Journal o f Petroleum Science and Engineering, 4 ( 1990 ) 273-289 273

Elsevier Science Publishers B.V., Amsterdam - - Printed in The Netherlands

Performance of a two-phase gas/liquid flow model in vertical

wells

C.S. Kabir and A.R. Hasan

Chevron Oil Field Research Co, P.O. Box, La Habra, CA 90633-0446 (U.S.A.)
Chemical Engineering Dep., University of North Dakota, Grand Forks, ND 58201 (U.S.A.)
(Received November 1988; revised and accepted May 8, 1989)

ABSTRACT

Kabir, C.S. and Hasan, A.R., 1990. Performance of a two-phase gas/liquid flow model in vertical wells. J. Pet. Sci. Eng.,
4: 273-289.

This paper presents application of a recently developed method for predicting two-phase gas/oil pressure-drop in ver-
tical oil wells. The new method, which is flow-pattern based, is capable of handling flow in both circular and annular
channels. Five principal flow regimes-bubbly, dispersed bubbly, slug, churn and annular - are recognized while develop-
ing appropriate correlations for predicting void fraction and pressure-drop in each flow regime.
Standard oilfield correlations are used for estimating PVT properties of oil and gas." Standing's correlation for solution
gas-oil ratio; Katz's correlation for oil formation volume factor; Standing's, and Chew and Connally's correlations for
dead and live oil viscosities, respectively; and Lee et al.'s correlation for gas viscosity. A finite-difference algorithm is
developed to compute pressure gradient in a wellbore.
Computations performed on 115 field tests, involving all the two-phase flow regimes, suggest that the new method
performs better than the Aziz et al. correlation. Further comparison of the new method's performance with other standard
methods, such as, Orkiszewski, Duns and Ros, Beggs and Brill, Hagedorn and Brown, and Chierici et al., reveals its
consistency and improved performance. The test data bank used in this study is that previously used by other authors;
thus, validation of the new method is demonstrated with an independent data set.

Introduction well can be traced back to the early 1950s when

Poettmann and Carpenter (1952) published
Predicting vertical multiphase flow behav- their predictive scheme. Since then many at-
ior in oil and gas-condensate wells is of great tempts have been made to predict the complex
practical significance and importance. Pres- flow behavior starting from modifications of
sure losses encountered during concurrent the Poettmann-Carpenter correlation to more
vertical flow of two or three phases enter into complex mathematical models capable of han-
a wide array of design calculations. Such de- dling flow hydrodynamics. To date, no single
sign considerations include: tubing size and correlation or model has emerged to success-
operating wellhead pressure in a flowing well; fully predict pressure drop under the wide
well completion or recompletion scheme; ar- range of operating conditions encountered in
tificial lift during either gas-lift or pump oper- wells around the world.
ation (submersible, sucker-rod, etc. ) in a low- The published correlation models can be
energy reservoir; liquid unloading in gas wells; classified into three categories based upon the
direct input for surface flow-line and equip- way the behavior is modeled. These models are
ment design calculations. briefly discussed in the order of increasing
Efforts to predict the pressure drop in an oil complexity.

0920-4105/90/$03.50 © 1990 Elsevier Science Publishers B.V.

274 C.S. K A B I R A N D A.R. H A S A N

Nomenclature
distribution parameter, dimensionless v~o terminal rise-velocity o f a gas bubble, m s-~ [ft
dp/dz pressure gradient, kPa m - ~ [psi f t - l ] S-l ]
D pipe diameter, m [ft] VolT terminal rise-velocity o f a Taylor bubble, m s - ~ [ft
De casing inside diameter, m [ft] S-I ]
Dt tubing outside diameter, m [ft] x mass-fraction o f the lighter phase, dimensionless
E entrainment o f liquid in gas-ore, dimensionless X Lockhart-Martinelli parameter for turbulent flow,
Eg in-situ gas volume-fraction, dimensionless dimensionless
in-situ gas volume-fraction in gas-core, e/D relative pipe roughness, dimensionless
dimensionless
Ym Fanning friction factor for the gas/oil mixture Greek symbols
g gravitational acceleration, m s -2 [ft s-~ ] A difference
g~ conversion factor, 32.2 Ibm ft lbf - l s -2 [ 1 in SI] pg gas viscosity, Pa s [cP ]
N~e Reynolds n u m b e r ( = OvmPm/J2m ), dimensionless /to oil viscosity, Pa s [cP]
p pressure at any location in flow string, kPa [lbf p~ gas density, kg m - 3 [ Ibm f t - 3]
ft -2 ] Po oil density, kg m - 3 [ Ibm f t - 3 ]
Ycgs critical gas velocity in the gas-core, m s -1 [ft s -~ ] a surface tension, N m - l [ dynes c m - l ]
Vgs superficial gas velocity, m s - l [ft s - l ]
I,~rn mixture (Vgs+ vow) superficial velocity, m s - l [ ft Subscripts
S-l ] c core-fluid in annular flow
YO in-situ oil velocity, m s - ~ [ ft s - J ] f fluid, single or multiphase
Vos superficial oil velocity, m s - l [ft s - ~] g gas
1 liquid
m mixture o f phases

Homogeneous flow model lations that are developed from laboratory ex-
periments on air-water systems. These corre-
lations have not found much application in the
This model assumes that a multiphase mix-
oil industry.
ture behaves much like a homogeneous single-
phase fluid, with property values that are some Separated flow or slip model
type of average (weighted, volumetric, etc. ) of This model assumes the phases to be segre-
the constituent phases. Thus the model im- gated with unequal velocities, known as slip.
plicitly assumes no slip or difference between Thus the slip velocity or the in-situ void frac-
in-situ phase velocities. tion of each phase needs to be known along
The pioneering work of Poettmann and Car- with the frictional interaction of the phases
penter (1952) and the subsequent modifica- with the wall and amongst themselves.
tions of that model made by Baxendell and Hagedorn and Brown (1965) reported a
Thomas ( 1961 ), Tek ( 1961 ), Fancher and correlation that requires estimation of an ef-
Brown (1963) and Hagedorn and Brown fective average in-situ void fraction. This cor-
(1964) fall into this category. Each modifica- relation has found wide application in the in-
tion of the Poettmann-Carpenter correlation dustry despite its empirical origin.
improved its applicability over the range of The Beggs and Brill (1973) correlation is
subsequent investigations. At the same time, based on extensive laboratory data and their
these studies revealed that the underlying as- own flow pattern map developed for horizon-
sumptions of the original work are severly lim- tal flow. This flow pattern map was used only
iting. More importantly, the effects of gas/liq- as a correlating parameter at other pipe orien-
uid ratio, total well flow-rate, liquid viscosity tations. The Mukherjee-Brill (1983) correla-
and tubing diameter are not properly handled tion was also developed empirically from their
in the model. large experimental database. A single equation
Govier and Aziz ( 1972 ) report other corre- for liquid holdup or void fraction was pro-
TWO-PHASE GAS/LIQUID FLOW MODEL IN VERTICAL WELLS 275

posed for al inclination angles from horizon- transition is predicted by the Griffith-Wallis
tal. The liquid and gas velocity numbers to- (1961) criterion, while the transitions from
gether with the inclination angle implicitly slug to churn, and churn to annular follow the
accounted for the flow regimes observed in Duns-Ros (1963) criteria. Orkiszewski's
their experiments. In general, both the Beggs- method has proven to be one of the reliable
Brill and Mukherjee-Brill correlations per- correlations in the oil industry.
form similarly. Thus, although flow patterns In the 1970s, four more correlations ap-
were recognized in these two correlations, no peared in the literature (Aziz et al., 1972;
independent equations were proposed to pre- Chierici et al., 1974a; Gould et al., 1974). Of
dict gas void fraction in each flow pattern as it these, Aziz et al.'s semi-mechanistic model
actually exists in the given conditions. based on the Govier-Aziz (1972) flow pat-
A large number of correlations have been tern map proved to be one of the better
developed using the slip model for horizontal methods.
multiphase flow. The pioneering work of The remaining two correlations of Chierici
Lockhart and Martinelli ( 1949 ) is based on the et al. (1974a) and Gould et al. (1974) were
slip model. A detailed work of Idsinga et al. both derived from the Orkiszewski (1967)
(1977) indicates that, for steam-water sys- model with minor modifications. For exam-
tems, the Lockhart-Martinelli correlation is ple, Chierici et al. made slight changes in the
applicable even for vertical upflow and down- void-fraction prediction during slug flow, while
flow when certain modifications are incorpo- Gould et al. made changes in predicting void
rated. However, application of such an ap- fraction in bubbly flow. Neither correlation has
proach has not been documented as yet. found wide application, however.
Although all these correlations, based on the
Flow pattern approach flow-pattern maps, are being used with vary-
In this approach, an attempt is made to de- ing degrees of success, no single method has
lineate the flow pattern at each pipe segment emerged as the most reliable. This uncertainty
and then use the appropriate void fraction and stems partially from the flow-pattern maps.
pressure-drop correlations for each flow pat- These maps, based on experimental data, are
tern. Although, in principle, this technique plotted on two-dimensional coordinate sys-
promises to be the most rigorous of all, the dif- tems delineating the transition boundaries. The
ficulty of identifying each flow regime has led dimensionless coordinate variables are gener-
to different flow-pattern maps and hence to ally chosen arbitrarily, thereby limiting the ac-
different correlations (Ros, 1961; Duns and curacy and scope of application of these maps
Ros, 1963; Orkiszewski, 1967; Aziz et al., 1972; beyond the original data base.
Chierici et al., 1974a; Gould et al., 1974). In their classical work, Taitel et al. (1980)
In the early work of Ros ( 1961 ) and Duns showed that the transition from one flow pat-
and Ros ( 1963 ), a flow pattern map with di- tern to another could be modeled from first
mensionless gas and liquid velocity numbers principles by properly incorporating the fluid
as coordinates was used. Other flow pattern properties and conduit size. Recently, Hasan
maps published in the 1960s were those of and Kabir (1988a,b) adopted an approach
Griffith and Wallis ( 1961 ) and Govier et al. very similar to Taitel et al. to model each flow-
( 1961 ). These maps were subsequently used pattern transition. Subsequently, the void
by others while developing pressure-drop fraction and pressure drop in each flow regime
correlations. was modeled.
Orkiszewski (1967) proposed a semi-me- In this work, we first briefly discuss the flow
chanistic approach in which bubbly to slug pattern transition boundaries. In particular, the
276 C.S. KABIR AND A.R. HASAN

departure from our previous approach for de- pipe carrying some liquids as droplets. The re-
lineating the slug/chum transition boundary is maining liquid flows up the wall through the
described. Second, we present both the com- annulus formed by the pipe wall and the gas
putational and application aspects of the model core.
(Hasan and Kabir, 1988a) using field In the following we briefly discuss the crite-
examples. ria for transition from one flow regime to an-
other. The approach adopted in this work is
Flow pattern determination the one pioneered by Taitel et al. (1980). This
approach examines each flow-pattern transi-
The chaotic nature of two-phase flow pre- tion individually and is more reliable than
sents a considerable challenge to classify the other empirical methods because it allows
flow regimes and ascribe transition criteria to physical modeling of transition boundaries.
them. However, with the recent attempts of
physical modeling (Taitel et al., 1980, for ex-
ample), a consensus appears to have emerged Bubbly/slug flow transition
with respect to the principal flow regimes. Various experimental and theoretical inves-
These flow regimes are identified as bubbly, tigations suggest that this transition occurs
dispersed bubbly, slug, churn and annular. when the gas void fraction reaches about 25
In bubbly flow, the gas phase tends to rise percent. This criterion translates into the fol-
through the continuous liquid column as small, lowing relationship in terms of the superficial
discrete bubbles. Because pure bubbly flow is velocities of the gas and liquid phases (Hasan
associated with low gas velocities, its occurr- and Kabir, 1988a):
ance is confined to low gas volume fractions. Vgs=0.429Vls +0.546 [gtr(pl --pg)/p~ ]0.25 ( 1)
At higher gas rates, the smaller bubbles collide
with each other, resulting in coalesence and At superficial gas velocities greater than that
formation of elongated (Taylor) bubbles. given by Eq. 1, transition to slug flow takes
These agglomerated bubbles are separated by place.
aerated liquid slugs and the flow regime is
called slug flow. With the rise in liquid veloc-
ity the turbulence increases, leading to the Transition to dispersed bubbly flow
breakup of large bubbles and formation of fully When large bubbles are dispersed into small
dispersed bubbly flow as dispersive forces bubbles at high liquid rates, transition to slug
overcome interfacial tension. flow is inhibited even though the gas void frac-
As the flow rates of both phases are further tion exceeds 25%. An examination of the shear
increased, the shear stress between the Taylor forces have led several investigators (Taitel et
bubbles and the liquid film separating them al., 1980; Bamea et al., 1982, 1985) to pro-
from the pipe wall increases, finally causing a pose the following equation for mixture veloc-
breakdown of these bubbles. The liquid slug ity for transition to dispersed bubbly flow:
also collapses and this liquid forms a new v1.12
m = 4.68 ( D )O.48[g(p I _pg)/rr]o.5
bridge after accumulating, only to be lifted
again by the oncoming, narrow Taylor bub- ( (T/p I ) O.6 (Pl/~Ul ) 0.08 (2)
bles. This oscillatory motion of the liquid slugs
leads to the transition to churn flow. The final When the gas void fraction exceeds 52%, bub-
flow regime, annular flow, occurs at extremely ble coalescence cannot be prevented and tran-
high gas-flow rates, which cause the entire gas sition to either slug, chum or annular flow must
phase to flow through the central portion of the occur.
TWO-PHASE GAS/LIQUID FLOW MODEL IN VERTICAL WELLS 277

Slug/churn flow transition on the liquid droplets and the gravitational

A very few reliable works exist to model this forces acting on them. The resulting expres-
transition. For instance, the approach of Tai- sion in terms of superficial gas velocity is given
tel et al. (1980) to view this transition as an by (Taitel et al., 1980; Hasan and Kabir,
entrance effect may not be applicable in long 1988a,b):
oil wells. In this work, we have adopted the vgs =3.1 [ag(pi _pg)/pg2 ]o.25 (3)
physical model of Barnea and Brauner ( 1985 )
and Brauner and Barnea (1986). They argued Equation 3 clearly suggests that this transition
that the transition from slug to churn flow oc- criterion is independent of the liquid rate.
curs when the liquid slug trailing a Taylor bub-
ble attains the m a x i m u m possible gas void Flow pattern maps
fraction of 52% as given by Eq. 2. In other The equations presented earlier can be used
words, this transition boundary is a locus of to construct the familiar flow-pattern maps,
constant mixture velocity, Vm, where the tur- with gas and liquid superficial velocities as the
bulent intensity is maintained at the same level coordinates. Two such maps were prepared for
as that in the dispersed bubbly flow. We clarify a 30°API oil flowing in a 3-inch ID pipe. The
this point further while discussing the flow first map was made by evaluating the fluid
pattern maps. properties at the standard conditions (14.7
psia, 60°F). The fluid properties used are:
Transition to annular flow a=0.0685 Ibm s -2, p1=54.67 Ibm ft -3,
In annular flow, the high gas velocities keep pg=0.0534 Ibm ft -3 a n d # g = 0 . 0 1 0 3 cP.
the liquid droplets in suspension. This transi- Figure 1 shows the flow pattern map at the
tion is modeled by balancing the drag forces standard conditions. Note that the dispersed

DISPERSED BUBBLY

,/ i0~_: A
c,q

I--
b_

/
rO BUBBLY
o CHURN
LuJ
10 ° -

ANNULAR

C)
b_ SLUG
LLI
D_
10-' -
U3

1 O-' 10 ° 10 ~ 10 2 10 3
S U P E R F I C I A L GAS V E L O C I T Y , FT S -1

Fig. l. Flowpattern map for a gas/oil system at standard conditions.

278 C.S. KAB1R AND A.R. HASAN

DISPERSED BUBBLY -~t

10'
'5
L/b

L
f BUBBLY
G
LtJ 10 °
>-
ANNULAR

/
o

LI_
Q~
LO
10-'
U3

• , . . . . ,,[
..... I ........ I ..... I ........ I
1 O-' 10 ° 10' 10 2 I0 ~
SUPERFICIAL GAS VELOCITY, FT S -~

Fig. 2. F l o w p a t t e r n m a p for a g a s / o i l s y s t e m at 1000 psia, 100 ° F.

bubbly flow pattern is represented by two 2 at 1000 psia and 100 oF. The property values
transition lines. The segment represented by used are: a = 0.03164 Ibm s -2, Pl - 54 Ibm ft- 3,
AB is given by Eq. 2, while the BC segment is pg= 3.77 Ibm ft -3 a n d s = 0 . 0 1 2 7 cP. At higher
based on the fact that the dispersed bubbly flow pressures, when the gas density is significant,
is limited by the maximum gas void fraction the transition to annular flow occurs at a much
criterion of 52%. The segment BD, represent- lower superficial gas velocity of 4.2 ft s- 1com-
ing the slug/churn transition, is indeed a con- pared to 44.4 ft s-1 at the atmospheric pres-
tinuation of the AB line. As discussed earlier, sure. Note also that the dispersed bubbly flow
the BD line represents the locus of constant area shrinks and the churn flow regime van-
mixture velocity for a given set of fluid prop- ishes. Consequently, there appears to be a
erties. Physically, this line implies that the ag- transition between bubbly and annular flow
glomeration of bubbles in the liquid slug regime. Physical explanation for a possible ex-
(present in between two Taylor bubbles) istence of such a transition is beyond the scope
reaches 52% leading to the destruction of the of this work.
liquid bridge, and the resulting churning mo-
tion. The BD line corresponds to the ratio le/D Pressure gradient estimation
of about 130 as given by the Taitel et al. ( 1980 )
equation:
The overall pressure gradient, in two- dp/dz,
Le/D=40.6(~gD+ 0.22) (4) phase flow can be written as the sum of the
gravitational or hydrostatic head (dp/dZ)H,
To examine the map at a condition likely to frictional (dp/dZ)F
and accelerational (dp/
be prevalent in an oil well, we constructed Fig. dz) A components:
TWO-PHASEGAS/LIQUID FLOWMODEL IN VERTICALWELLS 279

bubbly flow treatment could be used for esti-

dp/dz= ( dp/dz )u + (dp/dz)v + (dp/dZ)A
mating the gas void fraction.
2fmV2pm dvm
=(g/gc)Pm4 g¢D ~-pmVrn~z- g (5)
Slug and churn flow
where: The drift-flux model or Eq. 7 still applies in
slug flow with different values for Co and v~ as
pm =pgEg + pl(1--Eg) (6) given by:
In general, the accelerational component could
be neglected during all but the annular flow re- Co = 1.18 +0.90D~ (10)
gime. Equation 6 clearly suggests that an ac-
curate estimation of the gas void fraction is es- Voox= [0.30+0.22(Dt/Oc) ]
sential to the hydrostatic head computation [x/g(Oc -O,) (pt-pg)/pt ] (11 )
that accounts for most of the pressure drop oc-
curring in the bubbly and slug flow regimes. For churn flow, Co assumes a values of 1.15 to
Because of different hydrodynamics in each reflect the flatter velocity profile, whereas the
flow regime, estimations of void fraction, Eg, expression for VaT remains unchanged.
in-situ mixture density, Pm, and the friction
factor, fm, are made separately. Annular flow
In annular flow, estimation of gas void frac-
tion in the central core is pertinent rather than
Voidfraction estimation
the entire pipe cross-sectional area. The gas
void fraction for the gas-core, Egc, is given by:
Here, we present the appropriate equations
as given by Hasan and Kabir (1988a) for per- Egc _ Vgs ( 12 )
forming void fraction calculations in each of Vgs+ Evls
the five flow regimes. where the vapor entrainment, E, is estimated
from the following conditions.
Bubbly~dispersed bubbly flow If Vcgs× 104-<4:
In bubbly flow the expression for void frac-
E = 0.0055 (Vcgs× 104) 2.86 ( 13 )
tion, Eg, is given by:
If Vcgs×104>--4:
/)gs
Eg -Cov-~-+ voo (7) E=0.857 log [ VcgsX 104 ] --0.20 (14)

where: where the critical vapor velocity, Vcgs,is calcu-

lated from:
Co = 1.2+0.371D~ (8) v p,
Vcgs-- (15)
t7
voo = 1.50 [ga(p~ -pg)/p~ ] 0.25 (9)

and, Dt/Dc is the tubing to casing diameter ra- Frictional head estimation
tio when flow occurs in a tubing-casing annu-
lus. This term vanishes when the flow is in a The friction factor correlation used here is
circular pipe; thus, retaining the Co value of that of Chen (1979). In a recent study, Gre-
1.2. gory and Fogarasi (1985) observed that the
In the dispersed bubbly flow, where larger explicit form of f given by Chen is the best over
bubbles coexist without coalescing, the same a wide range of Reynolds number and relative
280 c.s. KABIR AND A.R. HASAN

pipe roughness. Chen's equation is given as: phase, while a thin liquid film creeps up the
wall of the pipe. Thus the frictional pressure
1 • F e/D 5.04521ogA]
~=4 log[ 3 y0~ 5 Re J (16) drop pertains to the gas interacting with the
wavy liquid film. The frictional pressure drop,
where: which is a large component of the total pres-
sure drop, can be written as:
A - (e/D)ll°98 F(7.149/Re) °'8981 (17)
2.8257 Apf _ 2f~pc( vgs/Eg)e (20)
gcD
~./D is the relative pipe roughness and Re is the
Reynolds number (=Dvp//a). While evaluat- where:
ing the Reynolds number in two-phase flow,
f ~ = O . O 7 9 [ l + 7 5 ( 1 - E g ) ] / R e °2~ (21)
the density and viscosity of the liquid phase
needs to be used (Aziz et al., 1972). This fric- v~vog + Epl Vl~
Pc - (22)
tion factor correlation is used for computing Vgs+ Evgs
frictional head in the bubbly, slug, and churn
The expressions for entrainment, E, are given
flow regimes. Use of the explicit form of the
by Eqs. 13 and 14.
Chen equation significantly enhances the
The total pressure gradient, including the
computation speed when compared with other
accelerational component, is given by:
implicit form of JZcorrelations, such as given
by Colebrook (1939). - d p l d z = 1 [gPc+(2fcPcV2/d)] (23)
gc [ 1 - (p¢v~/p&) ]
Bubbly flow
In bubbly flow we assume that the two-phase The gas void fraction in the gas core, E~c, in
frictional pressure drop is the same as the sin- Eq. 12 can be estimated from the Lockhart-
gle-phase liquid flow. Thus we can write: Martinelli ( 1949 ) parameter, X, which can be
written in terms of the gas mass fraction, x, as:
2fmV2pm
Apt-- (18) X= [ ( 1 --x)/x]°9(pg/pl)°S(ltl/ltg)°l (24)
g~D
In Eq. 18, fm indicates that the term should be
evaluated for the gas-liquid mixture Reynolds Fluid PVT properties computation
numbers.
Well-known empirical correlations were used
Slug and churn flow to obtain the pressure- and temperature-de-
In these two flow regimes, the wall friction pendent physical and transport properties of
is essentially experienced by the liquid film the two phases. For the oil phase, Katz's
flowing up the pipe with large bubbles occu- ( 1942 ) correlation is used to calculate the for-
pying most of the cross-sectional area of flow. mation volume factor, while Standing's
This fact requires a modification of the den- ( 1947 ) methods predicted the solution gas-oil
sity term as given by Eq. 18 given below: ratio and bubblepoint pressure. For dead-oil
viscosity we used Standing's (1962) correla-
zip, 2fm2
VmPl(1 - E g ) (19) tion and the Chew-Connally (1959) correla-
gcD tion for live-oil viscosity.
The gas viscosity was calculated from the Lee
Annular flow et al. (1964) method, whereas the gas-law de-
In annular flow, fine liquid droplets flow in viation factor or z-factor was obtained from the
the gas core with a velocity the same as the gas Standing's ( 1977 ) modification of Beggs-Brill
TWO-PHASE GAS/LIQUID FLOW MODEL IN VERTICAL WELLS 281

curve-fit of the Standing-Katz (1959) z-fac- During the pressure traverse calculation, a
tor chart. In computing surface tension, we "flow regime" sub-routine identifies the flow
used the Baker-Swerdloff (1956) correlation regime prevalent in a certain pipe segment,
for gas/oil system and the Hough et al. ( 1951 ) thereby permitting the use of an appropriate
correlation for gas/water system. set of equations for calculating the total zlp.
We recognize that the Beggs and Robinson The flow-regime transition equations pre-
(1974) correlations for the dead- and live-oil sented earlier pose some problems during
viscosities are perhaps superior to those used computation. The problems stem from sharp
in this work. However, two reasons prevent us rather than a smooth change at the transition
from using the Beggs and Robinson correla- boundaries. This apparent discontinuity prob-
tions. First, the correlations have a lower limit lem is pronounced at the bubbly/slug and
of 15 °API oil gravity, whereas we treated tests churn/annular boundaries. For the bubbly/
having as low as 8.3 °API oil. Second, we com- slug transition, a weighted-average rise veloc-
pared our results directly with the published ity is calculated using the terminal rise veloci-
results, which make use of the classical corre- ties of a single and a Taylor bubble spanning
lations (Standing, 1962; Chew and Connally, over +_ 15% of the transition velocity itself. The
1959). In any case, we point out that the oil void fraction in the transition zone is esti-
viscosity plays a minor role in pressure drop mated from Eq. 7 by simply replacing the ter-
computation. minal rise velocity with the weighted-average
Similar comments apply for not using the rise velocity. This scheme ensures a smooth
Vazquez-Beggs (1980) correlations for esti- transition across the bubbly/slug boundary.
mating oil formation volume factor and solu- Treatment of the churn/annular boundary is
tion gas-oil ratio. Besides the lower limit ofoil somewhat different, however. A zone of tran-
gravity ( 15 ° API ), the correlations require the sition is arbitrarily defined by assuming a
knowledge of gas gravity at the separator con- _+ 15% of the velocity given by Eq. 3. Void
ditions and of the operating pressure and tem- fractions are directly computed by assuming
perature of the separator. Unfortunately, sep- both the churn and annular flow, and weighted
arator conditions are rarely reported together by the relative "position" of the velocity with
with other data for two-phase flow calculations. respect to the transition velocity.

Computational method Model validation and comparison

An iterative finite-difference algorithm is
developed to calculate depth increments cor- The validity of our model is examined by
responding to the specified values of pressure comparing with measured pressure-drop data,
drop, zip. Gas void fraction and hence the hy- together with other methods. Some 115 test
drostatic head together with the frictional head data points are considered for statistical
are calculated after evaluation of the fluid comparison.
properties and volumetric flow rates of the In this study, the computation involved the
phases at the mid-point of the depth incre- proposed method and that of Aziz et al.
ment. When the calculated value of/lp is within (1972): the remaining results of other meth-
_ 0.10 psi of the assumed value, the calcula- ods are taken from various sources. The other
tion proceeds to the next step. Otherwise, the predictive schemes considered for comparison
depth increment is halved and the procedure are those of Orkiszewski (1967), Duns and
repeated. This iterative calculation is contin- Ros (1963), Beggs and Brill (1973), Hage-
ued until the sum of the depth increments dorn and Brown (1965) and Chierici et al.
equals the length of the flow string. (1974a).
282 C.S. KABIR AND A.R. HASAN

TABLE 1

Range o f well test parameters used in this study

Field Units SI units

Oil producing rate 44 -11 623 S T B / D 7 - 1838m3d -l

G a s - o i l ratio 143 - 9975 S C F / S T B 25 - 1776 m 3 m -3
Water-oil ratio 0 - 1.38
Oilgravity 8.3 - 46 ° API
Gas gravity ( a i r = 1 ) 0.57 - 1.7
Well Depth 3890 - 1 3 123 ft 1186 - 4000 m
Internal tubing diameter 1.995- 4.488 in 50.67- 114 m m
Wellhead pressure 20 - 2780 psig 138 - 1 9 168 kPa
Bottomholepressure 640 - 6630psig 4413 -45714kPa
Observed pressure-drop 474 - 4566 psi 3268 -31 4 8 2 k P a
Wellhead temperature 40 - 241 ° F 277.5 - 389K
Bottomhole temperature 95 - 297°F 308 - 420 K

Data sets from four independent sources are head's contribution was no greater than 2% of
used for the comparative study. Table 1 lists the total. Exception to this observation occurs
the range of well-test parameters used in this in some high flow rate wells. For example, well
study. We discuss each data set separately. no. 47 indicates a zJpf o f 16.4% of the total.
As pointed out by Aziz et al., some of the
Data of Aziz et al. data points are obviously suspect. For exam-
In their original work, Aziz et al. (1972) ple, although flow in well nos. 1 and 16 is es-
performed computation on 48 well tests, using sentially single phase, large errors are pre-
their own m e t h o d and that of Orkiszewski dicted by all the methods. These and other
(1967). They reported Espanol's (1968) cal- dubious data points are excluded from the sta-
culated results on 38 wells for Hagedorn and tistical analysis.
Brown ( 1965 ) and Duns and Ros ( 1963 ) cor-
relations. Table 2 displays those results to- Data of Chierici et al.
gether with calculated results from the pro- This data group is perhaps one of the most
posed method showing percentage error in reliable and complete sets reported in the lit-
predicting flowing bottomhole pressures. erature. Comparison of the various methods
The calculated percentage error in this study for the total pressure-drop is shown on Table
for both the total pressure drop and the flow- 4.
ing bottomhole pressure are very close to the Flow patterns in most wells are restricted to
reported values of Aziz et al. when their single phase, bubbly and slug. Churn and an-
method was used. Therefore, they are not re- nular flow regimes are rarely observed in these
peated here. wells. The proposed m e t h o d appears to per-
Table 3 compares the statistical results of form quite satisfactorily compared to the
five methods for predicting the total pressure Chierici et al. and Aziz et al. methods. Note
drop. The proposed m e t h o d performs just as that the Chierici et al. method is expected to
well as the methods of Aziz et al. and Orki- perform the best because their correlation was
szewski for this data group, dominated by the "tuned" to this data set.
single-phase, bubbly and slug flow regimes. In Our calculations with the Aziz et al. ( 1972 )
most cases, the hydrostatic head contributed m e t h o d are in good agreement with those re-
largely to the total pressure drop; the frictional ported by Gregory et al. (1980). As indicated
TWO-PHASE GAS/LIQUID FLOW MODEL IN VERTICALWELLS 283

TABLE2 TABLE 3

C o m p a r i s o n of results using Aziz et al. data Statistical results using Aziz et al. data

Error in calculated flowing b o t t o m h o l e pressure (%) Prediction Average Standard Remarks

method error deviation
(%) (%)
Well Aziz Orkiszewski Hagedorn- Duns- Proposed
No. et al. Brown Ros method Aziz et al. -0.57 10.45 44 points used
(well nos. 1, 11, 16,
1 29.7 29.7 - - 33.70
46 discarded)
2 2.0 2.0 2.1 1.7 2.10 Orkiszewski - 1.98 10.69 Same as above
3 -6.2 -6.3 -5.9 -7.0 -5.7 Hagedorn and - 17.52 22.98 36 points used
4 -4.9 -4.9 -4.7 -7.8 -4.9 Brown as indicated on
5 16.6 16.2 - - 17.0 Table 2
6 -8.8 -8.8 -8.6 -10.9 -8.7 Duns and Ros - 4.13 13.52 Same as above
7 -6.4 -6.4 -6.3 -8.7 -6.3 *Proposed method - 1.02 10.37 43 points used
8 -3.3 -3.3 -3.2 -8.4 -3.2 (well nos. l, 11, 16,
9 9.3 7.4 - - 10.4 18,24 discarded)
10 -11.9 -11.9 -11.7 -14.1 -11.8
*Computation performed for this study: the remaining results are
11 40.0 36.7 23.6 37.8 41.8
from Aziz et al.
12 -3.7 -7.6 -11.0 -5.0 -2.1
13 -11.1 -9.5 -51.3 -14.3 -9.6
14 - 1.7 - 1.7 - - - 1.6 TABLE4
15 -8.7 -7.2 -73.3 -13.2 -7.8
16 55.3 53.4 33.0 4.1 56.0 Statistical results using Chierici et al. data
17 -11.5 -12.1 -37.3 -14.4 -11.1
18 4.1 13.7 -59.3 -2.4 - Prediction Average Standard Remarks
19 -8.4 -8.5 -8.4 -10.2 -8.0 method error deviation
20 5.9 -4.8 -24.6 -0.2 4.1 (%) (%)
21 -4.5 -10.7 -30.0 -10.0 -7.2
Beggs and Brill 3.97 13.76 All 31 points
22 -13.6 -24.9 -41.6 17.9 -18.0
considered
23 4.8 2.6 - - 1.03
in all cases
24 14.8 1.9 - - -
Aziz et al. - 2.82 5.75
25 19.5 14.9 -9.3 15.2 17.80
Orkiszewski - 2.18 14.70
26 12.9 12.5 13.3 10.7 13.0
D u n s and Ros 6.94 15.43
27 -1,3 -3.6 -1.0 1.0 0.80
Hagedorn and Brown 1.06 17,01
28 10,2 9.6 10.4 8.0 10.8
+Chierici et al. 0.12 5,51
29 -10.3 -7.2 -84.2 -16.5 -8.5
*Aziz et al. - 2.45 7.01
30 -14.7 -14.8 -15.2 -15.4 -14.2
*Proposed m e t h o d -0.86 9.27
31 -15.1 -15.8 -31.6 -19.0 -15.6
32 -5.7 -6.8 - - -4.7 * C o m p u t a t i o n performed for this study: the remaining re-
33 12.3 11.7 12.7 11.0 12.4 sults are from Gregory et al. and Chierici et al.
34 -14.0 -14.6 -29.9 -17.2 -13.1 + Results from Chierici et al.
35 19.0 6.8 -27.5 46.0 13.8
36 -6.3 -9.5 -7.8 -6.4 -5.9
37 13.9 13.1 -23.7 11.1 12.1 in Table 4, the top five entries are reproduced
38 7.2 6.5 2.4 4.3 7.1 from Gregory et al.'s work for top-down cal-
39 -7.7 -5.2 -38.4 -12.4 -9.6
culations using the experimental bubblepoint
40 -1.8 -2.6 -3.6 -6.6 -1.8
41 -5.1 -5.1 -6.6 -6.5 -5.0 pressure.
42 -7.1 -7.6 -8.7 -8.4 -6.9 We observed an improvement in pressure-
43 -6.2 -6.4 -8.6 -8.4 -6.0 drop prediction when the reported bubble-
44 -11.1 -17.9 -3.4 -19.6 -16.4
45 9.5 -0.6 - - 10.6
point pressure was used instead of the Stand-
46 2.0 - - - 1.3 ing's predicted value. The same conclusion was
47 15.8 16.8 - - 17.1 reached earlier by Gregory et al. (1980). Thus,
48 21.0 23.3 5.0 23.1 18.2
whenever the producing bottomhole pressure
is above the oil's bubblepoint pressure -
meaning single-phase flow prevails in some
284 C.S. KABIRAND A.R. HASAN

segment of the pipe - pressure-drop prediction TABLE 6

becomes sensitive to the chosen bubblepoint Statistical results using Baxendell-Thomas data
pressure. Obviously, the input of measured
bubblepoint pressure is desirable in such a Prediction Average Standard Remarks
method error deviation
computational scheme. (%) (%)
Data of Orkiszewski Orkiszewski - 3.7 41.1 All 25 points
used
Results of statistical analysis are presented
Duns and Ros - 13.9 39.5 All 25 points
in Table 5 for 22 heavy-oil (9.5-18.7 ° API) used
wells. The proposed method appears to per- Hagedorn and Brown 6.0 11.6 All 25 points used
form reasonably well when compared with the *Axis et al. -38.8 26.8 24 points used,
test # 14
methods of Orkiszewski and Aziz et al. As in- discarded
dicated on Table 4, Lawson and Brill's ( 1974 ) *Proposed method - 8.8 27.3
calculated results are used for comparison. *Computation performed for this study: the remaining re-
Lawson and Brill reported a large discrep- sults are from Lawson and Brill.
ancy that exists between their own results and
those of Orkiszewski (1967). They attributed Data of Baxendell and Thomas
this anomaly to the definition of percentage This data set, obtained from a series of high-
error. However, the confusion resurfaces when rate experiments (upto 5082 STB/D), indi-
the same data are reexamined by Gould et al. cates presence of annular flow regime through-
(1974), who achieved good agreement with out the pipe in many cases. Results of the sta-
Orkiszewski's calculated values. tistical analysis reflecting the total pressure-
Because of the lack of necessary data on oil drop prediction are presented in Table 6. Test
and gas gravities, top and bottomhole temper- number 14 is discarded from analysis because
atures and the like, we assumed the same val- both Aziz et al. and the proposed method pre-
ues as reported for well 22 in Orkiszewski's dict over 200% error in pressure drop: experi-
original paper for all the wells. These neces- mental error is suspected.
sary assumptions could conceivably lead to The proposed method appears to perform
disagreements in results. In addition we found better than the other flow-pattern approaches
that some of the wells, five to be specific, be- when both the average error and standard de-
have erratically and were excluded from the viation are compared. This result probably re-
statistical analysis. flects the improved recent knowledge gained
in the annular flow regime as implemented in
TABLE 5
the model. We however note that the Hage-
Statistical results using Orkiszewski data dorn-Brown ( 1965 ) slip model performs sur-
prisingly well for this data set.
Prediction Average Standard Remarks
method error deviation
(%) (%) Discussion
Orkiszewski - 12.0 13.1 All 22 points
considered In this work we tested and compared results
Duns and Ros - 42.0 19.7 Same as above of the proposed model with those routinely
Hagedorn and Brown -29.7 42.0 Same as above
*Aziz et al. 24.92 13.3 17 points
used in the industry. As indicated in the pre-
considered vious section, the proposed model performs
*Proposed method 13.27 18.44 Same as above just as well as the Aziz et al. (1972 ) and Or-
*Computation performed for this study: the remaining re- kiszewski ( 1967 ) models when the flow is pre-
sults are from Lawson and Brill. dominantly in the bubbly and slug flow re-
TWO-PHASE GAS/LIQUID FLOW MODEL IN VERTICAL WELLS 285

gimes. This observation is not very surprising D of 0.0008 was used to retain the smooth pipe
because the proposed model does not substan- definition. Additionally, estimation of friction
tially differ from the Aziz et al. model. Bubbly- factor from the knowledge of Reynolds num-
slug transition criterion and the consideration ber and relative pipe roughness still retains
of dispersed bubbly flow are the major differ- empirical origin, especially in two-phase flow.
ences between the two models when bubbly Thus frictional pressure-drop prediction is not
and slug flow regimes are considered. How- rigorous.
ever, because of different transition criteria in Fortunately, the frictional pressure-drop
the two models, the proposed method predicts component is only a small fraction of the total
a longer pipe segment experiencing bubbly flow pressure drop in many cases. For example,
than that of Aziz et al. Gregory et al. (1980) showed that the fric-
When most of the pressure drop occurs in tional head amounted to only a maximum of
the churn and annular flow regimes, Aziz et 5% of the total head in 85 of their 105 well tests.
al.'s (1972) model appears to yield large er- The preceding observation implies that the
rors, as Table 6 indicates. We point out that hydrostatic head is by far the most dominant
the transition criteria for the churn and annu- component when the flow pattern can be de-
lar flow are adopted from the work of Duns scribed as either single phase, bubbly and slug
and Ros (1963), although not stated explic- or a combination of these. However, at higher
itly. Here, the proposed model appears to per- superficial fluid velocities, for example in an-
form well because of improved treatment of the nular flow regime, the dominating influence of
transition to annular flow and of the frictional the frictional head warrants a good handle on
pressure drop in annular flow. the friction factor itself.
Thus when we consider the 115 test data The second question pertains to the reliabil-
group spanning all the four flow regimes, the ity of data against which a model or a correla-
proposed model indicates an improvement tion is judged. Barring some obvious reporting
over the Aziz et al. method. Table 7 shows the errors from the field as eluded to earlier, pub-
comparative statistical data, while Figs. 3 and lished data do not state the well/reservoir flow
4 exhibit the graphical comparison of the two condition. In other words, if the surface flow
models with the measured data. measurements are made during the unsteady-
Despite the apparent improvement in pres- state flow period in the reservoir, serious prob-
sure-drop prediction, many uncertainties re- lems may arise. These problems relate to the
main. The first question concerns the relative reported values of flowing bottomhole pres-
pipe roughness. Certain crude constituents, sure (Pwf) and the corresponding flow rates.
wax for example, are likely to be deposited as The measured Pwf would change minimally
a thin film on the pipe wall, resulting in a when the pseudosteady-state flow is attained
smooth pipe. In all of our computations, a e/ in the reservoir, ideally, one would like to ob-
serve no change in the Pwf when steady-state
TABLE7 flow occurs in the reservoir. We point out that
stable flow condition in the entire flow string
Overall c o m p a r i s o n o f p r o p o s e d m e t h o d with Aziz et al., 115
tests
is required for application of the steady-state
pressure-drop correlations. Recognizing the
Prediction Average Standard difficulty of maintaining a constant wellhead
method error deviation
(%) (%) rate and given the current surface flow-rate
measurement practice, we recommend gath-
Aziz et al. - 5.34 25.42
ering Pwf and flow-rate data during either pseu-
Proposed method 1.25 19.05
dosteady- or steady-state flow in the reservoir.
286 CmS. KABIR AND A.R. HASAN

~oo
/
1 .............. i .............. ~ .............. ~ ............. ~ .............. i .............. i .............. ~ .............. ~ ........ ~ . . . . . . . . . . .

i ! i i i i i i iqlll

,ooo ............
;
........................................................................
i i i i
, .........................
) ~-'!, .........

v 3500- .............. i: . . . . . . . . . . . . . . ~: . . . . . . . . . . . . . .
~). . . . . . . . . . . . . .
0,. :! . . . . . . . . . . . . . . ~• . . . . . . . . . . . . . . .! . . . . . . . . . . . . . . ie ....... ="! .............. i ..............
0
,.,, i i :: i ie ,, =~== ::
o ~ o o o ............................. ::.............. ~......... ....~ ....... . - : .......... ~ - - i .............. i .............. i ..............
'" :: :: i i= ::= i :: :: i
:~ i ! i i !m i ! i i

03 2500-
O.
~ 2000-

.-J i aa:: =i ::
~D 1500 - ............................. ! ........ R : .............. i .............. i .........................................................................
U :: ~ ! i i i i :

1000 . . . . . . . . . . . . . . . . . . m . . . . ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

i,~i ,,=~' i
500 -t .......... '"fm- ,,t:.~..~..~ .............. i .............. ~ .......................................................... ~..............

O ~ / / ~ "" B : : :

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
MEASURED PRESSURE DROP(psi)

Fig. 3. C o m p a r i s o n o f m e a s u r e d a n d c a l c u l a t e d pressure drops: Aziz et al. m e t h o d .

5O00 ~ /

,coo- .............. :.............. ~.............. ::i.............. i::.............. ii.............. !i............. i~......... ~ i ~ ! i~. ........
4000 ............... i .............. i .............. i .............. i ............. i .............. i .............. :: . . . . . . . . . . . . . . . .

,.-... i i i i ! i ! . ill :

°
" i i i ~ Z. i .i )
3000-
O~
(/~
09 2500 ................................................................
.. m.-m .................................................................
o,,.
0..

2000-

..J
1500-

1000-

/U !
500 ...... ~ ~ = i ............................. i ............................. i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

J ='!" ~ i i
0 540 '
1000 1500 20~00 2500 30~00 3500 4000 4500 5000
MEASURED PRESSURE DROP(psi)
Fig. 4. Comparison of measured and calculated pressure drops: proposed method.
TWO-PHASE GAS/LIQUID FLOW MODEL IN VERTICAL WELLS 287

Flow rate measurement at the surface over a tity whenever a meaningful validation and
test duration gives only a time-average value, comparative study is attempted.
with no indication of instantaneous variations One of the major applications of a vertical
that might have occurred. Experience indi- multiphase flow model is to generate the tub-
cates that instantaneous downhole measure- ing performance curves under various operat-
ments of flow rate together with pressure, den- ing conditions of gas-oil ratio, tubing head and
sity and temperature can largely overcome the bottomhole pressures, and tubing sizes. Under
limitations associated with surface measure- a set of operating variables, even an empirical
ments. While the oil flow-rate measurement is correlation may indicate the least error com-
accurate, the same cannot be stated for gas and pared to a mechanistically based model, such
water. Our calculations indicate that some of as the one described here. However, with a
the discrepancies between the predicted and change in operating variables, an empirical
the observed pressure drops could be ex- correlation is unlikely to perform well even
plained by only ___15% change in gas-oil ratio. though a good match was indicated previously
Another source of uncertainty in the re- - a typical problem with an empirical correla-
ported data stems from the tubing diameter. A tion. For example, the Lawson and Brill
nominal diameter of, say 2] inch, does not ( 1974 ) study and the subsequent discussion by
specify the internal diameter required for cal- Chierici et al. (1974b) clearly expose the lim-
culations. Consequently, we observe a 5% dif- itations associated with the homogeneous
ference in the tubing diameter translating into model of Poettmann and Carpenter (1952)
a large error in the pressure-drop prediction - and other methods (Baxendell and Thomas,
upto 8% in some cases. Such an error is likely 1961; Fancher and Brown, 1963; Duns and
to be significant when the overall pressure drop Ros, 1963 ). Thus, for any design calculations,
is dominated by the frictional head. we strongly recommend use of a mechanistic
Additionally, our implicit assumption is that model for reliable performance over a wide
all the test data pertain to truly vertical wells. operating conditions.
This assumption is seldom strictly correct, es-
pecially for deep wells. In a deviated well, con- Conclusions
siderable problems may arise when slug flow is
present in the flow string. For example, the in- The general observations from this study are
crease in Taylor bubble rise velocity compared in good agreement with those reported in the
to that observed in a vertical well (Hasan-Ka- literature. In the following, we summarize the
bir, 1988b ) will translate into a lower gas void specific conclusions reached.
fraction and hence a higher overall zip. Thus, ( 1 ) When vertical multiphase flow is dom-
ignoring the true well deviation is a potential inated by any combination of single-phase,
source for error. bubbly and slug flow regimes, the new method
Even though a n u m b e r of uncertainties in- is consistent with the other mechanistic models
herent in field data may potentially prevent us of Aziz et al. and Orkiszewski.
from rigorously testing a model, comparison (2) In the churn and annular flow regimes,
with other models largely removes the valida- the new method appears superior to the exist-
tion problem. In this work, we considered some ing models.
115 well tests. Because a larger database is de- (3) For the 115 well tests considered in this
sirable from model validation viewpoint, we work, encompassing all the four two-phase flow
do not claim universality of the proposed regimes, the new method appears to be a letter
model.However, we point out that the quality overall performer than that of Aziz et al. Com-
of data is of greater importance than the quan- parison with other methods also indicates the
288 C.S. KABIR AND A.R. HASAN

new method's improved performance. correlation for gas-saturated crude oils, Trans. AIME,
(4) A number of uncertainties may cloud a 216-23.
Chierici, G.L., Ciucci, G.M. and Sclocchi, G., 1974a. Two-
model validation study. These uncertainties phase vertical flow in oil wells - prediction of pressure
include relative pipe roughness of the flow drop. J. Pet. Technol., (Aug.): 927-937; Trans. AIME,
channel, gas-oil ratio, water-oil ratio, well de- 257.
viation, oil bubblepoint pressure, and pressure Chierici, G.L., Ciucci, G.M. and Sclocchi, G., 1974b.
and flow-rate data (that is whether gathered Discussion - A statistical evaluation of methods used
to predict pressure losses for multiphase flow in verti-
during stable well and reservoir flow cal oilwell tubing. J. Pet. Technol., (Aug.): 913; Trans.
condition ). AIME, 257.
Chen, N.H., 1979. An explicit equation for friction factor
in pipe, Ind. Eng. Chem., Fundam., 18 (3): 296-297.
Acknowledgement Duns, H., Jr. and Ros, N.C.J., 1963. Vertical flow of gas
and liquid mixtures in wells. Proc. 6th World Petro-
Our appreciation is due to M. Fogarasi of leum Congr., Frankfurt, II, pp. 451-465.
Espanol, J.H., 1968. Comparison of three methods for
Neotec Consultants, Calgary, for providing calculating a pressure traverse in vertical multiphase
additional information on the 48-well dataset flow. M.S. thesis, Univ. Tulsa, Okla.
of their paper (Aziz et al., 1972). Fancher, G.H., Jr and Brown, K.E., 1963. Prediction of
pressure gradients for multiphase flow in tubing. Soc.
Pet. Eng. J., (Mar.): 59-69; Trans. AIME, 231.
References Govier, G.W. and Aziz, K., 1972. The flow of Complex
Mixtures in Pipes. Van Nostrand Reinhold, New York,
Aziz, K., Govier, G.W. and Fogarasi, M., 1972. Pressure N.Y.
drop in wells producing oil and gas. J. Can. Pet. Tech- Gould, T.L., Tek, M.R. and Katz, D.L., 1974. Two-phase
nol., (Jul-Sep.): 38-47. flow through vertical, inclined, or curved pipe. J. Pet.
Baker, O. and Swerdloff, W., 1956. Finding surface ten- Technol., (Aug.): 915-926; Trans. AIME, 257.
sion of hydrocarbon liquids. Oil & Gas J., 2 (Jan.): Gregory, G.A. and Fogarasi, M., 1985. Alternate to stan-
125. dard friction factor equation. Oil & Gas J., 1 (Apr.):
Barnea, D. and Brauner, N., 1985. Holdup of the liquid 120-127.
slug in two phase intermittent flow. Int. J. Multiphase Gregory, G.A., Fogarasi, M. and Aziz, K., 1980. Analysis
Flow, 11 : 43-49. of vertical two-phase calculations: crude oil-gas flow
Barnea, D., Shoham, O. and Taitel, Y., 1982a. Flow pat- in tubing. J. Can. Pet. Technol., (Jan.-Mar.): 86-91.
tern transition for downward inclined two-phase flow, Griffith, P. and Wallis, G.B., 1961. Two-phase slug flow.
horizontal to vertical. Chem. Eng. Sci., 37: 735-740. J. Heat Transfer, ASME, (Aug.): 307-320.
Barnea, D., Shoham, O., Taitel, Y. and Dukler, A.E., Hagedorn, A.R. and Brown, K.E., 1964. The effect of liq-
1982b. Gas-liquid flow in inclined tubes: flow pattern uid viscosity in vertical two-phase flow. J. Pet. Tech-
transition for upward flow. Chem. Eng. Sci., 40:131- nol., (Feb.): 203-210; Trans. AIME, 231.
136. Hagedorn, A.R. and Brown, K.E., 1965. Experimental
Baxendell, P.B. and Thomas, R., 1961. The calculation of
study of pressure gradients occurring during continu-
pressure gradients in high-rate flowing wells. J. Pet.
ous two-phase flow in small diameter vertical con-
Technol., (Oct.): 1020-1028; Trans. AIME, 222.
duits. J. Pet. Technol., (Apr.): 475-484.
Beggs, H.D. and Brill, J.P., 1973. A study of two-phase
flow in inclined pipes. J. Pet. Technol., (May): 607- Hasan, A.R. and Kabir, C.S., 1988a A study of multi-
617; Trans. AIME, 255. phase flow behavior in vertical wells. S.P.E. Prod. Eng.,
Beggs, H.D. and Robinson, J.R., 1975. Estimating the (May): 263-272.
viscosity of crude oil systems. J. Pet. Technol., (Sep.): Hasan, A.R. and Kabir, C.S., 1988b. Predicting multi-
1140-1141. phase flow behavior in a deviated well. S.P.E. Prod.
Brauner, N. and Barnea, D., 1986. Slug/churn transition Eng., (Nov.): 474-482.
in upward gas-liquid flow. Chem. Eng. Sci., 41: 159- Hough, E.W., Rzasa, M.J. and Wood, B.B., 1951. Inter-
163. facial tension at reservoir pressure and temperature;
Colebrook, C.F., 1939. Turbulent flow in pipes with par- apparatus and water-methane systems. Trans. AIME,
ticular reference to the transition region between 192: 57-60.
smooth and rough pipe laws. J. Inst. Civ. Eng., 11: 133. Idsinga, W., Todreas, N. and Bowring, R., 1977. An as-
Chew, J.-N. and Connally, C.A., Jr., 1959. A viscosity sessment of two-phase pressure-drop correlations for
TWO-PHASE GAS/LIQUID FLOW MODEL IN VERTICAL WELLS 289

steam-water systems. Int. J. Multiphase Flow, 3:401- Poettmann, F.H. and Carpenter, P.G., 1952. The multi-
413. phase flow of gas, oil and water through vertical flow
Katz, D.L., 1942. Prediction of shrinkage of crude oils. strings with application to the design of gas lift instal-
Drill. and Production Practice, API, p. 13. lations. Drill. and Production Practice, API, p. 257.
Katz, D.L. et al., 1959. Handbook of Natural Gas Engi- Ros, N.C.J., 1961. Simultaneous flow of gas and liquid as
neering. McGraw-Hill, New York, N.Y. encountered in well tubing J. Pet. Technol., (Oct.):
Lawsom J.D. and Brill, J.P., 1974. A statistical evaluation 1037-1049; Trans. AIME, 222.
of methods used to predict pressure losses for multi- Standing, M.B., 1947. A pressure-volume-temperature
phase flow in vertical oilwell tubing. J. Pet. Technol., correlation for mixtures of California oil and gases.
(Aug.): 903-913; Trans. AIME, 257. Drill. and Production Practice, API, p. 275.
Lee. A.L. et al., 1964. Viscosity correlation for light hy- Standing, M.B., 1962. Oil system correlation. In: T.C.
drocarbon systems. AIChE J., (Sep.): 694. Frick (Editor), Petroleum Production Handbook, Vol.
Lockhart, R.W. and Martinelli, R.C., 1949. Proposed II, Ch. 19. SPE of AIME, Dallas, Tex.
correlation of data for isothermal two-phase, two- Standing, M.B., 1977. Volumetric and Phase Behavior of
component flow in pipes. Chem,. Eng. Prog., 45: 39- Oil Field Hydrocarbon Systems. SPE of AIME, Tex.
48. Tek, M.R., 1961. Multiphase flow of water, oil and natu-
Mukherjee, H. and Brill, J.P., 1983. Liquid holdup cor- ral gas through vertical flow strings. J. Pet. Technol.,
relations for inclined two-phase flow. J. Pet. Technol., (Oct.): 1029-1036; Trans. AIME, 222.
(May): 1003-1008. Vazquez, M. and Beggs, H.D., 1980. Correlations for fluid
Orkiszewski, J., 1967. Predicting two-phase pressure drops physical property prediction. J. Pet. Technol., (Jun.):
in vertical pipe. J. Pet. Technol. (Jun.): 829-838; 968-970.
Trans., AIME, 240.