., .

A Study of Two~PhaseRow in
Inclined Pipes
H. Dale Beggs,* SPE-AIME, U.of Tulsa
James P. Brill, SPE-AIME, U. of Tulsa

Introduction
The prediction of pressure drop and liquid holdup lem in single-phase flow because the potential energy
occurring during two-phase gas-liquid flow in pipes is lost going uphill is regained in the downbili section.
of particular interest to the petroleum, chemical, and This is not the case for two-phase flow, because the
nucleat industries. In the nuclear industry, two-phase liquid holdup, and thus the mixture density, are
flow occurs in reactor cooling equipment, and liquid usually much lower in downhill flow. For this reason,
holdup greatly affects heat transfer. Two-phase flow pressure recovery in the downhill sections is usually
occurs frequently in chemical processing, and the neglected in the design of two-phase pipelines.
design of processing equipment and piping systems The number of directional or inclined welk is in-
requires knowledge of pressure drop, liquid holdup, creasing as the search for petroleum moves into pre-
and often flow pattern. In the petroleum industry, viously unexplored areas. In offshore drilling, several
two-phase flow occurs in pipelines and in oil and gas directional wells are usually drilled from one plat-
wells, More than one-half the natural gas gathered in form for economic reasons. Deviations of 35 to 45
the U. S. flows in two-phase flowlines. Most gas- from the vertical are common. In the permafrost
producing wells produce some liquid and most oil areas of Alaska and Canada, the cost of drilling-rig
wells produce some gas. As the natural reservoir foundations and the difficulty of transportation re-
energy is depleted, many wells are equipped with quire that several wells be directionally drilled from
artificial-lift systems such as gas lift. To design these one location, Existing vertical-flow correlations fre-
systems, a method of predicting two-phase-flow pres- quently fail to predict pressure gradients in these
sure gradients is required. wells within acceptable limits.
Although extensive research in two-phase flow has Gathering lines from offshore wells usually are
been conducted during the last 25 years, most of this laid along the sea floor that slopes up to the shore.
research has concentrated on either horizontal or The elevation pressure gradient in a pipeline with a
vertical flow, Several good correlations exist for pre- very small upward inclination from horizontal can
dicting pressure drop and liquid holdup in either hori- be much greater than the frictional pressure gradi-
zontal or vertical flow, but these correlations have ent. Therefore, in order to predict pressure drop, the
not been successful when applied to inclined flow. liquid holdup must be accurately predicted. The
Many gathering lines and long-distance pipelines pass ability to predict liquid holdup also is essential for
through areas of hilly terrain, This presents no prob- designing field processing equipment, such as gas-
*NOW with Louisiana Tech U., Ruston. liquid separators, When the flow rates of wells pro-

Gas-liquid flow in inclined pipes was investigated to determine the effect of pipe
inclination angle on liquid holdup and pressure loss, Correlations for liquid holdup
and friction jactor were developed for predicting pressure gradients for two-phase
flow in pipes at all angles for many flow conditions.

MAY, 1973 607

 .

ducing into gathering lines are changed, the amount earlier by Mamayeve in 1965 and odishariyal in
of liquid holdup in the lines may change and result 1966. Their data were obtained using 2-in. pipe
in overloading of processing equipment. at angles between + 9 and 9 fr;m hori~o;-
The two-phase flow problem is complicated by tal. Liquid holdup was found to be a function of
such phenomena as slippage between phases, change Froude number and input gas content, Am equation
of flow pattern, and mass transfer between phases. was given for holdup in uphill plug flow that was
The gas-liquid interface may be smooth or wavy and independent of angle, but the equations for downhill
energy may be transferred between phases. These flow included an inclination effect, A correlation was
factors result in a much greater pressure 10SS than given for the ratio of two-phase friction factor to
can be explained by the . ;duced area available to single-phase friction factor as a function of gas-input
flow for each phase. When angle of flow is added to content, with Froude number as a parameter. Their
such variables as fluid properties, flow rates, and friction factor correlation should be used with cau-
pipe diameter, the problem is indeed formidable. tion since the friction factor ratio becomes un-
Very little work has been published on the eflect bounded as the flow approaches all gas.
of inclination on two-phase flow in pipes. Pressure A study of slug flow in inclined pipes was reported
gradients in directional wells are usually calculated. by Singh and Griffith in 1970. They measured
using a vertica! flow correlation such as that of Hage- pressure drop and liquid holdup in pipe with dianle-
dorn and Brown or of 0rkiszewski.12 This is satis- ters of 0,626, 0,822, 1.063, 1.368, and 1.600 in., at
factory if the well is fnclined only slightly from ver- inclination angles of plus and minus 10 and 50 from
tical. Pressure gradients in pipelines may be calcu- horizontal, and at 0, An expression was derived for
lated using a horizontal-flow correlation such as that bubble-rise velocity from which liquid holdup could
of Dukler ef al.3 or Eaton, ~The pressure drop caused be calculated, Liquid holdup was found to be inde-
by elevation change may be calculated if liquid hold- pendent of inclination angle, which would result in
up can be determined. 100-percent calculated pressure recovery in the
In 1958, Flanigan proposed a method to calcu- downhill section of an inclined pipe, Frictional pres-
late two-phase pressure drop in pipelines in hilly sure drop was calculated using the Fanning friction
terrain. The method was developed using field data factor, the liquid density, and the mixture velocity.
obtained from a 16-in, pipeline and the data of Acceleration pressure drop was neglected,
Baker. Flanigan used the Panhandle equation for After the literature survey was completed, it was
gas flow with an dficiency factor to calculate the concluded that no correlation exists for predicting the
frictional pressure drop. The elevation pressure drop liquid holdup and pressure drop that occur during
was calculated using an elevation factor, the liquid two-phase flow in pipes at all angles. The purpose
density, and a summation of the uphill elevation of this study, then, was to find a correlation of this
changes. Flanagan concluded that the inclination type that would be particularly applicable in design-
angle of the hills had no effect on the elevation fac- ing pipelines for hilly terrain and tubing strings for
tor and that there was no pressure recovery on the inclined wells.
downhill side of the hill. A correlation was given for
elevation factor as a function of superficial gas Theory
velocity, This correlation includes, to some extent, The equation used to calculate pressure gradient
the effect of pressure recovery on the downhill side, when gas or liquid, or both, flow in a pipe is
since the over-all pressure drop was used in develop-
ing the correlation, A correlation for efficiency factor dp _
as a function of superficial gas velocity and gas- dZ
liquid ratio was given also.
f,,,G,),v,,,
In 1962, Sevigny studied the flow of air and ~ sin e[PJ+L +- P!,(1 H1.)j +
2grd ,
water mixtures through a 0.8245 -in. -diameter pipe
inclined at angles of plus and minus 90, 60, 30, 1. _ ~~r. + W( 1 HL)Iv,,?Vsg
15, 10, and 5 from horizontal, and at OO.Holdup gcP
was not measured, and the pressure drop not ac- . . . . ... . . . . . (1)
counted for by elevation change was included in a
friction-loss term. The no-slip or input density was The development of this equation is found in the
used to calculate elevation pressure drop, in a meth- Appendix, Eq. 1 reduces to the equation for single-
od similar to that used by Poettmann and Carpenterx phase liquid or single-phase gas flow as HL + 1 or
in their study of vertical flow. Sevigny presented a HL + O, respectively. Also, as the angle of the pipe,
correlation for two-phase friction factor as a function 0, becomes zero, + 90, or 90, Eq, 1 becomes
of input liquid content, gas Reynolds number, and applicable to horizontal or vertical flow.
liquid Reynolds number. The accuracy of this cor- Eq, 1 contains two unknowns: HL, which must be
relation is very questionable. For some conditions, determined to calculate the in-situ density, and jtP,
the method gives a pressure recovery of more which is used to calculate friction losses. The prir-
than 100 percent in downhill flow, which is clearly pose of this study was to develop correlations for
impossible. predicting HL and jtP from fluid and system properties
In 1967, Guzhov et al. presented results of a that are known. To accomplish this, an experimental
study of inclined two-phase flow. Their paper in- apparatus was designed and built so that flow rates,
cluded much of the information that was reported pressure gradient, inclination angle, and liquid hold-

608 JOURNAL OF PETROLEUM TECHNOLOGY

 up could be measured,
Elquid System
Experimental Study A schematic diagram of the test system is shown in
The data from which the correlations were developed Fig. 1. The liquid was stored in a 100-cu-ft tank
were taken in transparent acrylic pipes 90 ft long. equipped with a heat exchanger, The gas was passed
Sufficient length was provided between the gas-liquid through this heat exchanger to equalize the gas and
mixing point and the test section to eliminate en- liquid temperatures. The temperature of the two-
trance ef~ects before any measurements were made. phase mixture was measured at each end of the test
The parameters studied and their range of varia- section. Liquid flow was provided by a Pacific cen-
tion were ( 1) gas flow rate (O to 300 Mscf/D); (2) trifugal pump with an output capacity of 60 gal/rein
liquid flow rate (O to 30 gal/rein); (3) average at 95 psig. The liquid flow rate was controlled by a
system pressure (35 to 95 psia); (4) pipe diameter choke valve upstream of the meters. After Passing
( I and 1,5 in.); (5) liquid holdup (C to 0.870); through the flow system and separator, tht liquid
(6) pressure gradient (O to 0.800 psi/ft); (7) in- was returned to the tank and recirculated.
clination angle ( 90 to +90); and (8) flow The liquid-metering system consisted of a Foxboro
pattern. magnetic flowmeter and two rotameters. The mag-
Fluids used were air and water. For each pipe netic flowmeter was equipped with a Dynalog re-
size, liquid and gas rates were varied so that all flow corder and had two ranges, O to 30 and O to 100
patterns were observed when the pipe was horizontal, gal/rein, For low flow rates, Brooks rotarneters with
After a particular set of flow rates was set, the angle ranges of O to 1 and O to 5 gal/rein were used. The
of tbc pipe was varied through the range of angles meters were calibrated frequently, using a calibration
so :.::: the effect of angle on holdup and pressure tank.
drop ctiuki be observed. Liquid holdup and pressure
drop were measured at angles of plus and minus W2, Gas System
1.,~50, 75>, 550, 35~, 20[, ]50, 100, and 5 from Compressed air was provided by a Clark Model HO-
horizontal, and at OO. In all, 584 two-phase flow 6-4C four-stage, reciprocating compressor equipped
tests were run. Data for all the tests are in Ref. 2. with a tank having a volutne of 20 cu ft. The output

WIRELINE
/

A 4 *
PACE PRESSLIRE
- TRANSMJCERS
4 f

cwgf$$
sTRIP CHART
RECCROERS VES WITH
AcTuAToRS

~< ~-AcRyLlc TEST PIPE

c1

i?4LL VALVE - -..,-

PNE!JMATIC ACTUATOR -w I .1 WATER STORAGE
T..,K

Fig. lSchematic diagram of test system.

 capacity of the compressor was 150 Mscf/D at 3,500 45-ft sections of acrylic pipe, through which the two-
psig. For the high-gas-rate tests, the volume tank phase mixture flowed, were mounted on this frame
was charged to a high pressure and the air was and were joined together by a section of flexible pipe
passed through a regulator to provide the necessary of the same ID. The aluminum frame was mounted
flow capacity. Thegastlow rate wascontrolled bya on a track so that when the center of the frame was
choke valve upstream of the heat exchanger. After hoisted by an electrically powered winch to change
passing through the flow system and separator, the the angle of inclination, each end of the frame was
air was vented to the atmosphere. free to slide on the track. A wire line was strung
An orifice meter and a Cameo flow computer were over a 50-ft derrick and the flow system could be
used to measure the gas rate. Orifice plates were raised to any angle between horizontal and i- 90
easily changed and the optimum size was used for or 90.
each flow rate, The meters were calibrated with a
water manometer and dead-weight tester. Testing Procedure
The testing procedure, designed to emphasize the
Liquid-Holdup Measurement effect of inclination change on holdup and pressure
Liquid holdup was measured with pneumatically drop, consisted of the following steps:
actuated, quick-closing ball valves. This same type 1. The gas rate was set with a choke valve and
of system was used by Eaton and by Singh and was monitored with the Cameo flow computer and
Griftith. I Two valves were used in the uphill section the differential pressure recorder,
and two in the downhill section. A similar valve was 2, The liquid rate was set with a choke valve
placed in the bypass line. The valves were full- and bypass valve.
opcning ball valves, with inside diameters equal to 3. After steady-state conditions were reached, the
. pressure recording system was activated and the
the inside diameter of the pipe, so that the flow
stream was not disturbed by passing through the various pressures and pressure drops were recorded.
open valves. 4, After pressures were recorded for a sufficient
The valves were opened and closed by Ramcon length of time and all flow rates and temperatures
rotary pneumatic actuators. Air pressure was ad- were recorded, the holdup valves were actuated and
mitted to the actuator chamber through a solenoid the holdup was recorded for both the uphill and
valve and the five valves could be actuated simul- downhill sections, This step was repeated at least 10
taneously by a single switch. Thus the holdup could times and the holdup readings were averaged. By
be measured without disturbing the gas and liquid using the bypass systems, the flow ra.i?settir+y were
flow rate settings, After the valves were closed, the not disturbed by this step.
amount of liquid trapped between the valves was 5. The angle of the pipe was changed and Steps
measured to determine holdup. 1 through 4 were repeated for each angle tested.
6. The gas rate was changed and Steps 2 through
Pressure-Drop Measurement 5 were repeat~d.
Four pressure taps (see Fig. 1) were located in the 7. The liquid rate was changed and Steps 1
test section, twc on the uphill side and two on the through 6 were repeated.
downhill side, Pressure was transmitted through Each test took approximately 45 minutes, depend-
plastic tubing to pressure transducers located in an ing on the time required to reach steady state. Tests
instrument house. Small separators were provided at were first conducted with l-in. -ID pipe. The pipe
each pressure tap to keep liquid from entering the was then changed to 1,5-in. pipe, and tests were re-
pressure lines. The. pressure taps were numbered P,, peated at the same mass flux rates. A total of 584
P,, P,,, and PI, Absolute pressure was measured at average liquid-holdup and pressure-drop measure-
P,, and differential pressures were measured at P,-P,, ments were obtained.
P,-P,,, P,,-P,, and P,-P,.
Pressure drop was measured with variable reluc- Development of Correlations
tance pressure transducers, The output of these trans- The development of Eq. 1 revealed that correlations
ducers was fed to demodulators for amplification and must be devised for two ~arameters in order to cal-
then to strip-chart recorders. Diaphragms in the trans- culate pressure gradients in two-phase, inclined flow.
ducers could be changed to provide a differential These parameters are liquid holdup, Hr., and two-
pressure range of O to 1 psi and O to 50 psi. The phase friction factor, ftl,.
pressure-measuring system was designed so that the
transducers could be calibrated with a manometer Liquid-Holdup Correlation
before each test. In plotting liquid holdup vs inclination angle for
Transducers rather than manometers were used to constant flow rates, it was discovered that holdup
measure pressure drops. Because of the fluctuating has a definite dependency on angle. as can be seen
nature of two-phase flow, a recording of the pressure in Fig, 2. The reversal of these curves at approxi-
drop was made so that an average value could be mately 50, plus and minus, from horizontal was
determined, unexpected. This phenomenon may be explained by
considering the effects of gravity and viscosity on the
Test Section liquid phase. As the angle of the pipe is increased
The test section frame consisted of two 45-ft sec- from horizontal, gravity forces acting on the liquid
tions of aluminum pipe hinged at the center, Two cause a decrease in liquid velocity, thus increasing

610 JOURNAL OF PETROLEUM TECHNOLOGY

slippage and liquid holdup. As the angle is increased fit was applied to each curve, using a poIynominal
further, the liquid eventually bridges the entire pipe, in the sine of a multiple of angle; that is,
reducin~ slippage between phases and thus decreas-
ing holdup. It shoold be noted that segregated flow
was not observed at any angle greater than + 30 from
horizontal, In downhill flow, in which the flow pat- It was found that # could be predicted for all flow
tekn is almost always segregated, an increase in angle conditions by an equation of the form
in t}, negative direction results in increased liquid
velocity and decreased holdup. As the angle is further + = 1 i- B, sin (1.36) + B, sin (1.80) ,
increased in the negative direction, the flow pattern or
changes to semiannular and, finally, to annular. More
of the liquid is then in contact with the pipe surface, #=1-1- C(sin@-%sin3@), . . . (5)
and viscous drag causes a decrease in Iiquld velocity where # = 1.80. For uphill flow,
and an increase in liquid holdup,
These curves revealed also that the degree of c+=(#,),ax 1)1,5, . , . , , # (6)
holdup change with angle was different for different and for downhill flow,
flow rates, Many attempts were made to find a func-
tional relationship between holdup and flow rates, c-= (1 #,,,,,,) 1.5, , . , . . . (7)
pipe size, imgle, and other variables, It was finally
where ~r,laxand #,,,i,, are the maximum and minimum
decided to normalize the holdup by dividing the
values of t) for a particular set of flow conditions.
holdup at any angle by the holdup at OO. Thus,
To develop a method for predicting values of C,
~= it was necessary to divide the tests irito three regimes
(2)
m.(o)~ on the basis of the flow pattern of the test when the
pipe was in the horizontal position. If the horizontal
where
flow pattern was mist, bubble, or froth, the change
H,,(d) = holdup at angle 0, in holdup with inclination was insignificant.
H,,(O) = horizontal holdup, The tests were divided into the three flow regimes
# = inclination correction factor. described by Dukler et al, These are segregated flow,
intermittent flow, and distributed flow (see Fig. 4).
A plot of t) vs 0 for several flow conditions is shown
Regression analysis revealed that the C values could
in Fig, 3.
be correlated with input liquid content (A), Froude
To correlate horizontal holdup, a multiple, linear,
number (N~~), and iiquid velocity number (N,,,.). Dif-
stepwise regression analysis was performed, with
holdup as the dependent variable and with the fol-
lowing independent variables: liquid velocity num-
ber, N{,.,.;gas velocity number, N.,.; diameter num- ,50
ber, Nn; Reynolds number, NR,.; Froude number, A h=264
NI.R; pressure ratio, p/pa; gas-liquid ratio, qJqL; ,40 r] + x =.156
I
and input liquid content, A. o x =.010
This analysis revealed that the most significant in-
,30
dependent variables were Froude number and input
liquid content. These same two variables were used
.20 -
by Guzhov et d. and by Nezhilskii and Khodano-
vich. Various forms of the Reynolds number, in- 0
5 .10
cluding those of Dukler et aL,3 Hagedorn and Brown,7 c?
i
I
and EIughmark, were tried, but none proved to be
.00 1
significant. This may possibly be explained by the -60 -70 -30 -30 -1001030507090
fact that both liquid and gas phases were in turbulent ANGLE C+ PIPE FROM HORIZC+JTAL
flow in all tests, Therefore viscous effects were negli- Fig. 2Liquid holdup vs angle.
gible. The use of only two fluids air and water
2.00
could also tend to mask viscous effects. An equation
A A * .264
was obtained of the form,
* + A*.156

z?.(o) = AA@ N*.Rfl, . . . . . . , (3) g L50 >

1-

which predicted the experimental data farily well ex- g

cept at low values of holdup. This equation was in- : 1.00

dependent of flow pattern, but it was obvious that it ~

would nc ~~apply to all flow conditions. ~

At this point, efforts were directed to correlating ~ .50

the inclination correction factor, #. A study of the z

g
graphs of ~ vs O revealed that the curves were all of
.00
the same general shape and reached a maximum or -90 -70 -m -30 -IO OI03050T090

minimum at approximately the same angle, plus and ANGLE OF PIPE FROM HORIZONTAL

minus 50 from horizontal. A least-squares curve Fig. 3inclination correction factor vs angle.

MAY, 1973 611

 SEGREGATED FLOW ferent equations were obtained for segregated, inter-
(
mittent, and distributed flow regimes in uphill flow;
:>.-<------ -=
.=.
- s,.--=
____= .,___
but it was found that one equation could predict C
.=-- .+<=. .Z> -_.. ._ _ _

STRATIFIED
for all downhill flow, This may be explained by the
fact that the flow regime changes from intermittent
to segregated as the flow stream breaks over the hill,
- The equations obtained were of the form

m
WAVY

C=(I A) ln(D X6NFRNLv~),

where D, 8, e, and C are dif%rent for the various flow
ANNUAR conditions, As was stat% earlier, for uphill distrib-
INTERMITTENT FLOW
uted flow, holdup was not a function of angle, and
therefore C is zero i:j~ this condition.
Y/hen it was found necessary to resort to flow
~
PLUG regimes to correlate the j.nclination correction fac-
tor, the correlation for horizontal holdup was re-
examined, A different equation of the form of Eq.
=
9_,JG
3 was obtained for each of the three flow regimes,
DISTRIBUTED FLOW This improved considerably the accuracy of the pre-
diction for low holdup values. A plot of horizontal
liquid holdup vs input liquid content with Froude
~
SUBEi.E
number as a parameter is shown in Fig, 5, The sharp
breaks in the curves for N~R = 0,1 and N~.1,= 0.5
indicate a change in flow pattern from segregated to
intermittent. Onc of the problems caused by using
different equations for different flow regimes is that
Fig. 4-Horizontal flow patterns, a discontinuity appears in a curve of holdup vs input
liquid content at low values of Froude number. Al-
though the holdup does decrease greatly as the flow
changes from segregated to intermittent, the change
is not so rapid as is indicated by the equations, This
situation could possibly be eliminated by defining a
transition zone between flow regimes.
Since different holdup correlations are used for
different horizontal flow regimes, it was necessary to
develop a method to predict flow regime in horizontal
two-phase flow. When the Froude numbers were
plotted vs input liquid content on log-log paper the
ti.s:s fell into groups in different areas, These areas
are shown on the flow pattern map (Fig, 6). Because
of its simplicity, this map may or may not strictly
define the flow pattern into which a given flow condi-
tion will fall, but it does determine the equations to be
used in predicting liquid holdup using the present
correlation. Equations were fitted to the lines of Fig.
INPUT LIQUID CONTENT - ~ 6 so that the flow pattern could be determined with-
Fig. 5Florizontal liquid holdup vs input liquid content. out referring to the graph. The flow pattern may be
determined w follows.
loco. 1, If N~tt < L,, the flow pattern is segregated.
I

+
++*
i+ 1 i+ I

1 2. If NE~ > L, and > L,, the flow pattern is dis-

ZE DISTRISLITEO
tributed.
I 100. -
0 3. If L, < ~FR < L,, the flow pa:tern is inter-
6 00 mittent,
g
z
4. L, and L, are given by
10.
w 00 L, = exp(- 4.62 3,757X 0,481Xe
n
3
0 0
K SEGREGATE o 0.0207X3) , , . . . . . (9)
L 1. 0 SEGIifEGATEOFLW
x INTERMITTENT FLW and
+ OISTRWTED FLW
L, = exp(l.061 4.602X 1.609X2
1 ~Ll
.Cool .001 .01
INPUT LIQUIO CONTENT- A
.1 1, 0,179X3 + 0.635 X 1O-Y) ,
Fig. &,Flow-pattern map horizontal flow. . . . . . . . . . . . (10)

612 JOURNAL OF PETROLEUM TECHNOLOGY

where H,.(d), It is possible that the dependence on holdup
could have been eliminated by including holdup in
X = ln(~) . the Reynolds number, but after several fofis of
Equations have now been developed to predict Reynolds number were tried with little success, this
liquid +oidup in two-phase flow for all conditions. approach was abandoned. Regression analysis with
These equations are given in Table 1. After the values normalized friction factor as the dependent variable
for H,,(O) and C are determined, holdup at any angle and 1 and Hr,(6) as independent variables indicated
is calculated from a relationship of the type

H,,($) ==H,,(O) {1 + C[sin(l.80)

% sin3(l.8t9)]}, , . , . . (11)
k= f{[H,il)]2 1 + . . . .
(15)

The equation for friction factor is

with the restrictions that HI,(O) > A and O s ~L(@
<1. f+=es?,..,,, . . . . (16)
1,18
Friction-Factor Correlation
where
Values for the two-phase friction factor, jf,,, were
obtained by solving the pressure gradient equation, S = [in(y)]/{ 0.0523 + 3.182 in(y)
Eq. 1, for f~P:
0.8:25 [ln(y)]2 + 0,01853 [in(y)]
ft, =
[(
~
dp
) ~ _ Ptpvn,vsg
L?(,P
and

F
t? 1 sin6p[P ~
2gCd
mm
, . . (12) Y =
A
[~L(~)]2

The two-phase friction factcr was normalized by Eq. 16 becomes unbounded at a point in the interval
dividing it by a no-slip friction factor, ft,,, which 1 < y < 1.2; and for y in this interval, the function
would apply if the fluids were flowing at the same S is calculated from
velocity. The no-slip friction factor is obtained from S = 1n(2.2 y 1,2).
a Moody diagram or, for smooth pipe, from
As the flow approaches all gas,

A+o, S+o, and ff, + f.. + ~w.

. . . . . . . . . . . (13) As the flow approaches ail liquid,
where A
+ 1, S+o, and /0+ f.. + jsp.

N[u.,/, =
ijd + P,,(1 ~)]Vmd H,,
/JJ.~ + PV (1 ~, A plot of the normalized friction factor vs input
or liquid content, with holdup as a parameter, is shown
in Fig. 7.
G,),d
Nit,.,,. (14) The correlations for liquid holdup and friction
FA+/.L,, (l A) - factor were developed using dimensionless variables.
This Reynolds number approaches the Reynolds Both the holdup and friction factor correlations de-
number for liquid or gas as h approaches 1 or zrro, generate to single-phase conditions as the flow ap-
respectively. proaches all liquid or all gas, As the correlations
The normalized friction factor was found to be a were developed from data obtained using only two
functian of input liquid content, A, and liquid holdup, fluids -+water and air and two pipe sizes 1

TABLE 1EQUATIONS FOR PREDICTING LIQUID HOLDUP

Horizontal Horizontal
Flow Pattern Holdup ci- c

(Eq, 19) (Eq. 22) (Eq. 25)

0.8451MS1 2.95k0= Nm007s

Intermittent H,. (0) = ~WO.O,,, C+ = (1A) In Same as Segregated
NLVO.,*?J
[ 1
(Eq. 25)
(Eq. 20) (Eq. 23)

1.065A~
Distributed HL (0) = N,BO@~ C+=o Same as Segregated
(Eq 25)
(Eq. 21) (Eq. 24)

MAY, 1973 613

and 1.5 in, extrapolation to Iargex pipes and dif-
ferent fluids will have to be tested; but it has definitely
been established that pipe inclination has aconsid-
erable effect on liquid holdup in two-phase flow.
The procedure to calculate the pressure gradient at a
Discussion and Application of Results given pressure and temperature is outlined as follows,
1. Calculate PI,, pfl, v* I,, V,O, Vm, G~, A, NFR,
The foregoing correlations for calculating liquid
9 and NLVat p and T,
N R<.l!l!
holdup and pressure gradients occurring in two-phase
2, Calculate L, and L, using Eqs. 9 and 10.
flow are applicable to many situations in the petro-
3, Determine the flow pattern, using the following
leum and chemical industries. Designing well com-
criteria: (a) if N~R < L,, segregated; (b) if NF~ > L1
pletions and gas-lift systems in directionally drilled
and > L~, distributed; and (c) if Ll < ~FR < L~,
wells requires a knowledge of the effe?t of inclination
intermittent,
on two-phase flow. For example, the present work
4, Calculate H,,(O) using Eqs. 19 and 20 or Eq.
indicates that the pressure gradient occurring in a
21, Table 1,
well inclined at 150 to 20 from vertical can be
5, Calculate C using Eqs, 22, 23, 24, or 25,
greater than the pressure gradient in a vertical well.
Table 1.
One of the principal reasons for undertaking this
6. Calculate # =l,80and y=l+C(sin @%
study was to develop a method for predicting the
sins ~).
pressura drop occurring in two-phase pipelines con-
structed in hilly terrain. The degree of pressure re- 7. Calculate H,,(O) ==H,,(O) y, and p,n.
8, Calculate ~,,,/f,,. using Eq. 16.
covery in downhill two-phase flow has been a subject
9. Calculate j,,, using Eq. 13, or obtain it from a
of conjecture for many years, Most previous methods,
Moody diagram.
such as that of Flanigan,fi have assumed that there is
10, Calculate j(l) = (jti,/fn.J f,,,.
no pressure recovery in downhill two-phase flow. The
study dealt with here indicates that this is true in some 11. Calculate dp/dZ using Eq. 1.
This procedure is repeated with small pressure
cases, but that under many flow conditions the pres-
changes, and the pressure traverse is calculated using
sure recovery is considerable and must be taken into
account in designing pipelines. one of the methods described above.

Calculation Procedure Statistical Analysis of Results

The statistical analysis consisted of calculating the
The pressure gradient equation presented here is rela-
tively simple to apply and involves no trial-and-error errors between observed and predicted values of
solutions for isothermal flow. If temperature, and liquid holdup and pressure gradient and the standard
therefore z-factor, is a function of distance, a tnal- deviation of these errors. Errors were calculated on
and-error solution is necessary. Equations were de- a percentage basis as
veloped for all unknowns so that referring to graphs Percent error =
is unnecessary, making the procedure easily program-
mable for the computer. calculated value observed value
x 100,
The pressure gradient equation may be used to observed vaiue
calculate a pressure traverse in two ways. In the first so that a positive error indicates that the predl steal
method, a pressure decrement is chosen; and if tem- value is too high. The average error was calculated by
perature is a function of distance, a distance incre-
N
ment is estimated. The equation is salved for the ~ percent error
distance incremel:t corresponding to this pressure Average percent error = 0 ~
decrement.
The standard deviation of the percentage errors

When this method is used, all the barred variables

2.s -
in Eq. 17 must be evaluated at the average pressure
and temperature. If the calculated distance increment
2.0 -
is not close enough to the estimated value, a new &
vah~e is assumed and the calcul.~tion is repe?,ted until f n*
1!3
the two values agree. The accuracy of [hG calculation
depends on the size of the pressure decrement, 1.0
The second method, which would be applicable
for isothermal flow at a constant inclination angle, 0.5 -
consists of solving the equation for pressure gradient
at several pressure values, plotting the reciprocal of , + I 1

.01 .1
pressure gractient vs pressure, and integrating to INPUT LIJ.SD CDNTENT- A
obtain AZ; that is, Fig. 7Two.phase friction factor.

614 JOURNAL OF PETROLEUM TECHNOLOGY

 was calculated by in two-phase, air-water flow in 1- and 1,5-in, smooth,
circular pipes at any angle of inclination.
Standard deviation =
2 0.5 Nomenclature
N; (70 error) 1 $%0 error An = pipe area
0 ( 0 )] .
N C = coefficient
[ d = pipe diameter
Table 2 lists the average percentage errors and f = friction factor
standard deviations for each inclination angle and g = acceleration due to gravity
for all testa. g. = gravitational constant
G = mass flux rate = w/AP
Conclusions G. = mixture mass flux rate = G~ + G~
. The following conclusions are drawn from this study. h = vertical elevation
1. The inclination angle of a pipe in which two- 11~ = liquid hcldup fraction
phase flow is occurring definitely affects liquid hold- L, = correlation boundary, Eq. 9
up and pressure drop for most flow conditions. L, = correlation boundary, Eq, 10
2, In inclined two-phase flow, the liquid holdup M = molecular weight
reaches a maximum at an angle of approxitnately 0.5
-+ 50 and a minimum at approximately 50= horn
horizonta~. The fact that holdup is approximately
ND = diameter number = d ~
u () 0.23
equal at angles of + 90 and + 20 explains why
vertical holdup correlations can be used with some ()
gas velocity number = V80 @&
0.25
degret of success for horizontal flow.
3, Pressure recovery in the downhill section of a
twc-phase pipeline in hilly terrain can definitely exist Froude number = v~2/gd
()
liquid velocity number = v.I. ~

and should be considered in pipeline design, Reynolds number = pvd/p

4. The accuracy of a liquid holdup correlation for pressure
horizontal two-phase flow can be improved by con- in-situ volumetric flow rate
sidering flow pattern. gas constant
5. Input liquid content, A, and Froude number, correlating factor, Eq. 16
N~R, are very important correlating parameters in temperature
two-phase flow. velocity
6. Friction loss in two-phase flow is greatly af- mixture vebity = (qL + q$)/A,,
fected by liquid holdup. superficial gas vebcit y -- qq/AP
7. A correlation has been developed that can pre- superficial liquid velocity = q~/A j,
dict liquid holdup and pressure gradients or - wring mass flow rate

TABLE 2SUMMARY OF STATISTICAL RESULTS

Liquid Holdup Pressure Gradient
Inclination
Angle Nu~em& of Averake Standard Average Standard
(degrees) Percent Error Deviation Percent Error Deviation

0 58 0.87 5.51 +2.57 6.03
+5 33 1.46 7.07 +1.57 6.52
+10 32 0.34 5.70 +1 .43 5.84
+15 32 0.14 5.44 +1.30 15.27
+20 32 0.95 5.72 +2.03 6.64
+35 30 +0.63 6.83 + 0.97 6,92
+55 30 + 0.78 5,63 +1.22 5.50
+75 31 + 0.83 5.75 +1.77 7.59
+85 16 0.12 5.11 +0.24 6.32
+90 27 +0.18 7.27 -I-1.79 5.96

All uphill tests 263 0.09 6,19 +1.43 6.4!5

5 33 +0.13 7.45 + 1.% 9.32

10 32 0+84 8.90 1,51 17,73
15 32 +1.16 8.09 +0,57 15,94
-20 32 +0.29 7.17 + 1.47 10.30
35 30 4.17 10.59 + 2.95 11.24
55 30 3,13 13.57 1.83 11,85
75 31 0.75 9.13 +0.05 9,61
85 16 1.20 6.68 0,39 3,11
+ 5.47 11.23 +0.44 6.47
90 27 .
All downk.::1tests 263 0.34 9.82 + 0.47 11.89

All tests 584 0,28 7,98 +1.11 9.30

IWAY, 1973 615

 Wf = irreducible friction losses namics of Joint Transport of Gas arvf Liquid, Int.
Chem, En~. (April, 1965) 5, 318-322.
G = gas compressibility factor 10. Nezhilskii, B. M. and Khodanovich, I. E.: Experimen-
z. distance of axial flow tal Studies of True Gas Content in Descending and
~= specific gravity Ascending Sections of a Nonhorizontal Pipeline; Nauch-
~= difference Tekh. Sb. Ser. Ga.zovoe Delo (1970) 11.21-25.
g= angle from horizontal 11. Odishariya, G. E.: On Certa;n Problems in Gas Liquid
~$~9 Through Pipelines, Nefr Khoz (Sept., 1966) 9,
k= input liquid content = qL/(qL + qg) . .
= viscosity 12. Orkiszewski, J.: Predicting Two-Phase Pressure Drops
P
l-h = mixture viscosity = id + pa (1 ~, in Vertical Pipes; J, Pet. Tech. (June, 1967) 829-838.
= density 13. Poettmann, F. H. and Carpenter, P. G.: The Multi-
P phase F1OWof Gas, Oil, and Water Through Vertical
pm = no-slip density = pLx + PO(l A) Flow Strings with Application to the Design of Gas-Lift
ptp = two-phase mixture density = pLHL Installations, Drill. and Prod, Prac., APi ( 1952) 257.
+ PO(I H.J 14. %vigny, Rene: An Investigation of isothermal, Concu-
rrent, Two-Fluid Two-Phase F1OWin an Inclined Tribe;
liquid surface tension PhD dissertation, U. of Rochester, Rochester, N, Y.
;:
angle (1962).
+= inclination correction factor, Eq. 5 15, Singh, G. and Griffith, P.: Determination of the Pres-
sure Drop Optimum Pipe Size for a Two-P!lase Slug
Subscripts Flow in an Inclined Pipe; J. Eng, I//d. Trans., ASME
(Nov., 1970) 92, 717-726.
a = atmospheric
APPENDIX
acc = acceleration (kinetic energy)
el = elevation (gravity, potential energy) Development of Pressure-Gradient Equation
f= friction The basis for any fluid-flow calculation is an energy

gas balance for the flowing fluid between two points.
is liquid Assuming no external work is done on or by the
m. mixture fluid, a steady-state mechanical energy balance equa-
ns = no-slip (two-phase)
tion in differential form may be written for 1 lb,,,
sp = single phase of fluid as
tp = two phase
dp +:dh+ v,,,dv,,,
g.
+d(wf) = O, (A-1)
Arbitrary Constants pfp

A, B, D,cr, /3, y,tI, e,C where d(wf) represents the irreversible friction losses,
For flow in a pipe up or down an incline,
Acknowledgments
We wish to acknowledge the financial support for dh=sin OdZ, . . . . . . . (A-2)
this project given by Mobil Oil Corp., Field Research
Laboratory; Chevron Oil Co., California Co. Div.; where dh is the vertical distance moved, d is the angle
of the pipe from horizontal, and dZ is the axial dis-
and the U. of Tulsa Faculty Research Program.
Equipment was donated by Cameo, Inc.; Atlantic tance moved, Substitution of Eq, A-2 into Eq, A-1
Richfield Co.; C-E Natco; and Vinson Supply Co. gives
dv,,, d(w,)
1
@=_ V,,,
References ~ PIJJ sin 0 + Ptl, ~ ~ + Ptp ~
dZ [ .
1. Baker, Ovid: Discussion on How Uphill and Downhill
FIOW Affect Pressure Drop in Two-Phase Pipelines ,,, . . . . . . . . (A-3)
;ljr;-u;~ Hilly Country, Oii wd Gas J, (Nov, 11, 1957)
This may be written as
2. Beggs. H. D.: An Experimt ,:tal Study of Two-Phase
Flow in Inclined Pipes, PhD dissertation, U, of Tulsa
(1972).
3. Dukler, A. E., Baker, O., Cleveland, R, L., Hubtmrd,
M. G. and Wicks, M., III: Gas-Liquid Flow in Pipe-
-+= w+(:O.((
+(a ~A-)
that is, the total pressure drop is the sum of the pres-
lines, Part 1, Research Results, Monograph NX-28, U.
of Houston (May, 1969). sure drops due to pcential energy change, kinetic
4. Eaton, B, A.: The Prediction of Flow Patterns, Liquid energy change, and friction loss.
Holdup and Pressure Losses Occurring During Continu-
ous Two-Phase Flow in Horizontal Pipelines, PhD dis- Frictional Pressure Gradient
s ~tation, The U, of Texas at Austin ( 1966).
5.1 .migan, 0,: Effect of Uphill Flow on Pressure Drop By definition, the frictional pressure gradient is
in Design of Two-Phase Gathering Systems, Oil and ,given by
Gus J. (March 10, 1958) 56, 132.
ap = f(,)p),. v)),
()
6. Guzhov; A. I., Mamayev, V. A, and Odishariya, G. E.: = f,,) G,,,v~,
A Study of Transportation in Gas L~quid Systems, (A-5)
paper IGU/Cl 9-67 presented at 10th Intl. Gas Confer- mj 2 g,d 2g.d
ence, Hamburg, Germany, June 6-10, 1967.
7. Hagedorn, A. R. and Brown, K, E.: !Experimental Study
In two-phase flow, the no-slip density may be dif-
of Pressure Gradients Occurring During Continuous ferent from the in-situ density because of slippage
Two-Phase F1OV in Small-Diameter Vertical Conduits, between the phases. The no-slip density, p.,, is used
J. Pet. Tech. ( roil. 1965) 475-484.
in the friction-loss equation because the energy enter-
8. H~ghmark, C ~.: ;Holdup in Gas-Liquid Flows, Cltem.
Eng. Prog. ( .+pril, 1962 ) S8, 62, ing and leaving a differential element of the pipe by
9. Mamayev, V. A.: Some Problems in the Hydrody- way of the flowing fluid is a function of the proper-

616 JOURNAL OF PETROLEUM TECHNOLOGY

 1 dp 1 dM
ties of the fluid entering and leaving the differential
element, and not of the fluid in place.
$(PJ
Pg (
P

dZ
MdZ

Acceleration Pressure Gradient . 1 dz, dT 1

(A-9)
2,= dZ ) T
To analyze the kinetic energy or acceleration term,
which is negligible for most practical cases, several In analyzing the relative magnitudes of the remain-
simplifying assumptions are made. ing terms in Eq. A-9, it may be assumed that
1 dM
1 dz, 1 dT 1 dp
.
M dZ~~~~<< pdZ
therefore, Therefore,
dp
+(P9) =:77 . . . . . . . (A-1O)

Substitution of Eq, A-10 into Eq. A-7 gives

~ PfPvm
[+ (%)+* WI~ ap
() dp
f?. G, pg
Ptpvm
d G,, ~acc= gc ~, p dZ
While is not zero, it may be assumed to be
dZ @ or
d Go i3p
small compared with
dZ
ference in the compressibilities
because
p~
of the dif-

of the liquid and the

() E ,,. =
_ Pt,v.,v,g
gcP
dp
_=
. (A-n)

Gravity Pressure Gradient

gas. Therefore,
Calculation of the pressure gradient caused by eleva-
tion change,
ap
() me,
= ~
!
Ptp sin 6, . . , . . (A-12)

requires a procedure to determine the in-situ density

of the gas-liquid mixture, p~P. For this purpose, a
liquid holdup factor is defined as
~~ = Volume of liquid in an element
Volume of the element
It may be assumed also that the change in gas mass
flux is much smaller than the change in gas density, or The in-situ density of the fluid mixture may now be
represented by
ptp = @~ + Pg (1 HL) . . . (A-13)
Therefore, Eq. A-12 becomes
Incorporating this assumption into Eq. A-6 gives

()@ =_
z ac,
P,,vm Gg

4?C P!72
~
d
(PtJ . . . (A-7) By substitution of Eqs. A-5, A-11, and A-14 into
Eq, A-4, the total pressure gradient is
From the engineering gas law,
_ pM p g,g [f@L+fJo (1 ~.)]
dZ
sin~ +
ftpGwwm
2gcd
g ZORT
_ [p~L+ po (1 ~L)]%n V,g dp
and ~,
M

pM
$(Pg) =&j-
()

@lT
or

E sine [Pfifi+P~(l-HL)] +
ftpGmvm

= M dp
+
*%(M) dp _ 8.

2gcd
zgRT dZ dZ 1 {[PIflL +pg (1 ~L)l vmvstr}/gcP

pfbf . d (z,) * ~ . (A-8) . . . . . . . . . . (A-15)

Z92RT dZ
JPT
Dkiding Eq. A-8 by paper (sfJE waa presented at SPE.AI ME 47th Annual Fall
4007)
Moating, held In San Antonio, Tex., Ott, 8.11, 1972. @ Copyright
_ pM 1973 American Inatltuta of Mlnlng, Metallurgical, and Petroleum
Englneera, Inc.
o zJ?T This DaPer wIII be printed In Transact/o. I volume 255, which
gives
MAY, 1973