The Dynamic Two-Fluid Model OLGA:

Theory and Application

Knell H. Bendlkaen, Dag Mmlnes, Randl Moe, and Sven Nuland, Inst. for Energy Technology

JuMMM'. Dynamic two-fluid models have found a wide range of application in the simulation of two-phase-flow systems,
particu- larly for the analysis of steam/water flow in the core of a nuclear reactor. Until quite recently, however, very few
attempts have been made to use such models in the simulation of two-phase oil and gas flow in pipelines. This paper presents
a dynamic two-fluid model, OLGA, in detail, stressing the basic equations and the two-fluid models applied. Predictions of steady-
state pressure drop, liquid hold- up, and flow-regime transitions are compared with data from the SINTEF Two-Phase Flow
Laboratory and from the literature. Com- parisons with evaluated field data are also presented.

Introduction
The development of the dynamic two-phase-flow model Consewatian of Macs. For the gas phase,
OLGA started as a project for Statoil to simulate slow
transients associat- ed with mass transport, rather than the fast d 1 d ()
pressure transients well known from the nuclear industry.
dt
Problems of interest included ter- rain slugging, pipeline startup
and shut-in, variable pnxluction rates, and pigging. This implied For the liquid phase at the wall, d1 6
simulations with time spans ranging from hours to weeks in
extreme cases. Thus, the numerical method ap- plied would
have to be stable for long timesteps and not restricted by the
velocity of sound.
A first version of OLGA based on this approach was
working in 1983, but the main development was carried out in For liquid droplets,
a joint re- search program between the Inst. for Energy
Technology (IFE) and SINTEF, supported by Conoco Norway,
Esso Norge, Mobil Ex- ploration Norway, Norsk Hydro A/S,
Petro Canada, Saga Petro- leum, Statoil, and Texaco
Exploration Norway. In this project, the empirical basis of the
model was extended and new applications were introduced. In Eqs. 1 through 3, Pg, P¿, Kp ——gas, liquid-film, and
To a large extent, the present model is a product of this liquid- droplet volume p=density, v=velocity, p ——
project. pressure, and A =pipe cross-sectional area. Jg =mass-transfer
Two-phase flow traditionally has been modeled by separate rate between the
em- pirical correlations for volumetric gas fraction, pressure phases, be' id- the entrainment and deposition rates, and
drop, and flow regimes, although these are physically possible mass source of Phasef. Subscripts g, L, i, and D
indicate gas, liquid, interface, and droplets, respectively.
interrelated. In recent years, however, advanced dynamic Consewation o/ñfomenfum. Conservation of momentum is ex-
nuclear reactor codes like TRAC, 1 RRLAP-5, 2 and pressed for three different fields, yielding the following
CATHARE* have been developed and are based on a more separate ID momentum equations for the gas, possible liquid
unified approach to gas fraction and pressure drop. Flow droplets, and liquid bulk or film.
regimes, however, are still treated by separate flow- regime For the gas phase,
maps as functions of void fraction and mass flow only. In the
OLGA approach, flow regimes are treated as an integral part
of the two-fluid system.
The physical model of OLGA was originally based on ât
small- diameter data for low-pressure air/water flow. The 1983
data from the SINTEF Two-Phase Flow Laboratory showed that,
while the bubble/slug flow regime was described adequately,
the strati- fied/annular regime was not. In vertical annular flow,
the predicted
pressure drops were up to 50a too high (see Fig. 1). In —h;—p +V$pkg cos a+ f$ vg — F p......(4)
horizontal flow, the predicted holdups were too high by a 4N 2
factor of two in extreme cases.
These discrepancies were explained by the neglect of a For liquid droplets,
droplet field, moving at approximately the gas velocity, in the
early model. This regime, denoted stratified- or annular-mist
flow, has been in- corporated in OLGA 84 and later versions,
where the liquid flow may be in the form of a wall layer and a
possible droplet flow in the gas core.
This paper describes the basic features of this extended two-
fluid model, emphasizing its differences with other known
two-fluid
models. Eqs. 4 and 5 were combined to yiel‹i a combined momentum
equa- tion, where the gas/droplet drag terms, F p, cancel out:
OL0A—Tho Ext•nd•d Two•FIuId Mod•l
Physical Models. Separate continuity equations are applied for
gas,
liquid bulk, and liquid droplets, which may be coupled through
in- terphasial mass transfer. Only two momentum equations are
used,
however: a combined equation for the gas and possible liquid e e
droplets and a separate one for the liquid film. A mixture n r
gy- conservation equation currently is applied. Sg it
—h;—pg ’r ’r
2 4A 2 4d
Copyright 1001 Society of Petroleum Engineers

SPE Production Engineering, May 1991

171
 The droplet velocity is similarly defined by
-10’ e =vg —v /j cos a,...................................................(10)
” where vp0 is the fall velocity of droplets. The interphase
velocity,
v;, is approximated by v¿.
Prescure Equation. OLGA reformulates the problem before
dis- cretizing the differential equations to obtain a pressure
50 equation. This equation, together with the momentum
equations, may be solved simultaneously for the pressure and
phase velocities and thus allow a stepwise time integration.
The conservation-of-mass equations (Eqs. 1 through 3) may
be expanded with respect to pressure, temperature, and
composition, assuming that the densities are given as

where the gas mass fraction, s• is defined by Eq. 17.

Por the gas equation (Eg. 1), the left side may be expressed as

10
10*104
Measured pre ssure drop (N/T)
. (12)
Fig. 1—Predicted pressure drops compared with SINTEF Two•
Dividing the expansions (Eq. 12) for each phase by the
Phase Flow Laboratory data. Annular-mist dlesel/nitrogen
densitiesInclud-
flow at 0 MPa vertical. (- •) original OLGA; (X} OLGA and adding the three equations yields a volume-
ing a droplet field. conservation equa- tion (neglecting the last two terms in Eq. 12
because they normally are negligible in pipeline transport
problems owing to the slow tem- perature development):
For the liquid at the wall,

+— +— . ............................
(13) 1
- h¿ — +h,—p
P 4A 2 Inserting the mass-conservation equations for each phase and ap-
2 plying Kp + K¿ + Pp ——1 gives

...................................
. (7)
In Eqs. 4 through 7, n=pipe inclination with the vertical
and S , Sz, and St ——wetted perimeters of the gas, liquid, and
interface. be internal source, G/ , is assumed to enter at a 90°
angle to the pipe wall, carrying no net momentum.
v = v¿ for > 0....................................................(8a)
(and evaporation from the liquid film),
vq=vp for Jg > 0...........................................................(8b)
(and evaporation from the liquid droplets),
and v = rg for Jg < 0 .....................
(condensation).
The above conservation equations can be applied for all flow
re- gimes. Observe, however, that certain terms may drop out
for cer- tain flow regimes; e.g., in slug or dispersed bubble
flow, all the droplet terms disappear.
For slug flow, the frictional pressure-drop terms are
composite. They consist of three terms, owing to the liquid
slug, the slug bub-
ble and the film under it, and the liquid-film-acceleration
pressure drop.
The relative velocity, v„ is defined by the following slip
equation:
ttt
+G — +Gg— +Gp—........................................(ic)
Eq. 14 provides a single equation for the pressure and phase
flux- es. Note that if the phase transfer term, Jg, is a function
of pres- sure, temperature, and composition,
gf - g[p,TRR ),...................................................................(15)
then J$ may be expanded by a Taylor series in p, T, and
Jt„ as shown in Rq. 18. A strong effect of pressure on the
phase transfer (boiling) may then be incorporated into Eq.
14.
Energy Equation. A mixture energy-conservation equation is
applied:
2
— mlv$ +gh)
dz

where Up is a distribution slip ratio caused by an uneven 172

distribu- tion of phases and velocities across the pipe cross
section, as out- lined in the next section.
 . (16)

SPE Production Engineering, May 1991

where E -- mtem al energy per unit mass, /i ——elevation, HS —-
enthalpy from mass sources, and t/=heat transfer from pipe walls.

Thermal Calculations. OLGA can simulate a pipeline with a to-

tally insulated wall or with a wall composed of layers of
different thicknesses, heat capacities, and conductivities. The
wall descrip- tion may change along the pipeline to simulate, for
instance, a well surrounded by rock of a certain vertical
temperature profile, con- nected to a flowline with insulating
materials and concrete coat- ing, and an uninsulated riser.
OLGA computes the heat-transfer coefficient from the flowing m
fluid to the internal pipe wall; the user specifies the heat- *•• O’ ‘ ’ ‘ o

transfer coefficient on the outside. Circumferential symmetry is

assumed; if this is broken, for example, with a partly buried
pipe on the sea bottom, average heat-transfer coefficients
must be specified.
Special phenomena, such as the Joule-Thompson effect, are
in- cluded, provided that the PVT package applied to generate
the fluid- property tables can describe such effects.
Pluid Properties and Phace Transfer. All fluid properties
(den- sifies, compressibilities, viscosities, surface tension,
enthalpies, heat capacities, and thermal conductivities) are
given as tables in pres- sure and temperature, and the actual
values at a given point in time and space are found by
interpolating in these tables.
The tables are generated before OLGA is run by use of any
fluid- properties package, based on a Peng-Robinson, Soave-
Redlich- Kwong, or another equation of state, complying with
the specified table format.
The total mixture composition is assumed to be constant in
time along the pipeline, while the gas and liquid compositions
change with pressure and temperature as a result of interfacial
mass trans fer. In real systems, the velocity difference
between the oil and phases may cause changes in the total
composition of the mixture. This can be fully accounted for
only in a compositional model.
InteJacial ñfass Transfer. The applied interface mass-transfer
model can treat both normal condensation or evaporation and
retro- grade condensation, in which a heavy phase condenses
from the gas phase as the pressure drops. Defining a gas mass
fraction at equilibrium conditions as
R 'mz l{mz‘+ mz + Inc),.................................................. (17)
we may compute the mass-transfer rate as
+ —
bTp btbTp bz bt
The term {bR lbp) p {bplbt) represents the phase transfer
from a mass present in a section owing to pressure change in
that sec- tion. The term {bR ldp)T {bpfbz )(bz fbt) represents the
mass trans-
fer caused by mass flowing from one section to the next.
Because
only derivatives of RS appear in Eq. 18, errors resulting from
the assumption of constant composition are minimized.
Numerical Solution Scheme. The physical problem, as formu1at-
ed previously, yields a set of coupled first-order, nonlinear, 1D
partial differential equations with rather complex coefficients.
This nonlinearity means that no single numerical method is
optimal from all points of view. In fact, the codes
TRAC, RELAP,
CATHARE, 3 and OLGA all use different solution schemes.
De- tails are presented elsewhere. 4
Most two-ñuid models, including those listed above, apply
finite difference staggered mesh, donor cell methods. In explicit
integra- tion methods, the timestep, At, is limited by the Courant
Friedrich Levy criterion based on the speed of sound:
-j- c/j | }.......................................................(19)
Flg. 2—Schematic of stratlfled annular mlet and alug flow.
Because the speed of sound typically is about 102 to 103
Times larger than the average phase velocities, explicit integration
methods require timesteps up to 103 times smaller than
implicit methods. Traditionally, most nuclear reactor safety
analysis codes (e.g., NORA S) applied explicit methods because
they are simpler to for- mulate and code, and the time scales of
interest for typical prob- lems (pressure transient) were given
by the speed of sound. Because of stability problems, however,
and the need to simulate slow, small
breaks, implicit methods are now favored.
filow•RepIme Deacrlptlon
The friction factors and wetted perimeters depend on flow
regime. Two basic flow-regime classes are applied: distributed,
which con- tains bubble and slug flow, and separated, which
contains stratified- and annular-mist ñow. Because OLGA is a
unified model, it does not require separate user-specified
correlations for liquid holdup, etc. Thus, for each pipeline
section, a dynamic dow-regime pre- diction is required,
yielding the correct flow regime as a function of average
flow parameters.
Separated Eow. Stratified- and annular-mist flows are
character- ized by the two phases moving separately (Fig. 2).
The phase dis- tributions across the respective phase areas are
assumed constant. The distribution slip ratio, RD , in Eq. 9
then becomes 1.0. The transition between stratified and annular
flow is based on the wet- ted perimeter of the liquid film;
annular flow results when this perimeter becomes equal to
the film inner circumference.
Stratified flow may be either smooth or wavy. An expression
. (18) for the average wave height, lip, may be obtained by
assuming that the mass flow forces in the gas balance the
gravitational and sur- face tension forces, or
[112lpg(vg — y )2 -hp@ p —p g)g sin n +(‹r//tp)........(21)
2 2(p¿ —pg)g sin n
4‹r
+ . . (22)
Implicit methods are not limited by Eq. 19, but for dynamic
prob-
When the expression in the square root is negative, hq is flow at 1 bar, the onset of 2D waves corresponds very well
zero and stratified smooth flow is obtained. with the data of Andreussi and Persen6 (Fig. 3).
The onset of waves starts with capillary waves with Friction N‹irfors. The applied wall friction factors for gas
wavelengths of about 2 to 3 mm. As the mass-flow forces and liquid are those of either turbulent or laminar flow (in
increase, surface ten- sion becomes negligible and gravity practice, the maximum one is chosen), given as
dominates, resulting in longer wavelengths. For air/water pipe 2 x 1o4¿
106
lems a mass-transport criterion applies:
h,=0.0055 1 + + NR . . (23)
Atñ minNj(Az;/ | v/j |}.....................................................(20) dh

SPE Production Engineering, May 1991 173

 v, L = 0.028 m/sec
o Andieossi & Person*
OLGA

200Lockhart & ktartiIt

Ta te
Dukter

ONSET OF
GO2 -D WAYES

20

10
10 20 t0 100 kleasured liquid hold - up

Fig. 4—Comparison between measured and predicted IIquId

Fig. 3—Pressure gradient at the tranaltlon to 2D waves
{pt = 0.001 kg/(aec m), a= -0.65•].

Distributed Flow. As Malnes*1 showed, iR the general bUbbl -

or slug-flow case, the average phase velocities satisfy the
where e=abso1ute pipe roughness and d¿ ——hydraulic follow- ing slip relation:
diameter.
For stratified-mist flow, the wall liquid fraction, wetted vg -R ‹vp + v,),..............................................................(30)
perimeters, and other flow parameters are defined by the where v and RQ are determined from continuity requirements.
wetted angle, Q, as indicated in Fig. 2. For Eq. 30 reduces to the general expression for pure
Wallis7 pressed the following equation for slug
in annuldf flOW'
h;= 0.02[1+ 75( g)1t • • • • • • • • • • • -• .......... (25)
which has been applied for vertical flow. For inclined pipes, Eq. * L ‘ ‘............................(31)
26 is used for annular-mist flow: (1/CA ) — V$ ! + C (1 — Pg) ‘ ’ ‘ ‘ '
h, =0.02(1 + ¿),............................................................(26) For fully developed turbulent slug flow with sufficiently
where K is an empirically determined coefficient of the form large slug lengths (z 10D), Bendiksen'° showed that the velocity
of slug (or Taylor) bubbles, vb, may be approximated for all
inclinations by
R=R ’— * . ..........................(27)
I , V B ' (Vy + V ) + V@ 9 ... • . • • . • • • • • • • • • • • • .. • . • . • . (3 2)

d g@p -p )
2
For stratified smooth flow, the standard friction factors with where C = 1.05 +0.15 cos n for N FV < 3.5
zero . . (33)
pipe roughness are used; for wavy flow, the minimum value of 0
1.20 for Np, > 3.5
Eq. 26 and
p vqg cos n +
vp0 and sin a for Npr < 3.5
(34)
are used because Eq. 26 is assumed to yield an upper limit for wavy v0 y cos a for NFV > 3.5
flow. Eq. 28 then provides an improved description in the
region from smooth flow to higher velocities, where Eg. 26 where v0y and vp0 are the bubble velocities in stagnant liquid
applies. (ne- glecting surface tension) in vertical and horizontal pipes,
respec- tively:
Entrainmenti'Deposiâon. A droplet field was not incorporated
into the original version of OLGA. Compared with the
SINTEF
Two-Phase Flow Laboratory data, the predicted pressure drops in • =0.35 g . (35)
vertical annular flow were up to 50$‹ too high (Fig. 1). In . (36)
and v p=0.54 g .
horizontal flow, the pressure drop was well predicted, but the
liquid holdup was too high by a factor of two in extreme For pure bubble flow, Eq. 30 reduces to
cases. . (37)
For droplet deposition, the following equation for vertical vg 'ft(v£ ’1- vos),.......................................• •
flow may be obtained from Andreussi’s data:
where fi= (1 — P$)/(K— Kg$)...................................................(38)
VQ pp 0.8
2.3 x 10 —4 i and R= 1/Cl is a distribution parameter.
(29)
d Malnes*3 gives the average bubble-rise velocity as
Por inclined pipes another, extended correlation is applied.
A modified expression is proposed for liquid entrainment *os'1.18 .(39)
based on that of Dallman et al.° for vertical flow and Laurinat
et al. *0 for horizontal flow.
with positive values
upward.
174 SPE Production Engineering, May 1991

Predic ted liquid hold-

 20
10

VELOCITY
1
0G £ ¥ 0

LIOU
0.1

SUPERFICI
o@ o• • a•
0.01 o• • • pa

Fig. WPresaurs gradlent (- - -) and slug-flow boundades com- pared with OLGA predictions —âpIâx• --- alug-flow bound- arlec) for horizonta
0.00 0.02 1 10
0.t

8
SUPERFICIAL GAS VELOCIT Y (m is)

Flg. 6—Flow-reglme transitions from Ref. 17 compared wlth OLGA [2° upwar
Using Gregory and Scott’s'^ data, Malnes proposed the fol-
lowing equation for the void fraction in liquid slugs:

where C is a constant determined empirically and the void
fraction is limited upward.
This correlation (Eq. 40) is applied for small-scale systems
only. For high-pressure, large-diameter pipes, another set of
empirical correlations based on the data from the SINTEF Two-
Phase Flow Laboratory was used.
The total pressure drop in slug flow consists of three terms:
{bzfbp1 -- llL){Ip +dp$ + dp ),.....................................(41)
where ApS ——frictional pressure drop in the liquid slug, App=
frictional pressure drop across the slug bubble, and dpp=acceler-
ation pressure drop required to accelerate the liquid under the slug
bubble, with velocity v¿j, up to the liquid velocity in the slug, •
(Ap„=0 at present). L is the total length of the slug and bubble.
These terms are dependent on the slug fraction, the slug bubble
void fraction, and the film velocity under the slug bubble. The
void fraction in the slug bubble, Kg, is obtained by treating the
flow in the film under the slug bubble as stratified or annular
flow. This is further described by Malnes, who gives additional
equations. For slug flow, the wall friction terms will be more
complicated than shown in Eqs. 6 and 7 because liquid friction
will be depend- ent on vg and the gas friction on v¿.
Flow-Regime Transitions. As stated, the friction factors and
wet- ted perimeters are dependent on flow regime. The
transition be-

•.L '0.0083 mls

a = 0’
8

- - 0.20

300 - 0.15

200 - 0.10

100 - 0.05

0
0 10 20 30

Fig. 7—ComparIsons of pressure and llquld holdu

OLGA predictions(—) and Crouzler’c maaaursmanta(ID-
0.045 m; e etratlfled pressure drop; A alug pressure
drop; 0 stratI• fied holdup; A slug holdup).
 0.20

300 0.15

200 0.t0

100 0.05

0 0
0 \0 20 30

Flg. 8—Compadeona of pressure and Ilquld holdu

OLGA predlctlonc(—) and Crouzler’a
m•acurements(ID=0.045 m, e stratified pressure drop;
a slug pressure drop; O strati- fled holdup; A slug
holdup).
SPE Pixxluction Engineering, May 1991 175
 176
, L = 0.0083 mls
O = 2.86•
TOO 0.20

300 0.15
»
'
200 0.10

t00 0.05

0
0 10 20 30

Fig. 9—Comparisons of pressure and liquid holdu OLGA

predictions (—) and Crouzler’s measurements (ID =
0.045 m;
• stratified preeeure drop; A slug pressure drop; O
etratl- fled holdup; A alug holdup).

tween the distributed and separated flow-regime classes is based

on the assumption of continuous average void fraction and is
deter- mined according to a minimum-slip concept. That is, the
flow re- gime yielding the minimum gas velocity is chosen.
Wallis*° empirically found a similar criterion to describe the
transition from annular to slug flow very well. This criterion
covers the transitions from stratified to bubble flow, stratified to
slug flow, annular to slug flow, and annular to bubble flow.
In distributed flow, bubble flow is obtained continuously when
all the gas is carried by the liquid slugs (when the slug fraction,
FS , approaches unity). This occurs when the void fraction in the
liquid slug, Pg$, becomes larger than the average void fraction,
Kg.
The stratified-to-annular flow transition is obtained when the
wave height, 6q (Eq. 22) reaches the top of the tube (or S¿ -—
rD).

Comparison Wlth Steady•St•t• Experiments

OLGA has been compared with data from different experimental
facilities, covering a wide range of geometrical sizes, fluid types,
pressure levels, and pipe inclinations. The bulk of the data was
ob-
tained from experiments at the SINTEF Two-Phase Flow
Labora- tory. These unique data have increased our confidence
in the applied two-phase models. A detailed description of the
experimental fa- cilities is presented in Bendiksen er al. 1*
 Fig. 4 (from Ref. 16) compares predicted and measured
holdup for stratified flow for diesel and nitrogen at 3 x 10* Pa
and 30°C. The predictions are generally within -k 10a . SPE Pr‹xlcction Engineering, May t99l
Fig. 5 compares OLGA-predicted pressure gradients and
slug- flow boundaries with data from the SINTEF Two-Phase
Flow Lab- oratory. The predictions are quite good, although
the pressure drop in the aerated slugs is somewhat low.
Barnea ei al. ** measured flow panems for air/water flow in
horizontal and inclined tubes at near-atmospheric conditions
for 1.95- and 2.55-cm diameters. Fig. 6 shows a typical
example of predicted flow regime transitions compared with
the experimental flow map for +2° inclination.
Crouzier measured pressure drop and holdup in horizontal
and upwardly inclined pipes for air/water at near-atmospheric
pressures in a 4.5-cm tube.
In Figs. 7 through 9, OLGA predictions are compared to
Crou- zier’s data. The calculated and measured pressure drop
and hold- up generally agree quite well, considering the
spread in data. The flow-regime transitions are observed when
the holdup calculations for slug flow cross the predictions
from stratified/annular mist. The regime with minimum
holdup is that predicted by OLGA, and the agreement is quite
good.
The pressure drop from slug/bubble to stratified-/annular-
mist flow experiences a discontinuity at upsloping angles,
which is justi- fied partly by the experiments.

Compaz4aon With Dynamic Exp•rlm•ntc

The OLGA model has been compared with data from two
differ- ent types of experimental setups. Schmidt et al.
performed ex- periments at laboratory conditions with a
pipeline ID of 5.08 cm at atmospheric pressure. The SINTEF
Two-Phase Flow Labora- tory has been producing data for gas
and oil in 19-cm-ID pipes with total lengths of 450 m at
pressures up to 10 MPa (Fig. 10).

Terrain-Slugging Data. Terrain slugging is a transient flow

type associated with low flow rates. It may, for instance, be
observed in pipelines where a downsloping pipe terminates in
a riser. The slugging is initiated by liquid accumulating at the
low point. Refs. 19 and 20 give more detailed descriptions of
the phenomenon.
SINTEF T»'o-Phase fl ow Laboratory Data. The test
section of the SINTEF laboratory is sketched in Fig. 10;
further details are given in Refs. 16 and 21. In the reported
experiments, the fluids applied were diesel oil and nitrogen.
One terrain-slugging flow map obtained with OLGA is
shown in Fig. 11 for a system pressure of 3 MPa. The
agreement with the experimental map is quite good, although
the OLGA predic- tions are somewhat conservative for low
liquid superficial veloci- ties. Note that OLGA can properly
distinguish between the two types of terrain slugging (I, II) with
or without aerated slugs.
Data of Schmi& zt ef. This nitrogen/kerosene low-pressure
loop included a nearly horizontal pipe and a vertical riser test
section of 30- and 15-m lengths, respectively, and 5-cm
diameters. Schmidt er of.’s*° test matrix was reproduced for
downsloping pipes, par- ticularly for —5°, inclination. The
results are shown in Fig. 12, where the solid lines are the
results of the OLGA simulations. Fig.

ixng
point Pressure reference line
54m

334 m 65m

Fig. 10—Pips geometry of the SINTEF Two-Phaee Flow Laboratory (1963) (0 alngle-beam
dencltomster; ■ dlfierentlal pressure transmitter).
 3.0
O Terrain slugging I 240 -
2.5 —@' ^ A Terrain slugging II
« No terrain slugg g

PRESSURE P • 10*
210
velocity

2.0 -
\80

t50
liq

t.0 -
Superfici

120
0.5 - 90
300320340360380400
TIME (S)
0
0 0.51.01.52.02.53.0
Superficial gas velocity (mls)

Flg. 11—Terrain-sIugglng flow map from the SINTEF Two• 225 -

Phaee Flow Laboratory. Terrain alugglng I (—) and II (- - -)
boundarlea predicted by OLGA.
J95

PRESSURE P « 10
32
165

0 SE¥ERE St.UGGlNG I+ SLUG FLOU J35 105

o sEvERE sruscius y • xxHucc9 rrow
A tRANS TO SLUGGING 0 EUBBLE FLOW 75
3003203 0360380
TIME (S )

a
Flg. 13—Pressure oscillations of severe slugging'* for flow rates—hf , - 1

able to simulate the time evolution of holdup and the pressure

1 t0 response to the inlet conditions (Figs. 16 through 20). The pre-
Qaa velocity number, H dicted time response is good, but slightly too slow.

Flg. 12—RIaer flow-pattern map with - 5• pipeline IncIlna-Exesrptc of Comparlconc Against Field Data
tlon’^ (— OLGA predictions). To test the extrapolation capabilities of OLGA, a separate study
was performed, comparing this model with evaluated field day,
4 The results of one of these studies, the Vic Bilh-Lacq
13 compares experimental and predicted pressure oscillations in oil/associated-
the horizontal line and at the bottom of the riser for one gas line, are presented here. This is an onshore field of relatively
particular experiment. heavy crude in southwestern France.
The pipeline is 43.8 km long and has a 25.1-cm ID, and an ab-
Transient Inlet-Flow Data. The dynamic experiments in the solute roughness of 0.03 mm. Flow conditions are characterized
SIN- TEF laboratory with time-dependent inlet flow rates were by low superficial liquid velocities of about 0.17 mls and
performed with a completely horizontal flowline terminating into superfi- cial gas velocities ranging from 0.02 to 0.4 mls. See
the riser 22 (see Fig. 14). The fluids were naphtha and nitrogen. Refs. 23 and 24 for further details. The inclines are very steep,
The inlet liq- uid superficial velocity was kept constant at 1.08 resulting in up- sloping sections almost filled with oil and
mls, while the superficial gas velocity was increased from 1.0 to downsloping ones near- ly filled with gas. Because of the very
about 4.2 mls in a period of 20 seconds (Fig. 15). low velocities, the pressure drop is dominated by gravity in the
The increase in the gas flow rate results in a decrease in the upsloping parts, with a slight recovery in the downsloping parts.
liq- uid holdup. This discontinuity in liquid holdup tends to be Pressure-drop data are reported along the pipeline at four loca-
smeared out and broken up into slugs as it travels along the tions for a pressure at Lacq of 1.7 MPa and a flow rate of 700
pipeline. The OLGA model applies a mean-slug-flow m 3 /d, which has been assumed to represent a total mass flow of
description but clearly is 7.28 kg/s. In Table 1, the results from OLGA (Version 87.0) and

SPE Production Engineering, May 1991 177

 --4S7
Distance from mix ing point IU
010
--429
130 299 328
---60 7

Fig. 14--Teat section of the SINTEF Two-Phase Flow Laboratory for the dynamic Inlet ex- periments (e liquid holdu

*0

LIQUID
2.0
0 ***!!!!
5006fXl620 6t0660 6807IXI
TUE (S)

Fig. 15--Superficial-gae•velocIty recordings for the dynamic inlet-flow expedmenta at the SINTEF Two•Phase Flow Labo

Fig. 17--LlquId-holdup recordlnga In the deer compared wit

OLGA (--) at locations 7 and 28 m from the riser bottom.
PRESSURE I

t.0

550soo 6S0 800

Tll•iE i S )

Fig. 18--Absolute pressure recorded 10 m from the mixing

point compared with OLGA (--).
1.2
10- sure drop from PEPITE is even lower than the OLGA
0.8- values, or 28$ too low.
ttOUlD

0.6 . 'p The discrepancies were found to result from the terrain
0c- profile reported by Lagiére et al. being too coarse. Table 2
shows new, substan0ally improved results from OLGA based
$80 600 620 640 660 680 700
on a more de- tailed pipeline profile obtained from TOTAL.
Tot S)
Coneluclonc
he- 16••••Llquld-holdup recordings In the horlaontal compared The OLGA model was presented, with emphasis on the
with OLGA (--) at locatlone 49, 176, and 290 m from the particular two-fluid model applied and the flow-regime
mlx- description. The im- portance of including a separate droplet
field was discussed. Ne- glecting the droplet field in vertical
the steady-stete model PEPITE2* are compared with the annular flow was shown to overpredict pressure by 50a in
measured values using the terrain profile reported by Lagiére typical cases.
e2 @, 23 Closure laws are still, at best, semimechanistic and require
The pressure drop calculated by OLGA is 15$ too low, ex- perimental verification.
whereas the reported PEPITE calculations are extremely good. The OLGA model has been tested against experimental data
To check the input data, an available PEPITE version was run over a substantial range in geometrical scale (diameters from
with the same input data used in OLGA. As can be seen from 2.5 to 20 cm, some at 76 cm; pipeline length/diameter ratios
Table 1, the pres- up to 5,tXD;
178 SPE Production Engineering, May 1991
 Flg. 18—Pressure difference over a part of the horlzontal line compared with OLGA (—).

E 30.

20

Fig. 2AThe pressure dlfference over the rlaer compared with OGLA (—) (the pressure difference Is recalculated aa equiva- lent IIquId heig

and pipe inclinations of —15 to +90°), pressures from 100 kPa fi = slip ratio
to 10 MPa, and a variety of different fluids. Rp -- distribution slip ratio
The model gives reasonable results compared with transient
RS —— gas/oil mass ratio
data in most cases. The predicted flow maps and the frequencies
S —— perimeter, m
of ter- rain slugging compare very favorably with experiments.
Sf -- wened perimeter, Phase f, m
The model was also tested on a number of different oil and gas
field lines. The OLGA predictions are generally in good t -— time, seconds
agreement with the measurements. The Vic Bilh-Lacq at -- timestep, seconds
oil/associated-gas line was a good test case because of the 2’ = temperature, °C
extremely hilly terrain and low flow rates, which imply that, to U -— heat transfer per unit volume, I/m 3
obtain a correct pressure drop, the liquid holdup prediction must v = velocity, m/s
be very accurate. The pressure drop predicted by OLGA was v$ = velocity of large slug bubble, mls
6.85 MPa compared with a measured drop of 6.8 MPa. V F -— volumetric fractions (N=g, L, D)
The actual number of available field lines where the fluid com- z = length coordinate, m
position and line profile are sufficiently well documented for a A = mesh size, m
meaningful comparison is, however, still limited. Further verifi- a = angle with gravity vector, rad
cation of this type of two-phase flow models is clearly needed. Q = angle, rad
c = absolute roughness, m
k = friction coefficient
A = pipe cross-sectional area, rnz p = viscosity, kg/m- s
c = speed of sound, mls p -- density, kg/m°
C —- constant « = surface tension, N/m
Co = distribution slip parameter = mass-transfer term, kg/m*- s
d = diameter, m Subscripts
E —- internal energy per unit mass, J/kg ac = acceleration
Fp —- drag force, N/m° b -— bubble
F$ -- slug fraction, LS /{LB +L S ) d -- droplet deposition
g = gravitational constant, m/s2 D -— droplet
G = mass source, kg/s m° e —- droplet entrainment
â = height, m f —— Phase f(G,L,D1
hf —- average film thickness, m F -— friction
g gig
If = enthalpy, I/kg
/i = hydraulic
<s —— enthalpy from mass sources, I/kg
K -- coefficient, distribution slip parameter H -— horizontal
L -— length, m iii = hydraulic, interfacial
i = interfacial
£ = laminar
L -— liquid
r -— relative
N -— number
s = superficial
NFr — Froude number S -— slug
NR = Reynolds number
I = turbulent
p —- pressure, N/m2
Ap = pressure drop, N/m 2 K = vertical
u' = wave
 TABLE 1—COMPARISON OF MEASURED AND PREDICTED PRESSURE DROPS
ON THE VIC BILH-LACQ PIPELINE°

Length Measurement PEPITE OLGA

Section (km) (bar) Reported (bar) Actual (bar) (bar)
Vic Bilh-Claracq 8.7 18 16.9 15.5 17.7
Claracq-Morlanne 21.1 31 30.9 22.0 24.4
Morlanne-Lacq 14.0 19 18.4 12.7 15.9
Vic Bilh-Lacq 43.8 68 66.2 50.2 58.0
“Terrain profile from Lagiâre st e/. ”

SPE Production Engineering, May 1991 179

 10. Laurinat, J.E., Hanratty, T.1. , and Dallman, J.C.: “Pressure
$g.H. Bsndlkcen 1s Drops and Film Height Measurements for Annular Gas-Liquid
Flow,” Intl. 1. ñfufiiPhose f’/ow' (1984) 10, No. 3, 341-56.
head of the Dept.
for Fluid Flow and i I, Malnes, D.: Slug Flow in Venical, horizontal and Inclined Pipes,
IFE/KR/E-83/002 (1983).
Gas Technology at
IFE at K]eller, Nor- 12. Bendiksen, K.H.: “An Experimental Investigation of the Motion of
way. He holds an Long Bubbles in Inclined Tubes,” Intl. J. Multiphase Flow (1984)
MS degree In 10, No. 4, 467-83.
phys- ics and a t3. Malnes, D.: “Slip Relations and Momentum Equations in Two-
PhD degree In Phase Flow,” IFE Report EST-1, Kjeller (1979).
,fi, t4. £iregory, G.A. and Scott, D.S.: “Correlation of Liquid Slug
fluid mechanics
from the U. of Velocity and Frequency in Horizontal Co-Current Gas-Liquid Slug
Bendiksen Oslo, where he Is Flow, mCh Z. (1969) 15, No. 6, 933.
Malnes
also a professor of 15. Wallis, G.B.: One-Dimensional Two-Phase Flow, McGraw Hill
ap- plied fluid Book Co. Inc. , New York City (1969).
16. &ndiksen, K. ct ct. : “Two-Phase Flow Research at SINTEF and
mechan- ics. His IFE.
technical Interests
Include research,
develop- ment,
and promo- tion of
multlphase t ra n
spa rt at I an
technology In the
offshore all and
gas Industry. He
Nuland has been In
l¥loe
charge of the
OLGA devel-
opment project
TABLE 2—COMPARISON OF MEASURED AND PREDICTED
PRESSURE DROPS ON THE VIC BILH•LACQ PIPELINE
WITH NEW PROFILE

Section
Vic Bilh-Claracq
Claracq-Morlanne
Morlanne-Lacq
Vic Bilh-Lacq
alnce Ita etart In 19a0. Dag Matnoc, a senior research sclen• Theory,”
tlat at IFE, kas been working wlth two•phase flow, heat J. Basic Eng. (1970) 59.
trana• ter, and dynamlc simulation for 30 years. He holda 8. Andreussi, P.: “Droplet Transfer in Two-Phase Annular Flow,”
MS and Dr. Ing. degrees In mechanical engineering from the Intl. J. Multiphase Flow (1983) 9, No. 6, 697-713.
Nor- wegian Inst. of Technology. Randl Mao Is a aenlor 9. Dallman, J.C., Barclay, G.J. , and Hanratty, T.J.: “Interpretation of
physicist at IFE, working on physical and numerical Entrainments in Annular Gas-Liquid Flows,” Two-Phase Momentum,
modellng of mul- tlphaee flow. She holds an MS degree In fleet and ñf‹us Tranfer, F. Durst, G.V. Tsiklauri, and N.H. Afgan
fluid flow from the (xs.).
U. of Oslo. 6v•n Nuland la chief engineer at IFE, where
he Is worklng on multlphase flow problems, both
experiments and modeling. He has been working on the
OLGA project alnce 1980. He holda an MS degree from
the U. of Oalo.

IF = wall
0 = relative zero liquid flow at infinity t80

Referenesa
1. TRAC-PFI An Advanced Best Estimate Camputer Program for Pres-
surized Water Reactor Analysis, NUREG/CR-3567, LA-994-MS
(Feb. 1984).
2. RELAP5/MODI Code Manual Volume 1: System Models and
Numeri- cal ñfetfio&, NUREG/CR-1826, EG£i-2070 (March
1982).
3. MicaePi, J.C.: 1987 CATHARE an Advanced Best-Estimate Code for
PWR Safety Analysis, SETh/LEML-EM/87—58.
4. Bendiksen, K., Espedal, M., and Malnes, D.: “Physical and
Numeri- cal Simulation of Dynamic Two-Phase Flow in Pipelines
with Appli- cation to Existing Oil-Gas Field Lines,” paper
presented at the 1988 Conference on Multiphase Flow in
Industrial Plants, Bologna, Sept. 27—29.
5. Malnes, D. , Rasmussen, J., and Rasmussen, L.: A Snort
Description of the Blowdown Program, NORA SD-129, IPE,
Kjeller (1972).
6. Andreussi, P. and Persen, L.N.: “Stratified Gas-Liquid Flow in
Down- wardly Inclined Pipes,” Ann. J. ñfufripfiose fiiow (1987) 13,
565—75.
7. Wallis, G.B.: “Annular Two-Phase Flow, Part 1: A Simple
 Some Experimental Results and a Demonstration of the Dynamic
Two- Phase Flow Simulator OLGA,” paper presented at the 1986
Offshore Northern Seas Conference, Stavanger.
Barnea, D. et at.: “Flow Pattern Transition for Gas-Liquid Flow
in Horizontal and Inclined Pipes,” Intl. J. Multiphase Ftow
(1980) 6, 217—26.
t8. Crouzier, O.: “Ecoulements diphasiques gaz-liquides dams les
conduites faibleoient inclinées par rapport a l’horizontale,” MS
thesis, l’Univer-
sité Pierre et Marie Curre, Paris (1978).
19. Schmidt, Z., Brill, J.P. , and Beggs, H.D.: “Experimental Study
of Severe Slugging in Two-Phase-Flow Pipeline-Riser System,” fPEf
(Oct. 1980) 407—14.
20. Bendiksen, K., Malnes, D., and Nuland, S.: “Severe Slugging in
Two- Phase Flow Systems,” paper presented at the third Lecture
Series on Two-Phase Flow, Trondheim, IFE Report KR/E-6,
Kjeller (1982).
21. Norris, L. ct ct. : “Developments in the Simulation and Design of
Mul- tiphase Pipeline Systems,” paper SPE 14283 presented at the
1985 SPE Annual Technical Conference and Exhibition, Las Vegas,
Sept. 22—25.
22. Linga, H. and €btvang, D.: “Tabulated Data From Transient
Experi- ments With Naphtha,” SINTEF/IFE project, Report No. 41
(Feb. 1985).
23. Lagiére, M., Miniscloux, C., and Roux, A.: “Computer Two-
Phase Flow Model Predicts Pipeline Pressure and Temperature
Profiles,” lii ord M J. (April 1984) 82-92.
24. Corteville, J., Lagiére, M., and Roux, A.: “Designing and
Operating Two-Phase Oil and Gas Pipelines,” Prec. , llth World Pet.
Cong., Lon- don (1983) SP 10.

Metrlc Converclon Factors

bar x 1.0* E +05 = Pa
bbl x 1.589 873 E —01 =
m3
ft x 3.048* E —01 = m
ft2 x 9.290 304* E —02 = m2
n3 x 2.831 685 E —02 = m
3
°F (°F —32)/1.8 = °C
in. X 2.54* E +£D = cm
1bf x 4.448 222 E +£D = N
lbm x 4.535 924 E —01 = kg
fioneaonCaowizegx SPEPE
Original SPE manuscript reoelved for review Jan. 4, 1909. Paper (SPE t0451) accepted
for publication Dec. 11, 1990. Re•1eed manuscript received April 30, 1900.

SPE Production Engineering. May l99t