Documente Academic
Documente Profesional
Documente Cultură
Introduction erages properties based on the overall gas-liquid ratios does not
reflect the relevant physics of the situation.
Research on heat transfer estimation for nonboiling gas-liquid
Previous work on heat transfer prediction in multiphase sys-
flow in pipes dates back at least 50 yr. Nevertheless, no general
tems primarily consists of methods limited to specific flow re-
predictive methods exist for engineering applications, particu-
gimes. Approaches fall into three classes: 1) empirical correla-
larly for gas-oil systems. Increasingly, subsea petroleum pro-
tions (Rezkallah and Sims, 1987; Shah, 1981; Kudirka et al.,
duction systems are being engineered to transport full well-
1965; Davis and David, 1964; Fried, 1954); 2) modified single-
streams. How these multiphase systems behave thermally is
phase flow methods (Oliver and Wright, 1964; Knott et al.,
critical to the prevention of costly gas hydrate and wax deposi-
1959); and 3) momentum transfer-heat transfer analogies, par-
tion blockages (Furuholt, 1988). The goal of the present work
ticularly for annular and stratified flow (Davis et al., 1975,
is to develop a readily applied method suitable for the estimation
1976, 1979; Pletcher and McManus, 1972; Hughmark, 1963).
of inside heat transfer coefficients for two-phase petroleum
The correlation methods are usually in terms of the superficial
flowlines. Only nonboiling flows are considered in the present
velocity ratio Usg/Vji and/or Reynolds number based on the su-
work. Two-phase boiling heat transfer is reviewed by Butter-
perficial liquid velocity (Rezkallah et al., 1986).
worth and Hewitt (1977).
Considering the physical analogy between heat and momentum
Reviews of heat transfer in nonboiling two-phase flow sys- transfer, heat transfer estimation should be of similar complexity
tems are given by several authors (Kim et al., 1998; Rezkallah, to multiphase flow modeling, which has advanced far beyond
1987; Shah, 1981; Michiyoshi, 1978). Pubhshed experimental correlative methods to detailed mechanistic models (Dhulesia and
data are limited to primarily small-diameter pipes (<2.5 cm) Lopez, 1996; Govier and Aziz, 1972). Moreover, in large part
at low pressure ( < 6 bar). Moreover, much of the data are the features of two-phase flow which impact frictional pressure
for air-water systems. Any estimation method for engineering loss should similarly impact heat transfer. Highly detailed mecha-
application in gas-oil pipelines must incorporate sufficient phys- nistic models for multiphase flow heat transfer eventually should
ical theory to provide adequate predictions well beyond the and wiU be developed, but this is at least several years away given
range of the experimental pipe sizes and pressures. Empirical the problem's complexity, experimental data requirements, and
methods for frictional pressure loss in two-phase flow which current interest levels. Some strides in this direction have already
are based on small-diameter and low-pressure systems tend to been taken for specific flow regimes (Shohamet al., 1982). How-
be inaccurate for petroleum pipeline systems (Govier and Aziz, ever, the goal of the present work is not to undertake the large
1972). Similarly, it is likely that empirical heat transfer methods task of developing new and separate mechanistic models for two-
may be satisfactory over the range of existing data, but inaccu- phase heat transfer, but rather to satisfy better the engineering need
rate for large-pipe diameter and high-pressure systems. in the meantime for plausible and readily applicable prediction of
At first glance the problem of two-phase heat transfer may multiphase heat transfer, particularly in petroleum pipeline sys-
appear to have a trivial solution; simply average the properties tems. The approach taken here is to interface strongly with the
of the gas and liquid phases and apply standard single-phase extensive existing work on two-phase flow modeling. In the fol-
heat transfer calculation methods. The quality of predictions by lowing sections, methods are described which mathematically sep-
this approach, however, is in general quite poor (Vijay, 1978). arate the heat transfer physics from the two-phase flow behavior.
The poor agreement is not due to the details of the averaging The methods allow heat transfer to be predicted directly from
method, but rather a fundamental flaw in the approach. Specifi- estimations of frictional pressure loss, liquid holdup, and flow
cally, heat transfer behavior is largely a function of the fluid regime by whatever predictive methods are most appropriate to
conditions near the wall, especially for high Prandtl number determine the two-phase flow behavior given the pipe size, op-
fluids, such as oil. Hence, it is the properties of the liquid, which erating pressure, and fluids. In this way, two-phase flow predictive
wets the walls and has the bulk of the sensible heat, which methods known to be accurate for large, high-pressure systems
dominate heat transfer behavior. Thus, any approach which av- may be employed, and, hence, increase the confidence in heat
transfer estimations for such systems. The new methods are evalu-
Contributed by the Petroleum Division and presented at the Twentieth Annual
ated against literature data.
Energy-Sources Technology Conference and Exhibition, Houston, Texas, Febru- Experimental Database
ary 1-3, 1999, of THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Manu-
script received by the Petroleum Division, August 3, 1998; revised manuscript Table 1 shows the extent of the database used in the present
received January 18, 1999. Associate Technical Editor; C. Sarica. work. The database consists of 786 data points, all for low-
Journal of Energy Resources Technology Copyright © 1999 by ASME JUNE 1999, Vol. 121 / 75
cm cp dyne/ °C "C
cm
Count Source Liquid - Gas Orien- D ^1 CT| Pr, Tb T ^w/^b Regimes Re,i
tation (L/D)
183 Vijay Water-Air V 1,2 0.7 72 4.9 10 23 0.54 All 231
(52) 1.3 9.3 34 38 0.95 206637
61 Vijay Glycerine - Air V 1.2 718 63 6142 27 33 0.47 AH except 2
(52) 817 6965 28 38 0.61 ann.-mist 24
102 Vijay 75wt% Glycerine/ V 1.2 20 65 173 23 34 0.55 All except 6
25wt% water - Air (52) 31 258 31 38 0.78 ann.-mist 4568
57 Rezkallah Water - Air V 1.2 0.8 71 5.3 19 24 0.70 All 708
(52) 1.0 7.2 31 43 0.92 128071
158 Rezkallah Silicone Liquid - V 1.2 4.5 20 65 19 23 0.75 All 171
Air (52) 5.4 77 27 41 0.89 20991
108 Rezkallah 58wt% Glycerine/ V 1.2 6.2 67 47 21 25 0.57 All 70
42wt% water - Air (52) 9.2 70 31 47 0.90 13240
21 King Water - Air H 1.9 0.5 70 2.9 58 85 0.61 Slug 22500
(252) 0.7 3.1 63 100 0.70 119000
62 Knott et Oil - Nitrogen V 1.3 75 25 1116 25 40 0.41 Bubbly, 7
al. (119) 133 1917 40 59 0.80 Slug 156
30 Johnson and Water - Air H 2.2 0.4 70 2.6 47 80 0.61 All 14247
Abou-Sabe (207) 0.6 3.7 69 98 0.73 220517
4 Johnson Oil - Air H 1.9 17.8 25 292 53 98 0.25 Bubbly, 2118
(253) 19.9 321 56 100 0.28 Slug 2312
NOTE: Minimum and maximum values are given for |Xi, Pr,, T,,, T„, and n^/Rb- "H" Indicates horizontal orientation and "V"
Indicates vertical orientation,
pressure ( < 6 bar) systems. No data were located for signifi- steam was used as the heating medium), leading to non-negligi-
cantly higher pressures. The database represents the large bulk ble property changes over the pipe length. Additionally, some
of published data. Other data sources were considered, but not systems have relatively large pressure drops. In this work prop-
used, due to incomplete reporting of conditions or low L/D erties are evaluated at the average pipe temperatures and pres-
ratios. No data are included for inclined flow, of which the sures for calculation purposes.
only identified source is Hetsroni et al. (1998) for air-water. Vijay (1978) and Rezkallah (1987) report the most complete
Dorresteijn (1970) and Chu and Jones (1980) studied both and detailed data. Both used the same experimental apparatus.
upwards and downwards vertical flow, but the reported data Reported estimated uncertainties and repeatabilities of their heat
were incomplete for incorporation into the database and L/D transfer experimental data are ± 7 - 2 1 percent. Small pressure
was low ( — 15) for Dorresteijn's data. Data are limited for drops associated with very low flow rates could not be accu-
horizontal flow. No quality data were identified for laminar rately measured in their experiments, and such cases are not
horizontal flow. Johnson (1955) reports some for an oil-air included in the database used in the present work.
system, but due to very large differences in the bulk and wall
temperatures, the physical properties varied considerably across
the pipe. Except for the air-water data of Johnson and Abou- Heat Transfer in Gas-Liquid Flow With Turbulent
Sabe (1951), none of the sources report experimental liquid Liquid Flow
holdup fractions. Some of the systems in the database have For turbulent flow in a pipe, a laminar boundary layer exists
large temperature differences over the pipe length (usually since near the wall. Prandtl demonstrated that, to a reasonable approx-
Nomenclature
(b)
0.50
A...K>i.^..^>^.^.4
]6"§
! iS - y^t
-stratified
0 fTS ^S^^jiP ^ T - ^
A Bubble rn 8na Dr7,w, 4.°-" •
DSlug £ -0.25 4 - 4 , - - - - V a D O - - - ' ^ '
X Froth O ^ °° S+D • a
+Churn
O Annular -stratified A Bubble DSiug X Froth
• Annular-Mist + Churn O Annular •Annular-Mist
I I 11 =
0.8 1 0.00 0.20 0.40 0.60 0.80 1.00
Experimental: h(p/hs| Liquid Holdup Fraction (Est.)
0.75
tfo +
(d)
0.50
tu V-
C 0.00
.2 ^.mh^^Wr
+
2 -0.25 4-J.:-
X X
+Water
DOii
AS8wt% Giyoerine/42wt% water -0.50 - +Water nOil
X75wt% Giyoerine/25wt% water
A58wt%Glycerlne/42wt%water X76wt%Glycerine/26wt% water
—Glycerine
0 Silicone -Glycerine O Silicone
-0.75
0.8 1. 2. 5. 10. 20. 0.00 0.20 0.40 0.60 0.80 1.00
Experimental: h|p/lis| Liquid Holdup Fraction (Est.)
0.75
u X A * ^
,? -0.25
-0.50
• Turb;horiz XLann;vert ATurb;vert
- L a m ; 1-phase DTurb; 1-phase
-0.75
0.8 1 2. 5. 10. 0.00 0.20 0.40 0.60 0.80 1.00
Experimental: htp/hg)
Liquid Holdup Fraction (Est.)
Fig. 1 Comparison of heat transfer predictions and experimental data. Ail figures report the same data plotted In various ways. Dashed ilnes
indicate ± 3 3 percent. Fractional Error = (Xoaio - Xe,p)/iVIAX[Xcate, x.xp]-