Aic 16860
Aic 16860
Aic 16860
DOI10.1002/aic.16860
Investigating the effect of pressure on a vertical two-phase upward flow with a high
viscosity liquid
*
Abdulkadir Mukhtar1, Abolore Abdulahi2, Carol N. Eastwick2, Barry J. Azzopardi2, Ivar E. Smith3 and Tor E.
Unander3
1
Chemical Engineering Department, Federal University of Technology, Minna, Nigeria
2
Process and Environmental Engineering Division, University of Nottingham, Nottingham, United Kingdom
3
Multiphase Flow Laboratory, SINTEF Petroleum Research, Trondheim, Norway.
ABSTRACT
This paper presents void fraction and pressure gradient data for sulphur hexafluoride (SF6) with gas densities of 28
and 45 kg/m3 and oil (with viscosity 35 times that for water) in a 127 mm diameter pipe. The superficial velocities
of gas ranged from 0.1 to 3 m/s and those for liquid from 0.1 to 1 m/s, respectively. Measurements of void fraction
data was recorded using a capacitance wire mesh sensor (WMS) system, which permits the 3D visualization of the
flow patterns. All the data were obtained with a data acquisition frequency of 1000 Hz. A differential pressure
transducer was used to measure the pressure drops along the length of the pipe. The WMS provide time and cross-
sectionally resolved data on void fraction and from an analysis of its output, flow patterns were identified using the
characteristic signatures of probability density function (PDF) plot of time series of void fraction. The PDF plots
showed the single peak shapes associated with bubbly and churn flows but not the twin-peaked shape usually seen
in slug flows. This confirms previous work in larger diameter pipes but with less viscous liquids. For the bubble
and churn flows investigated, the pressure gradient was observed to decrease with an increase in gas superficial
velocity. Nevertheless, there was an insignificant observed effect of pressure on void fraction below certain
transitional flow rates, the effect however became significant beyond these values. In the present work, wisps appear
to be smaller, which might be due to the different fluid properties of the working fluids employed. In addition, wisps
are easily revealed as long as there is a transition between churn and annular flows regardless of the pressure.
Experimental data on void fraction and pressure gradient are compared against existing data. Reasonably good
This article has been accepted for publication and undergone full peer review but has not been through
the copyediting, typesetting, pagination and proofreading process which may lead to differences
between this version and the Version of Record. Please cite this article as doi: 10.1002/aic.16860
1. Introduction:
Effect of pressure has previously been investigated on two-phase gas–liquid flows especially
with water which has low viscosity. Most of the work however that has been carried out focused
mainly on air–water. Hence, most of the available industrial models and correlations have been
validated with experimental results from air–water data. In the current study, the effect of
pressure on two-phase high viscosity liquid and gas flows had been investigated in a large
diameter pipe. Using pressure gradient, rather than void fraction, has been a promising tool for
the identification of flow patterns as can be seen in Figure 1. Pressure gradients are easily
accessible to measurements and may contain sufficient information on the features of flow
patterns. The first attempt to relate the fluctuating pressure gradient to flow patterns was that of
1
Nishikawa et al. who investigated the statistical properties of static pressures of each flow
pattern in upward air–water flow in a 26 mm internal diameter pipe. Frequency distribution and
power spectrum density (PSD) function of static pressure signals for different flow patterns were
presented. However, the fluctuations of the static pressure signals consist of two parts. One part
of the fluctuations is due to the mechanical vibrations of the whole measuring system whilst the
other part is due to the flow fluctuation, or void fraction instability, in the measuring section. It
is obvious that only the latter fluctuation is related to the flow pattern.
Interestingly, more industries have recently been using large diameter vertical and steeply
inclined riser systems to reduce the pressure drop usually experienced with the use of small
|Page
great challenge for field applications especially in prediction of certain flow characteristics. In
effect, evidence shows that flow behaviour is greatly influenced by pipe diameters. Hence, in the
current study a 127 mm diameter pipe has been employed. In the work by 2 in a 150 mm diameter
pipe, no conventional large bubbles which normally occupy the majority of the pipe cross section
were observed over the range of flow rates where slug flow would normally appear. Hence,
Cheng et al. 2 inferred that there is a very gradual transition to a type of churn flow as the gas
rate is increased instead of the traditional slug flow in their pipe. In a similar investigation,
Ohnuki and Akimoto 3 reported that the churn flow is dominant in large diameter pipes under
the same conditions where small diameter pipes have slug flow. They classified the flow patterns
observed visually in a transparent pipe as undisturbed bubbly, agitated bubbly, churn bubbly,
churn slug and churn froth. However, this contradicts the previous investigation carried out by
4-6
other researchers especially with small diameter pipes. For instance, suggested that small
diameter pipes seem to exhibit bubble, slug, churn and annular flows for a two-phase gas–liquid
Figure 1: Dimensionless pressure gradient as s function of the dimensionless air flow rate in air–water flow in a
tube (data obtained by Owen 7 as reported by Hewitt 8)
2.Experimental Facility
2.1 Methodology
2.1.1 Procedure
A schematic diagram of the experimental facility used for the current study is as shown in Figure
2. The experiments were run with Sulphur Hexafluoride (SF6) and a mixture of Exxsol D80 and
|Page
non-toxic. Exxsol D80 is a de-aromatized aliphatic hydrocarbon oil commonly used in flow loop
experiments while Nexbase 3080 is a catalytically hydro-isomerized and dewaxed based oil
comprising of hydrogenated, highly iso-paraffinic hydrocarbons. The gas and oil flow rates as
well as the oil density were measured by two Coriolis meters. Measurement of the gas density
was done through manually weighing a gas sample obtained from the flow loop. A table for the
fluid properties is shown in Table 1. The medium flow loop is located at the SINTEF Multiphase
Flow Laboratory at Tiller, Trondheim, Norway. Oil in a storage tank was pumped into the test
pipe (internal diameter of 127 mm, height of 9 m) using a centrifugal pump while SF6 was
injected with the aid of a compressor into the mixer near the bottom and allowed to flow up
through the liquid. The pressure in the test section was initially at 7.5 bar, this was later lowered
to 4.5 bar after a series of experiments were carried out at the initial velocity conditions. The
velocity of the oil ranged from 0.1-1 m/s while that of the gas ranged between 0.1-3 m/s. The
mixer consisted of an annular section through which the liquid was introduced. This is discussed
further in Section 2.1.3. The gas was injected into the mixer through a series of 4 mm holes on
the wall of the capped central pipe. The two-phase mixer was mounted at the bottom of the test
pipe while the wire mesh sensor (WMS) was located at 6.5 m away from it.
In this facility, the gas used was SF6. The gas was initially compressed to a liquefied form at 80
bar in a cylindrical tank. This was sent into the separator (Figure 2) which already contains high
|Page
oil with viscosity of 1.76 mPa.s). The properties of the fluids are presented in Table 1. The two
In the two-phase mixer employed here (Figure 3(a)) which is similar to those used by 9-14, the
liquid was introduced into a short annulus from the bottom. Gas is then mixed through 4 mm
diameter holes at the inner wall of the annulus. The mixture then flows into the main pipe. This
design was chosen to: (i) produce small bubble sizes and (ii) to minimize the flow of liquid back
into the gas supply line. To avoid liquid from entering the gas injection holes, the shutdown
process involves stopping the liquid flow while gas flow is continuously circulated for a few
more minutes in order to expel all liquid in the test section. The gas flow is then stopped. The
holes were positioned in rings up the pipe with an axial spacing of 20 mm and a circumferential
spacing of 16 mm. Alternate holes are staggered to give a triangular array as can be seen in
Figure 3(b).
Figure 3: Two-phase mixing configuration showing (a) layout of the gas injection unit (c) Two-phase flow mixing
point
2.2 Instrumentation
Many liquid substances like organic liquids and crude oil are non-conducting, this imply that the
conductive WMS could not be used in such applications. Thus, a new WMS is required to
|Page
on the measurement of the electrical permittivity (capacitance), to cover the sensor's capabilities
to the detection of non-conducting fluids. It is worthy of mention that a number of papers have
been published since 1998 using both options of the WMS. For example, for the conductivity
WMS, studies were completed by 15-22. For the capacitance WMS, Da Silva et al. 23-26, Thiele et
27 10 28 11 29-31
al. , Hernandez-Perez et al. , Azzopardi et al. , Szalinski et al. , Abdulkadir et al.
presented various research studies. WMS have also been deployed for high pressure high
Local time varying void fractions were acquired by using the WMS measurement transducer
developed by Presser et al. 15, 17. In this study, a 32 × 32 wire configuration sensor was used that
had been hitherto applied for conductivity measurements but now adapted for capacitance
diameter, 4 mm wire separation within each plane, and 3 mm axial plane distance. The wires are
uniformly distributed over the circular pipe cross-section and in addition the spatial resolution
of the images created by the sensor is approximately 4 mm or 820 pixels across the full diameter,
which tallies to the wire separation within a single plane. The WMS works at frequencies of 5000
frames/seconds which permits small bubbles to be identified. In this work, data was acquired at
a frequency of 1000 frames/seconds for a 30 second experimental run period. An acrylic frame
supports the sensor and permits fixation into the text flow pipe section.
|Page
gaps of all crossing points by successfully applying an excitation voltage (sine wave of 5 MHz)
to each one of the sender electrodes at one wire plane whilst measuring in parallel the current
flowing toward receiver electrodes at the other wire plane. The non-activated transmitter wires
are grounded. This step guarantees that the electrical field distribution is concentrated along the
activated wire and permits for a sampling of only a well-defined section within the pipe, so that
the measured currents are explicitly related to the corresponding crossing point. For the
is subsequently applied. After digitizing, the measured data are sent to a computer where they
are processed and displayed. The method is able to generate up to 7,000 images per second.
Figure 4: (a) Wire mesh sensor (WMS) used during experimentation (b) 32 × 32 wire mesh sensor for pipe
measurement
There are two main intrusive effects of WMS on bubbles in the two-phase flow. The first one is
34
the bubble break-up. The second one is the bubble deceleration. Wangjiraniran et al.
investigated the change of bubble flow with liquid superficial velocities of 0.1 and 0.2 m/s using
a WMS. They found that the bubbles were decelerated by about 40–50% compared to the bubble
35
velocity upstream of the sensor. Later, studied the intrusive effects of a WMS in a counter-
current bubble flow, and they observed a similar bubble deceleration. Ito et al. 36 sustained the
research in this line by deploying a WMS to estimate the bubble velocity of an air–water bubbly
flow in a vertical channel with a square cross section. They found that the bubbles were strongly
|Page
They concluded that the effect decreases with increasing liquid velocity and finally turns into a
slight acceleration which corresponds to the degree of the cross-section obstruction by the wires.
Ito et al. 36 thereafter provided an excellent explanation for the acceleration and deceleration of
the bubbles in terms of the Weber number (We = inertia forces/surface tension forces). That the
inertia of the liquid stabilizes the bubble shape and presses the bubble through the wire grids
against the action of superficial forces. At low liquid superficial velocities, the surface tension
effects dominate over inertia forces. Accordingly, decelerating forces during the contact of the
gas–liquid interface with the solid surface of the electrode wires dominate and the bubble is
decelerated.
Based on WMS’s strong spatial resolution, it enables data such as bubble size distributions to be
extracted. Hence, WMS can provide details of the distribution of bubble sizes present; this can
Figure 2 shows a differential pressure transducer (DPT) with a range of 0–2.5 bar and nominal
pressure of 16 bar and an accuracy of ± 0.25% , output voltage of 1 to 5 volts and a data
acquisition frequency of 1000 Hz that was used to measure the pressure drop in this work. The
model number of the DPT is 730-E-2501-1-5-100-J00-1-000 and was made by Impress Sensor
and Systems. Both the sensitivity and range were taken into consideration in the selection of the
DPT. The DPT was used to measure the time varying two-phase total pressure drop across the
test section. In order to achieve this, two pressure tapping were provided, located in the vertical
|Page
the foot of the test section. Hence, the total pressure drop was measured concurrently with the
void fraction.
Before the commencement of making measurements of pressure drop, the DPT had to be
calibrated. One of the advantages is having zero span features that make them easy to calibrate.
Although the DPT used in the experiment were factory calibrated with a calibration certificate
provided to the user, a second calibration was conducted on the transducers to confirm the
accuracy of the previous calibration. The calibration exercise carried out established the link
between the output voltage of the DPT and the differential pressure.
The DP transmitters were calibrated against static liquid columns up to 3 mm in length by filling
a vertical transparent tube with water at a temperature and density of 20oC and 998 kg/m3,
respectively. The transparent tube used for calibration was made of polycarbonate material. It
was scaled from zero using a special scaled tape for that purpose. The high pressure inlet port
(DPT1) of the DPT was connected to the water column, whilst the lower inlet pressure (DPT2)
was sealed.
It is worthy of mention that for accurate pressure drop measurements, it is essential for the
pressure tappings to be purged frequently. Otherwise, pressure fluctuations will introduce a two-
phase mixture into the DPT lines triggering incorrectness in the measured pressure drop as a
consequence of surface tension effects and indefinite hydrostatic head. It was against this
|Page
line to the DPT from the pressure tapings. The output voltage from the DPT was recorded using
The calibration was started by zeroing the DPT at the lowest level of the water column. The
voltage recorded was 1 V at that point. The height of the water was then gradually increased by
adding small volume of water in steps to the column from the tap. Thus, the corresponding
changes observed of the water level to the output voltage were recorded until it reached the
highest level, i.e., output voltage of 5 V. The change of the output voltage was then plotted
against the water static head to obtain the calibration for each DPT. The calibration curve is
shown in Figure 6. Figure 6 shows the plot of the results from calibration which perfectly
matches those provided by the manufacturer. The equation in the plot was programmed into
Two T-type thermocouples, TCl and TC2, were inserted into both the liquid inlet and the two-
phase flow outlet of the pipe, respectively, as shown in Figure 2 to ensure that temperature data
are logged in order to monitor and control the temperature. As changes to temperature can affect
density, viscosity, surface tension and other physical properties of the fluid. It therefore became
The calibration of these thermocouples was done using ice, hot and boiling water. A thermometer
was inserted into each system (ice cubes, hot and boiling water) to measure the temperature. This
temperature measurement with the aid of a mercury thermometer which has an accuracy of ±1oC
|Page
shows the plot of the temperature versus voltage with a line of best fit. The equations of this lines
(calibration equations) were programmed into LabVIEW to obtain the temperatures of the two
phase flow at the liquid inlet at the bottom end as well as at the two-phase flow outlet of the pipe.
Figure 7: Calibration plots for the Thermocouples (a) at the liquid inlet at the bottom end of the pipe (TC1) (b) at
the two-phase flow outlet of the pipe (TC2)
The flow rates at which measurements were made correspond to gas superficial velocities in the
range 0.1–3 m/s and liquid superficial velocities of 0.2 to 1.0 m/s. Figure 8 shows the
experimental operating points of the present study described in a flow pattern map (Figure 8)
37
generated using the FLOPATN computer code developed by . From an analysis of the
37
observations it is concluded that the model predicts bubble and churn flows accurately. For
each run, data was taken at a rate of 1000 cross-sections per second for 30 s. Measurements were
taken with the pipe full of oil and full of gas before and after each batch of tests to obtain and
Figure 8: Flow pattern map under current experimental conditions predicted by 37.
Gregory and Scott 38 developed a correlation for predicting slug frequency based on the data by
39
. Owing to the fact that higher viscosity liquid has been used for the current study which shifts
the transition between flow regimes, a different approach has been used to obtain the dominant
frequency in order to account for changes in viscosity. To determine this dominant frequency of
40
periodic structures, the methodology of Power Spectral Density (PSD) as defined by was
|Page
separated, dispersed and intermittent flows for horizontal pipes. PSD according to 13 is a measure
of how the power in a signal varies over a range of frequency and thus, it defines how the power
(or variance) of a void fraction time series is distributed with frequency. Mathematically, it is
defined as shown in equation (1) as the Fourier Transform of the autocorrelation sequence of the
time series. The method presents the PSD functions in terms of direct Fourier Transformations
of the original data. The auto covariance function of a signal x(t) is given by:
Where T is the sampling period, k∆τ is the time delay, τ is the interrogating time delay and
T
1
x = ∫ x(t )dt (2)
T 0
τ / ∆τ −1
1
Pxx ( f ) = ∆τ Rxx (0) + ∑ Rxx (k∆τ )W (k∆τ )Cos (2πfk∆τ ) (3)
2 k =1
the spectrum leakage which frequently comes out as the sidelobes at high frequency end of the
spectrum. By using a suitable windowing function, the frequency contributing to the system
becomes obvious. In the analysis carried out here to determine the dominant frequency, a basic
πk∆τ
w(k∆τ ) = cos (4)
2τ
|Page
responses. It is known that for a chaotic motion, the power spectrum is a continuous process.
Since the intermittent region of a two-phase flow is regarded as a chaotic region, PSD analysis
is a useful tool in identifying flow regimes. In effect, its characteristic features at different
pressure levels can be identified when compared. Hence, the transition from smooth to chaotic
flows can be identified through the PSD signatures as shown in Figures 9(a)-(d). At low gas
superficial velocities, the harmonic signal response was more chaotic than at higher gas
superficial velocities. This is probably due to the response to changes from bubbly to intermittent
flows.
Figure 9: PSD and dominant frequency plots at liquid superficial velocity of 0.4 m/s and gas superficial velocities
(m/s): (a) 0.1, (b) 0.6, (c) 1.0, (d) 2.0
In general, the dominant frequency can be observed in Figure 10a to increase with an increase in
liquid superficial velocity. However, the trends for the variation of dominant frequency with
pressure are quite irregular in particular at low gas superficial velocities. These can be attributed
11
to changes in flow pattern. Similar observation have been reported by Szalinski et al. and
Figure 10: (a) Dominant frequency of periodic structures as derived from power spectral density (PSD) analysis:
closed symbols and x =High pressure, open symbols and ӿ=Low pressure (b) Gas based Strouhal number versus
Lockhart-Martinelli parameter (c) Gas based Strouhal number versus Inverse viscosity number and velocity ratio
A good way to describe the dominant frequency of periodic structures is in terms of Strouhal
number (fDt/u). The plot of gas based Strouhal number against Lockhart-Martinelli parameter as
depicted in Figure 10(b) show that both present data and most of the other data lie on a parallel
|Page
By plotting gas based Strouhal number against the product of the dimensionless inverse viscosity
number and velocity ratio as proposed by 42 to account for viscosity effect (see Figure 10c), the
data in the present study was compared with data of 11. The data shows very good agreement.
The flow structures, as revealed by the WMS are illustrated in Figures 11 (A-B) at high and low
pressures, respectively. The information is held in the form of pixels of 4 × 4 mm. These are
coloured according to the amount of liquid or gas within the area. In these, the reddish-brown
represents 100% gas whilst blue is 100% liquid. Light blue, orange and yellow represents
intermediate phases. It can be observed that at low gas superficial velocities, there are no
It can be observed from Figure 12 that at liquid and gas superficial velocities of 0.2 and 1.0 m/s,
respectively, a number of drops which appear in yellow/light green colour (low region) but do
not occupy the whole of the individual pixel can be seen. Also seen on the plot, are high regions
occupied by gas which appear in red color. They may be regarded as parts of churn flow (huge
waves). The low and high regions can be associated directly with the more easily recognizable
features in the time series and PDF of void fraction shown in Figure 12(a-b). On the matter of
time series of void fraction (Figure 12a), periodic fluctuations in void fraction though not regular
between 0.7 and 0.4 can be observed on the time trace. The probability density function (PDF)
of void fraction, Figure 12b, shows a single peak at a void fraction of 0.6, but with a base
|Page
the PDF of void fraction shows that we have large wisps of liquid and the product of incomplete
atomization. Wisps are continuous ligaments of liquid flowing in the centre of the pipe. A similar
observation is also seen in Figure 13 at same liquid superficial velocity of 0.2 m/s but at a higher
Figure 11: Visual side view of wire mesh sensor (WMS) at A-High pressure: liquid superficial velocity = 0.4 m/s;
gas superficial velocity = (a) 0.1, (b) 0.6, (c) 1.0 (d) 1.4, (e) 2.0 (f) 2.4, (g) 3.0 m/s and at B-Low pressure: liquid
superficial velocity = 0.4 m/s; gas superficial velocity = (a) 0.1, (b) 0.6, (c) 1.0 (d) 1.4, (e) 2.0 (f) 2.4, (g) 3.0 m/s
Figure 12: (a) Time series of void fraction at high and low pressures of 7.5 and 4.5 bars, respectively (b) PDF of
void fraction at high and low pressures of 7.5 and 4.5 bars, respectively (c) Visualisation of wire mesh sensor (WMS)
data at high pressure of 7.5 bar and (d) Visualisation of wire mesh sensor (WMS) data at low pressure of 4.5 bar.
Liquid and gas superficial velocities of 0.2 and 1.0 m/s, respectively. (Note, blue region represents liquid/oil while
the red region represents gas, SF6).
Figure 13: (a) Time series of void fraction at high and low pressures of 7.5 and 4.5 bars, respectively (b) PDF of
void fraction at high and low pressures of 7.5 and 4.5 bars, respectively (c) Visualisation of wire mesh sensor (WMS)
data at high pressure of 7.5 bar and (d) Visualisation of wire mesh sensor (WMS) data at low pressure of 4.5 bar.
Liquid and gas superficial velocities of 0.2 and 1.4 m/s, respectively. (Note, blue region represents liquid/oil while
the red region represents gas, SF6).
This is not to dispute the existence of wisps at low pressures but to inform the readers about how
wisps are easily revealed as long as there is a transition between churn and annular flows
10 43
regardless of the pressure. Interestingly, wisps were observed by and Sharaf et al. at the
same liquid superficial velocities. However, in the present work, the wisps appear to be smaller,
which might be due to the different properties of the working fluids employed. It was against
this background that a comparison between the wisps frequencies obtained from present work
against those obtained by 10 and Sharaf et al. 43 are presented and discussed in Section 3.3.
|Page
Since wisps were observed in the current work as earlier reported in Section 3.2, it therefore
became necessary to plot the frequencies of these wisps and compare against published data.
This plot is shown in Figure 14. The wisp frequency was obtained by counting the wisps
observed in the side view images reconstructed from the WMS. This is to avoid confusion with
the dominant frequency obtained while using the method of PSD, due to other flow structures
It can be observed from Figure 14 that the wisps frequency increases with an increase in gas
43
superficial velocity. This is in good agreement with those observed by but vary from those
observed by 10.
Figure 14: A plot of wisps frequency from present study against data from 43 and 10 at various gas superficial
velocities
Since the viscosities of the fluids for all cases compared in Figure 14 vary, it is more logical to
make comparison based on these viscosity changes. Figure 15 expresses wisps frequencies in
terms of dimensionless numbers with consideration given to viscosity changes. In general, the
observed trends as shown in Figure 15 are seen to show the same tendency especially against the
work of 42. This is not surprising as 43 and present work used same 127 mm diameter pipe.
Figure 15: Dimensionless numbers plotted and compared with data from 43 and 10
When the differences in gas superficial velocities are analysed in terms of Weber number, the
43
resulting plots show a good agreement with the work by as can be seen in Figure 16. It is
observed from Figure 16 that the value of the ratio of wisps frequency (fw) and the dominant
frequency (f) against Weber number is high at low Weber number for the work of 43 compared
|Page
coalescence process.
Figure 16: Ratio of wisps frequency to dominant frequency against Weber number for both low and high pressure
compared with data from 43
The bubble size distribution has been obtained from further analysis of the 3D void fraction data.
44
It is performed using an extended decremental recursive fill algorithm developed by . The
bubbles were identified by extracting the contribution of each statistical bubble diameter bin to
dε
h( Db ) = (5)
dDbubble
Using the bubble volume, the corresponding equivalent bubble diameter Dbubble is calculated as
6
Dbubble = 3 Vbubble (6)
π
Details on the methodology for extracting bubbles size information from WMS data can be found
16
in . Bubble size distributions for selected gas superficial velocities and at liquid superficial
velocities of 0.2, 0.4, 0.6, 0.8 and 1.0 m/s are shown in Figure 17. They are constructed by
summing the contribution of the bubbles of a given range of diameters to the integral volumetric
dε
gas fraction. The partial gas fractions are plotted against the equivalent bubble diameter
dDbubble
DBubble. It can be observed in Figure 17 that at the lowest liquid superficial velocity 0.2 m/s, the
|Page
7.5 bar. However, as the liquid superficial velocity is increased, the size of the bubble
Figure 17: Some selected bubble size distribution plots at the following conditions: liquid and gas superficial
velocities of (a) 0.2, 0.1 (b) 0.2, 0.6, (c) 0.4, 0.1, (d) 0.4, 0.6, (e) 0.6, 0.1, (f) 0.6, 0.6 m/s respectively
It can be observed from Figure18(a-b) that at low liquid superficial velocities, the effect of
pressure on mean void fraction is insignificant. This effect however begins to manifest with an
increase in liquid superficial velocity. This is in particular occurs at critical gas and liquid
superficial velocities of 1 m/s and 0.6 m/s, respectively. Beyond this liquid superficial velocity,
the effect of pressure on mean void fraction at high pressure is bigger than at low pressure.
It is worthy of mention that since pressure is directly proportional to density, it follows that an
Figure 18: Effect of pressure on mean void fraction at different liquid and gas superficial velocities
Figure 19: Comparison between experimental mean void fraction measured using Wire Mesh Sensor (WMS)
against mean void fraction calculated from Differential Pressure Transducers (DPT)
The calculated void fraction was derived from the relationship between total pressure gradient,
gravitational pressure gradient and frictional pressure gradient (assuming a steady state flow with
|Page
elevated, the difference became significant especially at low gas superficial velocities.
Where, g is the gravitational acceleration constant, DP/Dz is the pressure gradient and ρ L and
The frictional component is rather small and the pressure gradient is almost determined by the
gravitational one. On the other hand, when the pressure increases and velocity also increases, the
frictional component cannot be neglected. Then gravitational component decreases and on the
other hand frictional component increases. As expected, the pressure gradient increases with an
increase in gas superficial velocity. Figure 19 d suggests that the decrease in the gravitational
component is significant than the increase in the frictional one. In addition, because we are
operating in the bubble/churn flow regime, the gravitational component is expected to decrease
with an increase in void fraction as a consequence of an increase in gas superficial velocity. This
It can be observed from Figure 20 that the total pressure gradient decreases with an increase in
gas superficial velocity for all the liquid superficial velocities considered. On the other hand, as
expected, the total pressure gradient increases with an increase in liquid superficial velocity
owing to an increase in mixture density. The observed decrease in the total pressure gradient can
|Page
an increase in gas superficial velocity, will stimulate an increase in the void fraction, thereby
reducing the mixture density as a result of a decrease in the liquid hold up (1 – void fraction).
However, the velocities encountered are not high enough to cause high frictional pressure
gradients. Thus, the total pressure gradient decreases with an increase in gas superficial velocity.
In order to examine the effect of pressure on pressure gradient in greater details, 7 made a plot of
dimensionless pressure gradient (representing y-axis) against the dimensionless air flowrate (x-
axis). See Figure 1. He defined the dimensionless pressure gradient and dimensionless air flow
rate as follows:
dP
dP
− ρG g dz − ρG
dz g
Dimensionless pressure gradient = = (7)
(ρ L − ρG )g ( ρ L − ρG )
and
U2
Dimensionless air flow rate, VG*= (8)
gD
7
The concept of was applied in this work in order to find the effect of pressure on pressure
gradient. The plot showing this effect of pressure is depicted in Figure 21. It is worth mentioning
however that the dimensionless air flow rate was replaced in this work using the Kutateladze
|Page
Ku = (9)
([ρ L − ρG ]gσ )0.25
It can be observed from Figure 21 that the effect of pressure though not very significant is more
3.6 Comparison of measured void fraction against other void fraction prediction
models
The oil and gas industry has a great number of void fraction prediction models. Many of these
models have been modified to fit into transient flows to effectively predict flow assurance and
multiphase flows in terrain environments. For instance, OLGA for flow assurance as well as 1D
LedaFlow and Maximus use some of these models in flow assurance predictions. Hence, it is
imperative to carry out analysis based on these models in order to suggest ways of improving on
the performance of the existing models since some of them have been observed to either under-
predict or over-predict void fraction. In this section, comparisons will be made with 3 models
reported in the literature at the conditions set for the experimental campaign. The models
considered in this work are the Homogeneous, CISE/Friedel and the Beggs and Brill models.
Figures 22-23 show comparisons between the measured (experimental) void fraction against the
Homogeneous, CISE/Friedel and the Beggs and Brill predicted void fraction models. In the case
described in Figure 22, the liquid momentum was quite low and the predicted void fraction varies
largely from those obtained from experiments. However, the CISE/Friedel and homogenous
models over-predict void fraction at low liquid superficial velocity of 0.4 m/s and high gas
|Page
superficial velocity of 0.6 m/s. In Figure 23, the liquid momentum for the flow conditions is
enhanced to create better predictions for the various models. Both the Homogenous and
CISE/Friedel models perform well while Beggs and Brill’s model over-predicts the void fraction.
Figure 22: Comparison between the Homogeneous, CISE/Friedel and the Beggs and Brill models against
experimental data at liquid superficial velocities of 0.4 m/s and 0.6 m/s (a) High pressure (b) Low pressure
Figure 23: Comparison between the Homogeneous, CISE/Friedel and the Beggs and Brill models against
experimental data at liquid superficial velocities of 0.8 m/s and 1.0 m/s (a) High pressure (b) Low pressure
In summary, the 3 considered models namely Homogeneous, CISE/Friedel and the Beggs and
Brill are more effective especially at higher liquid superficial velocities. These are assumed to
be the points at which the liquid momentum is high enough to avoid reverse flows which may
occur at low liquid superficial velocities. In addition, the results show that the 3 models perform
better at lower pressure than at high pressure. Generally, the void fraction data from experiment
are within 15% of those predicted by the 3 models at low pressure but higher (between 20%-
25%) for high pressure. As reported by 11, these models perform well for oil of low viscosity but
poorly with water with a lower viscosity, however, their performance with highly viscous liquid
3. Conclusion
The paper has presented the experimental results to study the effect of pressure on two-phase
gas–viscous liquid flow at two pressure conditions, 4.5 and 7.5 bar in a vertical 127 mm internal
diameter pipe. The system fluid is sulphur hexafluoride (SF6) with densities of 28 and 45 kg/m3
and high viscosity oil mixture (mixture of Nexbase 3080 oil with viscosity of 89 mPa.s and
|Page
characterized using advanced instrumentation, including capacitance wire mesh sensor (WMS)
and a differential pressure transducer. That the following conclusions can be drawn:
1. Dominant frequency increases with an increase in pressure. The dominant frequency also
increases with liquid superficial velocity. This is in agreement with observations reported in the
literature;
2. The plot of gas Strouhal number against Lockhart–Martinelli lie on a parallel line compared with
most of the other data under consideration;
3. At low gas superficial velocity, there are no significant changes in bubble size distribution.
However, wisps were observed at higher gas superficial velocities and at elevated pressures;
4. In the present work, wisps appear to be smaller, which might be due to the different fluid
properties of the working fluids employed;
5. Wisps are easily revealed as long as there is a transition between churn and annular flows
regardless of the pressure;
6. The effect of pressure on average void fraction at low liquid superficial velocities was found to
be negligible while on the other hand, became significant at high liquid superficial velocities;
7. As the liquid superficial velocity increases, the calculated average void fraction shows some
disagreement with measured void fraction;
8. The dimensionless pressure gradient relationship used by 7 with a modification to the
dimensionless air flowrate was used to find the effect of pressure;
9. Correlations of CISE/Friedel, homogeneous and Beggs and Brill’s perform well at high liquid
superficial velocities when compared against measured void fraction. This could be attributed to
higher momentum at higher liquid velocities. Better agreement was however observed for higher
pressure than for lower pressure;
10. Viscosity effect can be moderated using dimensionless groups as these take into consideration
different parameters that can affect two-phase flow.
Acknowledgment
Abdulkadir, M., would like to express his sincere appreciation to the Nigerian government through the Petroleum
Technology Development Fund (PTDF) for providing the funding for his doctoral studies.
This work has been undertaken within the Joint Project on Transient Multiphase Flows and Flow Assurance. The
Author(s) wish to acknowledge the contributions made to this project by the UK Engineering and Physical Sciences
Research Council (EPSRC) and the following: - ASCOMP, GL Noble Denton; BP Exploration; CD adapco;
Chevron; ConocoPhillips; ENI; ExxonMobil; FEESA; FMC Technologies; IFP Energies nouvelles; Granherne;
Institutt for Energiteknikk; Kongsberg Oil & Gas Technologies; MSi Kenny; PDVSA (INTEVEP); Petrobras;
PETRONAS; SPT Group; Shell; SINTEF; Statoil and TOTAL. The Author(s) wish to express their sincere gratitude
for this support.
A. Abdulahi will like to thank University of Nottingham Graduate School for providing funding to travel to the
SINTEF Multiphase Flow Laboratory, Trondheim, Norway under the BESTS Scholarship Scheme.
|Page
Greek letters
References
1. Nishikawa K, Sekoguchi T, Fukano T. On the Pulsation Phenomena in Gas-Liquid Two-Phase Flow : Relationship
between Pulsating Pressure and Flow Pattern in Upward Two-Phase Flow. Bulletin of JSME 1969; 12: 1410-1416.
2. Cheng H, Hills JH, Azzorpardi BJ. A study of the bubble-to-slug transition in vertical gas-liquid flow in columns
of different diameter. International Journal of Multiphase Flow. 1998; 24: 431-452.
3. Ohnuki A, Akimoto H. An experimental study on developing air-water two-phase flow along a large vertical pipe:
effect of air injection method. International Journal of Multiphase Flow. 1996; 22: 1143-1154.
4. Jayanti S, Hewitt GF, Prediction of the slug-to-churn flow transition in vertical two-phase flow. International Journal
of Multiphase Flow. 1992; 18: 847-860.
5. Taitel Y, Barnea D, Dukler AE , Modelling flow pattern transitions for steady upward gas-liquid flow in vertical
tubes. AIChE Journal. 1980; 26: 345-354.
6. Costigan G, Whalley PB. Slug flow regime identification from dynamic void fraction measurements in vertical air-
water flows. International Journal of Multiphase Flow. 1997; 23: 263-282.
7. Owen D. An experimental and theoretical analysis of equilibrium annular flow [PhD Thesis]. Birmingham:
|Page
|Page
|Page
Figure 1: Dimensionless pressure gradient as s function of the dimensionless air flow rate in air–
water flow in a tube (data obtained by Owen 7 as reported by Hewitt 8)
Figure 3: Two-phase mixing configuration showing (a) layout of the gas injection unit (c) Two-phase
flow mixing point
Figure 7: Calibration plots for the Thermocouples (a) at the liquid inlet at the bottom end of the pipe
(TC1) (b) at the two-phase flow outlet of the pipe (TC2)
37
Figure 8: Flow pattern map under current experimental conditions predicted by at (a) low pressure
of 4.5 bar and (b) high pressure of 7.5 bar
Figure 9: PSD and dominant frequency plots at liquid superficial velocity of 0.4 m/s and gas
superficial velocities (m/s): (a) 0.1, (b) 0.6, (c) 1.0, (d) 2.0
Figure 10: (a) Dominant frequency of periodic structures as derived from power spectral density
(PSD) analysis: closed symbols and x =High pressure, open symbols and ӿ=Low pressure (b) Gas
based Strouhal number versus Lockhart-Martinelli parameter (c) Gas based Strouhal number versus
Inverse viscosity number and velocity ratio
Figure 11: Visual side view of wire mesh sensor (WMS) at A-High pressure: liquid superficial
velocity = 0.4 m/s; gas superficial velocity = (a) 0.1, (b) 0.6, (c) 1.0 (d) 1.4, (e) 2.0 (f) 2.4, (g) 3.0 m/s
and at B-Low pressure: liquid superficial velocity = 0.4 m/s; gas superficial velocity = (a) 0.1, (b) 0.6,
(c) 1.0 (d) 1.4, (e) 2.0 (f) 2.4, (g) 3.0 m/s
Figure 12: (a) Time series of void fraction at high and low pressures of 7.5 and 4.5 bars, respectively
Figure 13: (a) Time series of void fraction at high and low pressures of 7.5 and 4.5 bars, respectively
(b) PDF of void fraction at high and low pressures, at high and low pressures of 7.5 and 4.5 bars,
respectively (c) Visualisation of wire mesh sensor (WMS) data at high pressure of 7.5 bar and (d)
Visualisation of wire mesh sensor (WMS) data at low pressure of 4.5 bar. Liquid and gas superficial
velocities of 0.2 and 1.4 m/s, respectively. (Note, blue region represents liquid/oil while the red
region represents gas, SF6).
42 10
Figure 14: A plot of wisps frequency from present study against data from and at various gas
superficial velocities
Figure 16: Ratio of wisps frequency to dominant frequency against Weber number for both low and
high pressure compared with data from 42
Figure 17: Some selected bubble size distribution plots at the following conditions: liquid and gas
superficial velocities of (a) 0.2, 0.1 (b) 0.2, 0.6, (c) 0.4, 0.1, (d) 0.4, 0.6, (e) 0.6, 0.1, (f) 0.6, 0.6 m/s
respectively
Figure 19: Comparison between experimental mean void fraction measured using Wire Mesh Sensor
(WMS) against mean void fraction calculated from Differential Pressure Transducers (DPT)
Figure 22: Comparison between the Homogeneous, CISE/Friedel and the Beggs and Brill models
against experimental data at liquid superficial velocities of 0.4 m/s and 0.6 m/s (a) High pressure (b)
Low pressure
Figure 23: Comparison between the Homogeneous, CISE/Friedel and the Beggs and Brill models
against experimental data at liquid superficial velocities of 0.8 m/s and 1.0 m/s (a) High pressure (b)
Low pressure