兲 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18c)

1 qI = hI共TF − TC兲 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18d)

UM = . . . . . . . . . . . . . . . . . . . . . . . . . . . (12)
1 dI dO dI where UF is the local overall-heat-transfer coefficient between the
+ ln + liquid film and the outside pipe environment, UC is the local over-
hM 2kP dI hOdO
all-heat-transfer coefficient between the gas core and the outside
The temperature gradient in the flow direction can be obtained as pipe environment, and hI is the convective-heat-transfer coefficient

February 2006 SPE Production & Operations 115

between the liquid film and the gas core. Therefore, Eq. 17 can be and kC are the thermal conductivities of the liquid film and gas
rewritten as core, respectively, and dF and dC are the hydraulic diameters of the

冋 册
liquid film and gas core (Zhang et al. 2003a). The convective-
UF共TF − TO兲SF heat-transfer coefficient across the interface, hI, is assumed to be
B = UC共TC − TO兲SC + hI共TF − TC兲SI, . . . . . (19)
+ hI共TF − TC兲SI the same as that for the gas core, hC.
The Nusselt numbers for the liquid film and gas core are cal-
from which we can obtain the relationship between TC–TO culated using the correlations for single-phase convective heat
and TF–TO: transfer. The Petukhov (1970) correlation is used for turbulent
hISI共1 + B兲 + BUFSF liquid-film flow:

冉冊
TC − TO = 共T − TO兲 . . . . . . . . . . . . . . . (20)
hISI共1 + B兲 + UCSC F fF F F

冉 冊
N N
Then, the two-phase overall-heat-transfer coefficient for stratified 2 Re Pr ␮L 0.25

冑
NFNu = , . . . . . . . . . . (29)
or annular flow based on TF–TO and the entire inside circumfer- fF F2 Ⲑ 3 ␮LW
ence of the pipe is defined as 1.07 + 12.7 共N − 1兲
2 Pr
qFSF + qCSC UFSF UCSC关hISI共1 + B兲 + BUFSF兴 where fF is the friction factor at the wall in contact with the liquid
USA = = + .
␲ dI共TF − TO兲 ␲ dI ␲ dI关hISI共1 + B兲 + UCSC兴 film. NFRe and NFPr are the Reynolds number and Prandtl number,
. . . . . . . . . . . . . . . . . . . . . . . . . . (21) respectively, of the liquid film:
From ␳LvFdF
F
NRe = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (30)
1 1 dI dO dI ␮L
= + ln + , . . . . . . . . . . . . . . . . . . . . . . . . (22) and
USA hSA 2kP dI hOdO
the two-phase convective-heat-transfer coefficient can be obtained, cPL␮L
NFPr = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (31)
kL
1
hSA = . . . . . . . . . . . . . . . . . . . . . . . . . . (23) The Dittus and Boelter (1930) correlation is used for turbulent
1 dI dO dI gas-core flow,
− ln +
USA 2kP dI hOdO
␳GvCdC
C
NRe = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (32)
If the temperature difference between the gas core and the outside ␮G
pipe environment, TC–TO, is used to calculate the heat transfer, the
definition of the two-phase overall-heat-transfer coefficient for where NC C
Re and NPr are the Reynolds number and Prandtl number

stratified or annular flow will be different: of the gas core, respectively:

␳GvCdC
qFSF + qCSC UCSC UFSF关hISI共1 + B兲 + UCSC兴 C
NRe = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (33)
USA = = + . ␮G
␲ dI共TC − TO兲 ␲ dI ␲ dI关hISI共1 + B兲 + BUFSF兴
. . . . . . . . . . . . . . . . . . . . . . . . . . (24) cPG␮G
NCPr = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (34)
The two-phase overall-heat-transfer coefficient, USA, can also be kG
TF +TC For fully developed laminar flows of the liquid film and gas
defined on the basis of −TO to compare with Manabe’s core, the Nusselt number approaches a constant value. According
2
(2001) data for horizontal and vertical annular flows: to Shah and London (1978), the Nusselt numbers for fully devel-
oped laminar flows in a pipe and in a two-dimensional channel
qFSF + qCSC (between two parallel plates) are 3.657 and 7.541, respectively,
USA = 2 with a uniform wall temperature. Therefore, we suggest interpo-
␲ dI共TF + TC − 2TO兲
lating between these two values for the liquid film on the basis of
hISI共1 + B兲 + BUFSF the nondimensional film thickness (film thickness divided by pipe
UFSF + U S
hISI共1 + B兲 + UCSC C C ID), ␦F,

冋 册
=2 . . . . . . . . . . . . . . (25)
hISI共1 + B兲 + BUFSF 7.541 − 3.657
␲ dI 1 + NFNu = 3.657 + 共0.5 − ␦F兲 . . . . . . . . . . . . . . . . . (35)
hISI共1 + B兲 + UCSC 0.5
The temperature gradient along the pipe in the flow direction can For a laminar gas core,
be calculated from NCNu = 3.657 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (36)
⭸T UFSF共TF − TO兲 + UCSC共TC − TO兲
=− . . . . . . . . . . . . . . . . . (26) Slug Flow. Film Region. Fig. 3 illustrates slug flow and the tem-
⭸l A共␳LcPLvSL + ␳GcPGvSG兲
perature variation with time observed at a stationary position when
The local overall-heat-transfer coefficients for the liquid film
and gas core are given, respectively, by
1
UF = . . . . . . . . . . . . . . . . . . . . . . . . . . . (27)
1 dI dO dI
− ln +
hF 2kP dI hOdO
and
1
UC = . . . . . . . . . . . . . . . . . . . . . . . . . . . . (28)
1 dI dO dI
− ln +
hC 2kP dI hOdO
The convective-heat-transfer coefficients for the liquid film and
NFNukF NCNukC Fig. 3—Slug flow and temperature variation at a stationary ob-
gas core are obtained using hF = and hC = , where kF
dF dC servation point.

116 February 2006 SPE Production & Operations

the film region and slug body pass by. Taking a small segment in UFSF + UCSC
the liquid-film region as the control volume, the heat-balance a= , . . . . (46b)
A关共vT − vF兲␳LcPLHLF + 共vT − vC兲共1 − HLF兲␳CcPC兴
equation for the liquid film can be written as
共TF2 − TF1兲vFHLFA␳LcPL vFHLF␳LcPL + vC共1 − HLF兲␳CcPC

冤 冥 ⭸TF b= . . . . . . (46c)
lU关共vT − vF兲HLF␳LcPL + 共vT − vC兲共1 − HLF兲␳CcPC兴
− UF共TF − TO兲SF dl = ␳ c H Adl , . . . . . . . . . (37)
⭸t L PL LF
− hI共TF − TC兲SIdl Integrating Eq. 46, we can express the temperature change with
time as the film region passes by the observation point,
where

TF2 − TF1 = −
⭸TF
⭸l
dl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (38)
b b
冉 冊
TF = TO + ⌬TU + TST − TO − ⌬TU e−avTt , . . . . . . . . . . . (47)
a a

Similarly, the heat-balance equation for the gas core can be writ- where TST is the temperature at the slug tail (film head) measured
ten as at the observation point.
Slug Region. The heat-balance equation for a control volume
共TC2 − TC1兲vC共1 − HLF兲A␳CcPC

冤 冥 ⭸TC in a slug body (see Fig. 2) can be written as

− UC共TC − TO兲SCdl = ␳ c 共1 − HLF兲Adl ,
⭸t C PC ⭸TS
+ hI共TF − TC兲SIdl 共TS2 − TS1兲vM HLS A␳LcPL − US 共TS − TO兲␲ d dl = ␳ c H Adl ,
. . . . . . . . . . . . . . . . . . . . . . . . . . (39) ⭸t L PL LS
. . . . . . . . . . . . . . . . . . . . . . . . . . (48)
where
⭸TC where TS is the slug-body temperature, HLS is the liquid holdup in
TC2 − TC1 = − dl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (40) the slug body, and US is the local overall-heat-transfer coefficient
⭸l for the slug body,
The spatial and temporal derivatives of the film can be related
using the total derivative if we follow a control volume with the 1
US = , . . . . . . . . . . . . . . . . . . . . . . . . . . . (49)
translational velocity, 1 d dO dI
− ln +
dTF ⭸TF ⭸TF hS 2kP d hOdO
= + vT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (41)
dt ⭸t ⭸l where hS is the convective-heat-transfer coefficient of the slug
The total derivative is assumed to be the same for the entire body, identical to that for dispersed bubble flow given by Eq. 11.
slug unit, By use of

dTF ⌬TU ⌬TU ⭸TS

=− = −vT , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (42) TS2 − TS1 = − dl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (50)
dt tU lU ⭸l
where ⌬TU is the temperature drop across the slug unit and tU is the and
time for traveling the distance of the slug-unit length. Therefore,
⭸TF 1 ⭸TF ⌬TU ⭸TS 1 ⭸TS ⌬TU
=− − . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (43) =− − , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (51)
⭸l vT ⭸t lU ⭸l vT ⭸t lU

Similarly, Eq. 48 can be rewritten as

⭸TC 1 ⭸TC ⌬TU 1 ⭸TS

=− − . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (44) = f⌬TU − e共TS − TO兲 , . . . . . . . . . . . . . . . . . . . . . . . . . . (52)
⭸l vT ⭸t lU vT ⭸t
The spatial and temporal derivatives of the film and gas-core
temperatures cannot be set equal (as for stratified or annular flow) US␲dI vM
where e and f are constants: e = and f = .
if there is a difference between the two temperatures because they ␳LcPLHLSA共vT − vM兲 lU共vT − vM兲
possess the same boundary conditions at both ends of the film
region. To obtain the film and gas-core temperature profile in the By integrating Eq. 52, the temperature of the slug body can be
film region, one must perform step-by-step integration of the obtained as a function of time, t,

冉 冊
above equations. However, in this study, the gas-core temperature
is assumed to be equal to the film temperature. For a pipe flow f f
with uniform outside temperature, this assumption is reasonable. TS = TO + ⌬TU + TST − TO − ⌬TU eevT共tS−t兲 . . . . . . . . . . (53)
e e
Then, Eqs. 37 and 39 can be combined:
Slug Unit. The average temperature across the liquid film and
1 ⭸TF
A关共vT − vF兲HLF␳LcPL + 共vT − vC兲共1 − HLF兲␳CcPC兴 slug body of a slug unit is calculated by
vT ⭸t
tF tS
⌬TU
=
lU
A关vFHLF␳LcPL + vC共1 − HLF兲␳CcPC兴 兰 T dt + 兰 T dt
0
F
0
S

− 共TF − TO兲共UFSF + UCSC兲. . . . . . . . . . . . . . . . . . (45) TUA = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (54)

lU Ⲑ vT
This can be further expressed as
where tF and tS are the times for the liquid film and slug body to
1 ⭸TF pass the observation point, respectively. The two-phase overall-
= b⌬TU − a共TF − TO兲 , . . . . . . . . . . . . . . . . . . . . . . . . . (46a) heat-transfer coefficient will be defined on the basis of the differ-
vT ⭸t
ence between this average temperature and the pipe outside tem-
where a and b are local constants, perature, TO. After integration, we obtain

February 2006 SPE Production & Operations 117

 TUA − TO =
b
⌬T l −
a UF a
1
冉 b
TST − TO − ⌬TU 共e−alF − 1兲
a 冊
lU

+
f
⌬T l −
e US e
1
冉 f
TST − TO − ⌬TU 共1 − eelS兲
e 冊
, . . . . (55)
lU
where lF, lS, and lU are the lengths of the liquid film, slug body,
and slug unit, respectively. By applying Eqs. 47 and 53 at the slug
front (or film end), the temperature difference at the slug tail (or
film head) can be expressed as
b f
⌬T e−clS共1 − e−alF兲 + ⌬TU共1 − eelS兲
a U e
TST − TO = . . . . . . . . . (56)
1 − e−alF−elS
Substituting Eq. 56 into Eq. 55, Fig. 4—Schematic diagram of heat-transfer measurement sec-

冉 冊
tion (Manabe 2001).
b f
l + l ⌬TU
a F e S
TUA − TO = The in-situ hydrodynamic behaviors of the gas/liquid pipe flow
lU must be identified to perform the heat-transfer calculation. All

+
冉 冊冉 冊
b f
−
a e
1 1
− 共1 − e−alF兲共1 − e−elS兲⌬TU
e a
. . . . . (57)
the flow conditions, such as flow pattern, liquid holdups, local
fluid velocities of the liquid film and gas core, and slug charac-
teristics, are predicted by use of the unified hydrodynamic
lU共1 − e−alF−elS兲 model for gas/liquid pipe flow recently developed by Zhang et al.
The two-phase overall-heat-transfer coefficient for an entire slug (2003a, 2003b).
unit is then defined as The convective-heat-transfer coefficient outside the pipe, hO, is
estimated by use of the Petukhov (1970) correlation based on the
共␳LcPLvSL + ␳GcPGvSG兲A⌬TU
USU = glycol velocity and the hydraulic diameter of the annulus (≈2400
␲dI lU共TUA − TO兲 W/m2K). If an insulation layer is added to the outside (or inside)
dI ␳LcPLvSL + ␳GcPGvSG pipe wall, it is convenient to incorporate the heat resistance into the
=
冉 冊 冉 冊冉 冊
. corresponding equations (Eqs. 12, 22, 27, 28, 49, and 59).
4 b f b f 1 1 共1 − e−alF兲共1 − e−elS兲 Figs. 5 and 6 show the comparisons between model predictions
l − l + − −
a F e S a e e a 共1 − e−alF−elS兲 and experimental measurements of the convective-heat-transfer
. . . . . . . . . . . . . . . . . . . . . . . . . . (58) coefficients for single-phase gas and liquid flows, respectively.
The pipe inside-convective-heat-transfer coefficient can be further The Dittus and Boelter (1930) and Petukhov (1970) correlations
obtained as are used to predict the (turbulent) single-phase gas and liquid
convective-heat-transfer coefficients, respectively. The experi-
1 mental results represented by the hollow diamond symbols in
hSU = . . . . . . . . . . . . . . . . . . . . . . . . . (59)
1 dI dO dI Fig. 6 are based on the temperatures measured by Probes TT29 and
− ln − TT49 (Fig. 4). These two probes are close to the pipe wall, and
USU 2kP dI hOdO
the measured temperatures are obviously lower than the bulk tem-
From Eq. 57, the average temperature gradient of a slug flow in the perature, especially for single-phase liquid and bubbly flows.
direction of flow can be expressed as Therefore, the calculated convective-heat-transfer coefficients
⭸T ⌬TU are slightly higher. This difference can be corrected by use of
=−
⭸l lU the other temperature measurements inside the flow (e.g., TT44).
The solid-diamond symbols in Fig. 6 show the results after cor-
共TUA − TO兲 rection. The good agreement between model predictions and ex-
=−
冉 冊 冉 冊冉 冊
.
b f b f 1 1 共1 − e−alF兲共1 − e−elS兲 perimental measurements for single-phase gas and liquid flows
lF − lS + − − indicates that the instruments are reliable and the selected corre-
a e a e e a 共1 − e−alF−elS兲 lations are appropriate.
. . . . . . . . . . . . . . . . . . . . . . . . . . (60)
The experiments of Shoham et al. (1982) on heat transfer in
horizontal slug flows were carried out under conditions of uniform
heat flux (by electrically heating the pipe wall). The data analy-
sis showed that there was significant temperature difference be-
tween the top and bottom of the flow because of the difference of
the heat-transfer coefficients. In modeling of heat transfer under
such conditions, the nonuniformity across the pipe section needs to
be considered.
Comparisons With Experiments
The present unified heat-transfer model for gas/liquid pipe flow is
compared with Manabe’s (2001) experimental results. The experi-
ments were carried out with natural gas and crude oil flowing in a
52.5-mm-ID pipe. The two-phase flow was cooled by cold glycol,
which flowed counter-currently in the outside annulus with a flow
rate of 0.00368 m3/sec. The pipe orientations were horizontal and
upward vertical. The flow patterns included bubbly flow, stratified
flow, annular flow, and slug flow. The test pressure was approxi-
matelly 3 MPa. Fig. 4 schematically shows the heat-transfer mea- Fig. 5—Convective-heat-transfer coefficient for single-phase
surement section in the Manabe study. gas flow.

118 February 2006 SPE Production & Operations

Fig. 6—Convective-heat-transfer coefficient for single-phase Fig. 7—Convective-heat-transfer coefficient for bubbly flow.
liquid flow.

Table 1 lists the statistical parameters for the overall compari-

Fig. 7 shows the comparison for vertical bubbly flow. Although son between model predictions and experimental measurements of
the correction for bulk temperature has been made, the model the two-phase convective-heat-transfer coefficients. These param-
still slightly underpredicts the two-phase convective-heat- eters are defined with the relative percentage error,
transfer coefficients by approximately 10% compared to most of
the measurements. Qpre − Qexp
er = × 100 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (61)
Fig. 8 shows the comparison for horizontal stratified flow. Qexp
Because the experimental convective-heat-transfer coefficients are
based on the gas-core temperature measured by TT29 and TT49, where Qpre is the physical quantity predicted by the model, and
Eq. 24 in the model is used to calculate the two-phase overall- Qexp is the corresponding experimental measurement:
heat-transfer coefficient, USA. N

兺e
Figs. 9 and 10 show comparisons for horizontal and vertical 1
␧1 = rj , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (62)
annular flows, respectively. According to the model prediction of N j=1
the liquid-film thickness, the locations of probes TT29 and TT49
N

兺 |e | ,
are very close to the interface, sometimes in the film and some- 1
times in the gas core. Therefore, in the model, the two-phase ␧2 = rj . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (63)
N j=1
overall-heat-transfer coefficient, USA, is calculated using Eq. 25.
The model slightly underpredicts horizontal annular flow and

冪
N
slightly overpredicts vertical annular flow.
The comparisons for horizontal and vertical slug flows are 兺 共e
j=1
rj − ␧1兲2
shown in Figs. 11 and 12. It is seen that the model predictions ␧3 = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (64)
N−1
agree with experimental measurements quite well for slug flows.
The overall comparisons of the two-phase convective- The average percentage error, ␧1, is a measure of the agreement
heat-transfer coefficients between model and experiment for the between predicted and measured data. It indicates the degree of
different flow patterns discussed above are shown in Fig. 13. It overprediction (positive value) or underprediction (negative
is seen that most of the data points are located inside the ±20% value). The degree of agreement between predicted and measured
error band. Comparisons between model predictions and experi- data is reflected even more significantly by the absolute average
mental measurements of the temperature gradient in the flow di- percentage error, ␧2, because the negative and positive errors do
rection are not demonstrated in this study. They are almost the not cancel out in the calculation. The standard deviation, ␧3, indi-
same as the comparisons for the inside two-phase convective-heat- cates the scatter of the errors with respect to their corresponding
transfer coefficients. average percentage error, ␧1.

Fig. 8—Convective-heat-transfer coefficient for horizontal Fig. 9—Convective-heat-transfer coefficient for horizontal an-
stratified flow. nular flow.

February 2006 SPE Production & Operations 119

Fig. 10—Convective-heat-transfer coefficient for vertical annu- Fig. 11—Convective-heat-transfer coefficient for horizontal
lar flow. slug flow.

As seen in Table 1, the present unified heat-transfer model for The unified heat-transfer model is verified by comparison with
gas/liquid pipe flow gives good predictions of the convective-heat- Manabe’s (2001) experimental results for different flow patterns in
transfer coefficient compared with experimental measurements. a crude-oil/natural-gas system flowing in a 52.5-mm-ID pipe with
The average percentage error is 2.52%, the absolute average per- horizontal and upward-vertical inclinations. Good agreement has
centage error is 19.88%, and the standard deviation is 29.39%. been observed in the comparison.
Comparisons between model predictions and experimental There are different definitions of the two-phase heat-transfer
measurements are more scattered in some cases than in others. coefficients based on different representative inside temperatures.
This is attributed both to measurement uncertainties and to mod- When comparisons are made between the model and experimental
eling simplifications. The experimental convective-heat-transfer data, it is important to make sure that the same representative
coefficient is calculated using measurements of gas- and liquid- inside temperature is used.
flow rates, holdups, temperatures of fluids and the pipe wall. Un- In the model, the interfacial convective-heat-transfer coeffi-
certainties of these measurements can accumulate and result in cient between the liquid film and gas core is assumed to be the
significant error in the convective-heat-transfer coefficient, espe- same as that between the gas core and the pipe wall. This may not
cially when the base value of the measurement is small. The mod- be a concern if the temperature difference between the film and gas
eling simplifications include assumptions and closure relation- core is not significant.
ships used in both the unified hydrodynamic and heat-transfer
Nomenclature
models. Correlations for predicting interfacial friction factor and
an interfacial heat-transfer coefficient are among those that may a ⳱ local constant defined in Eq. 46b
need to be improved. A ⳱ cross-sectional area inside the pipe, L2, m2
b ⳱ local constant defined in Eq. 46c
Conclusions and Discussion B ⳱ local constant defined in Eq. 18
A unified heat-transfer model for gas/liquid pipe flow is developed cp ⳱ specific heat, L2/t3-T, J/kg-K
in conjunction with the unified hydrodynamic model of Zhang d ⳱ pipe diameter, hydraulic diameter, L, m
et al. (2003a, 2003b), which can predict flow-pattern transitions, dl ⳱ length
liquid holdups, pressure gradient, and slug characteristics in gas/ e ⳱ local constant
liquid pipe flow at all inclinations from –90° to +90° from hori- er ⳱ relative percentage error
zontal. The heat-transfer modeling is based on energy-balance f ⳱ Fanning friction factor, local constant
equations and analyses of the temperature differences and varia- h ⳱ inside convective-eat-transfer coefficient, m/t3-K,
tions in the liquid film, gas core, and slug body. W/m2-K

Fig. 12—Convective-heat-transfer coefficient for vertical Fig. 13—Overall comparison of two-phase Convective-heat-
slug flow. transfer coefficients.

120 February 2006 SPE Production & Operations

 References
Aggour, M.A.: “Hydrodynamics and Heat Transfer in Two-Phase Two-
Component Flow,” PhD dissertation, U. Manitoba, Canada (1978).
Davis, E.J., Cheremisinoff, N.P., and Guzy, C.J.: “Heat Transfer with
Stratified Gas-Liquid Flow,” AIChE J. (1979) 25, No. 6, 958.
Dittus, F.W. and Boelter, L.M.K.: “Heat Transfer in Automobile Radiators
of the Tubular Type,” U. California (Berkeley) Pub. Eng. (1930) 2,
443.
Kim, D. et al.: “An Evaluation of Several Heat Transfer Correlations for
hO ⳱ outside convective-eat-transfer coefficient, m/t3-K, Two-Phase Flow with Different Flow Patterns in Vertical and Hori-
W/m2-K zontal Tubes,” Proc. Natl. Heat Transfer Conference, Baltimore, Mary-
HL ⳱ liquid holdup land (August 1997).
Knott, R.F. et al.: “An Experimental Study of Heat Transfer to Nitrogen-
k ⳱ thermal conductivity, mL/t3-K, W/m-K
Oil Mixtures,” Ind. Eng. Chem. (1959) 51, No. 11, 1,369.
l ⳱ length, L, m
Kudirka, A.A., Grosh, R.J., and Mcfaddden, P.W.: “Heat Transfer in Two-
N ⳱ nondimensional number Phase Flow of Gas-Liquid Mixtures,” Ind. Eng. Chem. (1965) 4, No. 3,
q ⳱ heat flux, m/t3, W/m2 339.
Q ⳱ physical quantity Manabe, R.: “A Comprehensive Mechanistic Heat Transfer Model for
S ⳱ perimeter, L, m Two-Phase Flow with High-Pressure Flow Pattern Validation,” PhD
t ⳱ time, seconds dissertation, U. Tulsa, Tulsa (2001).
T ⳱ Temperature, K or °C Matzain, A.B.: “Multiphase Flow Paraffin Deposition Modeling,” PhD
U ⳱ overall-heat-transfer coefficient, m/t3-K, W/m2-K dissertation, U. Tulsa, Tulsa (1999).
v ⳱ velocity, L/t, m/sec Petukhov, B.S.: “Heat Transfer and Friction in Turbulent Pipe Flow with
␦ ⳱ nondimensional thickness Variable Physical Properties,” Advances in Heat Transfer, J.P. Hartnet
and T.V. Irvine (eds.) Academic Press, New York City (1970) 6, 505–
⌬ ⳱ difference
564.
␧1 ⳱ average percentage error
Ravipudi, S.R. and Godbold, T.M.: “The Effect of Mass Transfer on Heat
␧2 ⳱ absolute average percentage error Transfer Rates for Two-Phase Flow in a Vertical Pipe,” Proc. 6th Intl.
␧3 ⳱ standard deviation Heat Transfer Conference, Toronto (1978) 1, 505.
␮ ⳱ dynamic viscosity, m/L-t, kg/m-s Rezkallah, K.S. and Sims, G.E.: “An Examination of Correlations of Mean
␳ ⳱ density, m/L3, kg/m3 Heat Transfer Coefficients in Two-Phase and Two-Component Flow in
Vertical Tubes,” AIChE Symposium (1987) Series 83, 109.
Subscripts Shah, M.M.: “Generalized Prediction of Heat Transfer during Two Com-
1 ⳱ inlet of control volume ponent Gas-Liquid Flow in Tubes and Other Channels,” AIChE Sym-
2 ⳱ outlet of control volume posium (1981) Series 77, 140.
B ⳱ bulk Shah, R.K. and London, A.L.: Laminar Flow: Forced Convection in Ducts,
C ⳱ gas core Academic Press, New York City (1978) 77.
exp ⳱ experimental Shoham, O., Dukler, A.E., and Taitel, Y.: “Heat Transfer During Intermit-
F ⳱ liquid film tent/Slug Flow in Horizontal Tubes,” Ind. Eng. Chem. Fundamentals
G ⳱ gas (1982) 21, No. 3, 312.
Zhang, H.-Q. et al.: “Unified Model for Gas-Liquid Pipe Flow via Slug
I ⳱ interface, inside
Dynamics—Part 1: Model Development,” J. Energy Res. Technol.
j ⳱ index
(2003a) 125, No. 4, 266.
L ⳱ liquid Zhang, H.-Q. et al.: “Unified Model for Gas-Liquid Pipe Flow via Slug
M ⳱ mixture Dynamics—Part 2: Model Validation,” J. Energy Res. Technology.
Nu ⳱ Nusselt (2003b) 125, No. 4, 274.
O ⳱ outside
p ⳱ pipe
pre ⳱ predicted SI Metric Conversion Factors
Pr ⳱ Prandtl
mm × 3.937* E–02 ⳱ in.
Re ⳱ Reynolds
S ⳱ slug m3/sec × 5.435 E+05 ⳱ bbl/D
SA ⳱ stratified or annular flow MPa × 1.450 377* E+02 ⳱ Psi
SG ⳱ superficial gas W/m2K × 1.761* E–01 ⳱ Btu/(h–ft2–°F)
SL ⳱ superficial liquid *Conversion factor is exact.
ST ⳱ slug tail
SU ⳱ slug unit
T ⳱ translational Hong-Quan Zhang is an assistant professor of petroleum
engineering at the U. of Tulsa and Associate Director of TUFFP.
U ⳱ slug unit (e-mail: [email protected].) From 1998 to 2003, he
UA ⳱ slug-unit average was a senior research associate of TUFFP. Before joining the U.
W ⳱ wall of Tulsa in 1998, he was an associate professor and professor at
Tianjin U. In 1993 and 1994, as an Alexander von Humboldt
Superscripts Research Fellow, he conducted research at the Max Planck
C ⳱ gas core Inst. of Fluid Mechanics and the German Aerospace Research
Establishment in Göttingen, Germany. Zhang holds BS and MS
F ⳱ liquid film degrees from Xian Jiaotong U. and a PhD degree from Tianjin
M ⳱ mixture U., China. Qian Wang is a research associate at TUFFP and the
Tulsa U. Paraffin Deposition Projects. (e-mail: qian-wang@
Acknowledgments utulsa.edu.) From 1990 to 1999, she served as Assistant Re-
The authors wish to thank the TUFFP member companies for search Professor and Associate Research Professor at the Chi-
supporting this research project. nese Academy of Sciences. She holds BS (1982), MS (1984),

February 2006 SPE Production & Operations 121

and PhD (1989) degrees in mechanical engineering from Xian struction Award Committee, and is also a member of the SPE
Jiaotong U., China. Cem Sarica is a professor of petroleum Journal Editorial Board. James P. Brill is Professor Emeritus and
engineering and the Director of TUFFP and Tulsa U. Paraffin Research Professor of Petroleum Engineering at the
Deposition Projects at the U. of Tulsa. (e-mail: sarica@ U. of Tulsa and Director Emeritus of TUFFP. (e-mail: brill@
utulsa.edu.) He was an associate professor of petroleum and utulsa.edu.) He served as Editor of the ASME J. of Energy Re-
natural gas engineering at Pennsylvania State U. before joining sources Technology (JERT) during 1996–2000 and was an asso-
the U. of Tulsa. Sarica holds BS and MS degrees in petroleum ciate technical editor for JERT during 1983–1989. He became
engineering from Istanbul Technical U. and a PhD degree in an ASME Fellow in 2000 and was elected to the Natl. Academy
petroleum engineering from the U. of Tulsa. He currently serves of Engineering in 1997. Brill holds a BS degree from the U. of
as a member of the SPE Facilities and Construction Standing Minnesota and a PhD degree in petroleum engineering from
Committee, is the chairperson of the SPE Facilities and Con- the U. of Texas at Austin.

122 February 2006 SPE Production & Operations