Effects of High Oil Viscosity on Oil/Gas

Flow Behavior in Horizontal Pipes

B. Gokcal, SPE, Q. Wang, SPE, H.-Q. Zhang, SPE, and C. Sarica, SPE, University of Tulsa

Summary pattern prediction model and separate models to calculate the flow
Oil/gas pipe flows are expected to exhibit significantly different variables, such as pressure drop and liquid holdup, for individual
behavior at high oil viscosities. Effects of high-viscosity oil on flow patterns.
flow pattern, pressure gradient, and liquid holdup are experimen- Zhang et al. (2003) conducted a detailed review of existing
tally observed, and differences in flow behavior of high- and low- two-phase flow prediction models, and developed a unified hydro-
viscosity oils are identified. The experiments are performed on a dynamic model to predict flow-pattern transitions, pressure gradi-
flow loop with a test section of 50.8-mm ID and 18.9-m-long ent and liquid holdup, and slug characteristics in gas/liquid pipe
horizontal pipe. Superficial liquid and gas velocities vary from flows for all inclination angles from –90° to 90° from horizontal.
0.01 to 1.75 m/s and from 0.1 to 20 m/s, respectively. Oil viscosi- The model was based on the dynamics of slug flow and was
ties from 0.181 to 0.587 Pa⭈s are investigated. The experimental applicable to all pipe geometries and fluid physical properties.
results are used to evaluate the performances of existing models Weisman et al. (1979) experimentally studied the effects of
for flow pattern and hydrodynamics predictions. Comparisons of fluid properties and pipe diameter on two-phase flow patterns in
the data with the existing models show significant discrepancies at horizontal pipes. Air-glycerol water solutions having viscosities of
high oil viscosities. Possible reasons for these discrepancies are 0.075 Pa⭈s and 0.15 Pa⭈s were used as test fluids. They concluded
carefully examined. Some modifications are identified and imple- that liquid viscosity had little effect on flow pattern transition
mented to the closure relationships employed in the Zhang et al. boundaries. Taitel and Dukler (1987) conducted an investigation of
(2003) model. After these modifications, the model predictions the effect of pipe length on flow pattern transition boundaries for
provide better agreement with experimental results for flow pattern high-viscosity liquids. In their study, the liquid viscosity ranged
transition, pressure gradient, and liquid holdup. from 0.09 Pa⭈s to 0.165 Pa⭈s. Contrary to Weisman et al.’s claim,
Introduction Taitel and Dukler (1987) concluded that pipe length can have a con-
siderable effect on transition boundaries for high-viscosity liquids.
Gas/liquid two-phase flow in pipes is a common occurrence in the
Andritsos et al. (1989) experimentally studied the effect of liquid
petroleum, chemical, nuclear, and geothermal industries. In the
viscosity on the initiation of gas/liquid slug flow in horizontal
petroleum industry, it is encountered in the production and trans-
portation of oil and gas. Accurate prediction of the flow pattern, 25.2-mm and 95.3-mm ID pipes. They proposed a mechanism for
pressure drop, and liquid holdup is imperative for the design of viscous liquids that slugs arise from small-wavelength Kelvin-
production and transport systems. Helmhotz (KH) waves. The mechanism showed good agreement
High-viscosity oils are discovered and produced all around the with experimental results. They concluded that the new mecha-
world. High-viscosity or “heavy oil” has become one of the most nism was applicable to liquids with viscosities above 0.02 Pa⭈s.
important future hydrocarbon resources, with ever-increasing Barnea (1991) proposed a combined model of viscous and in-
world energy demand and depletion of conventional oils. viscid KH stability analysis to determine the transition to slug and
Almost all flow models have viscosity as an intrinsic variable. annular flows. The results were compared with the Taitel and
Two-phase flows are expected to exhibit significantly different Dukler (1976) model. It was concluded that the combined model
behavior for higher viscosity oils. Many flow behaviors will be gave good results for different experimental data, and could be
affected by the liquid viscosity, including droplet formation, sur- used to determine annular flow or roll waves. It was also shown
face waves, bubble entrainment, slug mixing zones, and even that the Taitel and Dukler (1976) model was valid for liquid vis-
three-phase stratified flow. Furthermore, the impact of low- cosities up to 0.1 Pa⭈s.
Reynolds-number oil flows in combination with high-Reynolds- Barnea and Taitel (1993) investigated the KH stability cri-
number gas and water flows may yield new flow patterns and teria for stratified flow. Inviscid (nonviscous) and viscous KH
concomitant pressure-drop behaviors. analyses were used for this study. It was shown that the results of
The literature is awash with two-phase studies addressing the inviscid and viscous analyses were quite different for low
mainly the flow behavior for low-viscosity liquids and gases. liquid viscosities. On the other hand, they were similar for high
However, very few studies in the literature have addressed high- liquid viscosities.
viscosity multiphase flow behavior. In this literature review, the An experimental study conducted by Nadler and Mewes (1995)
state-of-the-art of two-phase flow is first summarized. Then, the investigated the effect of the liquid viscosity on the phase distri-
studies addressing the effects of liquid viscosity on two-phase bution in slug flow for horizontal pipes with the other fluid physi-
oil/gas flow behavior are reviewed. cal properties being kept constant. The viscosity range for their
Taitel and Dukler (1976) proposed the first mechanistic model experimental study was from 0.014 Pa⭈s to 0.037 Pa⭈s. Experi-
to predict flow pattern transitions for horizontal and near- mental results indicated that the average liquid holdup within the
horizontal gas/liquid flow. Later, Barnea (1987) proposed a unified slug unit and the elongated bubble region increased with increas-
model for predicting steady-state transition boundaries for the ing liquid viscosity.
whole range of pipe inclinations. Newton et al. (1999) performed an experimental study to in-
Xiao et al. (1990) developed a comprehensive mechanistic vestigate the effect of liquid viscosity on gas wall and interfacial
model for two-phase flow in horizontal and near-horizontal pipe- shear stress in horizontal pipes. It was observed that the known
lines. The comprehensive mechanistic model incorporated a flow- correlations for interfacial friction factors were not adequate when
the liquid viscosity was increased. By increasing the liquid vis-
cosity, the onset of large amplitude wave motion on the gas/liquid
interface was delayed. However, the effect of viscosity became
Copyright © 2008 Society of Petroleum Engineers
unimportant at lower liquid heights. After the flow became wavy,
This paper (SPE 102727) was accepted for presentation at the 2006 SPE Annual Technical the interfacial wavy structure was viscosity independent. They
Conference and Exhibition, San Antonio, Texas, 24–27 September, and revised for publi-
cation. Original manuscript received for review 15 June 2006. Revised manuscript received
proposed new correlations for interfacial friction factor for smooth,
21 November 2007. Paper peer approved 26 November 2007. rippled, and wavy interfaces.

June 2008 SPE Projects, Facilities & Construction 1

 Fig. 1—Schematic of indoor high-viscosity test facility.

Furukuwa and Fukano (2001) conducted an experimental the data with the range of more viscous liquid phase is limited. The
study to analyze the effect of liquid viscosity on flow patterns of first objective of this study is to investigate the effects of high-
gas/liquid two-phase flow with a 19.2-mm ID vertical pipe. Water viscosity oil on flow pattern, pressure drop, and liquid holdup
and two different concentrations of aqueous glycerol solutions experimentally and to identify the differences in flow behavior of
were used. The kinematic viscosity of these liquids changed from high- and low-viscosity oils. Next, the performance of existing
1 centistoke to 14.7 centistokes. Flow patterns for different liquid flow-pattern, pressure-gradient, and holdup-prediction models
viscosities were proposed. They concluded that flow-pattern tran- is evaluated using the data acquired in the experimental part of
sitions are dependent on liquid viscosity. this study.
Colmenares et al. (2001) studied pressure-drop models for hori-
zontal slug flow for viscous oils. According to their experimental Experimental System
results, the slug-flow pattern enlarged when the oil viscosity in- Test Facility. The indoor high-viscosity test facility is comprised
creased. In their study, the authors evaluated existing slug models, of an 18.9-m-long, 50.8-mm ID pipe with a 9.15-m-long trans-
and concluded that the Taitel and Barnea (1990) model was the parent acrylic pipe section to visually observe the flow. Inclination
best candidate to be used for high-viscosity oils. A modified model angle can be changed from –2° to 2° from horizontal. A 76.2-mm
was developed by using experimental data for a liquid viscosity of ID return pipe is connected to the test section with a flexible hose.
0.48 Pa⭈s. They concluded that slug frequency and liquid film The return pipe goes to the oil storage tank. A schematic of the
height increased, and the slug length decreased as the liquid vis- facility is given in Fig. 1. Viscosity is the parameter of interest for
cosity increased. this study. Viscosity is a strong function of temperature. Therefore,
McNeil and Stuart (2003) experimentally studied the effects of temperature control is important for the success of the experi-
a highly viscous liquid phase on two-phase flow in vertical pipe. ments. Heating and cooling systems were designed and con-
Water and glycerin solutions were used as test liquids. The liquid structed to achieve good temperature control. A chiller is used for
viscosities were 0.001, 0.05, 0.2, and 0.55 Pa⭈s. Mostly, annular cooling, and a heater is used for heating. Prior to each test, the
flow was observed in this study. They concluded that low-viscosity temperature of the oil tank is set to the desired temperature. Small
correlations for entrained liquid fraction and interfacial friction fluctuations in the oil temperature during testing are adjusted by
factor were not suitable for the highly viscous range. A new cor- using either the chiller or the heater.
relation for the interfacial friction factor was developed based on Air and viscous oil were used as test fluids. The properties of
the high-viscosity oil data. the oil are given in Table 1. The most important characteristic of
Rosa and Netto (2004) investigated experimentally the influ- the oil is its large range of viscosity, owing to strong temperature
ence of liquid viscosity on gas/liquid structures of horizontal slug dependence. Compressed air was used as the gas phase, and was
flow. Air/water (0.001 Pa⭈s) and air/glycerin (0.027 Pa⭈s) were supplied by a dry rotary screw air compressor. The oil was pumped
used as the two pairs of test fluids. Bubble shape, velocity and void by a 20-hp screw pump from the oil storage tank. A motor fre-
fraction, bubble length and slug lengths, slug frequency, and co- quency drive was installed to provide better flow rate control and
alescence rate were analyzed in this study. They concluded that the reduce the amount of heat generated. The oil storage tank con-
average slug, bubble length, and coalescence rate decrease with tained 3.03 m3 of oil. Both air and oil flow rates were metered by
increasing liquid viscosity. The bubble front velocity and slug Micro MotionTM mass flowmeters. The fluids were mixed at the
frequency increased with an increase of the liquid viscosity. mixing tee, flowed through the test section, and returned to the oil
From the literature review, most of the experimental data come storage tank. The oil storage tank was also used as a separator. The
from air and water/glycerin solutions as test fluids. Moreover, separated air was discharged outside through a ventilation system.

2 June 2008 SPE Projects, Facilities & Construction

 Fig. 2—Schematic of test section.

The test section consists of two separate pipe sections. The first limits of superficial velocities are attributable to the accuracy lim-
pipe section following the mixing tee is a 9.75-m-long steel pipe its of the Micro MotionTM flow meters. The higher limits are
with an inner diameter of 50.8-mm. The use of steel pipe allows determined by the pressure gradient and facility limits. The ex-
operating the facility at inlet pressures up to 100 psig. The second periments are performed at temperatures of 21.1, 26.7, 32.2, and
pipe section consists of a 9.15-m-long transparent acrylic pipe with 37.8°C. The oil viscosities corresponding to these temperatures are
the same diameter as the steel pipe. Transparency of the acrylic 0.587, 0.378, 0.257, and 0.181 Pa⭈s, respectively. The pipe was
pipe allows visual observations of the flow behavior. A high-speed kept in a horizontal position.
video system is used to observe the characteristics of high-
viscosity oil and gas two-phase flow. A schematic of the test Experimental Results
section is shown in Fig. 2. The flow patterns, pressure gradients, and liquid holdups are
After the mixing tee, a pressure relief valve is installed which briefly discussed in following subsections, respectively. The ex-
can withstand pressures up to 110 psig. Two differential pressure perimental results are available in their entirety in Gokcal (2005).
transducers on the steel pipe (DP1 and DP2) are used to monitor
the development of the flow before it reaches the transparent part Flow Patterns. A high-speed video camera was used to iden-
of the test section. Pressure transducers are located after the mixing tify the flow patterns. Stratified wavy (SW), elongated bubble
tee and at the start of the acrylic pipe section. The acrylic test (EB), slug (SL), and annular (AN) flow patterns, and the slug-
section is instrumented with pressure and differential pressure dispersed (SL-DB) and slug-annular (SL-AN) transition bound-
transducers (DP3 and DP4), thermocouples, capacitance sensors, aries were observed. SW flow was only seen at flow patterns
and quick closing valves. The DP3 differential pressure transducer, that had low viscosity values (0.181 Pa⭈s and 0.257 Pa⭈s) and at
which covers 3.05 m of the transparent pipe, is mainly used for low superficial oil and high superficial air velocities of 0.01 m/s
high flow rates. Another differential pressure transducer, which and 10 m/s, respectively. From high-speed video recordings, it
covers 6.55 m of the transparent pipe, is used for low flow rates. was observed that the bubble shape changed with increasing vis-
Quick closing valves are used for flow control and liquid trapping. cosity for elongated bubble. Also, the frequency of elongated
The measurements of thermocouples are imperative to determine bubbles increased with increasing viscosity while the bubble
the viscosity of the oil during experiments. length decreased. Another observation is that slug frequency and
liquid holdup increased, while slug length decreased as oil viscos-
Experimental Range. Gas and oil flow rates and oil tempera- ity increased.
ture are varied. Superficial liquid and gas velocities range from Figs. 3 and 4 show the experimental observations for flow
0.01 m/s to 1.75 m/s and from 0 to 20 m/s, respectively. The lower patterns at 0.181 Pa⭈s and 0.587 Pa⭈s. As the viscosity increased

Fig. 3—Experimental observation of flow patterns (0.181 Paⴢs).

June 2008 SPE Projects, Facilities & Construction 3

 Fig. 4—Experimental observation of flow patterns (0.587 Paⴢs).

from 0.181 Pa⭈s to 0.587 Pa⭈s, the data do not suggest a significant Liquid Holdups. Quick closing valves are used to measure the
effect of high oil viscosity on the location of transition boundaries. liquid holdup. The drainage of the trapped volume was not
Constant gas/oil ratio curves that are typical of high-viscosity oils efficient because to the viscous nature of the oil. Therefore,
were shown in the graphs. It is concluded that the observed flow the trapped volume of oil was measured with a tape from out-
patterns can be seen in high viscosity and gas flow. side of the pipe. The liquid holdup was measured three times for
each test in order to increase the accuracy of the result at different
Pressure Gradients. Fig. 5 presents the measured pressure gra- pipe locations.
dients at different oil viscosities and superficial gas velocities The liquid holdup results are shown for different oil viscosities
when superficial liquid velocity is 0.5 m/s. As expected, the pres- and superficial gas velocities when the superficial liquid velocity is
sure gradients increase with increasing superficial gas velocity. At 0.5 m/s in Fig. 6. The decrease in liquid holdup with an increase
the same superficial air and liquid velocities, the pressure gradient in superficial air velocity exhibits a similar trend for all viscosity
increases with the increase of oil viscosity. Similar results are values. No significant effect of high viscosity on liquid holdup is
obtained for different superficial liquid velocities. It is seen that the observed. Similar results are obtained for different superficial liq-
viscosity plays an important role. uid velocities.

Fig. 5—Pressure gradient comparison for vSL=0.5 m/s.

4 June 2008 SPE Projects, Facilities & Construction

 Fig. 6—Liquid holdups for vSL=0.5 m/s.

Model Comparison bubble-slug flow boundary and slug-dispersed bubble boundary

The effects of high viscosity on two-phase flow were analyzed for predictions by Barnea model matched well with the experimental
different flow patterns, liquid holdups, and pressure gradients. Ex- observations, while the slug-annular flow boundaries were under-
perimental results from high-viscosity two-phase flow tests are predicted. Fig. 8 shows the Zhang et al. (2003) unified model flow
compared with the flow pattern model of Barnea (1987) and the pattern map predictions and the observed flow patterns. The Zhang
Zhang et al. (2003) model. Experimental results for liquid holdup et al. (2003) unified model underestimated the transition bound-
and pressure drop were compared with predictions by the Zhang aries of intermittent-dispersed bubble and intermittent-separated
et al. (2003) unified hydrodynamic model and the Xiao et al. flows (both stratified SW flow and AN flow).
(1990) mechanistic model. Figs. 9 and 10 show Barnea (1987) and Zhang et al. (2003)
unified flow-pattern maps, respectively, and the experimental ob-
Flow-Pattern Predictions. The experimental flow pattern obser- servations for a liquid viscosity of 0.587 Pa⭈s, the highest viscosity
vations were compared with predictions of the models. Fig. 7 value investigated in this study. The model predictions for the
shows the Barnea (1987) flow-pattern map and the experimental transition boundaries did not match the experimental observations.
observations for a liquid viscosity of 0.181 Pa⭈s. The elongated It is apparent that the discrepancies between experimental obser-

Fig. 7—Barnea (1987) flow pattern map comparison (0.181 Paⴢs).

June 2008 SPE Projects, Facilities & Construction 5

 Fig. 8—Zhang et al. (2003) flow pattern maps (0.181 Paⴢs).

vations and model predictions of flow-pattern transition become • The slug liquid holdup model used in the unified model was
more significant with the increase of oil viscosity. developed based on the competition between turbulent energy of
The possible reasons for these discrepancies in the Zhang et al. the liquid phase and the surface free energy of the dispersed
(2003) unified model have been carefully examined. Some modi- bubbles. The turbulent energy is estimated using the shear velocity
fications have been identified and implemented into the model: of the pipe flow. This approach is not suitable for high-viscosity
• Mostly laminar flow is observed in the experimental study. and laminar flow, where the shear stress is high because of the high
However, some of the closure relationships adopted in the unified viscosity but the turbulent energy is minimal because of laminar
model were either developed based on turbulent flow assumption flow. Therefore, the momentum term for gas entrapment is modi-
or did not include the option for laminar flow. One is the corre- fied by a factor of Re/5000 when the Reynolds number is less
lation for slug translation velocity, which is dependent on the than 5,000.
velocity profile of a pipe flow. Therefore, modification has been Zhang et al. (2003) flow-pattern maps for different viscosity values
made to the slug translational velocity to include both the turbulent were compared with experimental flow-pattern observations
and laminar flow options. again after the modifications are implemented to the unified

Fig. 9—Barnea (1987) flow pattern map comparison (0.587 Paⴢs).

6 June 2008 SPE Projects, Facilities & Construction

 Fig. 10—Zhang et al. (2003) flow pattern maps (0.587 Paⴢs).

model. In Figs. 11 and 12, the new slug-dispersed bubble and culated as −2.4% and −259 Pa/m, respectively. Fig. 13 gives the
slug-annular transition boundaries are shown in the solid lines for comparison of the Zhang et al. (2003) unified model predictions
liquid viscosities of 0.181 Pa⭈s and 0.587 Pa⭈s, the lowest and against the measured pressure gradients before and after the modi-
the highest viscosity values for this study. The model predictions fications, respectively.
provide better agreement with experimental results for flow- When the entire data set is compared with the Xiao et al. (1990)
pattern transition. mechanistic model, the average percentage relative and actual
errors are –25.1% and –954 Pa/m, respectively. Fig. 14 shows
Pressure-Gradient Predictions. When the entire data set is com- the comparison of experimental data against Xiao mechanistic
pared against the Zhang et al. (2003) unified model, the average model predictions.
percentage relative and actual errors are 32.4% and 1623 Pa/m, In summary, Zhang et al. (2003) unified and Xiao et al. (1990)
respectively. The reasons of high errors for the unified model have mechanistic models are compared with the experimental data for
been analyzed. After the modifications mentioned previously are pressure gradient. The performance of the modified Zhang et al.
implemented, agreement with experimental results is significantly unified model is shown to be better than that of the Xiao et al.
improved. The average percentage relative and actual errors cal- (1990) mechanistic model.

Fig. 11—Zhang et al. (2003) flow pattern map comparison (0.181 Paⴢs).

June 2008 SPE Projects, Facilities & Construction 7

 Fig. 12—Zhang et al. (2003) flow pattern map comparison (0.587 Paⴢs).

Liquid Holdup Predictions. When the entire data set is compared predictions. These numbers indicate that the Xiao et al. (1990)
against the Zhang et al. (2003) unified model, the average percent- model performs better than the modified Zhang et al. (2003) uni-
age relative and actual errors are –39.8% and –0.2, respectively. fied model. This can probably be attributed to the use of different
As the modifications mentioned before are implemented, agree- closure relationships in both models.
ment with experimental results is improved. The average percent- The model predictions are still not satisfactory for flow pat-
age relative and actual errors calculated are –14.6% and –0.05, terns, pressure losses, and liquid holdup for higher-viscosity oils.
respectively. Fig. 15 presents comparison of predictions of the It is speculated that all are caused by two possible factors. First, the
Zhang et al. (2003) unified model against the measured liquid modified closure relationships may not represent the right physics
holdups before and after the modifications. of the high-viscosity oil/gas flow. An example is that the drift
When the entire data set is compared against the Xiao mecha- velocity term in the translational velocity equation does not con-
nistic model, the average percentage relative and actual errors are sider the effect of the liquid viscosity. All the other closure rela-
–8.1% and –0.04, respectively. Fig. 16 shows the comparison of tionships for every flow pattern need to be reconsidered and modi-
experimental data against Xiao et al. (1990) mechanistic model fied for higher viscosity oils. Secondly, the flow description given

Fig. 13—Comparison of Zhang et al. (2003) unified model predictions and measured pressure gradients.

8 June 2008 SPE Projects, Facilities & Construction

 Fig. 14—Comparison of Xiao et al. (1990) model predictions and measured pressure gradients.

in the conventional mechanistic models, such as the Zhang et al. fications are identified and implemented to the closure relation-
(2003) unified model, may not represent the true flow configura- ships employed in the Zhang et al. (2003) unified model. Com-
tion of higher-viscosity oil/gas flows. This can be supported by paratively, the Zhang et al. (2003) unified model gives better pre-
visual observation of thick and slowly draining liquid film in the dictions for pressure gradient than the Xiao model. For liquid
film region of slug flow. This thick film is completely ignored in holdup, the Xiao et al. (1990) model performs better than the
the current mechanistic models. Zhang et al. (2003) unified model. The existing models need to be
modified or new models should be developed to predict flow pat-
Conclusions terns, pressure losses, and liquid holdup more accurately for
All flow patterns can exist for typical gas/oil ratios within the higher-viscosity oils.
range of the viscosities studied. The comparisons of the data with
the existing models show significant discrepancies at high oil vis- Nomenclature
cosities. The comparison of experimental data for pressure gradi- dP/dL ⳱ pressure gradient
ent and liquid holdup against the Zhang et al. (2003) unified and HL ⳱ liquid holdup
Xiao et al. (1990) mechanistic models show that the performances ID ⳱ inner diameter
of the models are not sufficient for high-viscosity oils. Some modi- L ⳱ length

Fig. 15—Comparison of Zhang et al. (2003) unified model predictions and measured liquid holdups.

June 2008 SPE Projects, Facilities & Construction 9

 Fig. 16—Comparison of Xiao et al. (1990) model predictions and measured liquid holdups.

P ⳱ pressure Nädler, M. and Mewes, D. 1995. Effects of the liquid viscosity on the
Re ⳱ Reynolds number phase distributions in horizontal gas-liquid slug flow. International
vSG ⳱ superficial gas velocity Journal of Multiphase Flow 21 (2): 253–266. DOI:10.1016/0301-
vSL ⳱ superficial liquid velocity 9322(94)00067-T.
Newton, C.H., Behnia, M., and Reizes, J.A. 1999. The effect of liquid
viscosity on gas wall and interfacial shear stress in horizontal
Acknowledgments two-phase pipe flow. Chemical Engineering Science 54 (8): 1071–
The authors wish to thank the TUFFP member companies for 1079. DOI:10.1016/S0009-2509(98)00423-0.
supporting this research project. Rosa, E.S. and Netto, J.R.F. 2004. Viscosity Effect and Flow Development
in Horizontal Slug Flows. Paper 306 presented at the International
References Conference on Multiphase Flow, Yokohama, Japan, 30 May—4 June.
Taitel, Y. and Barnea, D. 1990. Two Phase Slug Flow. In Advances in Heat
Andritsos, N., Williams, L., and Hanratty, J. 1989. Effect of liquid viscos- Transfer, Volume 20, ed. J.P. Hartnett and T.F. Irvine, 83–132. London:
ity on the stratified-slug transition in horizontal pipe flow. International Academic Press.
Journal of Multiphase Flow 15 (6): 877–892. DOI:10.1016/0301- Taitel, Y. and Dukler, A.E. 1976. A model for predicting flow regime
9322(89)90017-7. transitions in horizontal and near-horizontal gas-liquid flow. AIChE
Barnea, D. and Taitel, Y. 1993. Kelvin-Helmholtz stability criteria for Journal 22 (1): 47–55. DOI:10.1002/aic.690220105.
stratified flow: viscous versus non-viscous (inviscid) approaches. Taitel, Y. and Dukler, A.E. 1987. Effect of pipe length on the transition
International Journal of Multiphase Flow 19 (4): 639–649. boundaries for high-viscosity liquids. International Journal of Multi-
DOI:10.1016/0301-9322(93)90092-9. phase Flow 13 (4): 577–581. DOI:10.1016/0301-9322(87)90023-1.
Barnea, D. 1987. A unified model for predicting flow-pattern transitions Weisman, J., Duncan, D., Gibson, J., and Crawford, T. 1979. Effects of
for the whole range of pipe inclinations. International Journal of Fluid Properties and Pipe Diameter on Two-Phase Flow Patterns in
Multiphase Flow 13 (1): 1–12. DOI:10.1016/0301-9322(87)90002-4. Horizontal Lines. International Journal of Multiphase Flow 5 (6): 437–
Barnea, D. 1991. On the effect of viscosity on stability of stratified 462. DOI:10.1016/0301-9322(79)90031-4.
gas—liquid flow—application to flow pattern transition at various pipe Xiao, J.J., Shonham, O., and Brill, J.P. 1990. A Comprehensive Mecha-
inclinations. Chemical Engineering Science 46 (8): 2123–2131. nistic Model for Two-phase Flow in Pipelines. Paper SPE 20631 pre-
DOI:10.1016/0009-2509(91)80170-4. sented at the SPE Annual Technical Conference and Exhibition, New
Colmenares, J., Ortega, P., Padrino, J., and Trallero, J.L. 2001. Slug Flow Orleans, 23–26 September. DOI: 10.2118/20631-MS.
Model for the Prediction of Pressure Drop for High Viscosity Oils in a Zhang, H.-Q., Wang, Q., Sarica, C., and Brill, J.P. 2003. Unified model for
Horizontal Pipeline. Paper SPE 71111 presented at SPE International gas-liquid pipe flow via slug dynamics—Part 1: model development.
Thermal Operations and Heavy Oil Symposium, Porlamar, Margarita ASME Journal of Energy Resources Technology 125 (4): 266–273.
Island, Venezuela, 12–14 March. DOI: 10.2118/71111-MS. DOI:10.1115/1.1615618.
Furukawa, T. and Fukano, T. 2001. Effects of liquids viscosity on flow
Bahadir Gokcal is a research assistant at the University of Tulsa
pattern in vertical upward gas-liquid two-phase flow. International Fluid Flow Projects (TUFFP) and is pursuing his PhD degree in
Journal of Multiphase Flow 27 (6): 1109–1126. DOI:10.1016/S0301- Petroleum Engineering at University of Tulsa. Gokcal is a SPE
9322(00)00066-5. member and holds a BS degree in petroleum and natural gas
Gokcal, B. 2005. Effects of High Oil Viscosity on Two-Phase Oil-Gas engineering from Middle East Technical University in Turkey and
Flow Behavior in Horizontal Pipes. MS thesis, University of Tulsa, a MS degree in petroleum engineering from the University of
Tulsa, Oklahoma. Tulsa. Qian Wang is a currently a consulting engineer with
Scandpower Technologies in Houston. She previously served as
McNeil, D.A. and Stuart, A.D. 2003. The effects of a highly viscous liquid a research associate with TUFFP and Tulsa University Paraffin
phase on vertically upward two-phase flow in a pipe. International Deposition Projects (TUPDP). Wang is an SPE member and holds
Journal of Multiphase Flow 29 (9): 1523–1549. DOI:10.1016/S0301- BS, MS, and PhD degrees in mechanical engineering from
9322(03)00122-8. Xi’an Jiaotong University, China. Hong-Quan Zhang is an Assis-

10 June 2008 SPE Projects, Facilities & Construction

tant Professor of Petroleum Engineering at the University of gree from Tianjin University, China. Cem Sarica is a professor of
Tulsa and Associate Director of TUFFP. e-mail: hong-quan- petroleum engineering and the director of both TUFFP and
[email protected]. From 1998 to 2003, he was a Senior Re- TUPDP at the University of Tulsa. He is a member of SPE. He
search Associate of TUFFP. Before joining the University of Tulsa currently serves as a member of the SPE Projects, Facilities &
in 1998, he was an Associate Professor and Professor at Tianjin Construction Advisory Committee. He served as a member of
University. In 1993 and 1994, as an Alexander von Humboldt the SPE Journal Editorial Board between 1999 and 2007. His
Research Fellow, he conducted research at the Max Planck research interests are multiphase flow in pipes and flow assur-
Institute of Fluid Mechanics and the German Aerospace Re- ance. Sarica holds BS and MS degrees in petroleum engineer-
search Establishment in Göttingen, Germany. Zhang holds BS ing from Istanbul Technical University and a PhD degree in
and MS degrees from Xian Jiaotong University and a PhD de- petroleum engineering from the University of Tulsa.

June 2008 SPE Projects, Facilities & Construction 11