International Journal of Multiphase Flow 67 (2014) 37–51

Contents lists available at ScienceDirect

International Journal of Multiphase Flow

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / i j m u l fl o w

Investigation of two-phase flow pattern, liquid holdup and pressure drop

in viscous oil–gas flow
Hatef A. Khaledi a, Ivar Eskerud Smith b, Tor Erling Unander b, Jan Nossen c,⇑
a
SPT Group Norway AS, P.O. Box 113, N-2027 Kjeller, Norway
b
SINTEF Petroleumsforskning AS, Dept. of Wellstream Technology, P.O. Box 4763 Sluppen, 7465 Trondheim, Norway
c
Institute for Energy Technology (IFE), P.O. Box 40, 2027 Kjeller, Norway

a r t i c l e i n f o a b s t r a c t

Article history: An experimental investigation was carried out on viscous oil–gas flow characteristics in a 69 mm internal
Received 14 October 2013 diameter pipe. Two-phase flow patterns were determined from holdup time-traces and videos of the flow
Received in revised form 3 June 2014 field in a transparent section of the pipe, in which synthetic commercial oils (32 and 100 cP) and sulfur
Accepted 18 July 2014
hexafluoride gas (SF6) were fed at oil superficial velocities from 0.04 to 3 m/s and gas superficial velocities
Available online 21 August 2014
from 0.0075 to 3 m/s.
Ó 2014 Elsevier Ltd. All rights reserved.
Keywords:
Two-phase flow
Gas–oil flow
Holdup
Pressure drop
Flow pattern
Point model

Introduction an empirical correlation for gas void fraction in the slug body.
Andritsos and Hanratty (1987) performed experiments in 10 m
Multiphase flow simulators are important tools used in the long horizontal pipes with inner diameters 2.54 and 9.53 cm, using
design and operation of oil and gas fields. The flow regime (flow air and liquids with viscosities of 1, 4.5, 16 and 70 cP. They pro-
pattern), liquid holdup and pressure drop in pipelines are of great posed a new correlation for the interfacial friction factor.
interest for design of pipelines and equipment. The current state of Andritsos et al. (1989) performed experiments in a 25 m long hor-
the art multiphase flow simulators are mainly one-dimensional izontal pipe with inner diameter 9.53 cm, using water/glycerine
models, which require empirically based correlations for parame- mixtures with viscosities of 1, 20, and 100 cP. The study was
ters like slug velocity, void fraction in slugs and interfacial friction focused on the stratified to slug transition at different viscosities.
factors. Nädler and Mewes (1995) investigated the effect of liquid viscosity
High viscosity oils are among the most important future hydro- on the phase distribution of slug flow in a horizontal pipe with an
carbon fuels due to the increasing world energy demand and the inner diameter of 59 mm. Air as the gas phase, and water and oil
depletion of conventional oil resources. Since most of the closure with viscosities in the range of 14–37 cP as the liquid phases were
laws implemented in the commercial simulators are based on used. Multi-detector gamma densitometers were used to measure
experiments using low viscosity oils, experiments using oils with the liquid holdup, which was found to increase with increasing
higher viscosities are crucial to improve the existing closure laws. liquid viscosity. Gokcal (2005, 2008) and Kora et al. (2011) per-
Among the earliest studies using oils with a high viscosity are the formed air–oil experiments, using a 50.8 mm ID horizontal pipe.
experiments performed by Kago et al. (1986), who conducted slug The oil viscosity was varied by varying the temperature, and ran-
flow experiments in a 51.5 mm ID horizontal pipe. Water with ged from about 200 cP to 600 cP. Gokcal proposed new correlations
polymers and water slurries were chosen as the liquid phases for the Taylor bubble drift velocity and the slug frequency based on
and air was chosen as the gas phase. The viscosity of the liquid his experiments, while Kora proposed a new correlation for the
phase ranged from 0.8 cP (water) to 55 cP (slurry). They proposed void fraction in slugs. Smith et al. (2011) performed experiments
in a 69 mm ID horizontal pipe in a 52 m long test section. 2 cP
and 100 cP oils were used, and the gas phase was SF6 at about
⇑ Corresponding author. Tel.: +47 63 80 61 99. 8 bara pressure. The slug flow region was found to be much smaller
E-mail address: [email protected] (J. Nossen). than in the experiments of Gokcal and Kora, most likely due to the

http://dx.doi.org/10.1016/j.ijmultiphaseflow.2014.07.006
0301-9322/Ó 2014 Elsevier Ltd. All rights reserved.
38 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51

higher gas density. The current commercial multiphase flow simu- let was connected to a DN 200 vertical pipe (downcomer) that con-
lators were also found to significantly overpredict the pressure nected to the inlet of the separator 4 m below. A separate DN 50
drop. Jeyachandra et al. (2012) recently performed more experi- line connected the separator’s gas layer to the top of the downcom-
ments in the same flow loop as Gokcal and Kora, and proposed a er. This way, pressure fluctuations in the downcomer due to any
new correlation for the slug drift velocity as function of viscosity siphoning effects were minimized.
and pipe diameter. For further studies involving high viscosity liq- To enable visual observation, two transparent pipe sections
uids see for example the studies reported by McNeil and Stuart were added to the test section. One section was installed between
(2003), Schmidt et al. (2008), Valle (2000). Most of the above men- 14.0 and 14.8 m and one section was installed between 34.8 and
tioned studies used low pressure air or nitrogen as the gas phase, 36.8 m. The sections were made of polycarbonate pipe with OD
which results in a gas density significantly lower than that of high 75 mm and ID 69 m.
pressure natural gas in most industrial cases. In general there is a The hydraulic roughness of the test section was determined to
lack of available experimental data using high viscosity oils, partic- be 7 lm from single phase experiments.
ularly in combination with a high density gas phase.
The focus of this experimental campaign was to provide new Fluid system
experimental data for a two-phase gas–oil pipe flow using SF6 as
gas phase, and both a medium (nominal viscosity 32 cP) and high Sulfur hexafluoride (SF6) and synthetic oils were used as test
(nominal viscosity 100 cP) viscosity oil. For each oil system, exper- fluids. The experiments were conducted with Nexbase 3080, which
iments were run at both 4 and 8 bara pressure, resulting in gas was later diluted with Exxsol D80 to obtain a lower viscosity mix
densities of approximately 25 and 50 kg/m3 respectively. It is fluid. Exxsol D80 oil is a dearomatized aliphatic hydrocarbon oil
worth mentioning that the density of SF6 at a pressure of 8 bara commonly used in flow loop experiments due to its non-flamma-
is equal to that of methane at 71 bara, a value highly relevant for ble, nontoxic and transparent properties. Nexbase 3080 is a cata-
multiphase hydrocarbon transport. The focus areas have been bub- lytically hydroisomerized and dewaxed base oil comprising of
bly flow, the stratified to slug flow transition, and slug flow. The hydrogenated, highly isoparaffinic hydrocarbons. Nexbase 3080
obtained experimental results are compared with the predictions was chosen since it is non-flammable, transparent, and has a high
from a two-phase gas–liquid point-model, which was also devel- viscosity at room temperature. The physical properties of the oils
oped as part of the study. This model was presented in another are shown in Table 1.
publication by Smith et al. (2013). The viscosities of the oils were measured using a Micromotion
The paper is organized as follows. Following the introductory Visconic 7829 viscosimeter connected to a bypass on the oil-feed
section, in which earlier investigations are summarized, the pres- line. The viscosity values were recorded regularly together with
ent experimental procedure is presented in Section ‘Experimental temperature in the viscosimeter. The viscosity models based on
procedure’. The test section, fluid systems, instrumentation, data these measurements for the 32 cP mix fluid and 100 cP Nexbase
sampling and uncertainties are presented and discussed in detail. 3080 oil are shown in Eqs. (1) and (2). For Nexbase 3080:
The two-phase gas–liquid point-model is briefly reviewed in Sec-
tion ‘Two-fluid gas–liquid model’, with details referred to Smith
lo ¼ 196  5T ð1Þ
et al. (2013). The observed flow patterns, pressure drop, liquid and for the 32 cP mix:
holdup and other flow characteristics and experimental results,
in combination with model comparisons are shown and explained lo ¼ 59:5  1:39T ð2Þ
before conclusions are drawn in Section 4. The resulting viscosities from these models are in cP, and the
parameter T is the temperature in °C.
Experimental procedure The oil density was measured by manually recording the value
shown by the coriolis flow meter. To measure the gas density, a
Test section representative gas sample was taken from the loop using a 1 L
sample flask. The mass of the gas sample (weight of filled bottle
A set of two-phase gas–oil experiments was carried out in SIN- – weight of empty bottle), ms, was noted together with sample
TEF’s Medium Scale Flow Loop at Tiller, Norway. The experiments pressure ps and temperature Ts (in Kelvin). The gas density for a
were performed in a straight 51.8 m long horizontal test section, certain pressure p and temperature T are calculated using extrapo-
with an inner diameter of 69 mm. lation based on the ideal gas law. A density parameter b is calcu-
The test section was constructed by 5 m steel pipe section sec- lated from the measurements,
tions connected with flanges. The pipe sections were made of stain-
less steel 316 L with dimensional tolerances according to EN ISO ms T s
b¼ ð3Þ
1127/D3/T3. These tolerances imply that inner diameter was V s ps
ID = 69 ± 0.5 mm. The flange connections were standard EN 1092-
where Vs is volume of the sample flask. Density at a pressure p and
1 DN65 PN16 weld neck flanges (Type 11). Special precautions were
temperature T close to the pressure and temperature of the mea-
taken to ensure alignment of the flanges, and 1 mm thin intra flange
surement are calculated using
gaskets were used to minimize the spacing between the flanges.
Oil and gas were mixed in a full diameter T-junction. The gas p
qðp; TÞ ¼ b ð4Þ
entered through the branch connection, and this connection was T
oriented upwards. The test section positions are relative to the cen- The density parameter was measured regularly, at least once for
ter of this T-junction. For low gas rates (<8 m3/h) gas was injected each change in nominal pressure and temperature. The measure-
through a 3=4 ’’ nozzle on the side of the test section. The nozzle was ments show that there is no significant change in beta over the
located at 3.0 m. duration of the campaign and for all experiments in this campaign
The outlet of the test section was connected to a 4 m long hose. an average beta value of 1757 K kg/m3/bara has been used.
The inlet and outlet of this hose were at the same elevation, but the The ranges of flow rates employed are given in Table 2. Based on
middle of the hose was lifted one pipe diameter. This was done to the superficial oil velocity, the oil-phase Reynolds number
avoid gravitational draining of the test section at low flow rates.  
Reo ¼ qolUso D varies from 25 to 5000.
The nominal inner diameter of the hose was 75 mm. The hose out- o
 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51 39

Table 1 time-series and a subset of the analysis results were stored in a

Oil density and viscosity at the two nominal line pressures, and temperature 20 °C. database at the flow laboratory.
Oil Density, nominal Density, nominal Viscosity (cP) The model for the uncertainty is based on the method recom-
pressure = 4 bar pressure = 8 bar mended by the Guide to the Expression of Uncertainty in
(kg/m3) (kg/m3) Measurement (International Organization for Standardization,
32 cP mix fluid 845 844 32 1993). The measurement uncertainties of the main instruments
Nexbase 3080 850 846 96 are listed in form of standard deviation in Table 3.

Pressure gradient uncertainty

Table 2 The uncertainty in the calculated pressure gradient was esti-
Range of nominal gas and liquid flow rates. mated from the deviation of the individual pressure measurements
Max Min from a linear pressure gradient. To obtain values for the best mea-
m/s m3/h m/s m3/h surement capacity this was done in single phase flow to avoid the
larger natural variance in multiphase flow. For the current setup,
Gas 3 40 0.0075 0.1
Oil 3 40 0.04 0.56
the uncertainty is the largest of 3 Pa/m and 0.5% of the measured
value. (Note that this model describes the uncertainty as one stan-
dard deviation.) This model is shown together with uncertainty
Instrumentation estimates from the single phase experiments with the 32 cP oil
in Fig. 1 These estimates cover the uncertainty due to the uncer-
Both gas and oil flow rates were measured using coriolis flow tainty in the pressure measurements; the inhomogeneity in the
meters. Six differential pressure cells were used to calculate the pipe (diameter variations, obstructions, etc.) and the natural scat-
pressure drop through the test section. The dp-cells were con- ter in pressure drop. In particular, note that experiments from
nected in one end to pressure taps located at the side of the pipe, the transition region between laminar and turbulent flow show
and to a reference-line in the other. This way, the absolute pressure excess scatter in the pressure gradient estimate. This transition
p at one cell can be calculated by: occurs around 400 Pa/m.
p ¼ dp þ pref ð5Þ To quantify the reproducibility in multiphase flow a different
analysis must be done. Fig. 2 shows the deviations in pressure gra-
where dp and pref are the recorded differential and reference pres- dient in groups of experiments with identical conditions (flow
sure. Along the test section four narrow-beam gamma densitometers rates, viscosity, pressure, etc. within 4%). The deviation has been
were used. Narrow-beam gamma densitometers can be used to mea- calculated as follows: One of the experiments in each group is cho-
sure the liquid line fraction in two-phase flows. Using appropriate sen as reference, and the deviations for this group are then set to be
assumptions regarding the geometry of the liquid, the liquid line the deviation between the reference experiment and the others in
fraction can be used to calculate the liquid holdup. In addition, a tra- the group, subtracted by the similar deviations from LedaFlow
versing gamma densitometer was used to measure the vertical time- simulations (Kongsberg Oil and Gas Technologies, 2011) of the
averaged density profile in the pipe. This was done by vertically tra- same experiments. By subtracting the deviations from the
versing a gamma source on one side of the pipe and a gamma photon LedaFlow simulation one removes any differences due to slightly
detector on the other side. There was also one broad beam gamma different flow rates, viscosities, etc.
densitometer installed in the test section. The broad beam gamma Fig. 2 shows that the scatter in the pressure gradient is mostly
densitometer has a source that is collimated into a fan shaped beam within ±10 Pa/m, though larger deviations are encountered in
of photons, so that it covers the entire pipe cross section. Hence, this some cases, particularly in slug flow. Though larger deviations up
instrument can be used to measure the oil volume fraction directly. to 40 Pa/m might occur, the relative scatter is still within 2% for
Note however, the reading of this instrument is also dependent on most experiments, and within 4% for the largest deviations.
fluid distribution, and a similar assumption about liquid distribution
must be used as for narrow beam instruments. Holdup uncertainty
The traversing gamma, the broad beam gamma and the narrow The holdup values reported in this article are either the values
beam gammas can all give a measure of the liquid holdup in the from the broad beam gamma or from one of the narrow beam gam-
pipe. Inasmuch as, in most experiments, the traversing gamma mas, depending on which measurement was available. The typical
was stationary and positioned near the top of the pipe (to help dis- deviation from the actual holdup reported here is based on experi-
tinguish slugs from large waves), it is not used to report the holdup ence, and is approximately ±0.025 in absolute value for both the
in this study. The holdup values in this article are set to be the broad beam gamma and the narrow beam gammas. Note that this
holdup measured by the broad beam gamma, and for experiments is the typical deviation from actual holdup, and that larger devia-
with no broad beam gamma measurement, the holdups from the tions might occur in some cases. Fig. 3 shows the holdup from
narrow beam gammas are used instead.
Videos of each experiment were recorded using a high-speed
video camera. The camera was a Weinberger SpeedCam Minivis Table 3
Uncertainty in the measurements.
camera with maximum resolution of 512  512 pixels in black-
and-white. At this resolution, the camera is able to record videos Measurement Uncertainty
up 2500 frame/s (fps). These videos together with line fraction Pipe diameter ±0.5 mm
time traces were used to determine the flow regimes. Straightness of loop (per measured 1 m) ±2 mm
Pressure gradient ±10–40 Pa/m
Holdup ±0.025
Data sampling and uncertainties Oil mass flow rate 0.035%
Gas mass flow rate 0.4%
All instruments were sampled at either Fs = 50, 100 or 200 Hz, Gas density 1.6%
depending on the gas and liquid phase velocities. During the post Oil density 0.2 kg/m3
Temperature 0.21 °C
processing, ‘‘slow’’ instruments like temperature sensors, flow
Pressure 0.06–0.12%
meters and pressure sensors were downsampled to Fs = 2 Hz. All
40 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51

1 Transitional
Stratified
0.9 Stratified wavy
Slug
0.8 Bubble
Plug flow

Narrow beam gamma [−]

0.7 Roll−waves

0.6

0.5

0.4

0.3

0.2

0.1

Fig. 1. Estimated pressure gradient uncertainty in single phase experiments with 0

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
32 cP oil.
Broad beam gamma [−]

the broad beam gamma compared to the holdup from the narrow Fig. 3. Comparison of holdup measurements from the broad beam gamma and the
beam gamma. narrow beam gamma. Note that a stratified flow geometry is assumed when
converting the narrow beam line fraction to holdup. The red dotted lines show a
Like for the pressure gradient measurements, Fig. 4 shows the 10% error margin. (For interpretation of the references to color in this figure legend,
deviations in holdup in groups of experiments with identical con- the reader is referred to the web version of this article.)
ditions (flow rates, viscosity and pressure within 4%). The deviation
has been calculated the same way as pressure gradient uncertainty
was presented. Fig. 4 indicates that the assumed scatter in holdup
of ±0.025 is reasonable, though larger deviations might occur. equations describing two-phase stratified flow of gas and liquid.
A unit-cell model is applied for slug flow with the assumption of
Two-fluid gas–liquid model stratified and bubbly flows as special cases with slug to bubble
fractions of zero and unity respectively. The schematic description
A two-fluid model for gas–liquid flow has been developed, with of slug flow is depicted in Fig. 5.
focus on improving the predictions of flow regime, pressure drop The model assumes fully developed flow, which is periodical in
and holdup in two-phase gas–oil flows with high viscosity oils. a frame of reference moving with the bubble nose velocity Ub. The
The model consists of a standard stratified model, and a unit-cell slug front and the bubble nose are thus assumed to have the same
model for slug flow. Thus, the model can be used to simulate strat- velocity. The mass flow of each phase is constant along the pipe,
ified flow, slug flow and bubbly flow. The model has been imple- and by also assuming constant densities along the pipe the conti-
mented in Matlab. In the following subsections the model and nuity equations for the volume flux in the slug and bubble regions
implemented empirical correlations are presented. read:

as U gs þ ð1  as ÞU ls ¼ U m
Conservation of mass and momentum for the stratified model ð6Þ
ab U gb þ ð1  ab ÞU lb ¼ U m
The model presented in this section is referred to as a two-fluid where as and ab are the void fractions in the slug and bubble
model. The model consists of mass and momentum conservation regions, Uls and Ugs are liquid and gas velocities in the slug, and

Fig. 2. Repeatability of pressure drop measurements, the pressure drop is plotted against the maximum deviation in groups of experiments with almost identical conditions
(flow rates, viscosity and pressure within 4%).
 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51 41

Fig. 4. Repeatability of holdup measurements. The holdup is plotted against the maximum deviation in groups of experiments with almost identical conditions (flow rates,
viscosity and pressure within 4% deviation).

Fig. 5. Schematic description of slug flow in a circular pipe.

Ulb and Ugb are liquid and gas velocities in the bubble region. Conti-
nuity of each phase across the bubble nose gives:
as ðU gs  U b Þ ¼ ab ðU gb  U b Þ
ð7Þ
ð1  as ÞðU ls  U b Þ ¼ ð1  ab ÞðU lb  U b Þ Fig. 6. Stratified flow in an inclined pipe.

where Ub is the velocity of the bubble nose. The empirical correla-

tions used in our model for the slug void fraction and bubble nose
velocity are described in Sections ‘Void fraction in slug’ and ‘Bubble Table 4
Description of flow regimes.
nose velocity’ respectively. Finally one can combine Eqs. (6) and (7)
and solve for the superficial velocities in the bubble region. The Name Description
superficial velocities in the slug and bubble regions thus read: Transitional Unstable flow, switching between stratified and slug.
Only a few experiments behaved like this
U sgs ¼ as U m Stratified smooth The gas and liquid are separated by a undisturbed
U sls ¼ ð1  as ÞU m smooth interface
ð8Þ Stratified wavy The gas and liquid are separated by a clear interface,
U sgb ¼ U sgs þ U b ðab  as Þ with waves. The waves do not reach the top of the
U slb ¼ U m  U sgb pipe
Slug flow The flow consists of alternating liquid plugs that fills
The slug fraction Sf which describes the ratio of the lengths of the the pipe completely, and gas Taylor bubbles
bubble and slug regions can then be calculated by: Bubbly flow/ All of the gas is transported dispersed in the liquid
dispersed bubble phase
U sg  U sgb flow
Sf ¼ ð9Þ Plug flow/elongated Elongated gas bubbles at the top of the pipe are
U sgs  U sgb
bubble flow transported by the liquid. The diameter of the
Fig. 6 shows an idealised stratified gas and liquid in stratified flow bubbles is much smaller than the pipe diameter
Roll-waves Large waves occur which are too large for the flow to
with geometrical and physical measures.
be characterized as stratified wavy, but too small to
Further, the steady state momentum equation in the bubble be characterized as slug flow
region can be expressed by Eq. (10). Assuming constant holdup,
equal pressures in both phases and neglecting acceleration and
convection terms gives:
where @p
@x
is the pressure gradient, and h is the inclination relative to
@p horizontal (positive for upward inclination), A is the phase cross-
Al þ swl Sl  si Si  ql Al g sin h ¼ 0
@x ð10Þ sectional area, s is the shear stress, S are the perimeters over which
@p the shear stresses act, and q are the densities. The subscripts i, l and
Ag þ swg Sg þ si Si  qg Ag g sin h ¼ 0
@x g represents the interface, liquid and gas respectively.
42 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51

Fig. 7. Video snapshots of different flow regimes. The upper part of the figures shows the top view while the lower part depicts the side view of the pipe, (a) stratified (b)
stratified wavy (c) bubbly flow (d) plug flow (tail) (e) roll-wave (f) slug flow (slug front).

To obtain an equation for the pressure gradient, the momentum By eliminating the pressure gradient @p@x
between the two momen-
equations for two phases are added to eliminate the gas–liquid tum balances in Eq. (10), an implicit equation for the liquid holdup
 
interfacial momentum term, Hl ¼ AAl is obtained,
 
@p Sg Sl Sl Sg 1 1
 ¼ swg þ swl þ ðHl ql þ ag qg Þg sin h ð11Þ f ðHl Þ ¼ swl  swg  swi Si þ þ Dqg sin h ¼ 0 ð12Þ
@x A A Al Ag Al Ag
 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51 43

Fig. 8. Flow regime map for experiments and the point-model for the 32 cP liquid and 4 bara pressure. (a) Experiment and (b) point-model.

The wall shear stress for the phase p (p = g, l) in question is defined as for the liquid and gas phase respectively. In the transitional region
between laminar and turbulent flow, we interpolate the friction fac-
1
sp ¼  f p qp U 2p ð13Þ tor between the upper limit for laminar flow (Rel) and fully turbu-
2
lent flow (Ret):
where Up is the mean velocity. The wall friction factors are calcu-
lated using the Fanning and Haaland formulas (White, 2011): Rep  Rel
w¼
"  1:11 # Ret  Rel ð16Þ
1=2 6:9 e=Dhp f p ¼ wft þ ð1  wÞf l
fp ¼ 3:6log10 þ Rep > 4000
Rep 3:7
ð14Þ In our model we have chosen Rel = 1700, and Ret = 4000. Similarly,
16
fp ¼ Rep < 1700 the interfacial shear stress is defined as:
Rep
1
Here, Rep is the Reynolds number and e is the hydraulic roughness. si ¼ f i qg ðU g  U l ÞjU g  U l j ð17Þ
2
We define the hydraulic diameters as
As interfacial friction factor we use a modified version of the corre-
4Al 4Ag
Dhl ¼ ; Dhg ¼ ð15Þ lation by Andreussi and Persen (1987), where fi0 is equal to the gas-
Sl Sg þ Si
wall friction factor with zero roughness;
44 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51

Fig. 9. Flow regime map for experiments and the point-model for the 32 cP liquid and 8 bara pressure. (a) Experiment and (b) point-model.

dAl
!12 the gas phase is modified by a factor of up to 10. To make a contin-
qg dH
F ¼ Ug l uous transition from low viscosity to high viscosity, the wall-rough-
ðql  qg ÞAg g cosðhÞ ness for the gas-phase is modified as follows:
 
fi Hl ð18Þ 0  2 1
¼ 1 þ 20ðF  F 0 Þ0:67 F > F0 l0
f i0 D 100
B l C
fi eg B
¼ e0 @1 þ 9e C ð19Þ
¼1 F 6 F0 A
f i0

where Hl is the liquid film height. The coefficient 29.7 in the work of with l0 = 0.00175 Pa s.
Andreussi and Persen (1987) has been changed to 20 in the present
work Another difference is that, similar to the correlation proposed Average liquid holdup in slug flow
by Wallis (1969), we use a linear relationship with the line fraction
 0:2 In slug flow regimes the average holdup is calculated as the
Hl
D
, instead of HDl as proposed by Andreussi and Persen. To weighted sum of holdup in the bubble and slug parts,
account for the significant liquid film on the wall as observed in
our experiments with high viscosity oils, the wall-roughness in H ¼ Sf ð1  as Þ þ ð1  Sf Þð1  ab Þ ð20Þ
 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51 45

Fig. 10. Flow regime map for experiments and the point-model for the 100 cP liquid and 4 bara pressure. (a) Experiment and (b) point-model.

The holdup in the bubble section (1  ab) is calculated by solv- where k = (FrgFrv)0.2.
ing the holdup equation in the steady state model for stratified
flow as described in Section ‘Conservation of mass and momentum Pressure drop in slug flow
for the stratified model’. The void fraction in slug as is found by the
correlation proposed by Kora et al. (2011). This correlation is Pressure drop calculation in slug flow is divided into separate
described in Section ‘Void fraction in slug’. The slug fraction is also modeling and calculation of pressure difference in each parts of a
calculated from the continuity equations in the unit cell model. slug unit. The average pressure drop is found by weighting the con-
tributions of each part by using the slug fraction. The total pressure
Void fraction in slug drop across a slug unit is
Dp ¼ Dpslug þ Dpbubble ð22Þ
For the liquid holdup in the liquid slug, we use the correlation
proposed by Kora et al. (2011): The pressure drop in the bubble region can be calculated from Eq.
(11),
Hls ¼ 0:9473e0:041k k P 1:5
 
Hls ¼ 1:012e 0:085k
0:15 < k < 1:5 ð21Þ @p Sg Sl ðql Al þ qg Ag Þg sin h
 ¼ swg þ swl þ ð23Þ
@x bubble A A A
Hls ¼ 0 k 6 0:15
46 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51

Fig. 11. Flow regime map for experiments and the point-model for the 100 cP liquid and 8 bara pressure. (a) Experiment and (b) point-model.

The slug body is modeled as dispersed bubble flow. The pres- where Frg is the gravitational Froude number. For continuity we
sure drop equation for the slug region reads use the maximum value of U0 obtained for high and low Froude
  number. The constant Frv = 0.351, while constant Frh is normally
@p S
 ¼ swl þ qs g sin h ð24Þ set to be 0.542, a value calculated by Benjamin (1968). However
@x slug A
when Benjamin obtained the value of 0.542, the viscosity was
where qs = qgas + ql(1  as) is the gas–liquid mixture density. not taken into account in his calculations. Jeyachandra et al.
(2012) recently measured the slug drift velocity using different
Bubble nose velocity oil viscosities, diameters and inclinations, and developed a correla-
tion where the horizontal drift velocity decreases exponentially
For the bubble-nose velocity Ub in slug flow, we use a modified with the inverse of the Archimedes and Eötvös numbers, defined
version of the correlation proposed by Bendiksen (1984) of the by
form Ub = C0Umix + U0: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffi
2
U b ¼ ðC 0 þ 0:15sin ðhÞÞU mix þ ðF rv sinðhÞ þ F rh cosðhÞÞ gD F rg < 3:5 D1:5 ql ðql  qg Þg
pffiffiffiffiffiffi Ar ¼ ð26Þ
U b ¼ C 0 U mix þ F rv sinðhÞ gD F rg P 3:5 ll
ð25Þ and
 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51 47

Fig. 12. Holdup comparison between the experiment and the point model, the 100 cP liquid and 8 bara pressure.

€¼
ql gD2 whereas the lower part shows the side view (a mirror was placed
Eo ð27Þ above the pipe at a 45° angle). At moderately low oil to gas super-
rgl
ficial velocity ratios (Usl/Usg) the fluids are separated into distinct
respectively. The drift velocity thus depends on the ratio of gravita- layers, with the lighter gas phase flowing smoothly above the
tional and viscous forces, and also on the surface tension forces for much heavier oil phase. This flow regime is called stratified
smaller diameters. We thus apply the following modification to the smooth, and is shown in Fig. 7(a). As the superficial gas velocity
constant Frh in the drift velocity: is increased waves start forming on the oil surface. Stratified wavy
0:46 €0:1 Þ
Eo flow, which is depicted in Fig. 7(b), includes a range of different
F rh ¼ 0:53eð13:7Ar ð28Þ
wave patterns. Further increase in the gas velocity leads to the for-
This model will give a value close to that of Benjamin for low mation of larger waves rolling over the interface, almost filling the
viscosity oils, and lower values with increasing oil viscosity. entire pipe cross-section. A video snapshot of the roll-wave flow
Following Nuland (1998). The value of C0 has been set to 2 for pattern, which basically was seen at higher oil rates, is shown in
laminar flow, and 1.2 for fully turbulent flow. In the transitional Fig. 7(e). If the oil flow rate is much larger than the gas rate, the
region between laminar and turbulent flow, we use an interpola- gas tends to disperse into the oil phase as fine bubbles. Fig. 7(c)
tion method proposed by Nuland (1998): shows a video snapshot of a dispersed bubble flow where the indi-
vidual bubbles can be clearly seen. For somewhat lower oil flow
Res  Rel
w¼ rates the gas void fraction increases. The bubbles get closer to each
Ret  Rel ð29Þ
other and eventually merge to form Taylor bubbles. A video snap-
f p ¼ wC 0t þ ð1  wÞC 0l shot of a slug flow case can be seen in Fig. 7(f). Slug flows are inter-
where Res is the Reynolds number in the liquid slug. Based on our mittent in nature and in most cases a high mixing of oil and gas
experimental data, we choose Rel = 500, and Ret = 3000. The coeffi- occurs especially at the slug front. Fig. 7(d) shows the nose of an
cient C0l = 2, while C0t is defined as follows: elongated bubble. In this flow regime, which is referred to as plug
flow in this article, most of the gas moves as elongated bubbles
ðn þ 1Þð2n þ 1Þ 1 being transported by the oil phase. The diameter of the bubbles
C 0t ¼ ; n ¼ pffiffiffiffi ð30Þ
2n2 fs is smaller than the pipe diameter and the liquid phase flows con-
tinuously below the gas phase. Compared to slug flow, plug flows
The friction factor fs is the friction factor in the liquid slug, and
are less turbulent and the level of mixing is much lower.
we set a lower limit on C0t at 1.2.
In modeling gas–liquid two-phase flow the concept of flow
regime has been used to characterize the interfacial structure of
Results and discussion the flow field. Nearly all correlations and relations that provide
some flow characteristics data in two-phase flows are flow regime
Flow pattern dependent. Therefore, predicting the correct flow regime is essen-
tial for properly calculating the flow features such as pressure drop
As mentioned, the video recordings together with the time and liquid holdup. The flow regime prediction in the current flow
traces from the gamma densitometers were used to determine model is based on the slug fraction calculation
the flow regimes. Several flow patterns were observed at different  
U U
SF ¼ U sg UsgBubble . According to this criterion, bubbly flow is pre-
flow conditions. The different flow regimes used in this campaign sgSlug sgBubble

are listed in Table 4. dicted when the slug fraction is larger than or equal to unity, and
Fig. 7 show snapshots from the video recordings of different stratified flow when the slug fraction is smaller than or equal to
flow regimes observed in this experimental campaign. The upper zero. Similarly, the flow is predicted to be slug flow when the slug
part of each figure represents the top view of the cross section fraction is between zero and unity.
48 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51

Fig. 13. Holdup comparison between the experiment and the point model, the 100 cP liquid and 4 bara pressure.

Fig. 14. Holdup comparison between the experiment and the point model, the 32 cP liquid and 8 bara pressure.

Figs. 8 and 9 show the flow regime map for the experiments, dicted relatively well for the experiments at 4 bara. However,
together with the predicted flow regimes from the point-model for the experiments at 8 bara the model predicts some of the
for the 35 cP oil. The flow regime is in general well predicted, espe- stratified experiments to be slug flow, and some of the slug flow
cially for the experiments at 4 bara. For the experiments at 8 bara experiments to be in the stratified flow regime. It can also be
the model predicts too much slug flow for Usg larger than 1.5 m/s seen that there is significantly less slug flow observed in the
and low Usl, and also predicts some stratified experiments to be experiments performed at 8 bara due to the increased gas den-
slug flow at low superficial velocities. Most of the experiments cat- sity. The plug flow experiments are again mainly predicted to
egorized as roll-waves are predicted as slug flow. In addition most be in bubbly flow, and the roll-wave experiments are predicted
of the plug flow experiments are predicted to be in the bubbly flow as slug flow. One should also have in mind that there might be
regime. Due to a higher gas density giving more efficient liquid uncertainties in the experimental flow regimes, especially in
transport in stratified flow it can be seen that the slug flow region the transitional regions from one flow regime to another.
is somewhat smaller in the 8 bara experiments compared to the Although we believe that the presented experimental flow
experiments at 4 bara. regimes are quite accurate, the determination of the flow regimes
Figs. 10 and 11 show the flow regime from the point model in the experiments was done by manual inspection, which will
compared to the flow regime obtained from the 100 cP oil exper- be somewhat subjective and might differ slightly from person
iments. As for the 32 cP experiments, the flow regime is pre- to person.
 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51 49

Fig. 15. Holdup comparison between the experiment and the point model, the 32 cP liquid and 4 bara pressure.

Fig. 16. Pressure drop comparison between the experiment and the point model, the 100 cP liquid and 8 bara pressure.

Holdup measurements pressure drop. In general, the pressure drop is predicted well,
though for some experiments the pressure drop is underpredicted.
In this section we present the holdup predictions from the It can be seen in Figs. 18 and 19 that the pressure drop is underpre-
point-model compared with the experimentally measured holdup. dicted in some experiments with the 32 cP oil. The reason is that
The measured liquid holdups in the current experiments are plot- the model predicts laminar or transitional flow in the slugs for
ted against the calculated data at different oil viscosity and line these experiments, and thus a coefficient C0 for the slug velocity
pressure in Figs. 12–15. Fig. 12 shows the holdup comparison at between 1.2 and 2. The slug velocities calculated from a cross-cor-
nominal oil viscosity of 100 cP. It can be seen that the holdup relation of the adjacent narrow beam gamma densitometers on the
obtained from the point-model at these flow conditions compares other hand follows the fully turbulent version of the Bendiksen
quite well with the experimental data as 99% of the experiments fit correlation in Eq. (20) in Smith et al. (2013) with C0 = 1.2, though
within 20% deviation. A similar trend can be seen for the holdup the Reynolds numbers in the slug become as low as approximately
comparison at ll = 32 and pressure = 4 and 8 bara, where the 800. By using the maximum value of the Hagen-Poiseuille and Haa-
holdup is also predicted quite well (see Figs. 14 and 15). land friction factors instead of the interpolation method shown in
Eq. (14) in Smith et al. (2013) for the friction factor in the slug, and
Pressure drop measurements forcing the coefficient C0 to be 1.2 (turbulent), the predicted pres-
sure drop is in better agreement with the 32 cP experiments. The
Figs. 16–19 show the comparisons between the point-model agreement with the 100 cP experiments on the other hand
pressure drop predictions and the experimentally measured becomes worse when using this modification, and better by using
50 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51

Fig. 17. Pressure drop comparison between the experiment and the face-point model, the 100 cP liquid and 4 bara pressure.

Fig. 18. Pressure drop comparison between the experiment and the face-point model, the 32 cP liquid and 8 bara pressure.

the models presented in the paper of Smith et al. The reason for gas–liquid flow regime (flow pattern) is not so good and leaves
this difference in slug velocities between the 32 cP and 100 cP oil room for improvement. There is also room for improvement in
is not known, but possible effects could be that the onset of turbu- the transition from laminar to turbulent flow, which is still incom-
lence occurs at very low Reynolds number in slug flow. pletely understood, and much more so for multiphase flow than for
single phase flow.
Conclusions and recommendations Droplet transport in the gas, bubble entrainment in the liquid
layer in stratified flow and effects of a non-planar gas–liquid inter-
In this paper we have presented a new set of experimental data face in stratified flow are all phenomena that have been neglected
for two-phase gas–liquid pipe flow using high density SF6 gas and in the present model and need further experimental and theoreti-
two different oils with viscosities of 32 and 100 cP respectively. cal study.
The new data were compared to a new steady state model for Experimental data for multiphase flow with high viscosity oils
two-phase gas liquid flow with viscous oil, Smith et al. (2013). In in large diameter pipes are still lacking. This implies that the up-
the work of Smith et al., existing closure laws were identified and scaling of models to large diameter pipes will be associated with
improved where required, particularly for the bubble front velocity significant uncertainty until such data are available. Detailed mea-
and for the gas–liquid interfacial friction factor in stratified flow. surements of the spatial and temporal distribution of phase frac-
The agreement between the model and the new experimental tions and velocity fields would be particularly valuable to
data for average liquid holdup and pressure drop is very good in improve the understanding of the fundamental flow physics in
most of the cases. The ability of the model to predict the observed high viscosity systems.
 H.A. Khaledi et al. / International Journal of Multiphase Flow 67 (2014) 37–51 51

Fig. 19. Pressure drop comparison between the experiment and the face-point model, the 32 cP liquid and 4 bara pressure.

Acknowledgements Jeyachandra, B.C.G.B., Al-Sarkhi, A., Al-Sarica, C., Sharma, A., 2012. Drift-velocity
closure relationships for slug two-phase high-viscosity oil flow in pipes. SPE J.
17, 593–601.
The authors acknowledge the partial financial support from The Kago, T., Saruwatari, T., Kashima, M., Morooka, S., Kato, Y., 1986. Heat transfer in
Multiphase Flow Assurance Innovation Centre (FACE). FACE is a horizontal plug and slug flow for gas–liquid and gas-slurry systems. J. Chem.
Eng. Jpn. 19, 125–131.
research cooperation between IFE, NTNU and SINTEF. The centre
Kongsberg Oil & Gas Technologies, 2011. LedaFlow.
is funded by The Research Council of Norway and by the following Kora, C., Sarica, C., Zhang, H.-Q., Al-Sarkhi, A., Al-Safran, E., 2011. Effects of High Oil
industrial partners: Statoil ASA, GE Oil & Gas, SPT Group (a Schlum- Viscosity on Slug Liquid Holdup in Horizontal Pipes, Canadian Unconventional
Resources Conference. Society of Petroleum Engineers, Alberta, Canada.
berger company), FMC Technologies, CD-adapco and Shell Technol-
McNeil, D.A., Stuart, A.D., 2003. The effects of a highly viscous liquid phase on
ogy Norway. vertically upward two-phase flow in a pipe. Int. J. Multiph. Flow 29, 1523–1549.
Nuland, S., 1998. Bubble front velocity in horizontal slug flow with viscous
References Newtonian-, shear thinning- and Bingham fluids. In: Third International
Conference on Multiphase Flow. Lyon, France.
Nädler, M., Mewes, D., 1995. Effects of the liquid viscosity on the phase
Andreussi, P., Persen, L.N., 1987. Stratified gas–liquid flow in downwardly inclined distributions in horizontal gas–liquid slug flow. Int. J. Multiph. Flow 21, 253–
pipes. Int. J. Multiph. Flow 13, 565–575. 266.
Andritsos, N., Hanratty, T.J., 1987. Interfacial instabilities for horizontal gas–liquid Schmidt, J., Giesbrecht, H., van der Geld, C.W.M., 2008. Phase and velocity
flows in pipelines. Int. J. Multiph. Flow 13, 583–603. distributions in vertically upward high-viscosity two-phase flow. Int. J.
Andritsos, N., Williams, L., Hanratty, T.J., 1989. Effect of liquid viscosity on the Multiph. Flow 34, 363–374.
stratified-slug transition in horizontal pipe flow. Int. J. Multiph. Flow 15, 877– Smith, I.E., Krampa, F., Fossen, M., Brekken, C., Kjølaas, J., Unander, T.E., 2011.
892. Investigation of horizontal two-phase gas–liquid pipe flow using high viscosity
Bendiksen, K.H., 1984. An experimental investigation of the motion of long bubbles oil: comparison with experiments using low viscosity oil and flow simulations.
in inclined tubes. Int. J. Multiph. Flow 10, 467–483. In: Group, B. (Ed.), 15th International Conference on Multiphase production
Benjamin, T.B., 1968. Gravity currents and related phenomena. J. Fluid Mech. 31, Technology. Cannes, France 15–17.06.2011, pp. 293–307.
209–248. Smith, I.E., Unander, T.E., Nossen, J., 2013. Improved holdup and pressure drop
Gokcal, B., 2005. Effect of High Oil Viscosity on Two-Phase Oil–Gas Flow Behavior in predictions for multiphase flow with gas and high viscosity oil. In: 15th
Horizontal Pipes. University of Tulsa, Oklahoma, Tulsa, p. 97. International Conference on Multiphase Production Technology. Cannes, France.
Gokcal, B., 2008. An Experimental and Theoretical Investigation of Slug Flow for Valle, A., 2000. Three Phase Gas–Oil–Water Pipe Flow, Department of Chemical
High Oil Viscosity in Horizontal Pipes. University of Tulsa, Oklahoma, Tulsa, p. Engineering and Chemical Technology. Imperial College of Science, Technology
146. and Medicine, London.
International Organization for Standardization, 1993. Guide to the Expression of Wallis, G.B., 1969. One Dimensional Two Phase Flow. McGraw-Hill, New York.
Uncertainty in Measurement. International Organization for Standardization, White, F.M., 2011. Fluid Mechanics. McGraw Hill, New York, NY.
Geneva.