Liquid Holdup Correlations for

Inclined Two-Phase Flow

Hemanta Mukherjee, sPE, Johnstm-Maccco Schl.tnbwgp

James P. Brill, SPE, u. of Tulsa

summary

Liquid holdup behavior in two-phase inclined flow was The Beggs and Brill correlation was developed from
studied in an inclined pipe-flow simulator. Two sets of data obtained in an air/water flow system with 1- and
empirical equations, one each for uphill and downhill 1,~.in. (2.5- and 3.8-cm) diameter pipes. They con-
flow, are presented. For downhill stratified flow, a third sidered a rsnge of inclination angles from O to & 90.
equation is presented. The liquid holdup equations arc Use of the correla~on requires first detemnining the
functions of dimensionless liquid and gas velocity holduv for horizontal flow according to cmdicted
numbers in addition to liquid viscosity number and angle horiz&tal flow patterns. The horizontarholdu~ is then
of inclination. These four parameters uniquely define the corrected for angle of inclination. Palmer6 found that the
flow-pattern transitions in inclined two-phase flow. Con- Beggs snd Brifl liquid holdup was overpredicted for both
sequently, the hoIdup equations are also implicitly flow- uphill and dowrdill flow and suggested proper correction
--
pattern dependent. factors.

For uphill flow from O to 9, Guzhov et al. 7 pmposed

Introduction a holdup correlation that is independent of inclination
An accurate prediction of liquid holdup is rcquied to angle. Hughmark and Pressburg g developed a general
compute the hydrostatic head loss in two-phase inclined holdup correlation for gaslliquid flow covering a wide
flow. In this case, hydrostatic head maybe the most im- range of physical properties and diarnetem. This correla-
portant of the pressure gradient components. There are tion is based on data taken in 1-in. (2,54-cm) diameter
many liquid holdup correlations in the Iitcramre, but pipe for vertical uphill flow of air, water, oils of different
nfmost d are for horizontnf or vertical uphilI flow. viscosities, and carcfolly selected data of other
Eaton et al. 1 proposed a holdup correlation based on investigators.
dnta for natnral gas, water, crude oil, and distillate oil In addition to these empirical liquid holdup correla-
mixtures in 2- and 4-in. (5- and 10-cm) diameter tions, at least two analytical holdup correlations deserve
horizontal pipes. This correlation is based on tive dimen- mention. Bonnccaze et al. 9 developed a slug flow model
sionless groups reflecting various physicaf properties, for inclined pipe based on a mass and force balance
flow rates, system pressures, and pipe diameters. A around a simplified slug unit. The pressure drop con-
study by Vohra et al. 2 showed that MIS correlation per- tributions caused by the Iiqnid film and the gas bubble
formed best on a collection of horizontal data taken .by were negfected, Using this holdup correlation, they cor-
Eaton et al. 1 and Beggs and Brill. 3 Cunliffe4 found that related pressure dmp data obtained in 1 x-in. (3.8-cm)
using the Eaton et al. correlation to predkt the total liq- diameter pipe inclined at various angles around + 10 to
uid volume in a wet gas pipeline was quite successful for find an expression for friction factor. The holdup equa-
determining the incremental volume of liquid removed tion and friction factor correlation were compared with
from rak increases. field dsta taken in a 6-in. (15.24-cm) dinrneter, 10,000-ft
Using dyn~ic similarity annlysis, Dukler et at. 5 (3C48-m) long pipe with a maximum deviation of 5 %.
developed a boldup corm?lation for horizontal two-phase With a very similar mechanistic approach, Singh and
flow. This holdup correlation is implicit in Iiquid Griffith 10 proposed a simple model for two-phase sIug
holdup, ~quiring an iterative calculation. Experience flow in inclined pipes. Most of their mndel pammetera
has shown that most wet gas-transmission applications were experimentally determined using five different
will result in a nwslip liquid holdup calculation when the diameters of copper pipe at 5, 10, and 15 inclinations
Dukfer et at. correlation is used. with an air/water system. For stratified tlow, the authors
developed a holdup model based on Chezys open-

0149-25 36/S3{0041 .0923500 ,25

channel flow equation. This equation shows that the liq-
CoPYrlglIt ,983 socie>y of Petrde.m E@nws of AJ.4E
uid holdup is independent of gas flow, which is contrnry

MAY 1983 1003

.
 . . . .. . k,.,. 4

q-:?----- . ...- TEST S3CT10N ,

v TO TRANSDUCER

. .
+

DETAIL A [>.:4
SOLENOID

DRAIN VALVE

-X- GATE VALVE

- CHECK VALVE

+ ACTUATED BALL VALVE

+ MOTOR VALVE

~ ROTAMETER

@ OFJF,C,METER

- OIL F[LTER

Ml, TANK

a.
Fig. lSchematic of inched flow simulator.

to accepted phenomena. They also suggested a. method The two-phase mixmm flowed through the test sections
of calculating liquid holdup for annular flow that is and into a horizontal separator. The gas (air) was vented
iterative in nature. Their liquid holdup models repro- to the atmosphere and the liquid passed through a filter
duced their data with rcasoaable accuracy. and into a storage tank.
Kerosene and lube oil were used as the liquid phas~s.
Experimental Program The surface tension, density, and viscosity of the

An experimental facility was designed and constructed to kerosene at 60F (15 .56C) were 26 dyne{cm (26

obtain the desired test data. Fig. 1 is a schematic of the mN/m), 51 lbm/cu ft (816.9 kg/m3) and 2 cp (0.002

test facility. The test sections consisted of an inverted U- Pa. s), respectively. Corresponding values for the lube

shapc 1.5-in. (3.8-cm) ID nominal steel pipe. The closed oil were 35 dyne/cm (35 mN/m), 53. lbmlcu fr (849

end of the U-shape test sections could be raised or kg/m3 ) and 29 cp (O .029 Pa.s). Temperatures between

lowered to any angle from O to +90 from horizontal. 18 and 132F (7.8 and 55.56C) were encountered

Each leg of the U was 56 ft (17 m) long with 22 ft (6.7 during the tests.

m) entrance lengths followed by 32- ft (9.8-m) long test

sections on both uphill and downhill sides. Each test sec- Phase Slippage and Liquid Holdup
tion could be isolated from the rest of the piping by Ia inclined two-phaae pipe flow, a substantial part of the
pneumatically actuated ball valves tint could be opened totrd pressure losses may be contributed by, the
or closed simultmeously when calibrating holdup sen- hydrostatic pressure difference. The relative contribu-
sors. Pressure taps 30.5 ft (9.3 m) apmt were located in tions of friction gradient and hydrostatic gradient maybe
each test section to permit measuring absolute and dlf- dictated by the prevailing flow patterns, angle of inclina-
fercntiaJ pressures using Vnlidyne transducers. A 7-ft tion, and dwction of flow. Many of the current design
(2-m) long transpammt Lexin pipe section was located in pmcedurcs used for two-phase pipelines fiil to account
each test section to permit flow pattern observations and for these effects with any rigor. Part of the problem. ii
mounting of capacitance-type holdup sensors shown in most design procedrm?a is the assumption that the void
F@ 2. The outputs from tiese two sensors were record- fraction is a unique function of quality and physical
ed on an oscillograph as a time-varying trace. For ob- properties of the fluids. This is prnbably tme where
taining am integnited value of holdup over a particular homogeneous flow can be assumed or during bubble
period of time, a digital multimeter was used to note the flow at very low gas flow rates. Similar situations may
gain in voltage output over that particular time. This also arise where the phase velocity ii very Klgh, so that
voltage gain can easily be converted to a liquid holdup friction pressurs drop governs the total pressure loss. But
fraction using a linear interpolation over the calibrated in the remaining cases erirm may arise from. neglecting
values of voltage gain for O and 100% oil in the pipe. the slip velocity between phases. Thk concept of slip
The oil and gas phases were carefully metered before velocity comes from the physical phenomenon called
mixing, turbbre meters, orifice meters, or rotametera Slippage.
wece used. depending on the phase and the flow rates. The term slippage is used to describe a natural

1004 JOLRNAL OF PETROLEUM TECHNOLOGY

 Fig. 2Capacitance sensor test cell and local electmics board.

phenomenon of one phase slipping past the other in two- presence of slippage between phases in two-phase pipe
phase pipe flow. There are several causes for slippage flow is unavoidable at any angle of inclination. In both
between phases, Frictional resistances to flow or irrever- uphill and downhill bubbie or slug flow, when the liquid
sible energy losses in the direction of flow arc much less phase is continuous and is capable of being suppmted by
in the gas phase than in the liquid phase. This makes the itself, buoyant forces generate bubble rise velocity caus-
gas more transmissible than liquid in two-phase flow, ing slippage between phases. Near the slug and annula-
even in the absence of strong buoyancy effects such as in mist flow transition or when the slug length becomes
horizontal flow. This effect can be ve~ prenounccd in long [> 1.5 to 3 ft (0.5 to 1 m)], the phases become
ariy segregated flow regime such as stratified flow. The discontinuous. During this type of flow, broken liquid
large difference ~ compressibilities between gas snd Iiq- slugs or rippIes incapable of bridging the pipe are seen to
uid causes the expanding gas to travel at a higher veloci- fall back against the direction of uphill flow. Vecy
ty and to slip past the liquid when pressure decreases in similar flow phenomena occur in dowabill stratified flow
the direction of flow. Slippage between phases is also when the liquid falls back and accelerates until the liquid
premoted by the difference in buoyant forces acting on kinetic energy is balanced by the shear energy around the
the phases. In a static liquid medium, less-dense gas liquid layer. In stratified flow, large ir-situ velocities at-
tends to rise with a veIocity proportional to the density tained by the liquid as a result of acceleration by grsvity
difference. Zukoski 11 studied the effect of pipe inclina- normally causes a very smaU liquid holdup. This
tion angle on bubble rise velocity in a stagnant liquid. He phenomenon is shown in Fig. 3, where, in stratified flow
concluded that, depending on the pipe diameter, surface at O. 363-ft/sec (O. 11-m/s) liquid superficial velocity,
tension and viscosity of fluids may appreciably affect the void fmction rises rnpidly to nearly 97% when very little
bubble rise velnci~. His findings also showed that for air flows simultaneously. At bigher liquid velocities in
some conditions an inclimtion angle as small as 1 from bubble or slug flow, the void fraction builds up mop
the horizontal can cause the bubble rise velocity to be sIowly. An important deduction at this point is that in
mo= than 1.5 times the value obtained for horizontal uphill flow, slippage causes liquid velocity to SIOW
pipes. This establishes a strong dependence.between in- down, resulting in a net accuqndation of liquid in the
clination angle and phase slippage. In the absence of any flow channel or pipe and increasing the in-situ liquid
annlyticzd formulation, the phenomenon of slippage fraction. The in-situ liquid fraction is commonly called
caused by bubble rise velocity is studlcd empirically. liquid holdup. In downMll flow, slippage causes the
Cheater gravitational forces on the more-dense liquid in-situ liquid velocity to increase, resulting in a decrease
phase promotes fallback of liquid when shear forces and in liquid holdup. AU these causes of phase slippage and
buoyant forces fail to support the liquid in upward flow. the resulting flow patterns will occur as soon as one end
For downward flow it causes the liquid to travel faster of the pipe is raised about one pipe diameter from the
than the gas. Thus, while buoyancy aIways causes the other end, regardless of the angle of inclination. Thus,
gas phase to rise dative to the liquid phase, gravity depending on the Iengtl of the pipe snd direction of
always tends to cause the liquid to faU faster than the flow, characteristic flow patterns or liquid hoIdup for in-
gas. cliied flow should be observed even at extremely low
A few impommt conclusions can be made from the angles. For exsmple, at my low uphiU angle, the
preceding discussion. Except for homogeneous flow, the s~titied flow pattern should never be observed.

MAY 1983 11305

.
 Development of Liquid HoIdup Correlation

ArmIytical expressions for liquid holdup have been at-

tempted for uphill two-phase slug flow in vertical pipes
and for downhill flow at low angles of inclination in the
range of O to 15. Considering the complex slippage
mechanism, a global liquid holdup model for any pipe
inclination has not been attempted before.
More than 1,500 liquid holdup measurements at uphiIl
and downhill inclination angles from O to +90 from
horizontal were obtained in this stndy. Attempts to cor-
relate these data into a global empirical liquid holdup
correlation are presented in the following. At each uphill
and downhii angle, void fraction was plotted as a func-
,.o~
,, ,0 ,0 m ,00 ,,0 tion of superficial gas velocity for fixed supeflcial liquid

S ,EX,,C,AL 0.4, VS,OCITY (FT/$EC1

velocity. Each of these pIots was continuous within the
error tolerance of the holdup measurements. Example
Fig. 3Void fraction vs. V%g at different values of V,L fOr plots are shown in Figs. 4 through 6. At very high gas
- 30~ angle. rates, the curves almost become asymptotic with the
100% void fraction (O% liquid holdup). For downhiil
stratified flow at very low gas rates, the void fraction
rises rapidly and then almost linearly increases with in-
creased gas rates. However, the void fraction plot for
horizontal ffow is sirniIar to the uphiIl plot, even in the
stratified flow regime. The geneml shapes of these plots
prompted sekction of a nonlinear regression equation of
the focm

HL =exp (cl +c2sin@+cssin26 +c4NZ )

[

. NC. 5
,,,
,,,
.,,
,,,
,,. . . . . .
NL, 6 1 ...(1)

Subsequently, three liquid holdup correlations were at-

tempted, one for uphiII and horizontal flow and the other
.. two for downhill stratified flow and the other downhill
0 20 40 ,0 m IOD 120
flow patterns. The regression coefficients are given in
SUPERFICIAL0,$ vEmcnY (fTmc)
Table 1. The coefficients were obtained by using the
nonlinear BMDP regrcsaion programs. 12 In each of the
Fig. 4Void fraction vs. Vw at different values of V,L for
regression analyses, the outliers in the residual plot w, w
horizontal flow.
deleted from the data set and the anaIysis was repeate~.
The seIection of phase velocity numbers as the in-
dependent variables instead of the phase supefllcial
velocities as shown in Figs. 4 through 6 was done to
! .0
make the variables dimensionless. These numbers were
also suggested as correlating parameters by Duns and
Ros, 13 Hagedorn and Brown, 14 and Eaton et al. 1 The
velocity numbers, together with the inclination angle,
also formed the independent variables defining the flow
patterns. Hence, inclusion of all these variables implicitl-
: y makes the holdup conflation flow-regime-dependent.
2 ,,4 The use of dimensionless numbem should not affect the
:. shapes of curves shown in Figs. 4 through 6 since, for a
, s, = p..:,M,;sEc

. s!. : ,.,6, FV5,c

fixed oil, convecting superficial velocities to dimen-
0,1 ~ ,,L, f.;.,+!,:;,,=
sionless form reqnires multiplication by a nearly constant
(mm ./s1
quantity of appmxima.tely 2.5 for this study.

,o~o Effects of In&nation Angle and Viscosity

,PERFIc,.+L . AS ,LOC,T, (FvwC)
The second-degree polynomial function of the form
c1 +c2sinO+c3sin2@ was selected by plotting liquid
Fig. 5Void fraction vs. Vw at different values of V,L for
+90 angle. holdups for differsnt angles of inclination at fixed liquid
and gas velocity numbers. This relation was also con-
firmed by comparing results of other equation forms in
trisl runs of the regmsion analysis. The best error as in-

1006 JOURNAL OF PETROLEUM TEcHNOLOGY

 dicated by the sum of squares was obtained using the ,.0

second-degree relation. The equation is ako consistent ?- .-

*
with the Beggs and Brill discove~ that liquid holdup
0.,
passes through maximums and minimums at fixed in-
clination angles of approximately +5o 0 and 500,
~spectively, for their data. Eq. 1 shows that the liquid 0.,

holdup should increase as the uphill angle of inclination

increases: Thk can be shown graphically by comparing
,.4
liquid holdup values obtained fmm the plots in F@. 7
and 8, where void fractions arc plotted for the iame oil at e
.

s,
= W%s=
= ,.,,, ,,/s,.
three simif.w &pirOciaI liquid velocities for horizontal 0.2 ,,, . 0.55;.gj,,c

and uphill 30 pipe inclination. (0.09 .,s

Intuitively, increased liquid viscosity should increase

viscous shear, causing irrcrcased liquid holdup regardless
of iriclination angle. Positive coet%cients of the liquid
viscosity n~ber in all the holdup correlations support
this hypothesis, Fig. aVoid fraction vs. V,g at different values of v~L for
-90 angle,
In general, any force that creates drag on any phase
against the direction of flow tends to increase the in-situ
fraction of that phase. As a result, viscous dtag on the
liquid will always tend to increase the liquid hoIdup, ir-
respective of inclination angle. However, gratihy forces !.,

on the mo~ dense phase will tend to increase the holdup

of that phase for uphill flow and decrease it for downhill
,.8
flow. Similarly, buoyant forces will tend to decrease
void fraction for uphill flow, while increasing it for
downhill flow,
6 0,
~
Discussion of Results . s,
~ [FT/?.Ecl

The proposed fiquid holdup corrdatirm was tested with $ ,., . ,,,,4 (mm ,s1

o ,.,6, (0.,, km
the observed data to check the reproducibility of the s

$.9 (u, MI,)

obsewed holdup values. For both the oils at different
. 7.i 12.2s MIS)
0,2
angles .of inclination, the relative percerit errors were 1,.0 13.66 MB)

calculated for individual experimental observations. For

each angle and each oil, the average percent errors md
their standard deviations (Table 2) were also calculated.
The capacitance sensor for liquid holdup rneasurcment
was found to be less. accurate in the fwtge of liquid
Fig. 7Void fraction vs. superficial gas velocity at fixed
holdups less than 10% or more than 90%. In these
superficial liquid velocity for horizontal flow.
ranges of liquid holdups, ve~ high percent errors were
observed (more than 30%). More than 80% of these data
points wiib high percent errors wee found to be in an-
nular or stratified flow regime, III both of these flow
regimes, the contribution from hydrostatic head to the
total p~ssure loss is quite insignificant. As such, those
data points with more than 30% relative percent error
were not used for the calculation of either average per-
cent errors or the standard deviations.
In the development of these liquid holdup correlations,
the BMDP nordirrcar regression package was used. This

regression method minimizes the residual sum of squares

to calculate the regression coefficients. The observation
correspondkrg to the otitf ien in the residual plot were ex-
cluded in the development of the holdup correlations.
Normally, these outliera indicated erroneous observa-
tions. This criterion for cuHing data does not correspond
to minimizing the average percent error. Hence, when
these corrdations were applied to the observed data, a
fu~er culling of data based on average percent error
was required. Normally, dependbrg on the value of the Fig. 8Void fraction vs. superficial gas velocity at fixed
absolute relative error, the sensitivity of the holdup superficial fiquid vefocity for uphill flow at 30.

measurement techniques reflects a great deal on the

average percent error. For very small values of obsemed
holdup, even with acceptable absolute error, relative er-

MAY 1983 1007

-.
 TABLE 1COEFFICIENTS OF LIQUID HOLOUP EQUATION

Flow
Flow Direction Pattern e, C2 Cs c. C5 C6
__ _ _ z .-

Uphill flow all -0.380113 0,129875 -0.119788 2.343227 0.475686 0.288657

downhll flow stratified -1,330282 4. S08139 4.17< 584 56.262268 0.079951 0.504S87

other -0.5i6644 0.78s805 0.551627 15.519214 0.371771 0.393952

TABLE 2STAT13TICAL PARAMETERS FOR tiOLDUp NLP =liqtidvelocity nuniber, U$L[P@T)]025

CORRELATIONS APPLIEO TO OBSERVED DATA
P,g = supefIcisJ gas velocity, ft/5ec (m/s)

VSL = supefllcial liquid veloci~, ft/sec (m/s)

Average Standard
Number of Deviation
x = viscosity,, cp (Pas)
Anale Error
(degkes) Data (%) (%) p = density, lbm/cu ft (kg/m3)

Kerosene
~ = s<~ace tension, dynelcm (mN/m)

5 35 2.79 13.64 @ = pipe inclinatiori angle from horizontal,

20 48 0.04 14.25
degree (rad)
30 57 4.71 13.40
45 5.49 1.77
50 2 - ? .96 12.30 . . . . . ----
60 -2.17 3.16
1. Eaton, B.A. t-r cd.: ..The Prediction of FIOW Patterns, Liquid
70 6: 4,98 11,92
Holdup and Pressure Lasses Occurring During Continuous Two-
80 49 -2,77 i3.55
Phase F70w in Horkontal Pipelines, Tram., AJME (1967) 240,
90 35 2,95 16.80
815-28.
0 42 -1,86 13.95
2. Vohra, LR. et al,: Comparison of Liq.id Holdup and J+iction-
40 2.44 13.79
Facmr Correlations for Gas-Liquid Flow,,, J. Per, Tech. (May
-x 41 -0.33 25,95
1975) 564-68.
-30 42 6.81 18.73
3. Beggs, H.D. and Brill, J. P.: A,$fudy or Two-Ph&se F30w in In-
-50 31 -1.71 20.32
clined Pipes, ?, J. Pa. Tech. (t&y 1973) 607-17.
-70 33 5.35 18.91
4, Cm[iffe, R, S,: Prediction of Condensate Flow Rates in LarEe
-80 29 -0.05 19.s9
Diameter High Press.= Wet Gas Pipelines, APEA J. (197~)
-90 29 6.19 21.20
171.
5. Dukler, A. E., Wicks, M. HI, md Cleveland, R, G.: Frictional
Lube al
Pressure Drop in Two-Phase Flow: B. An Approach Through
30 33 -1.01 15.01 3imilirity Analysis,, AIChE J. (Ian., 1964) 10, M-51.
90 43 -7.52 8,22 6. Palmer, C. M,: Evaluadon of Inclined Pipe Two-Phase Li@id
0 38 -4.34 13,58 Holdup Correlations Using Experirnmaal Data,, MS thesis, U. of
-30 23 -0.26 15.43 7.1s (1975).
-90 37 -7,15 15.83 7. Guzho., A.1., Marnaycv, V.A., and OdisbariYa, G. E.: A Study
of Tmspadadon in Gas-Liquid Systems,,, Pro.., 10tb Jnd. Gas
Conference, Hamburg, West Germany (1967).
8, Hughmwk, G.A. ad Pressburg, B. S.: .Holdup and pressure
Dmp with Gas Liquid Flow in a Vertical Pipe,, AIChE J. (Dec.
1961) 7, 677.
ror may be very kuge. This is often caused by division of 9. Bamecaze, R. H., Erskine, W., and Greskovich, E. J,: Holdnp
a small quantity in the calculation of percent error. and PEssure Dmp for, Two Phase Slug Flow in Inclined

Values of average percent error and sfsndard deviations Pipelims,., AJCIUE J. (Sept. 1971) 17, 1109,
10. Sigh, G, and Griftifh, P.: Determination of PEssure Drop Op-
for liquid holdup for each oil at different angles of in-
tinium Pipe Size for a Two:Phase Slug Flow in an Inclined P!pe,
clination are shown in Table2. J. En*. fo, Ind, (My. 1970); Tram., ASME 92, 717-26.
11. Zukoski, E. E.: %fl.erice of Viscosity, Surface Tension and In-
Conclusions clination Angle on Motio of !-ong Bubbles in Closed Tubes, J.
Fluid ,?dech. (1966) 25,821-37.
Ari empirical model forinclined two-phase flow liquid Dixon, W .J.: BMDP-Biomedical Compuler Programs, P-
12.
holdup is proposed. The proposed model enables. the Series,v, U, of California Press, Los Angeles (1977).

determination of liquid holdup regardless of the angle of 13. 0..s. H. Jr. and Ros. N. C. J.: Vetical Flow of Gas and Licwid
Mi.m& in Wells,,, Pro.., SMh Wodd Pi. Cong., Frank~mt
inclination and the direction of flow. The set of holdup
. .. ---- 451.
(19S3>
correlations is dependent on the snrne dlmension168s
14. Haged?m, A.R, and Bmym, K. E.: .Experimerdal Study of
painmeters that control the flow pattern transitions in PIEssure Gradients OcmnrirIE Dwig CoMin.ous Two-Phase
two-phase. flow. Except for downhll stratified flow, the Flow i. Small Diameter Verdcal Ccmduits,,, J. Pet Tech. (Apdf

liquid holdup correlations are continuous across flow- 1965) 475-84.

pattem transitions.

Nomenclature S1 Metric Conversion Factors

c = ~~pirial constant in. x 2.54* E+OO = cm

ft X 3.048* E01 = m
g =mvitationd accelemtion, ft/secz (m/sz)
.Conversion factor k exact. m
HL =Iiq!iid holdup
N8V =gas velocity number, vfcbL/(g@]O25 Original manuscript received in Sociely of Petroleum Engine.m o!hm March 22,

1 S82. Revised ma.uscr;pl received Jan. 21, 19S3. Paw acwted for Pubkaao.
NL =liquid viscosity number, pL~/(pLn3)]0Z Ocl 8, 1SS2.

1008 JOURNAL OF PETROLEUM TECHNOLOGY

-.. .
time. This voltage gain can easily be converted to a liquid holdup fraction

using a linear interpolation over the calibrated values of voltage gain for

zero percent and 100 percent oil in the pipe.

The oil and gas phases were carefully metered before mixing; turbine

meters, orifice meters or rotameters were used depending on the phase and the

flow rates. The two-phase mixture flowed through the test sections and into a

horizontal separator. The gas (air) was vented to the atmosphere and the

liquid passed through a filter and into a storage tank.

Kerosene and lube oil were used as the liquid phases. The surface

tension, density and viscosity of the kerosene at 60F were 26 dynesjcn,

51 lb#ft3 and 2 cp, respectively. Corresponding values for the lube oil were

35 dynes/cm, 53 lb~ft3 and 29 CP, Temperature between 18 and 132F were

encountered during the tests.

PHASESLIPPAGEANDLIQUID HOLDUP

In inclined two-phase pipe flow a substantial part of the total pressure

losses may be contributed by the hydrostatic pressure difference. The relative

contributions of friction gradient and hydrostatic gradient may be dictated by

the prevailing flow patterns, angle of inclination and direction of flow. Many

of the current design procedures used for two-phase pipelines fail to account

for these ef}ects with any rigor. Part of the problem Inmost design

procedures is the assumption that the void fraction is a unique function of

quality and physical properties of the fluids. This is probably true where

homogeneous flow can be assumed or during bubble flow at very low gas flow

rates, Similar situations may also arise where the phase velocity is very

high, so that friction pressure drop governs the total pressure loss. But in

the remaining cases errors may arise due to neglect of the slip velocity

between phases. This concept of slip velocity comes fr@m the physical
phenomena called slippage.

The term slippage is used to describe a natural phenomena of one phase

slipping past the other in two-phase pipe flow. There are several causes for

slippage between phases. Frictional resistances to flow or irreversible energy

losses in the direction of flow are much less in the gas phase than in the

liquid phase. This makes the gas more transmissible than liquid in two-phase

flow, even in the absence of strong buoyancy effects such as in horizontal

flow. This effect can be very pronounced in any segregated flow regime such as

stratified flow. The large difference in compressibilities between gas and

liquid causes the expanding gas to travel at a higher velocity and slip past

the liquid when pressure decreases in the direction of flow. Slippage between

phases is also promoted by the difference in buoyant forces acting on the

phases. In a static liquid medium, less dense gas tends to rise with a
15
velocity proportional to the density difference. Zukoski studled the effect

of pipe inclination angle on bubble rise velocity in a stagnant liquid. He

concluded that, depending on the pipe diameter, surface tension and viscosity

of fluids may appreciably affectthe bubble rise velocity. His fixidings also

showed that for some conditions an inclination angle as small as one degree

from the horizontal can cause the bubble rise velocity to be more that 1.5

times the value obtained for vertical pipes. This establishes a strong

dependence between inclination angle and phase slippage. In the absence of any

analytical formulation, the phenomenon of slippage caused by bubble rise

velocity is studied empirically.

Greater gravitational forces on the more dense liquid phase promotes fall

back of liquid when shear forces and buoyant forces fail to support the liquid

in upward flow. For downward flow it causacth liquid to travel faster than

the gas. Thus, while buoyancy always causes the gas phase to rise relative to

the liquid phase, gravity always tends to cause the liquid to fall faster than

the 8as,
 A few important conclusions can be made from the precedin8 discussion.

Except for houiogeneous flow the presence of slippage between phases in two-

phase pipe flow is unavoidable at any angle of inclination. In both uphill and

downhill bubble or slug flow, when the liquid phase is continuous and is

capable of being supported, buoyant forces generate bubble rise velocity

causing slippage between phases. Near the slug and annular-mist flow

transition or when the slug length becomes long (more than about two to three

ft) the phases become discontinuous. During this type of flow, broken liquid

slugs or ripples incapable of bridging the pipe are seen to fall back against

the direction of uphill flow. Very similar flow phenomena occur in downhill

stratified flow when the liquid falls back and accelerates until the liquid

kinetic energy is balanced by the shear ner8y around the liquid layer, In

stratified flow, large insitu velocities attained by the liquid as a result of

acceleration due to 8ravity normally causes a very small liquid holdup, This

phenomenon is shown in Fig. 3 where in stratified flow at 0.363 ft/sec liquid

superficial veloci:y, void fraction rises rapidly to almost 97% when very

little air flows simultaneously. At higher liqutd velocities in bubble or slug

flow, the void fraction builds up more slowly. An important deduction at this

point is that in uphill flow, slippage causes liquid velocity to slow down

resulting in a net accumulation of liquid in the flow channel or pipe and

increasing the insitu liquid fraction, The insitu liquid fraction is commonly

called liquid holdup. In downhill flow slippage causes insitu liquid velocity

to increase resultlng in a decrease in liquid holdv,p. AM these causes of

phase slippa8e and the result$n8 flow patterns will occur as soon as one end

of the pipe i8 raised about one pipe diameter from the other end regardless of

the an81e of inclination. Thus, dependins OR the lmgth of the pipe and

direction of flow, characteristic flow patterns or liquid holdup for inclined

flow should be observed even at extremely low angles. For example, at any low

uphill angle, the stratified flow pattern should never be observed.


1.0

E

o
I I

000
0
0,8 lB o
.

006 w
3
)

O*4

m VSL = 12,041FT/SEC
O VSL = 3,910 FT/SEC
0.2 I
vSL = 0,363 FT/SEC

O*O t I I 1 1
n mn .-
4U 80 100 120
superficial GASvELOcITY(FT/sEc) ~

Fig. 3- Void Fraction w v8g at

Different Values of VSL
for -30 Degrees Angle
 Development of Liquid Holdup Correlation

Analytical expressions for liquid holdup have been attempted for uphill

two-phase S1U8 flow in vertical pipes and for downhill flow at low angles of
,.
Inclination ln,the range of O to 15 degrees. Considering the complex slippage

mechanism, a global liquid holdup model for any pipe inclination has not been

attempted previously.

More than 1500 liquid holdup measurements at uphill and downhill

inclination an81es from 0 to f90 from horizontal were obtained in this

study. Attempts to correlate these data Into a global empirical liquid holdup

correlation are presented below. At each uphill and downhill an81e, void

fraction was plotted as a function of superficial 8ae velocity for fixed

superficial liquid velocity. Each of these plots was continuous within the

error tolerance of the holdup measurements. Example plots are shown tn Fi8s. 4

throu8h 6. At very high 8as rates the curves almost become asymptotic with the

100% void fraction (O% liquid holdup). For downhill stratified flow at very

low8as rates the void fraction rises rapidly and then almost linearly

increases with increased gas rates. However, the void fraction plot for

horizontal flow Is similar to the uphill plot, even in the stratified flow

reeime. The general shapes of these plots prompted selection of a non-linear

\
regression equation of the form,

2 N C5
HL=EXF(Cl+C2 Sin e+C3Sin 0+C4NL) ~.. .o~**. .~C(l)*..C(l)
NLV 6

Subsequently, three liquid holdup correlations were attempted, one for uphill

nd horizontal flow and the other two for downhill strat~fied flow and the

other downhill flow patterns. The re8ress&on cwdikients are given In Table

1. The coefficients were obta~ned by using the non-ltnuar BIOMED4regression

pro8ramao In each of the regression analysee, the outliers in the res$dual

plot were deleted from the data set and the analyees wao repeated.
100

0,8

0.6

0.4
+ VSL = 7.325 FT/SEC
VSL = Oc363FT/SEC
0.2 .
x VSL = 0.094 FT/SEC

0.0
o 20 40 60 80 100 120

SUPERFICIALGASVELOCITY(FT/SEC)~
. .

100 I

0.8 .

0.6

0.4 m

VSL = 10,679FT/SEC
VSL = 0.363 FT/SEC
0.2 m

x VSL = 0.094 FT/SEC

I i 1 I 1 1
0 20 40 80 80 100 120
superficial GASVELOCITY(FT/sEC) ~

Fig. 6- Void Fraution w Vsg at

Different Values of V~L for
+90 Degrees Angle
 I
~. A

I m VSL = 10.679FT/SEC
A VSL = 1.817 FT/SEC
VSL = 0.305 FT/SEC

I I I I I
o 20 40 60 80 100 120
SUPERFICIALGASVELOCITY(FT/SEC)~

Fig. 6- Void Fraction vs VW at

Different Values of vsL for
-90 Degrees Angle
 II

Table 1

Coefficients of Liquid Holdup Equation

I I
Values of Coefficients
i
Flow Direction I Flow Pattern ! i
1 1 I 1 B I 1 @ 1
I 1 I IC I 1 I I
1
1 I
! Ic
I II lc~cvlc
I I j I~
~ Uphill Flow ~ All ! I I 0.475686 ~ 0.288657 I
~ -0.380113 ~ 0.129875 ~ -0.119788 ~ 2.343227 ~
1 I I
1 ! 1 I 1 ~ I ~
1
t Downhill Flow
I
1I Stratified
i
! 4.808139
-1.330282 ~ ~ 4.171584 i 56.262268 ~ 0.079951 i 0.5048W !
.1 i~Other ~ -0.51644 ~ 0.789805 ~ 0.551627 ~ 15.519214 ~ 0.371771 \I 0.393952 iI
~
I i I I I I I
 The selection of phase velocity numbers as the independent variables

instead of the phase superficial velocities as shown in Figs. 4-6 was done to

make the variables dimensionless. These numbers were also suggested as

correlating parameters by Duns and Ros 6, Hagedorn and Brown9, and Eaton
et al. The velocity numbers, together with the inclination angle, also formed

the independent variables defining the flow patterns. Hence, inclusion of all

these variables implicitly makes the holdup correlation flow regime dnpendent.

The use of dimensionless numbers should not affect the shapes of curves shown
. in Figs. 4-6 since, for a ~xed oil, converting superficial velocities to
/

dimensionless form requ~.- _. ~lication by a nearly constant quantity of

approximately 2.5 for this study.

Effects of Inclination Angle and Viscosity

The second degree polynomial function of the form Cl + C2 Sin e +C3 Sinz e

was selected by plotting liquid holdups for different angles of inclination at

fixed liquid and gas velocity numbers. This relation was also confirmed by

comparing results of other equation forms in trial runs of the regression

analysis. The best error as indicated by the sum of squares was obtained using

the second degree relation. The equation is also consistent with the Beggs and
-
.

Brill discovery that liquid holdup passes through maximums and minimums at

fixed inclination angles of approximately +50 and -50, respectively for

their data. Eq. 1 shows that the liquid holdup should increase as the uphill

angle of inclination increases. fiis fact can be shown graphically by

comparing liquid holdup values obtained from the plots in Figs. 7-8 where void

fractions are plotted for the same oil at three similar superficial liquid

velocities for horizontal and uphill 30 pipe Inclination.

Intuitively, increased liquid viscosity should increase viscous shear

causing increased liquid holdup irrespective of inclination angle. Positive

coefficients of the liquid viscosity number in all the holdup correlations

support this hypothesis,

 .0.094
1.0

3.9
0.8 . 7.3
//
1200

006
g
-.
VSL
(FT/SEC)
0.4 , A 0.094
0 0.363
3.9
x 7.3
0.2
12.0

000
1 1 I 1 1
0 20 40 60 80 100 120
SUPERFICIALGASVELOCITY(FT/SEC)~

Fig. 7-, Figure ShowingVoidFraction w Superficial (3as Velodty

at Fixed Superficial LiquidVelocity for Horizontal Flow
 o

VSL
(FT/SEC)
A o 094

3*9
i, 12.0

0 20 40 60 80 100 120
SUPERFICIALGAS VELOCITY(FT/SEC)~

Fig. 8- Figure Showing Void Fraction vs Superficial Gas Velocity at Fixed

Superficial LiquidVelocity for Uphill Flowat 300
 In general any force which creates drag on any phase against the

direction of flow tends to increase the insitu fraction of that phase. As a

result, viscous drag on the liquid will always tend to increase the liquid

holdup, irrespective of inclination angle. However, gravity forces on the more

dense phase will tend to increase the holdup of that phase for uphill flow and

decrease it for downhill flow. Similarly, buoyant forces will tend te decrease

void fraction for uphill flow while increasing it for downhill flow.

DISCUSSIONOF RESULTS

The proposed liquid holdup correlation was tested with the observed data

to check the reproducibility of the observed holdup values. For both

experimental oils at different angles of inclination the relative percent

errors were calculated for individual experimental observations. For each

angle and each oil the average percent errors and their standard deviations

were calculated. The data points with more than 30 percent relative percent

error were not used for the calculation of either average percent errors or

the standard deviations. A majority of these data were for liquid hcldups less

than 10 percent or more than 90 percent. In this range of liquid holdup the

capacitance sensor was found to be less accurate and high percent errors in

the measured values were expected. However, for most of these very low or very

high liquid holdup cases, the sensitivity of total pressure loss to the liquid

holdup was greatly reduced.

In the development of these liquid holdup correlations, the BMDPnon-

linear re~ression package was used, Thts regression method minimizes the

residual sum of squares to calculate the regression coefficients, The

observations corresponding to the outliers in the res&dual plot we~= c~cludei

in the development of the holdup correlations. Normally, these outliers

indicated erroneous observations, This criterion for culli~ data does not

correspond to minimizing the average percent error. Hence, when these

correlations were applied to the observed data, a further culling of data

based on average percent error was required. Normally, depending on the value

of the absolute relative error, the sensitivity of the holdup measurement

techniques reflests a great deal on the average percent error. For very small

values of observed holdup, even with acceptable absolute error, relative error

may be very large. This is often caused by division of a small quantity in the

calculation of percent error. Values of average percent error and standard

deviations for liquid holdup for each oil at different angles of inclination

are shown in Table 2.

CONCLUSIONS

An empirical model for inclined two-phase flow liquid holdup is proposed.

The proposed model enables the determination of liquid holdup irrespective of

the angle of inclination and the direction of flow. The set of holdup

correlations is dependent on the same dimensionless parameters that control

the flow pattern transitions in two-phase flow. Except for downhill stratified

flow, the liquid holdup correlations are continuous across flow pattern

transitions.
 Table 2

Statistical Parameters for Holdup Correlations

Applied to Obeerved Data

oil Angle i Standard i

(degrees) I~ &~f ,/ Average
% Error Deviation I
~
i I

35 i 2.79
5
20 48 I O*O4
13,64
14.25
57 I 4.71 13040
:; I 5*49 1.77
!5J 4; I\ -1.96 12.30
-2.17 3.16
70 6; I~ 4.98 11.92
80 -2.77 13.55
90 % ~ 2.95 16.80
42 -1.86 13.95
-; 40 / 2.44 13079
-20 . I -O*33 25.95
-30 :; ~ 6.81 10.73
-50 31 -1.71 20.32
-70 33
I 5*35 18,91
-80 29 I -os05 19.89
-90 29 1 6.19 21.20
1

I -1 01 15.01
:; II -7.52 I 8.22
-4.34 I 13.58
: I -Oe26 I 15*43
I 15.87
37
I -7.15
NOMENCLATURE

c empirical constazs

l~quid holdup
L
gas-velocity number, v
gv Sg {pL/(gu)}25
liquid-velocity number, vsl {@@}25
*Lv
liquid viscosity number, pL {!j/(pLu3)}*25
L
superficial gas velocity, ft/sec
Sg
superficial liquid velocity, ft/sec
s1
u viscosity, Cp

P density, lbm/ft3

u surface tension, dynes/cm

e pipe Inclination angle from horizontal

S1 METRICCONVERSION
FACTORS

Cp x 1*O* E-03 = Pa.S

dyne x 1.0* E-02 = mN

- 32)/1.8
F *F = c
ft X 3.048* E-01 = m

inx 2.54* E+OO = cm

lbm X 4.535924 E-01 = KS

* Conversion factor is exact.

 .

1. Beggs, H.D. and Brill, J.P.: A Study of Two-Phase Flow in Inclined

Pipes, J. Pet. Tech. (May 1973), 607-617.

2. Bonnecaze, R.H., Erskine, W. and Greskovich, E.J.: Holdup and Pressure

Drop for No Phase Slug Flow in Inclined Pipelines, AIChE J. (Sept. 1971)

17, 1109*

3* Cunliffe, R.S.: Prediction of Condensate Flow Rates in Large Diameter

High Pressure Wet Gas Pipelines, APEA J. (1978), 171.

4. Dixon, W.J.: BMDP- Biomedical Computer Pro8rams, P-Series, Univ. of

California Press (1977).

50 Dukler, A.E., Wicks, III, M. and Cleveland, R.G.: Frictional Pressure

Drop in Two-Phase Flow: B. An Approach Throu8h Similarity Analysis,

AIChE J. (Jan. 1964) 10, No. 1.

6. Duns, H., Jr. and Ros$ N.C.J.: Vertical Flow of Gas and Liquid Mixtures

in Wells, ~roc. 6t~ World Pet. Cong. (1963), 451.

7. Eaton, B.A., et al: The Prediction of Flow F~ttems, Liquid Holdup and

Prescure Lo.sacs 0ccurrin8 Durin8 Continuous Two-Phase Flow in Horizontal

Pipelines, Trans. AIME, (1967)$ 815.

8. Guzhov, A.I., Mamaye~?,V*A. and Odishariya, G.E.: A Study of Transport-

ation in Gas-Liquid Systems, 10th Int, Gae Conference, Hamburg, Germany

(1967).

9. Hagedorn, A.R. and Brown, K.E.: Experimental Study of Pressure Gradients

Occurring During Continuous Two-Phase Flow in Small Diameter Vertical

Conduits, J. Pet. Tech. (April 1965), 475-484.

10* Hu8hmark, G.A, and Pressbur8, B.S.: Holdup and Pressure Drop with Gas

Liquid Flow in a Vertical Pipe, AIChE J. (Dec. 1961) ~, 677.

llQ Mukherjee, H.: An Experimental Study of Inclined Two-Phase Flow, Ph.D.

Dissertation, The Univ. of Tulsa (1979).

12. Palmer, CoM@: Evaluation of Inclined Pipe Two-Phase Liquid Holdup

Correlations Usin8 Experimental Data, M.S. Thesis, The U. of Tu1s8(1975).

13. Singh, G. and Griffith, P.: Determination of Pressure Drop Optimum Pipe

Size for a Two-Phase Slug Flow In an Inclined P$pe, J, Eng. for Ind.

(Nov. 1970), Trans. ASME, 92, 717-726.

14. Vohra, I.R., et al: Comparison of Liquid Holdup and Friction-Factor

Correlations for Gas-Liquid Flow$t, J. Pet. Tech. (May 1975), 564-568.

15. Zukoski, E.E.: Influence of V?.scosity, Surface Tension and Inclination

Angle on Motion of Long Bubblas in Closed Tubes, J. Fluid Mech. (1966)

2S, Part 4, 821-837.

e,