SPE-191543-MS

Pressure Effects on Low-Liquid Loading Oil-Gas Flow in Slightly Upward

Inclined Pipes: Flow Pattern, Pressure Gradient and Liquid Holdup

Hendy Rodrigues, The University of Tulsa, Petrobras; Eduardo Pereyra and Cem Sarica, The University of Tulsa

Copyright 2018, Society of Petroleum Engineers

This paper was prepared for presentation at the 2018 SPE Annual Technical Conference and Exhibition held in Dallas, Texas, 24-26 September 2018.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
This paper studies the effects of system pressure in oil-gas low-liquid loading flow in a slightly upward
inclined pipe configuration using new experimental data acquired in a high-pressure flow loop. Flow rates
are representative of the flow in wet gas transport pipelines. Results for flow pattern observations, pressure
gradient, liquid holdup and interfacial roughness measurements are presented and compared to available
predictive models. The experiments were carried out at three system pressures (1.48, 2.17 and 2.86 MPa)
in a 0.155 m ID pipe inclined at 2° with the horizontal. Isopar-L oil and nitrogen gas were the working
fluids. Liquid superficial velocities ranged from 0.01 to 0.05 m/s while gas superficial velocities ranged
from 1.5 to 16 m/s. Measurements included pressure gradient and liquid holdup. Flow visualization and
Wire-Mesh Sensor (WMS) data were used to identify the flow patterns. Interfacial roughness was obtained
from the WMS data.
Three flow patterns were observed: pseudo-slug, stratified and annular. Pseudo-slug is characterized as
an intermittent flow where the liquid does not occupy the whole pipe cross-section as the traditional slug
flow does. In the annular flow pattern, the bulk of the liquid was observed to flow at the pipe bottom in a
stratified configuration, however, a thin liquid film covered the whole pipe circumference. In both stratified
and annular flow patterns, the interface between the gas core and the bottom liquid film presented a flat
shape. The superficial gas Froude number, FrSg, was found to be an important dimensionless parameter to
scale the pressure effects on the measured parameters. In the pseudo-slug flow pattern, the flow is gravity-
dominated. Pressure gradient is a function of FrSg and vSL. Liquid holdup is independent of vSL and a function
of FrSg. In the stratified and annular flow patterns the flow is friction-dominated. Both pressure gradient and
liquid holdup are functions of FrSg and vSL. Interfacial roughness measurements show a small variation in
the stratified and annular flow patterns. Model comparison gives mixed results, depending on the specific
flow conditions. A relation between the measured interfacial roughness and the interfacial friction factor is
proposed, and the results agree with existing measurements.

Introduction
Low liquid loading gas-liquid flow in near-horizontal pipes is a common phenomenon in wet gas transport
pipelines. It is characterized by the large ratio of gas to liquid volume flow. It is a common occurrence in
2 SPE-191543-MS

offshore operations, since processing plants are less efficient due to size limitations, and the gas transport
pipelines have a small positive inclination when connecting the offshore unit to onshore facilities. Although
the liquid volume is small, the liquid distribution can dramatically change the overall pipe pressure drop
when compared to single-phase gas flow (Meng, 2001; Badie et al., 2000; Vuong, 2016).
Proper pressure gradient and liquid holdup predictions are crucial in the design of pipeline systems. The
choice of pipe diameter, boosting units, receiving and processing facilities are largely impacted by these
predictions. During the operation, flow assurance issues such as corrosion and hydrate formation occurs.
Chemical inhibitors are used to avoid these issues, and the correct description of the flow pattern in which
the phases are arranged is critical for the efficiency of these chemical treatments.
Within the normal operational ranges of low liquid loading flow, the phases are expected to be arranged in
a segregated pattern, either stratified or annular. Under these conditions, two of the main topics for modeling
the flow are the geometrical distribution of phases at the pipe cross-sectional area and the interfacial shear
stress between liquid and gas. At earlier and later stages of the field development, lower gas flow rates can be
expected. In inclined pipes, this may lead to the formation of intermittent flows (large waves, pseudo-slugs
or slug flow). In this case, flow parameters are changed dramatically when compared to segregated flow.
One of the most used models for calculating flow parameters in segregated flow was proposed by
Taitel and Dukler (1976). They assume that the phases are segregated in a stratified condition with flat
interface, as shown in Figure 1a. A force balance in a differential control volume in the axial direction
results in the combined momentum equation. After solving the combined momentum equation for the liquid
holdup, pressure gradient is calculated with either liquid or gas phase momentum equations. Several closure
relationships are needed for the shear stresses at the liquid and gas walls and at the interface. For the
wall shear stresses, the usual approach is to use single-phase friction factor models based on the hydraulic
diameters. The interfacial friction factor is one of the most important parameters and was subject of several
studies. Other closure relationships are related to the geometric distribution of the phases in the cross-section
area. Since Taitel and Dukler (1976) assumed a flat interface shape, the interfacial length and liquid and gas
wetted wall fractions relation with the holdup is straightforward.

Figure 1—Liquid distribution geometry

Experimental studies of low-liquid loading in small diameter pipes showed that the interface was not flat,
but rather curved. Hart et al. (1989) studied low-liquid loading flow in a horizontal, 0.051 m diameter pipe
and proposed a liquid distribution geometry with constant liquid thickness, such as described in Figure 1b.
They proposed the ARS (Apparent Rough Surface) model where only the gas momentum equation is solved
and the liquid phase is treated as a modified pipe roughness. Correlations for the wetted wall fraction, SL,
and the liquid holdup, HL, were proposed. Grolman and Fortuin (1997) used the same liquid distribution
SPE-191543-MS 3

geometry, but the model solved the combined momentum equation. A new correlation for the wetted wall
fraction was given that performs better for non-horizontal pipes.
Chen et al. (1997) studied low-liquid loading flow of air-kerosene in a horizontal, 0.0779 m diameter
pipe. The authors proposed a new liquid geometry distribution with a curved interface called the double-
circle, represented in Figure 1c. The model uses the combined momentum equation, and the correlation
proposed by Hart et al. (1989) for the wetted wall fraction. Later, Fan et al. (2007) used both a 0.05 m ID
and a 0.155 m ID facility with inclinations ranging from ±2° from horizontal flowing air and water for their
experimental investigation. The double-circle geometry was used, and a model for the wetted wall fraction
calculation based on the liquid gravity-center shift as a function of the liquid Froude number was proposed.
Zhang and Sarica (2011) extended Fan et al. (2007) model to unify the wetted wall fraction predictions and
the transition to annular flow.
Meng et al. (2001) used a 0.05 m diameter facility with inclinations ranging between ±2° from the
horizontal with air and light oil as working fluids and measured several flow parameters. The authors
measured the droplet entrainment fraction and observed the flow pattern change from stratified to annular
flow due to droplet deposition on the upper pipe wall. A two-fluid model and an interfacial friction model
were proposed. The model uses the combined momentum equations, with double-circle geometry, and
includes the droplets entrainment fraction, fE, into the properties of the gas core flow.
More recently, studies focused on experiments with larger diameters and higher pressures. Diameter
scale-up has been attempted by Karami et al. (2016) using a 0.155 m ID low pressure facility to study three-
phase flow under low liquid loading in horizontal conditions. Vuong et al. (2017) studied the effects of
pressure on two-phase low liquid loading flow in a 0.155 m ID, high pressure (200 to 400 psig) facility
with horizontal configuration and Isopar L oil and nitrogen as flowing fluids. The observed flow patterns
were stratified wavy and annular flow, and measured flow characteristics included liquid holdup, pressure
gradient, wave characteristics and wetted wall fraction. Both Karami et al. (2016) and Vuong et al. (2017)
observed that the interface in larger pipe diameters retain the flat shape, rather than curved. Moreover,
annular flow pattern was obtained when the droplets are deposited on the pipe walls. The interface between
the bottom liquid film and the gas core was still flat-shaped even in annular flow.
Tayebi et al. (2000) and Viana et al. (2016) focused on the entrainment fraction measured, but also
presented results for pressure gradient and liquid holdup for low liquid loading flows at higher pressures.
Rodrigues et al. (2018) presented a model to improve the pressure gradient calculations when the flow is
in the stratified to annular transition region. The model takes into account the thin film that is formed in the
gas-wall interface. Predictions are compared to high-pressure, large diameter data available in the literature
and show improvements when compared to the results of other models.
At lower gas flow rates in slightly inclined pipes, low-liquid loading flows present an intermittent
flow behavior. Alsaadi et al. (2015), Fan et al. (2015) and Fan et al. (2018) studied the onset of liquid
accumulation in inclinations ranging between 2° and 30° from horizontal in a 0.0762 m ID pipe with air
and water as flowing fluids. Alsaadi et al. (2015) measured several parameters including the critical gas
velocity where liquid starts to accumulate. He also compared the predictions of available flow pattern
transition models against his observations finding a poor agreement. Fan et al. (2015) investigated the onset
of liquid accumulation and the characteristics of pseudo-slug flow and showed that pseudo-slugs travel with
a velocity greater than roll waves but lower than the expected translational slug velocity. Fan et al. (2018)
proposed a model for calculating the onset of liquid accumulation based on the film reversal at the bottom
of the pipe.
Langsholt and Holm (2007) studied the conditions where the liquid holdup exhibited an abrupt increase
with decreasing superficial gas velocities in slightly inclined pipes (0.5° to 5°). The authors used a 0.1 m
ID pipe facility with three-phase flow of water, Exxsol D80 oil, and SF6 gas at system pressures of 3.5 and
7.1 bar. At these conditions, the corresponding gas densities were 22.6 and 46.9 kg/m3, respectively. The
4 SPE-191543-MS

superficial liquid velocity was kept at 0.001 m/s, while the gas superficial velocity was varied from 1.5 to 4
m/s. They presented results comparing the effects of pipe inclination, water fraction and system pressure, and
provided a comparison of their results with OLGA commercial software predictions. The critical velocity
was found to be lower for higher pressures systems, lower pipe inclinations and lower water-cuts.
Kjolaas et al. (2015) focused on the investigation of the existence and characteristics of multiple solutions
for the liquid holdup in low liquid loading flows. The authors used the SINTEF (independent research
organization based in Norway) large scale multiphase flow loop with diameters of 0.2 and 0.3 m and
inclination angles of 1°, 2.5° and 5° with the horizontal. Nitrogen was used as gas phase and naphtha, Exxsol
D40, water and glycerol-water mixture were used as the liquid phases. The nominal system pressure was
855 psig. The critical superficial gas velocity when liquid starts to accumulate was obtained. One of the
main conclusions is that for a given pipe inclination and liquid flow rate, liquid accumulation occurs at a
constant gas Froude number.
In the present study, an experimental program was performed to acquire data for low-liquid loading flow
in near-horizontal pipes at system pressures up to 2.86 MPa, which is considered high when comparing to
previous experiments (more than 20 times higher than atmospheric conditions). The utilized pipe diameter,
D = 0.155 m, is larger than the previous studies. The flowing conditions covered stratified flow and
transitions to annular flow at higher gas flowrates and pseudo-slug flow at lower gas flowrates. Isopar L oil
is used as the liquid phase, since its properties are similar to the properties of light condensate oils. Basic
and advanced instrumentation were used to obtain detailed information of the flow parameters.
Although experimental conditions presented in this study are still far from the real operational conditions,
the results can indicate trends that help properly scale the effects of pressure and diameter. The unique
data obtained in this study is useful for validation and modification of models and simulators to account
for different system pressures and diameters. A relation between the wavy interface roughness and the
interfacial friction factor is proposed.

Experimental Program
The experimental program was conducted at the high-pressure, large-diameter facility of the Tulsa
University Fluid Flow Projects (TUFFP). The facility was designed to perform upscaling studies on low-
liquid loading flows. The experimental conditions are presented in the following.

Fluids
The two fluids used in this experimental study were Isopar L mineral oil as the liquid phase and nitrogen
as the gas phase. Isopar L has low density and viscosity similar to gas condensate systems. The relevant
properties are considered constant and given as: density (ρL) = 760 kg/m3, viscosity (μL) = 0.0013 Pa.s and
surface tension (σ) = 0.024 N/m.
Nitrogen was used instead of natural gas. It presents a lower Health, Safety and Environmental risk
since it is inert. During the experimental program, system pressure and temperature were varied, and the
nitrogen properties were calculated for each flowing condition. At different pressures and temperatures,
nitrogen density was calculated with the equation given by Span et al. (2000). According to Span et al., his
equation presents a maximum deviation of ±0.001% for pressures up to 30 MPa. Nitrogen viscosity was
calculated with the correlation proposed by Seibt et al. (2006). The authors conclude that the uncertainty of
this correlation is about ±0.2% when compared to their own experimental data.
The effects of pressure and temperature on the interfacial tension was considered negligible, based on the
studies of Nino Amezquita et al. (2010) and Zolghadr et al. (2013) on nitrogen and hydrocarbon mixtures.

Experimental facility
The facility was operated in a closed loop at pressures up to 2.86 MPa (400 psig), and is divided into two
main parts: the flow loop and the processing plant.
SPE-191543-MS 5

The flow loop is schematically shown in Figure 2. It is made of a stainless steel pipe with ID of 0.155
m (6.1 in.). The main test section is inclined at 2° with the horizontal and is 85 m long, which corresponds
to an L/D = 548. The total length is 160 m, including the horizontal section before and the downward and
horizontal sections after the main section. Most of the instrumentation is installed at the downstream location
of the main test section. Exceptions are the pressure gradient measurements taken in the middle of the main
test section, and capacitance probes that are distributed along the test section. Pipe roughness was estimated
by Conner (2013) as ε/D = 3.61e-5 when flowing single-phase gas and comparing the measurements with
Colebrook's friction factor equation.

Figure 2—Flow loop schematics

After flowing through the flow loop, the fluids return to the processing plant, shown schematically in
Figure 3. The gas and liquid first arrives at a vertical separator and from there they follow different paths.

Figure 3—Schematic of the processing facility

A level-controlled valve allows the liquid to flow from the bottom of the separator into the tank. From
the bottom of the tank, the liquid is pumped through a Moyno model 2000 G1 progressing cavity pump.
After the pump, the control system operates 2 valves: one for recirculation and the other into the flow loop
6 SPE-191543-MS

to achieve the desired liquid flow rate. Before getting into the flow loop, the liquid flow rate is measured
with a Micro Motion CMF200 Coriolis mass flow meter.
The gas flows from the top of the separator into the suction receiver. From suction receiver, it flows
through a control valve before reaching the Sundyne BMC-343 EF gas compressor, which is a single-stage
centrifugal compressor. After the compressor, a heat exchanger is used to lower the gas temperature before
it reaches the discharge receiver. A recirculation valve controls anti-surge flow back from the discharge to
the suction receiver. The remaining gas passes through a Micro Motion CMF300 Coriolis mass flow meter
before it is mixed with the liquid.

Test matrix
The test matrix is representative of the low-liquid loading flow of natural gas with condensate in production
pipelines. Three system pressures, 200 psig (1.48 MPa), 300 psig (2.17 MPa) and 400 psig (2.86 MPa) were
used. The liquid superficial velocities varied between 0.01 m/s and 0.05 m/s in increments of 0.01 m/s.
The gas superficial velocities varied from 1.5 m/s to 16 m/s. The maximum and minimum gas superficial
velocities varied depending on the systems pressure since the gas compressor has mass flow rate limitations

Measured parameters
Several instruments were used to measure flow parameters. Figure 2 shows their position along the main
test section, while Figure 4 shows the detailed location of some of the measurement equipment that are
located inside the quick-closing valve (QCV) section. The first QCV is located at a distance of around 65
m from the beginning of the main inclined test section.

Figure 4—Instrumentation inside the trapped section

System pressure was measured with a Rosemount 3051-CG pressure transmitter with systematic
uncertainty of ±0.15% of the transmitter full range. System temperature was measured by a Resistance
Temperature Detector (RTD) connected to a Rosemount 3144P temperature transmitter. Measurement
uncertainty is the highest between ±0.25% of the measurement range and ±0.25 °C. During a test, pressure
and temperature variations through the main test section were lower than 3.5 psi (0.024 MPa) and 2
°C at extreme conditions. These variations are small in comparison to the system absolute pressure and
temperature; therefore, it was considered that fluid properties did not change when flowing through the main
test section. However, between different test points, pressure and temperature change based on ambient
SPE-191543-MS 7

conditions and variation of liquid volume inside the pipe. Therefore, the system pressure and temperature
were used to correct the fluid properties for each data point.
A Rosemount 3051S differential pressure transmitter located in the middle portion of the main test section
was used to measure the pressure gradient. The differential pressure reading was corrected by the transmitter
zero value – when there is no flow and both tubes connecting the transmitter to the pressure taps are filled
with oil – and then divided by the distance between the two pressure taps, 7.7 m. During the experiments, it
was noted that the transmitter's zero value varied with temperature. Collecting data points for the zero value
under different system pressures and temperatures, its variation with temperature followed a linear trend as
shown in Figure 7 (left) for the 3 different system pressures. For each pressure, a corresponding equation
was used to correct the measured differential pressure before calculating the pressure gradient

(1)

where DP0 is in inches of H2O, T is the system temperature in °F, and P is the system pressure.
Pressure gradient measurements for single-phase gas flow were made and compared with standard
calculation methods. Figure 5 presents the comparison between the measured pressure gradient and
calculated using the friction factor equation proposed by Churchill (1977). The pipe roughness used for this
calculation was ε/D = 3.61e-5.

Figure 5—Comparison of measured and calculated pressure gradient for single-phase gas flow

Two quick-closing valves are located at a downstream portion of the main test section to measure
the liquid holdup. When the two-phase flow condition is stable, the two valves are closed and the local
instantaneous liquid and gas volume are trapped between the valves. Liquid holdup is obtained by measuring
the trapped liquid volume and dividing by the total volume of the trapped section. A differential pressure
transmitter was used to measure the liquid volume. One pressure tap was located at the bottom of the lower
end of the trapped section, and the other was located at the top of the higher end, as shown in Figure 4.
Calibration was done to relate the liquid holdup to the measured differential pressure, as shown in Figure 6
for 8 different calibration runs. The final equation obtained to calculate the liquid holdup is given as
(2)
8 SPE-191543-MS

Figure 6—Calibration of liquid holdup as afunction of the measured pressure differential

where, HL,QCV is the liquid holdup measured through the QCV system and DPT is the pressure differential
measured at the trapped section.
During the experiments the measured differential pressure was corrected with the differential pressure
when there is no liquid in the trapped section (DPT zero). Similar to the pressure gradient measurement, it
was noted that the DPT zero varied with temperature. A collection of data for DPT zero at different system
pressures and temperatures is shown in Figure 7 (right). In this case, the system pressure did not affect the
DPT zero, however temperature did. The following fitting equation was used to correct the liquid holdup
calculation:
(3)

Figure 7—Variation of the differential pressure value with temperature for all system
pressures for the pressure gradient (left) and liquid holdup (right) measurements

A wire-mesh sensor (WMS) is used to measure the liquid and gas distribution at a cross section of the
pipe over time. It is made of a bundle of 32 parallel wires in the vertical direction and another bundle of
32 parallel wires in the horizontal direction. The two wire bundles are offset from each other, as shown in
SPE-191543-MS 9

Figure 8. An electronic signal is transmitted through the vertical wires and measured by the receiver through
the horizontal wires. The returning signal has information about the capacitance of the fluid between each
of the 1024 wire crossings. The frequency of acquisition is 2500 Hz, and the data was recorded for 4.5 min
for the pseudo-slug pattern and 30 s for stratified or annular flow pattern.

Figure 8—Schematic of the wire-mesh sensor

Calibration files were obtained with the pipe full of liquid and full of gas, which gave the capacitance
at each of the 1024 points when there was only liquid or gas, and were used to interpolate the measured
data and obtain a local liquid holdup at each of the points. A MATLAB code was used for further post-
processing of the data. Reconstruction of the interface was achieved by using a threshold for the liquid
holdup at each of the measured points. The interface was defined at the boundaries of liquid and gas points.
After the interface was reconstructed, all other variables were calculated over time such as liquid holdup,
liquid height, interfacial length, wetted wall fraction and interfacial roughness.
The Canty visualization system is composed of a high-speed camera and a high-resolution camera
mounted into a swivel, as schematically represented in Figure 9. The two cameras are positioned at 180°
from each other with two light sources located at the other quadrants. The swivel allows for rotation of the
camera around the pipe perimeter. The high-resolution camera shows a wider image and is the main way
to visualize the flow, while the high-speed camera is used to analyze details of the flow. The high-speed
camera can record videos up to 5000 frames per second. The cameras were positioned at a 45° angle from
vertical with the high-speed camera focusing at the liquid-gas interface, and the high-resolution camera
focusing on the upper wall of the pipe, as shown in Figure 9.
10 SPE-191543-MS

Figure 9—Schematic of the Canty visualization system rotated at 45° from vertical

The overall measurement uncertainty is summarized in Table 1.

Table 1—Measurement uncertainty summary

Measured variable Uncertainty

Liquid superficial velocity U = ±0.0003 [m/s]

Gas superficial velocity U = ±0.01 × vSg [m/s]

Pressure gradient U = ±2.7 [Pa/m]

QCV liquid holdup (no pseudo-slugs) U = ±0.0021 [-]

WMS liquid holdup (pseudo-slugs) U = ±0.0109 ln(HL)+0.0714

Experimental results
This section presents the experimental results: flow pattern analysis based on the flow visualization and
WMS data, pressure gradient, liquid holdup and interfacial roughness measurements. Comparison of the
measurements to existing predictive models is also presented. Further results and details were presented
by Rodrigues (2018).

Flow pattern
Three distinct flow patterns were observed during the experiments. At moderate superficial gas velocities,
vSg, stratified flow pattern is observed. At lower vSg, the gas momentum is not large enough to overcome
gravity to drag the liquid in a stratified configuration resulting in an intermittent behavior. Liquid has
accumulated at several locations in the test section until it is pushed downstream of the pipe by the gas.
These liquid accumulations do not touch the upper part of the pipe wall; hence, they are called pseudo-
slugs. For the higher range of superficial gas velocities, the deposition of entrained liquid droplets creates
a thin liquid film along the entire pipe wall circumference. At some point, this thin film becomes thicker
and faster, and this pattern is defined as annular. Even for the annular pattern, it is still noticeable that the
majority of the liquid flows at the pipe bottom in a stratified configuration.
The flow patterns were identified based on video observations. Images are included here in an attempt
to describe the observed flow patterns. In the figures, both cameras are positioned as depicted in Figure 9,
with the high-speed camera focusing on the interface and the high-resolution camera focusing on the upper
part of the pipe wall.
Figure 10 shows images of the flow in the stratified configuration. The high-speed camera image shows
the liquid film at the bottom and the gas flowing over the liquid. Two red lines are included; the continuous
SPE-191543-MS 11

line represents the main liquid-gas interface, while the dashed line represents a small film that is created
due to the passage of waves. Above the dashed line there is no noticeable liquid flow. In the high-resolution
camera image, there is a red arrow pointing to the liquid-gas interface, which is easy to distinguish. A red
circle shows a position at the top of the pipe. In the videos, this position is observed to identify any moving
liquid film. In the case of stratified flow, there is no noticeable film moving at this position.

Figure 10—High-speed (left) and high-resolution (right) cameras image

of flow in stratified pattern (p = 400 psig, vSL =0.03 m/s, vSg = 5 m/s)

Images of the flow in annular flow pattern are shown in Figure 11. This figure is for the lowest superficial
gas velocity classified as annular flow (for p = 400 psig and vSL = 0.03 m/s), while Figure 12 represents the
largest superficial gas velocity data-point obtained (for the same pressure and superficial liquid velocity).
Both are classified as annular flow due to the visible moving liquid film at the upper pipe wall – position
highlighted with a red circle in the high-resolution camera image. The high-speed camera shows one
continuous red line representing the interface between the bottom liquid film and the gas core. However,
above the red line there is another liquid film, much thinner, but still moving. This thin film is created by
liquid droplets that deposit on the pipe wall. The flow pattern is called annular due to this thin film covering
the entire circumference. However, the main liquid flow is still at the pipe bottom. The differences in the
images of Figure 11 and Figure 12 are due to the thin film that is thicker and faster and there are more
entrained droplets at higher vSg.
12 SPE-191543-MS

Figure 11—High-speed (left) and high-resolution (right) cameras image of flow

in the beginning of annular pattern (p= 400 psig, vSL = 0.03 m/s, vSg = 8 m/s)

Figure 12—High-speed (left) and high-resolution (right) cameras image

of flow in annular pattern (p = 400 psig, vSL =0.03 m/s, vSg = 11 m/s)

For the pseudo-slug flow pattern, one image cannot completely describe the flow. Therefore, Figure 13
shows a sequence of images of a pseudo-slug passage. The first image (from left to right) shows a moment
before the pseudo-slug arrival when the flow is similar to the stratified flow. However, the recorded high-
speed videos show that, at some point, the bottom layers of the liquid film start flowing backwards. The
second image shows the moment a pseudo-slug arrives, and the third image shows the pseudo-slug body. The
fourth image shows the flow after the pseudo-slug is gone. A stratified-like pattern is reestablished with a
higher liquid height than the previous one. With time, the liquid height decreases until the cycle starts again.
SPE-191543-MS 13

Figure 13—Sequence of images of a pseudo-slug passage from high-speed (top)

and high-resolution (bottom) cameras (p = 400 psig, vSL = 0.03 m/s, vSg = 2.5 m/s)

Analysis of the wire-mesh sensor (WMS) data is also useful for the flow pattern definition. Figure 14
presents the time-averaged interface geometry for the same flowing conditions as presented in the previous
images. Although only a few experimental conditions are shown here, the results are representative of other
experimental conditions considered in this study.

Figure 14—Time-averaged interface geometry obtained with the wire-mesh sensor for p = 400 psig and vSL = 0.03 m/s

The most important feature in Figure 14 is the characteristic flat geometry of the interface. The WMS
resolution is 4 mm, which is not enough to detect the thin film at the pipe wall in the annular pattern. Previous
studies, in smaller diameter pipes, assumed that the transition from stratified to annular flow pattern is due
to phenomena such as wave spreading or secondary flow. For this case, the main interface should become
curved (such as the ARS or double circle models) until the curvature was large enough to cover the whole
pipe wall perimeter. The results of Figure 14 show a different configuration for the conditions considered
in this study. The main interface retains the flat geometry even when the thin film is flowing at the pipe
upper wall. This leads to the conclusion that the transition from stratified to annular flow patterns in larger
diameter pipes is due to the entrained droplets that flow to the top of the pipe and deposit at the wall.
In Figure 14, the pseudo-slug interface shows an increase in liquid towards the sides. This is due to the
pseudo-slug passage, when a great portion of liquid wets the wall and slowly flows back to the bottom.
14 SPE-191543-MS

Moreover, the figure shows a time-averaged value, which is not the best way to represent the pseudo-slug
interface.
Based on the flow pattern definition previously mentioned, the entire data-set was classified and plotted
in a flow pattern map shown in Figure 15. The horizontal scale is represented by the superficial gas Froude
number, given as , where ρg and ρL are the gas and liquid densities, vSg is the superficial
gas velocity, D, is the pipe diameter, β is the pipe inclination angle, and g is the gravity acceleration constant.

Figure 15—Experimentally observed flow pattern map

The results in Figure 15 show that the superficial gas Froude number is a good dimensionless number
to represent the flow pattern transitions at different system pressures for a given superficial liquid velocity.
The Froude number represents the ratio of the gas drag force over the gravity force acting in the liquid
phase. Both flow pattern transitions are well represented by this ratio. The stratified to pseudo-slug transition
occurs when the gravity acting at the bottom liquid layer is dominant over the gas inertia that drags the
liquid through the interface resulting in liquid accumulation. Al-Sarkhi et al. (2017) used Kelvin-Helmholtz
linear instability in oil-water stratified flow and obtained the mixture Froude number, FrM, as the correlating
number when making the equations dimensionless. They used several flow pattern data to show that the FrM
correlates well with the flow transitions from stratified to non-stratified. Since the vSL is much smaller than
vSg in the present study, the mixture Froude number and superficial gas Froude numbers are practically the
same. The stratified to annular transition occurs when the entrained droplets reach the top of the pipe. Gravity
acts on the liquid droplets to keep them at the bottom, but when gas inertia is dominant, the associated
turbulence pushes the droplets to the top.

Pressure gradient
Figure 16 (left) shows the measured pressure gradient as a function of vSg and vSL for the system pressure
of 300 psig, and is representative of the other pressures. The dashed lines represent the flow pattern
SPE-191543-MS 15

transitions, pseudo-slug to stratified (PS-ST), and stratified to annular (ST-AN). The calculated single-phase
gas pressure gradient is also included for comparison.

Figure 16—Experimentally observed pressure gradient as a functions of vSg for the system pressure of
300 psig (left) and as a function of the superficial gas Froude number for all system pressures (right)

The pressure gradient results show two different trends, depending on the flow pattern. In the pseudo-
slug flow pattern, the flow is gravity dominated. A decrease in the gas superficial velocity represents a lower
gas fraction in the flow, which increases the pressure gradient. In stratified and annular flow patterns, the
flow is friction dominated, and an increase in the superficial gas velocity increases the pressure gradient. In
both regions, vSL has a minor effect on pressure gradient when compared to vSg.
Comparing the values with the calculated single-phase gas flow confirms the fact that even the small
amount of liquid greatly increases the pressure gradient. The errors in assuming a single-phase gas flow are
very high for the pseudo-slug pattern and are considerable for the highest vSg in the segregated pattern.
Figure 16 (right) shows the pressure gradient for all experimental conditions as a function of the
superficial gas Froude number. The results show a linear trend for stratified and annular flow patterns. Data
at different system pressures collapse for the same vSL, which shows that the superficial gas Froude number
is an important parameter to scale the pressure gradient at different system pressures.
A useful way to combine the data and analyze the effect of system pressure, superficial liquid velocity
and pipe diameter was given by Al-Sarkhi and Sarica (2010). The authors proposed the definition of the
dimensionless parameters and .
Figure 17 presents the pressure gradient results for all system pressures using these dimensionless
parameters. In the segregated flow pattern, the data for all system pressures and vSL collapse into one line
with negative slope, represented by the black dashed line. This agrees with the findings of Al-Sarkhi and
Sarica (2010) when proposing the scaling parameters. In the pseudo-slug flow pattern, the data for different
system pressures collapse into a line for each vSL. Although the lines for different vSL do not collapse with
each other, they all present the same slope, as represented in the figure by the colored dashed lines. This
observation is surprising, since the scaling parameters were obtained from a stratified flow model. Finally,
the colored arrows indicate the X* that correspond to the liquid film reversal as modeled by Fan et al. (2018)
16 SPE-191543-MS

for each vSL. The film reversal is related to the onset of pseudo-slug flow. The figure shows that the onset
calculated by Fan et al. (2018) model is in good agreement with the observed initiation of pseudo-slug flow.

Figure 17—P* as a function of X* for this study data. Open symbols indicate data
in segregated pattern andfilledsymbols indicate data in the pseudo-slug pattern

Figure 18 presents a comparison of this study data with other available data obtained with different pipe
diameters, fluid properties and flow rates. The scaling variables X* and P* are used. Only this study and
Fan (2005) data were obtained with a pipe inclination of 2°. The other data sets correspond to horizontal
configurations but with different pipe diameters and fluids properties. The data from this study and from Fan
(2005) present similar trends. Vuong (2016) used the same facility and similar flow conditions, however in
a horizontal configuration. Therefore, the difference in the results is due to the pipe inclination. Data from
Viana (2017), Karami (2015) and Tayebi et al. (2000) seem to agree with each other, however, present a
greater P* than the present study data, which was not expected.
SPE-191543-MS 17

Figure 18—P* as function of X* for this study and other data

Liquid holdup
For the high vSg range, where the flow pattern is either stratified or annular, the liquid holdup was measured
with the trapped liquid by the quick-closing valves (QCV). For the lower vSg range, when pseudo-slugs
occur, the use of QCV became questionable owing to the comparable size of the pseudo slugs and the
trapped section. For this case, the trapped volume does not represent a good sample of the flow. Thus, the
liquid holdup was measured through the post-processing of wire-mesh sensor (WMS) data for the pseudo-
slug region.
Results are shown in Figure 19 (left) for the liquid holdup variation as a function of both superficial
velocities and system pressure in the range where the QCV are used. In the stratified and annular flow
patterns, the liquid holdup is largely affected by both superficial liquid and gas velocities. The superficial
liquid velocity effect is larger than the pressure effect since data points for the same vSL at different pressures
are closer as compared with same pressure but different vSL case.
18 SPE-191543-MS

Figure 19—Liquid holdup measured with the QCV (left) and WMS (right) for all system pressures

The measured liquid holdup for the lower range of vSg, measured with the WMS, is shown in Figure 19
(right) for the three system pressures. When the flow pattern is pseudo-slug, the holdup is largely affected
by the superficial gas velocity but not much by the superficial liquid velocity. Results are similar for the
same system pressure regardless of the superficial liquid velocities.
The combined effects of pressure and superficial velocities are seen in Figure 20, where all the liquid
holdup results are plotted as a function of the superficial gas Froude number. There are three different trends
that are related to the flow patterns. For the pseudo-slug pattern, the data collapsed into one curve. That
means that the superficial liquid velocity has less influence in the liquid holdup, and the superficial gas
Froude number is a good dimensionless number to capture the pressure effect. When the flow is in the
segregated pattern, the superficial gas Froude number is still good in capturing the pressure effects. The
data for the same vSL at different pressures collapse in a single line (in log-log scales). However, different
vSL show different trends of liquid holdup. The lines have similar slope. At very high Froude numbers, the
HL does not follow the linear relation anymore. This occurs earlier at lower vSL. One possible reason for this
result is that the resolution of the liquid holdup measurements is lower at lower holdups.
SPE-191543-MS 19

Figure 20—Liquid holdup for all system pressures as a function of FrSg. Open symbols
indicate data in segregatedpattern and filled symbols indicate data in the pseudo-slug pattern

Interfacial roughness
Interfacial roughness is obtained by analyzing the time series of the liquid height obtained through the
WMS. Roughness may be defined in several ways. Here, the root-mean square, ,
is used, where n is the number of measure liquid height values, xi is the ith measured liquid height value,
and is the average liquid height.
Figure 21 (left) shows the RRMS for the system pressure of 300 psig as a function of vSg and vSL and is
representative of the other system pressures. The lower vSg represents the pseudo-slug flow pattern where
large liquid structures increase the roughness. Once the flow pattern turns into stratified and annular, the
interfacial roughness tends to be constant and independent from vSg and vSL. This is an important observation
that the roughness of the waves does not change with both parameters. Figure 21 (right) highlights this
effect by showing the roughness for all data as a function of the superficial gas Froude number. A constant
value of around 2 mm is reached for almost all conditions. The exception is the lowest vSL at higher vSg.
This is due to the very small amount of liquid flowing at the bottom liquid film. The WMS resolution is not
enough to detect the waves, and the roughness calculation is compromised.
20 SPE-191543-MS

Figure 21—Measured interfacial RMS roughness as a functions of vSg for the system pressure of 300
psig (left) and as a function of the superficial gas Froude number for all system pressures (right)

A comparison between current and the horizontal flow data reported by Vuong (2016) is shown in Figure
22. The main difference is noted in the low vSg range, since the horizontal data show a transition from a
smooth interface (RRMS ~ 0) whereas the inclined data shows the very high pseudo-slug roughness. However,
both data reach a similar value for stratified and annular flow, at FrSg > 0.5. This result shows that the interface
roughness is constant with a value of RRMS ~ 2 mm regardless of vSg, vSL, system pressure and inclination.

Figure 22—Comparison between interfacial roughness measurements in

this study and the results of Vuong (2016)for all system pressures and vSL
SPE-191543-MS 21

Comparison to existing models

A comparison between the experimental results for flow pattern, pressure gradient and liquid holdup against
other models is presented in this section. The TUFFP's Unified Model 2017, commercial flow simulator
OLGA version 7.3.5 and Rodrigues et al. (2018) model are used for this comparison. The model proposed
by Fan et al. (2018) for the onset of pseudo-slugs is also included in the flow pattern comparison.
Figure 23 shows the experimentally obtained flow pattern map for the system pressures of 200 and 400
psig, along with the model predictions. The Unified Model predicts the transition from stratified to annular
at gas superficial velocities much higher than experimentally observed. The stratified to annular transition
criteria in the Unified Model is based on the increase of the wetted wall fraction with the gas Froude number.
The double circle model is utilized to identify when the wetted wall fraction reaches the whole perimeter.
The experimentally observed mechanism for transition is the droplet entrainment, which occurs earlier than
the increase in wetted wall fraction. This transition is not shown for the OLGA simulations. The simulator
uses the term segregated flow for these conditions, and the specific annular term is used only for pipes with
larger inclinations.

Figure 23—Observed and predicted flow pattern map for the system pressures of 200 and 400 psig

Both models, OLGA and Unified Model, do not consider pseudo-slug as a flow pattern. The pseudo-
slug flow is included within the slug flow pattern once stratified flow is not obtained. Experimentally, the
transition to slug flow was not achieved in this study. This transition will occur either at lower superficial
gas or larger superficial liquid velocities. Here, a comparison is made for the beginning of the intermittent
flow region (pseudo-slug flow in the experiments and slug flow in the models). The Unified Model gives
a good prediction for this transition, especially for lower vSL. OLGA prediction shows a larger effect of vSL
in the transition; however, it always predicts the transition at lower vSg than obtained experimentally. Both
models show better prediction at higher system pressures. Fan et al. (2018) model for the onset of pseudo-
slug shows very good predictions for the considered experimental data.
Figure 24 (left), Figure 25 (left) and Figure 26 (left) show a comparison between the measured and
calculated pressure gradients using the Unified Model, OLGA simulator and Rodrigues et al. (2018) model,
respectively. The comparison shows two different results for segregated and pseudo-slug flows. When the
flow is in the segregated pattern, the Unified Model underpredicts the pressure gradient for the whole
range. OLGA presents good agreement for the lower and higher pressure gradient ranges, and slightly
22 SPE-191543-MS

under predicts the data in the mid-range of pressure gradient. Rodrigues et al. (2018) model results are only
compared to the data in the stratified and annular flow patterns. The prediction is good, except for a few
points in the very low range of vSg when the flow is starting to transition to pseudo-slug flow. When the
flow is in the pseudo-slug pattern, the Unified Model shows a good agreement while OLGA slightly over
predicts the data.

Figure 24—Comparison of measured pressure gradient and liquid holdup to the calculated with the Unified Model
2017. Open symbols indicate data in segregated pattern and filled symbols indicate data in the pseudo-slug pattern

Figure 25—Comparison of measured pressure gradient and liquid holdup to the calculated with the OLGA 7.3.5.
Open symbols indicate data in segregated pattern and filled symbols indicate data in the pseudo-slug pattern
SPE-191543-MS 23

Figure 26—Comparison of measured pressure gradient and liquid holdup to the calculated with Rodrigues et al. (2018) model

Figure 24 (right), Figure 25 (right) and Figure 26 (right) show the liquid holdup comparison between
experiments and the calculations of the Unified Model, OLGA and Rodrigues et al. (2018). For lower liquid
holdups, when the flow pattern is stratified and annular, the Unified Model under predicts the data while
OLGA gives a good agreement. At the highest experimental holdups, the flow is similar to slug flow and
both models give good predictions. In the intermediate region, the flow pattern is transitioning to the pseudo-
slug. For this case, both models consider slug flow relationships, which overpredicts the holdup. Rodrigues
et al. (2018) comparison is only made for the stratified and annular flow patterns and the agreement is good.
Model predictions are reasonable for stratified and annular flows, while they present larger errors for the
pseudo-slug pattern. However, when the flow is in the pseudo-slug pattern, it is better to use a slug-flow
model than a stratified flow model.

Interfacial friction factor

The shear stress between the liquid and gas phases is one of the most important closure relations in separated
two-phase flow, and it has been intensively studied by several authors. However, since direct measurement
is not available, the studies used different methods to calculate the friction factor. In the earlier models such
as Taitel and Dukler (1976), it was assumed that the interfacial friction factor is equal to the gas-wall friction
factor. This assumption gives good results when the interface is smooth, i.e. there are no interfacial waves.
One of the first detailed studies on the interfacial friction factor in wavy interfaces was performed by
Cohen and Hanratty (1968). The authors used a channel 0.3 m wide and 0.0254 m high and measured the
pressure gradient, gas local velocities at different vertical positions and interfacial wave characteristics.
They presented the RMS (root-mean squared) wave roughness as a function of superficial gas velocities
and the measured interfacial friction factor. The interfacial friction factor was obtained by fitting a turbulent
velocity profile to the measured local velocities in the gas phase. When evaluating the variation of interfacial
friction factor as a function of the increase in superficial gas velocity, their results showed a rapid increase
in friction at small gas velocities; after that, the friction reached a constant value for large gas velocities.
Later, Shoham and Taitel (1984) used Cohen and Hanratty's results and assumed a constant value for the
friction factor when small interfacial waves occur. By converting the original number to Fanning's friction
factor, they obtained a value of fi = 0.0142. Later, Newton et al. (1999) and Liu et al. (2007) presented
back-calculated interfacial friction factor results based on the pressure gradient measurements. Both studies
24 SPE-191543-MS

follow the same trend as presented by Cohen and Hanratty of increasing interfacial friction factor with
increasing flow rates up to a value close to 0.0142. Analysis of the data presented by Karami et al. (2016)
and Vuong et al. (2017) also agrees with this constant value.
Andritsos and Hanratty (1987) proposed a model for the interfacial friction factor based on the wave
regime. When the gas velocity is lower than a transition velocity, vSg,t = 5 m/s (ρg0/ρg)0.5, the interface is
smooth and fi=fg. For higher velocities, waves are formed and the friction factor is increases with the gas
velocity. Andritsos et al. (2008) revisited this concept and divided the model into three regions: smooth
interface, 2-D waves, and Kelvin-Helmholtz (K-H) waves. When the interface is smooth, fi=fg is used. After
a critical gas velocity vSg,2D, small, 2-D waves are formed and the friction factor is increased. Further increase
of the superficial gas velocity leads to the generation of K-H waves and the friction factor is calculated
with another function.
Another approach is to consider the interface as a solid wall with the waves represented by a sand-grain
equivalent roughness. The earlier study of Nikuradse (1932) with roughened-wall pipes is a cornerstone on
friction factor in pipes studies. The author carried out a comprehensive experimental study with different
pipe diameters, sand grain sizes and flow rates, and measured the velocity profile and pressure gradient.
The obtained results showed a unique graphical relationship between friction factor and Reynolds number.
Although his results are mainly used for small-scale roughness in pipes, it is important to realize that his
experiments included sand grains with diameters up to 1.6 mm. He concluded that at some point the friction
factor is not a function of Reynolds number anymore, but only of the dimensionless pipe roughness. This
happens when the laminar sublayer height inside the turbulent boundary layer is smaller than the roughness
itself, and is called fully-rough flow. In this case, the friction factor was given as a unique function of the
pipe equivalent sand-grain roughness, ks. The pipe roughness was assumed as the average projected area
formed from rounded sand particles glued to the pipe walls. Nikuradse proposed Eq. (4) which relates the
friction factor with the sand-grain roughness, ks:

(4)

Paras et al. (1994) and Vlachos et al. (1997) studied horizontal stratified flow with measurements of
the interfacial RMS roughness. They proposed models for the friction factor based on the fully rough flow
concept and showed an equation for the equivalent sand-grain roughness.
Afzal (2007) and Adams et al. (2012) proposed correlations to obtain equivalent sand-grain roughness
in pipes from measurable parameters, such as the RMS roughness. However, these studies were concerned
with the small-scale commercial pipes roughness. The proposed correlations may not be able to give good
results for larger waves roughness.
Other authors studied experimentally and numerically the effect of wave-like walls in the pipe friction
factor. Hudson et al. (1996) carried out experiments in a rectangular-shaped channel with a cross section
of 5.08 × 61 cm with the bottom wall following a sinusoidal wave geometry with wave length of 5.08 cm
and amplitude of a = 2.54 mm. They measured the velocity at different distances from the wall and fitted
a logarithmic equation following the turbulent law of the wall. Pressure gradient measurements were also
conducted. The authors concluded that the friction factor given by Eq. (4) could be used if the equivalent
sand-grain roughness was equal to the total amplitude (peak-to-peak), ks = 2a = 5.1 mm. This means that the
equivalent sand-grain roughness is, indeed, related to the average projection of the roughness as Nikuradse
stated. In the case of wave-like shapes, the average projection of roughness is the total amplitude 2a.
Cherukat et al. (1998) used direct numerical simulation to relate a wavy-wall to the wall roughness. They
used the same geometrical configurations of the work of Hudson et al. (1996). The authors presented several
parameters and velocity distributions along the simulation domain. However, the most important result for
SPE-191543-MS 25

this study was that the equivalent roughness also matched the same results as presented by Hudson et al.
(1996), and Eq. (4) could be used if ks = 2a.
In the experimental results from the present study, the interfacial RRMS roughness reaches a constant value
of approximate RRMS = 2 mm. This occurs at FrSg number above 0.5 at all system pressures and vSL. If the
interfacial waves are assumed as sinusoidal, the total amplitude is related to the RMS as 2a = 2.83RRMS.
For a value of 2 mm, the total amplitude is 2a = 5.66 mm. Using the value of ks = 2a = 5.66 mm in Eq.
(4), the value of fI = 0.0155 is obtained. This is very close to the common used value of 0.0142. Since the
experiments show a constant interfacial roughness, the interfacial factor is expected to be constant as well.
This result shows that the interfacial friction factor can be related to the measured roughness through the
fully rough regime equation.

Conclusions
Two-phase gas-liquid flow experiments were carried out in a 6 in ID pipe with inclination of 2° with
the horizontal at three different system pressures (200, 300 and 400 psig). The superficial liquid and
gas velocities are that of low-liquid loading flow, which is common in gas/condensate pipelines. Several
instruments were used to measure relevant flow parameters.
Three distinct flow patterns were observed during the experiments: pseudo-slug, stratified and annular.
In the stratified flow pattern, the two phases are segregated with the liquid flowing at the bottom and the gas
on the top of the liquid. The interface between the two phases is flat with roll waves in the axial direction.
The flow is gas-dominated and much more affected by vSg changes than vSL. Liquid holdup variations are
small and affected by both superficial velocities. Pressure effects on the pressure gradient and liquid holdup
are well captured by the superficial gas Froude number. In annular flow, a thin liquid film covers the whole
pipe wall circumference. However, the bulk of the liquid still flows segregated at the pipe bottom and the
flow is very similar to the stratified pattern.
Transition from the stratified to the annular flow pattern occurs when the liquid droplets entrained in
the gas core reach the upper part of the cross-section. The droplets are deposited at the upper pipe wall
creating a liquid film that covers the entire pipe wall circumference. The effect of pressure in this flow
pattern transition is well captured by the superficial gas Froude number.
For the stratified and annular flow patterns, the interfacial roughness is measured and found to reach
a constant value for larger vSg. It is shown that this constant roughness value can be used to estimate the
interfacial friction factor by converting the wave roughness into an equivalent sand-grain roughness.
Transition from stratified to pseudo-slug flow pattern occurs at lower superficial gas velocities. The
intermittent flow is different from the conventional slug flow pattern since the liquid slugs do not reach the
top of the pipe. Therefore, the pattern is called pseudo-slug. The effect of pressure on this transition is well
captured by the superficial gas Froude number.

Acknowledgements
The authors would like to thank the Tulsa University Fluid Flow Projects (TUFFP) member companies and
Petrobras for their support.

References
Adams, T., Grant, C. and Watson, H. 2013. A simple algorithm to relate measured surface roughness to equivalent
sand-grain roughness. Int. J. of Mechanical Engineering and Mechatronics, 1 (1): 66–71. https://dx.doi.org/10.11159/
ijmem.2012.008.
Afzal, N. 2007. Friction factor directly from transitional roughness in a turbulent pipe flow. J. of Fluids Engineering, 129
(10): 1255–1267. https://dx.doi.org/10.1115/1.2776961.
Alsaadi, Y., Pereyra, E., Torres, C. et al. 2015. Liquid loading of highly deviated gas wells from 60° to 88°. SPE Annual
Technical Exhibition. Houston, TX, USA, 28-30 September. SPE-174852-MS. https://doi.org/10.2118/174852-MS.
26 SPE-191543-MS

Al-Sarkhi, A. and Sarica, C. 2010. Power-law correlation for two-phase pressure drop of gas/liquid flows in horizontal
pipelines. SPE Projects, Facilities & Construction, 5 (4): 176–182. SPE-138516-PA. https://doi.org/10.2118/138516-
PA.
Al-Sarkhi, A., Pereyra, E., Mantilla, I. et al. 2017. Dimensionless oil-water stratified to non-stratified flow pattern
transition. Journal of Petroleum Science and Engineering, 151: 284–291. https://doi.org/10.1016/j.petrol.2017.01.016.
Andritsos, N. and Hanratty, T. J. 1987. Influence of interfacial waves in stratified gas-liquid flows. AIChE Journal, 33
(3): 444–454. https://doi.org/10.1002/aic.690330310.
Andritsos, N., Tzotzi, C. and Hanratty, T. J. 2008. Interfacial shear stress in wavy stratified gas-liquid two-phase flow.
5th European Thermal-Sciences Conference, the Netherlands.
Badie, S., Hale, C. P., Lawrence, C. J. et al. 2000. Pressure gradient and holdup in horizontal two-phase gas-
liquid flows with low liquid loading. Int. J. of Multiphase Flow, 26 (9): 1525–1543. https://doi.org/10.1016/
S0301-9322(99)00102-0.
Chen, X. T., Cai, X. D., and Brill, J. P. 1997. Gas-Liquid Stratified-Wavy Flow in Horizontal Pipelines. ASME J. Energy
Res. Tech., 119 (4): 209–216. https://dx.doi.org/10.1115/1.2794992.
Cherukat, P., Na, Y., and Hanratty, T. J. 1998. Direct numerical simulation of a fully developed flow over a wavy wall.
Theoretical and Computational Fluid Dynamics, 11: 109–134. https://doi.org/10.1007/s001620050083.
Churchill, S.W. 1977. Friction Factor Equations Spans all Fluid-Flow Regimes. Chem. Eng., 7: 91–92.
Cohen, L. S. and Hanratty, T. J. 1968. Effect of waves at gas-liquid interface on a turbulent air flow. J. Fluid Mechanics,31
(3): 467–479. https://doi.org/10.1017/S0022112068000285.
Fan, Y., Wang, Q., Zhang, H., et al. 2007. A model to predict liquid holdup and pressure gradient of near-horizontal wet-
gas pipelines. SPE Projects, Facilities & Construction, 2 (2): 1–8. SPE-95674-PA. https://doi.org/10.2118/95674-PA.
Fan, Y., Pereyra, E. and Sarica, C. 2015. Experimental study on the onset of intermittent flow and pseudo-slug
characteristics in upward inclined pipes. 17th International conference on Multiphase Production Technology, Cannes,
France, 10-12 June.
Fan, Y., Pereyra, E. and Sarica, C. 2018. Onset of liquid film reversal in inclined pipe. SPE Journal. SPE-191120-PA (in
press). https://doi.org/10.2118/191120-PA.
Hart, J., Hamersma, P.J., and Fortuin, J.M.H. 1989. Correlations predicting frictional pressure drop and liquid holdup
during horizontal gas-liquid pipe flow with a small liquid holdup. Int. J. Multiphase Flow, 15: 947–964. https://
doi.org/10.1016/0301-9322(89)90023-2.
Hudson, J. D., Dykhno, L. and Hanratty, T. J. 1996. Turbulence production in flow over a wall. Experiments in Fluids,
20: 257–265.
Karami, H., Torres, C.F., Pereyra, E., et al. 2015 Experimental investigation of Three-Phase Low Liquid Loading Flow.
Oil & Gas Facilities, 5 (02): pp. 45–56. SPE-174926-MS. https://doi.org/10.2118/174926-MS.
Kjolaas, J. Unander, T. E., Wolden, M., et al. 2015. Experiments for low liquid loading with liquid holdup discontinuities in
two- and three-phase flows. 17th International Conference on Multiphase Production Technology, 10-12 June, Cannes,
France.
Langsholt, M. and Holm, H. 2007. Liquid accumulation in gas-condensate pipelines – an experimental study. BHR Group
Multiphase Production Technology 13.
Liu, Y., Zhang, H. and Wang, J. 2007. Estimation of liquid wall and interfacial shear stress in horizontal stratified gas-
liquid pipe flow. AIP Conference Proceedings, 914 (1): 610–616. https://doi.org/10.1063/1.2747488.
Meng, W., Chen, X. T., Kouba, G. E. et al. 2001. Experimental study of low-liquid-loading gas-liquid flow in near-
horizontal pipes. SPE Production & Facilities, 16 (4): 240–249. SPE-74687-PA. https://doi.org/10.2118/74687-PA.
Newton, C. H, Behnia, M. and Reizes, J. A. 1999. The effect of liquid viscosity on gas wall and interfacial
shear stress in horizontal two-phase pipe flow. Chem. Eng. Science, 54: 1071–1079. https://doi.org/10.1016/
S0009-2509(98)00423-0.
Nikuradse, J. 1933. Stromungsgesetze in rauhen Rohren. VDI-Forschungsheft 361. Beilage zu "Forschung auf dem
Gebiete des Ingenieurwesen"s Ausgabe B Band 4, July/August.
Nino Amezquita, O. G., Enders, S., Jaeger, P. T. et al. 2010. Interfacial properties of mixtures containing supercritical
gases. The Journal of Supercritical Fluids, 55: 724–734. https://doi.org/10.1016/j.supflu.2010.09.040.
Paras, S. V., Vlachos, N. A. and Karabelas, A. J. 1994. Liquid layer characteristics in stratified-atomization flow. Int. J.
Multiphase flow, 20 (5): 939–956. https://doi.org/10.1016/0301-9322(94)90103-1.
Rodrigues, H. T., Pereyra, E. and Sarica, C. 2018. A model for the thin film friction factor in near-horizontal stratified-
annular transition two-phase low liquid loading flow. Int. J. Multiphase Flow, 102: 29–37. https://doi.org/10.1016/
j.ijmultiphaseflow.2018.01.017.
Rodrigues, H. T. 2018. Pressure effects on low-liquid loading two-phase flow in near-horizontal upward inclined pipes.
PhD dissertation, The University of Tulsa, Tulsa, Oklahoma.
SPE-191543-MS 27

Shoham, O. and Taitel, Y. 1984. Stratified turbulent-turbulent gas-liquid flow in horizontal and inclined pipes. AIChE
Journal, 30 (3): 377–385. https://doi.org/10.1002/aic.690300305.
Seibt, D., Vogel, E., Bich, E. et al. 2006. Viscosity measurements on nitrogen. J. Chem. Eng. Data, 51: 526–533. https://
dx/doi.org/10.1021/je050399c.
Span, R., Lemmon, E., Jacobsen, R. et al. 2000. A reference equation of state for the Thermodynamic properties of nitrogen
for temperatures from 63.151 to 1000 K and pressures to 2200 MPa. Journal of Physical and Chemical Reference
Data, 29 (6): 1361–1433. https://doi.org/10.1063/1.1349047.
Taitel, Y. and Dukler, A. E. 1976. A model for predicting flow regime transitions in horizontal and near horizontal gas-
liquid flow. AIChE Journal, 22 (1): 47–55. https://doi.org/10.1002/aic.690220105.
Tayebi, D., Nuland, S. and Fuchs, P. 2000. Droplet transport in oil/gas and water/gas flow at high gas densities. Int. J. of
Multiphase Flow, 26: 741–761. https://doi.org/10.1016/S0301-9322(99)00054-3.
Viana, F., Mantilla, I., Mohan, R. et al. 2016. Liquid entrainment in gas at high-pressure – Part 1: Experimental approach
and initial testing. 10th North American Conference on Multiphase Technology, Banff, Canada, 8-10 June.
Vlachos, N. A., Paras, S. V. and Karabelas, A. J. 1997. Liquid-to-wall shear stress distribution in stratified/atomization
flow. Int. J. of Multiphase Flow, 23 (5): 845–863. https://doi.org/10.1016/S0301-9322(97)00007-4.
Vuong, D. 2016. Pressure effects on two-phase oil-gas low-liquid loading flow in horizontal pipes. PhD dissertation, The
University of Tulsa, Tulsa, Oklahoma.
Vuong, D., Sarica, C., Pereyra, E. 2017. Pressure effects on pressure gradient and liquid holdup in two-phase oil-gas low-
liquid-loading flow in horizontal pipes. SPE Annual Technical Conference, San Antonio, Texas, USA, 9-11 October.
SPE-187229-MS.
Vuong, D., Sarica, C., Pereyra, E. et al. 2018. Liquid droplet entrainment in two-phase oil-gas low-liquid-
loading flow in horizontal pipes at high pressure. Int. J. Multiphase Flow, 99: 383–396. https://doi.org/10.1016/
j.ijmultiphaseflow.2017.11.007.
Zhang, H. Q. and Sarica, C. 2011. A model for wetted-wall fraction and gravity center of liquid film in gas/liquid pipe
flow. SPE Journal, 16 (3): 692–697. SPE-148330-PA. https://doi.org/10.2118/148330-PA.
Zolghadr, A., Riazi, M. Escrochi, M. et al. 2013. Investigating the effects of temperature, pressure, and paraffin groups
on the N2 miscibility in hydrocarbon liquids using the interfacial tension measurement method. Ind. Eng. Chem. Res.,
53: 9851–9857. https://dx/doi.org/10.1021/ie401283q.