Accepted Manuscript
Oil-water dispersion formation, development and stability studied in a wheel-shaped
flow loop
Heiner Schümann
PII:
S0920-4105(17)30844-6
DOI:
10.1016/j.petrol.2017.10.066
Reference:
PETROL 4390
To appear in:
Journal of Petroleum Science and Engineering
Received Date: 11 May 2017
Revised Date:
19 October 2017
Accepted Date: 25 October 2017
Please cite this article as: Schümann, H., Oil-water dispersion formation, development and stability
studied in a wheel-shaped flow loop, Journal of Petroleum Science and Engineering (2017), doi:
10.1016/j.petrol.2017.10.066.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Title:
Oil-water dispersion formation, development and stability studied in a wheel-shaped flow loop
Abstract
M
AN
U
SC
RI
PT
The work describes how a quasi-endless wheel flow loop can be used to study dispersion formation,
development and stability at different pressures and temperatures. The so-called Wheel Flow Loop
consists of a wheel-shaped closed pipe section that can be filled with gas, oil and water. By rotating
the device in vertical orientation, a gravity driven flow is initiated. Several different flow conditions
were simulated by varying the rotational speed. Onset of dispersion and the stability of emulsions
formed was identified by interpreting torque profiles supported by visual observations. Two fluid
systems, a realistic fluid system consisting of a light crude oil, synthetic natural gas and brine, and a
model system consisting of a mineral oil, tap water and nitrogen as gas phase were investigated.
Furthermore, the effect of temperature, pressure and water cut on dispersion properties was
demonstrated. The sensitivity of the results indicated that the wheel qualifies as a characterization
tool for dispersion properties. Furthermore, agreement in dispersion behaviour was found when
straight pipe flow experiments were compared with the wheel matching the energy dissipation rate.
1. Introduction
TE
D
Water is often co-produced during oil production, originating from formation water, water flooding
or reinjection for pressure support. Although immiscible, oil and water can be dispersed as droplets
of one phase in the other. Such dispersions can be formed during all stages of production like in the
reservoir, production lines, in high shear pumps and valves. Factors affecting the dispersion process
include relative phase fractions, interfacial tension, viscosity and shear rates. Moreover, interfacial
active molecules and solids may play an important role in stabilizing dispersions, leading to what is
defined as emulsions. Aged and stabilized emulsions may further affect break-up, coalescence and
rheology of the droplets in the fluid system produced.
AC
C
EP
Dispersions are, during oil production, often deemed as unfavourable as they increase the effective
viscosity of the fluid system resulting in increased pressure drop. Furthermore, emulsions lead to oilwater separation challenges. However, in some cases, like for highly viscous oils, pressure drop
reductions can be achieved by dispersing the oil in water (Yang, Velthuis et al. 2013) leading to a
mixture viscosity closer to water than oil. Oil continuous dispersions can be preferable as effective
measures against corrosion keeping the steel pipe oil wetted during production.
However, correctly predicting dispersion behaviour in the production system is a challenge. Reliable
dispersion models are not yet available for commercial flow simulators. This results in rather high
uncertainties in predictions of the pressure drop. Experiments may be performed to provide detailed
information about dispersion properties then used as input for simulations and further development
of models. Previous experimental studies compared separated and dispersed flow at otherwise
similar conditions, where inlet mixing was used to create dispersed flow (Angeli 1996, Soleimani
1999, Schümann, Tutkun et al. 2016). The results from these studies showed possible increases in the
pressure gradient by up to factor three. This demonstrates the importance of correctly measuring,
understanding and ultimately predicting dispersed flow behaviour.
ACCEPTED MANUSCRIPT
TE
D
M
AN
U
SC
RI
PT
Pipe flow measurements are commonly performed in flow loops. Flow rates, pressure drops, flow
patterns, turbulence or local phase distributions are measured with high accuracy under controlled
flow conditions. However, two major disadvantages of such flow loops should be pointed out. First,
unless very large amounts of fluids are available flow loops are designed as recirculation loops
requiring a separation process which often can lead to changes in the fluid system over time. Such
changes include recirculation of very small droplets, not being able to separate, and accumulation or
depletion of components at the interface in the separator and in the bulk phases. The second
disadvantage is related to the restricted test section length in a flow loop compared to several
kilometres of production line. Long term flow development cannot be measured in such flow loops
and fully developed flow will not be reached. In the current work, the application of an autoclave in
form of a 2" stainless steel pipe bent into a wheel shape was used as a tool for characterization of
dispersion formation and stability for various oil-water systems. A similar set-up has also previously
been used to study the viscosity of dispersions. Urdahl, Fredheim et al. (1997) and Johnsen, Førdedal
et al. (2001) estimated the effective viscosity of dispersions formed in such a wheel shaped flow loop.
In a later work Johnsen and Rønningsen (2003) compared this method with theoretical models for
rheology. A comparison of friction factors and shear obtained in a wheel flow loop with pipe flow
experiments was demonstrated by Krampa, Ytrehus et al. (2005). Friction factors compared well for
fully turbulent flow and Reynolds numbers Re > 10000. However, only liquid-gas experiments with a
single liquid phase were exploited, without the presence of oil-water dispersions. Roca, Carneiro et
al. (2014) and Johansen, Mo et al. (2015) used wheel flow loop experiments to evaluate the
performance of numerical flow models. The wheel has previously been used for the study of hydrate
formation in production systems (Urdahl, Lund et al. 1995). The main advantage with this set-up is
that no recirculation is necessary so the fluid will see an "infinite" pipe line allowing for reaching
steady state conditions even when the transient is several hours or even days. Furthermore, the
volume is relatively small and live oil systems in terms of pressure, temperature and composition can
be studied. An alternative technique for studying dynamic dispersion stability was presented by Patil,
Arnesen et al. (2017). A stirred tank approach with varying rotational speed was used to produce
dispersions. In-situ droplet measurements were used to monitor changes in the droplet size as result
of changing the rotational speed. Measurements were used for evaluating dispersion development
and stability for different mineral oil and crude oil systems.
AC
C
EP
The objective of this work was to evaluate to what degree Wheel Flow Loop experiments are suitable
for testing dispersion properties of different fluid systems, namely conditions required for dispersion
formation, dispersion stability and flow development time scales. Furthermore, it was evaluated how
the obtained information can be used to predict flow behaviour in real pipe flow systems.
2. Experimental setup and procedure
2.1 Experimental setup.
The Wheel Flow Loop used (the Wheel) as shown in Figure 1 consisted of a wheel shaped steel pipe
with a radius of R = 1 meter in the center of the pipe and an inner pipe diameter of d = 52.5 mm. The
wheel was installed in a vertical orientation and mounted on a shaft, which again was connected to
an engine. The maximum operating pressure of the set-up used was 250 bar. Instrumentation
consisted of a temperature sensor, a pressure sensor and a combined position and torque sensor.
The flow inside the wheel could be visually inspected through a sapphire glass window by use of a
camera mounted to and rotating with the wheel. The setup was placed inside a temperature
controlled climate chamber, which could be adjusted between -10°C and +80°C. With the pressure
ACCEPTED MANUSCRIPT
range and materials used, recombined crude oil systems, or live oils can be used for the study of
transportability with regard to gas hydrates, emulsions and high viscous systems.
RI
PT
Figure 1: Left: The wheel flow loop placed in a climate chamber and seen from the front. A camera is mounted on the wheel
facing a sapphire window section. A counter weight is mounted on the opposite to balance the wheel. The picture also
shows valves for filling and draining the wheel. Right: The schematic drawing showing the wheel setup from the back with a
cut through the climate chamber wall. Driving and control unit of the wheel flow loop are placed behind the wall outside the
chamber.
SC
Figure 2: Schematic drawing of the filled wheel. Left: At rest, the phases distribute evenly. Right: When rotating at a high
velocity, the flow becomes fully dispersed. A liquid tail is drawn up the pipe walls.
M
AN
U
Figure 3: Pictures from the video camera mounted on the transparent window. a) Crude oil - water interphase b) Crude oil water dispersion c) Crude oil - water dispersion with thin layer of free water at the bottom d) Mineral oil - water interphase
e) Mineral oil – water dispersion f) Mineral oil and water with dispersion at the interphase.
2.2 Test Procedure
AC
C
EP
TE
D
For normal operation the liquid filling degree is 40% by volume of a total volume of 13.4 l. When
filled with fluids of different densities gravity will keep the liquid in place while the wheel rotates,
meaning the flow is gravity driven. The relative movement between liquid and pipe will result in a
shear force between the liquid and the pipe wall. The liquid will be slightly dragged upwards on the
side where direction of rotation is upwards. The resulting imbalance in total liquid distribution, as
result of the friction force, can be measured as torque. At sufficiently high rotational velocities,
depending on the fluid system and water cut, dispersions were formed as illustrated in Figure 2.
Pictures captured from the mounted video camera are shown for different flow situations in Figure 3.
A velocity profile with characteristic step changes was defined and used for all tests in order to be
able to compare results. Instant velocity changes followed by constant velocity periods were chosen
instead of smooth velocity gradients. This made it possible to identify critical velocities required for
dispersion onset and transition to fully dispersed flow. Typical ramp up or ramp down times from
one velocity to another were 15-20 seconds. This was rather short compared to typically several
minutes required for complete flow development. Therefore, it is justified to talk about instant
velocity changes. In reality the rotational speed in rounds per minute, RPM, is controlled. For the
purpose of easier comparing results with pipe flow experiments, we will report velocities, U, in m/s.
The conversion is U = RPM * R. Figure 4 shows the velocity profile used with baseline velocity of 0.5
m/s. For systems with early dispersion onset 0.1 m/s was used as baseline velocity. From the baseline
velocity the velocity was increased to a higher velocity (mixing velocity) before ramping down again
to the baseline velocity. Mixing velocities were increased stepwise and held for 30 minutes while
flow at baseline velocity lasted for 1 hour. Ramping down to the baseline velocity allowed for
identification of dispersion onset and evolution of the dispersions (i.e. coalescence and separation),
as will be explained below. The tested mixing velocities covered a range from 0.2 m/s to 1 m/s with
steps of 0.1 m/s followed by two higher velocities of 1.5 m/s and 2 m/s respectively. The velocity
range was sufficient to fully disperse the flow in all cases. At the end of the velocity profile a velocity
step change from 2m/s down to 1m/s was undertaken. The idea was to have a comparison with a
ACCEPTED MANUSCRIPT
previously performed step change from baseline velocity of 0.5 m/s up to 1 m/s for identifying
possible deviations with regard to emulsion viscosity and stability due to the higher mixing energy at
2 m/s.
Figure 4: Predefined velocity profile used for the wheel experiments.
M
AN
U
SC
RI
PT
Calibration measurements were performed with the wheel empty (depressurized nitrogen
atmosphere and no liquid), 40% water filled (nitrogen as gas phase) and 40% oil filled (nitrogen as gas
phase). The torque from the empty wheel experiments can be ascribed to mainly mechanical friction
and was subtracted from the torque measured with the filled wheel. Calibration measurements with
one liquid only can be used for normalizing experimental results in order to demonstrate relative
increase in torque. The time averaged torque measurements at increasing velocities for these
calibration tests are plotted in Figure 5. As seen, the average torque for the empty wheel is very close
to zero at velocities below 0.5 m/s and increases slightly upon increasing velocities. For the water
and oil filled tests the torque values increase in a nonlinear way, with the values almost, but not
entirely identical.
Figure 5: Calibration measurements for oil, water and the empty wheel. The filling content was 40% for the oil and water
calibration.
2.3 Measurement uncertainty
AC
C
EP
TE
D
The systematic uncertainty of the torque measurements is dominated by the absolute uncertainty of
the torque transducers resulting from system accuracy, linearity and hysteresis of the sensor. An
absolute systematic uncertainty of 0.135 Nm can be assumed for all measurements. A random
uncertainty results from the averaging process used for processing the results. Torque is logged with
a sampling frequency of 10 Hz. These high frequency measurements suffer from large fluctuations
mainly caused by the dynamic movement of the liquids in the pipe. In order to smoothen results
measurements are averaged over a full rotation of the wheel, resulting in one data point per
rotation. Depending on the rotational speed of the wheel results where then averaged over at least
31 sample points (from the maximum speed at 2 m/s). Total measurement uncertainties for the
water-gas and oil-gas calibration experiments are shown for different velocities in Figure 6. Both
curves for water-gas and oil-gas are very similar. High peaks at velocity steps can be ascribed to
acceleration and deceleration of the liquids. Due to the short time of occurrence this phenomenon is
not further considered here. It can be seen, that the total uncertainty is governed by the systematic
uncertainty for velocities of 0.7 m/s and below. For higher velocities, the total uncertainty is
governed by the random uncertainty. Uncertainties calculated for the later presented 3-phase
experiments (not shown) resulted in very similar values. In general, absolute total uncertainties are
below 0.5 Nm for the measurements presented in this work. The uncertainty could be drastically
reduced by averaging over a larger number of samples. However, sampling over only a single rotation
of the wheel was chosen to catch the dynamics of the flow.
Figure 6: Total measurement uncertainties shown for the water-gas and oil-gas calibration experiments.
ACCEPTED MANUSCRIPT
3. Test matrix and fluids
M
AN
U
SC
RI
PT
Two different fluid systems were tested. The first fluid system represented a crude oil system with
brine and synthetic natural gas, consisting of methane, ethane, propane, CO2 and iso-butane. The
second system was a typical laboratory fluid system, consisting of a light mineral oil Exxsol D80, tap
water and Nitrogen. For both systems, tests at one high and one low water cut forming water and oil
continuous dispersions respectively were performed. The water cut describes the water fraction,
⁄
Fliquid, of the total liquid fraction, Fliquid, of the system,
=
. Other parameters that
were varied were the system pressure and temperature. A test matrix summarizing the tested
conditions and including the fluid properties is given in Table 1. Temperatures where controlled
during the experiments with fluctuations of the fluids typically below ±1°C. Two system pressures
were tested, namely high (approximately 50 bar) and low (close to atmospheric pressure). The wheel
was first filled with liquids at room temperature, followed by pressurization with the gas phase.
Three experimental runs at the three temperatures (50°C, 25°C and 10°C), were performed. Being a
closed system, pressure changed when temperatures were changed. This was not adjusted for.
Following the high pressure test the system was depressurized and the procedure repeated for the
three temperatures. For the high pressure experiments typical pressure variation was within 10 bar
when changing the temperature from 50°C to 10°C.
4. Results and discussion
TE
D
Table 1: Test matrix summarizing tested conditions and fluid properties.
4.1 General behaviour of the system
AC
C
EP
Figure 7 shows a segment of the velocity profile and corresponding torque profile for fluid system #1
with 75 vol% water cut at 25°C and high pressure. The results from this experiment were well suited
to explain the system behaviour when dispersion forms and separates as well as interesting aspects
of flow development.
From video recordings, it was observed that oil and water were separated up to a velocity of 0.5 m/s.
In Figure 7 the measured torque is indicated by a blue line. The red dashed line indicates the baseline
torque value produced at 0.5 m/s for fully separated flow. A velocity of 0.6 m/s marks the onset of
dispersion. A typical dispersion behaviour can be observed from the torque curves belonging to
mixing velocities up to 1.0 m/s. When ramping up the velocity from baseline to mixing velocity the
torque rises immediately to a higher value followed by a further gradual increase until a final value is
reached. The immediate rise can be ascribed to the velocity increase since the flow was still
separated. The dispersion process of the flow did occur rather slowly and was observed by the
increased torque with increasing amount of entrainment of one phase into the other. The change in
torque was a result of a change in proportion of the water and oil wetted wall area as one phase
losses its continuity. Furthermore, the effective viscosity of the dispersion will exceed the continuous
phase viscosity as function of dispersed phase fraction. Similarly, when the velocity was ramped
down to baseline velocity, the flow separated rather slowly. After a sudden drop, due to the lower
ACCEPTED MANUSCRIPT
velocity, the torque tended to decrease further with time until the baseline level was reached. The
development length seemed to increase as a consequence of increased velocity. This can be
explained by smaller droplet sizes produced at higher velocities needing longer time to separate.
M
AN
U
SC
RI
PT
For the mixing velocity of 1.5 m/s the behaviour was somewhat different. After a gradual increase,
the torque suddenly increased rapidly by approximately 50%. Furthermore, when returning to the
baseline velocity the torque profile showed a notable shoulder before dropping once more after
several minutes and only slowly returning to the baseline value. Similar behaviour was found for a
mixing velocity of 2 m/s. Such operating conditions are disadvantageous for both transport where
increased pressure drop will be the result, and a subsequent separation process where retention
times may become too high. Two possible explanations were put forward; the first was that mixing
intensity controlled phase inversion from water continuous to oil continuous flow could cause a
sudden increase in torque. A higher viscosity of oil compared to water now being the continuous
phase and higher dispersed phase fraction would explain increasing torque. The inversion point
previously obtained by other fluid characterization methods, however, was around 60% water cut,
which is below the 75% water cut present in the experiment. In the second possible explanation, the
long time span the fluid is exposed to a higher mixing intensity leads to creation of very small
droplets. A stable dispersion (emulsion) forms. Smaller droplet sizes would explain long separation
times and a high effective viscosity (Pal 1996). Instrumentation allowing for droplet size
characterization and identification of the continuity of the fluid system would be required to verify
such a behaviour. The resolution of the currently installed video camera did not allow for droplet size
analyses.
EP
TE
D
The last part of the experiment presented in Figure 7 shows a history behaviour of the flow. The
second period with mixing velocity 2 m/s (from approximately t = 650min) was ramped down to a
mixing velocity of 1 m/s. The measured torque in this case was more than twice that of the first
period ramping up from 0.5 m/s to 1 m/s (at approximately t = 370min). This can be explained by the
presence of the stable dispersion formed at the higher mixing energy at 2 m/s, as described above.
The torque profile did not indicate any recovery of the flow over the time period of one hour at 1
m/s indicating hysteresis of the flow.
AC
C
Figure 7: Tested velocity profile and measured torque profile for fluid system #1, WC = 0.75, T = 25°C and high pressure.
4.2 The effect of single system properties
The effect of both temperature and pressure on the dispersion behaviour was measured in order to
quantify changes in onset of dispersion, stabilities and viscosity effects. Figure 8 shows the results for
fluid system #1 at 75% water cut and high pressure for the three test temperatures.
Compared to tests at 25°C a higher mixing velocity of 0.9 m/s was required at 50°C to initiate
dispersion of the phases. Also the stability of the dispersion formed was reduced, as indicated by a
rather short development length after returning to the baseline velocity. Formation of a stable
dispersion leading to hysteresis, as observed when ramping down from 2 m/s to 1 m/s for 25°C, was
not observed for 50°C.
ACCEPTED MANUSCRIPT
In contrast, at 10°C the flow was already dispersed at 0.5 m/s and formation of a stable dispersion
was found at a mixing velocity of 0.9 m/s as indicated by the remarkably high torque values obtained
after ramping down to baseline velocity. The baseline period of 60 minutes was insufficient to reach
the reference torque.
RI
PT
In summary, reduction of temperature gave earlier onset of dispersion with respect to mixture
velocities. Furthermore, dispersions were more stable at lower temperatures compared to
dispersions produced at similar velocities at higher temperatures.
SC
Figure 8: Tested velocity profile and measured torque profile for fluid system #1, WC = 0.75, high pressure at three different
temperatures: T = 10°C, 25°C and 50°C. The torque values increase upon decreasing temperatures for all flow rates.
M
AN
U
For fluid system #1 the effect of system pressure was found to be similar to the temperature effect,
meaning that lower pressure gave earlier onset of and higher stability of the dispersion formed. This
is demonstrated in Figure 9 where results for the low and high pressure experiment with fluid system
#1 at WC = 0.3 and 50°C are compared. While dispersion onset was found at a mixing velocity of 0.6
m/s for the low pressure experiment, the high pressure experiment stayed separated up to 0.9 m/s.
Moreover, when both systems were dispersed, torque measurements were equal indicating that the
dispersion mechanisms did not govern the emulsion properties as such.
TE
D
Figure 9: Tested velocity profile and measured torque profile for fluid system #1, WC = 0.3, T = 50°C at low and high
pressure.
4.3 Sensitivity to the fluid system
EP
A second fluid system was tested to demonstrate how sensitive the presented characterization
method is when fluids with different dispersion characteristics are tested. However, a detailed
analysis of the two fluid systems and the effects causing the different dispersion behaviour will not
be studied in detail and rather be part of future work.
AC
C
Below, a comparison is given between fluid system #1 (crude oil based system), and fluid system #2
(Exxsol D80 mineral oil) which was similar to systems used elsewhere (Angeli 1996, Soleimani 1999,
Valle 2000, Elseth 2001, Kumara, Halvorsen et al. 2009, Amundsen 2011).
Results from tests with fluid system #2 are shown in Figure 10. The test conditions in terms of
pressure, temperature and water cut were the same as before for fluid system #1 (high pressure, P ≈
50 bar, T = 10°C, 25°C and 50°C, WC = 75%).
Variations in the measured torque profiles for three different temperatures were much smaller for
fluid system #2 than for system #1. As seen in Figure 10 the torque varied only by a few percent for
the model system while for the crude oil based system the difference was much more significant,
especially at the higher velocities where complete dispersion can be assumed. Furthermore, for
system #1, no formation of a stable dispersion, noticed by a distinct increase in the torque, was
found. The crude oil system contains wax and asphaltenes which may be responsible for both
increased and non-Newtonian viscosity behaviour and increased emulsion stability.
ACCEPTED MANUSCRIPT
SC
RI
PT
A difference in the general dispersion behaviour for the two systems was observed. Both systems
gradually started to entrain one of the phases once the mixing velocity exceeded the point of
dispersion onset. However, once dispersion onset was reached fluid system #1 always ended up as
fully dispersed while for fluid system #2 partly dispersed flow patterns were present at the end of
several of the mixing velocity periods. The reason for this may be contents of surfactant in the crude
oil system being a driving force for dispersion and stabilizing droplets once created. Similar to fluid
system #1 flow development times after returning to baseline velocity increased with decreasing
temperature and increasing mixing velocity. However, torque profiles indicated shorter development
times for fluid system #2. Fluid system #2 was much less stable than #1, and showed still increasing
torque measurements at the end of a cycle for some of the mixing cycles in Figure 10. In fact, partly
dispersed flow patterns as indicated by these results, also known as dual continuous flow, have
previously been reported for a wide range of mixing velocities in pipe flow experiments with similar
types of fluid system (Angeli 1996, Elseth 2001, Lovick and Angeli 2004).
M
AN
U
Figure 10: Tested velocity profile and measured torque profile for fluid system #2, WC = 0.75, high pressure at three different
velocities: T = 10°C, 25°C and 50°C.
In Figure 11 below the mixing velocities required for onset of dispersion as function of temperature
are summarized for the tested conditions. The three lower curves belong to fluid system #1, while
the upper curves belong to fluid system #2. Mixing velocities at onset of dispersion were
approximately twice as high for fluid system #2.
TE
D
Figure 11: The figure shows the mixing velocity required for onset of dispersion as function of temperature for fluid system
#1 and #2.
4.4 Relative increase in torque and effective viscosity
AC
C
EP
⁄
A torque ratio,
+1=
, is defined to compare torque for dispersed flow,
Mdispersed, and separated flow Mseparated at the same velocity. A ratio larger unity indicates a relative
increase in torque due to dispersion of the flow, Mrel. The torque ratio was determined at the
baseline velocity of 0.5 m/s. The torque values selected corresponded to measurements obtained
shortly after returning from mixing velocity to baseline velocity although long enough after not to be
influenced by the de-acceleration. Even for the highest mixing velocity of 2 m/s, acceleration and
deceleration took less than 20 seconds. Only for fluid system #2 at 25% water cut, forming very
unstable dispersions, this time was too long to keep the flow dispersed and these cases were not
evaluated further.
The torque ratios, as defined above, are plotted in Figure 12 with the mixing velocity for each case
given at the abscissa. For smaller mixing velocities, e.g. below 1m/s, results were more scattered.
Reasons for this could be incomplete dispersion at the end of a mixing cycle, instantaneous settling
of droplets or coalescence of droplets as part of the separation process in the pipe. At mixing
velocities larger than 1 m/s however a rather constant torque ratio of approximately 1.7, was found
for most of the experiments. A first approach was made to relate torque measurements to an
effective viscosity of the dispersion. The method used for computing the effective dispersion is
described in more detail in Johnsen, Førdedal et al. (2001). Using the results shown in Figure 12 as a
ACCEPTED MANUSCRIPT
starting point, relative viscosities in the range of mainly two to three times the continuous phase
viscosity were found. This is in good agreement with crude oil measurements by Johnsen and
Rønningsen (2003) considering comparable dispersed phase fractions. Similar results can also be
predicted by effective viscosity models. For example, the often used Pal and Rhodes (1989)
correlation predicts a relative viscosity of 2.23 for 25% dispersed phase fraction.
RI
PT
For fluid system #1, at 75% water cut, a sudden increase in the relative torque was found for the
highest mixing velocities, as shown in Figure 12 (left). These points belong to velocities at which also
the formation of stable dispersions was observed, as described earlier. Notice also that the velocity
at when this sudden increase occurred, increased with increasing temperature clearly stating the
temperature effect for crude oil systems, while this behaviour was not observed for the #2 mineral
oil system shown to the right in Figure 10 and neither for the system #1 at 30% water cut (middle)...
SC
Figure 12: The plots show the torque ratio against the mixing velocities at which the respective dispersions were created.
Torque measurements were taken directly after ramping down from the mixing velocity to the baseline velocity of 0.5 m/s.
M
AN
U
4.5 Scaling of results
Considering the different geometry of the wheel flow loop and considerably smaller pipe diameter
compared to a production pipeline, transferability of results to pipe flow conditions is an issue. One
parameter of importance is the energy dissipation rate which is a measure for the turbulence
intensity governing droplet production. Thus, energy dissipation was used as a scaling parameter for
prediction of dispersion formation in pipelines, based on results from the wheel experiments.
TE
D
In analogy with (Johnsen, Førdedal et al. 2001) the energy dissipation rate, Ɛ, for the wheel flow loop
is determined based on torque and thus friction factor measurements applying equations 1 to 3
below;
2M
LLπ dD
(1)
2τ
ρU 2
(2)
U3
ε =2f
d
(3)
AC
C
EP
τ=
f=
where τ is the wall shear stress, M is the measured torque, LL is the liquid column length, d is the
inner pipe diameter and D the diameter across the wheel. The fanning friction factor f was calculated
considering a mixture density ρ for dispersed flow and a liquid bulk velocity, U, equal to the
rotational velocity of the wheel. This required the assumptions of a homogeneous dispersion and no
liquid carry-over when the wheel rotated.
For pipe flow, equation 3 was used as before, while the friction factor was determined using
equation 4;
ACCEPTED MANUSCRIPT
f=
1 dP d
2 dx ρU 2
(4)
with the measured pressure gradient dP/dx.
EP
TE
D
M
AN
U
SC
RI
PT
Analogy of friction factors obtained in the wheel to friction factors in pipe flow was demonstrated by
Krampa, Ytrehus et al. (2005). It was therefore reasonable to assume that also shear induced droplet
formation occurs at similar conditions. Wheel experiments for fluid system #2 were compared with
pipe flow experiments using comparable fluid systems given in Table 2. As the wheel tests were
performed with 25% and 75% water cut, data from the pipe flow experiments with similar water cuts
were used for comparison. These data are presented in Figure 13 where energy dissipation rates are
plotted versus mixture velocity for the mentioned experimental campaigns. High water cuts are
presented as filled symbols, while low water cats are presented as empty symbols. As expected from
Equation 3, the mixture velocity required to obtain a certain energy dissipation rate decreases with
decreasing pipe diameter. For each data set a simplified division into three flow regions was
proposed. The first point of occurrence of a certain flow region was marked for each dataset. The
flow regions were defined as either separated flow, partly dispersed flow or dispersed flow.
Separated flow covers the flow patterns stratified smooth, stratified wavy and stratified wavy with
single drops according to the work by Angeli (1996) and Elseth (2001). Partly dispersed flow covers
the flow patterns stratified mixed, three-layer flow (with a distinct dispersion layer separating oil and
water) and dense packed layer flow. Dispersed flow covers inhomogeneous and homogeneous
dispersed flow (summarized as mixed flow by Angeli (1996)). Comparing the different datasets from
the pipe flow experiments with the current data from the wheel tests indicates that results from the
wheel experiments are comparable with pipe experiments in terms of flow pattern transition.
Moreover, this comparison also indicates that energy dissipation rate might be a suitable scaling
parameter for predicting flow transition in different pipe geometries. These results shows a strong
potential for using the wheel system to measure the rheology of non-Newtonian systems like
emulsions, waxy crude oils and systems containing particles like sand. Furthermore, the velocity and
torque data can be used to determine energy dissipation, which in turn is a relevant parameter for
up-scaling of results and prediction of straight pipe systems.
AC
C
Table 2: The table lists pipe material, pipe diameter and fluid system for the experimental data used for comparison. Test
conditions were low pressure and ambient temperature (≈20°C).
Figure 13: Energy dissipation rate versus mixture velocity. Onset of partly dispersed flow and transition to fully dispersed
flow are marked for each data set. Filled symbols present high water cuts. Empty symbols present low water cuts.
5. Conclusion
A wheel flow loop was used to test the dispersion behaviour of two different fluid systems, namely a
crude oil and a mineral oil based system. A velocity profile with a baseline velocity of 0.5 m/s and
mixing velocities up to 2 m/s was tested to determine onset of dispersion, dispersion development
and friction factor for water cuts of 25% (for Exxsol D80), 30% (for crude oil) and 75% (both systems).
The tests were performed at atmospheric pressure (nitrogen for Exxsol D80 and natural gas for the
ACCEPTED MANUSCRIPT
crude oil) and high pressure of approximately 50 bar. Temperatures of 10 °C, 25°C and 50°C were
tested. It was demonstrated that the onset of liquid-liquid dispersion formation and its stability could
be identified by means of measured torque supplemented by video recordings. Results verified that
lower temperatures lead to earlier onset of and more stable dispersion. Furthermore, higher
pressure went along with later onset of dispersion and lower dispersion stability, especially for the
crude oil system.
RI
PT
Typical differences between a realistic crude oil system and a common model fluid system were
demonstrated. For the model system, high mixing velocities were required to create dispersions,
which were rather unstable. In contrast, the crude oil system was dispersed at lower flow rates and
the dispersions were more stable. This shows the value of performing experiments with as realistic
fluid system as possible. Model fluid systems might let critical production challenges be overseen.
SC
Flow transitions found for the wheel experiments with the model fluid system agreed well with
experimental flow loop data when energy dissipation rate was used as scaling parameter. Upscaling
of results to predict flow behaviour in large diameter pipes seems possible.
M
AN
U
A method by Johnsen, Førdedal et al. (2001) was used to estimate effective viscosity based on torque
measurements. A comparison with a model by (Pal and Rhodes 1989) works well. However, more
work has to be done to fully understand the torque-viscosity relation in the wheel.
TE
D
The use of the wheel flow loop made it possible to study the evolution of a dispersion, both
formation and separation without the interference of pumps. The possibility of such long-term
observation of dispersion and emulsion behaviour with realistic temperature and pressure conditions
is one of the main advantages of this type of quasi-infinite pipe flow loops. Oppositely, restricted test
section lengths in traditional flow loops do only allow investigating of flow development for a few
tens of seconds or some minutes. Any long-term effects will be overseen.
EP
Even if different fluid systems were used, the presented dispersion characterization method gives
qualitatively similar results in terms of different flow development time scales and hysteresis effects
for different fluid systems when compared with a stirred tank approach by Patil, Arnesen et al.
(2017). However, the method based on droplet size characterization requires fully dispersed flow and
cannot be used to indicate dispersion onset from initially separated flow.
AC
C
A drawback of the current wheel flow loop is the insufficient measurement instrumentation required
for in-situ dispersion characterization. A future upgrade should include instrumentation indicating
the phase continuity and allow for in-situ droplet size characterization.
6. References
Amundsen, L. (2011). An experimental study of oil-water flow in horizontal and inclined pipes.
Doctoral Thesis, Norwegian University of Science and Technology.
Angeli, P. (1996). Liquid-liquid dispersed flows in horizontal pipes Ph.D. Thesis, Imperial College of
Science, Technology and Medicine.
Elseth, G. (2001). An Experimental Study of Oil / Water Flow in Horizontal Pipes Ph.D. Thesis,
Telemark University College.
Johansen, S. T., S. Mo, E. Meese, J. E. S. Oliveira, J. F. R. Reyes and J. N. E. Carneiro (2015). CFD
Simulations of Multiphase Flows Containing Large Scale Interfaces and Dispersed Phases with
Selected Production Technology Applications, Offshore Technology Conference.
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
Johnsen, E. E., H. Førdedal and O. Urdahl (2001). "A Simplified Experimental Approach for Measuring
Viscosity for Water-in-Crude-Oil Emulsions Under Flowing Conditions." Journal of Dispersion Science
and Technology 22(1): 33-39.
Johnsen, E. E. and H. P. Rønningsen (2003). "Viscosity of ‘live’ water-in-crude-oil emulsions:
experimental work and validation of correlations." Journal of Petroleum Science and Engineering
38(1–2): 23-36.
Krampa, F., J. D. Ytrehus and E. Straume (2005). Analogy between the SINTEF Wheel-Shaped Flow
Loop and Pipe. 5th International Conference on Multiphase Flow, Tampa, Florida.
Kumara, W. A. S., B. M. Halvorsen and M. C. Melaaen (2009). "Pressure drop, flow pattern and local
water volume fraction measurements of oil-water flow in pipes." Measurement Science and
Technology 20(11): 1-18.
Lovick, J. and P. Angeli (2004). "Experimental studies on the dual continuous flow pattern in oil-water
flows." International Journal of Multiphase Flow 30(2): 139-157.
Pal, R. (1996). "Effect of droplet size on the rheology of emulsions." AIChE Journal 42(11): 3181-3190.
Pal, R. and E. Rhodes (1989). "Viscosity/Concentration Relationships for Emulsions." Journal of
Rheology 33(7): 1021-1045.
Patil, A. V., K. Arnesen, A. Holte, T. Størseth, U. Farooq, A. Brunsvik, T. Skjetne and S. T. Johansen
(2017). Development of an advanced characterization technique for dynamic emulsion stability.
Tekna Conference.
Roca, J. F., J. N. E. Carneiro, J. E. S. Oliveira, S. Mo, M. Fossen and S. T. Johansen (2014). CFD
simulation of the two-phase flow of different mixtures in a closed system flow wheel. 10th
International Conference on CFD on Oil & Gas, Metallurgical and Process Industry 17-19th June 2014.
Trondheim, Norway.
Schümann, H., M. Tutkun, Z. Yang and O. J. Nydal (2016). "Experimental study of dispersed oil-water
flow in a horizontal pipe with enhanced inlet mixing, Part 1: Flow paterns, phase distributions and
pressure gradients." Journal of Petroleum Science and Engineering 145: 742-752.
Soleimani, A. (1999). Phase distribution and associated phenomena in oil-water flows in horizontal
tubes Ph.D. Thesis, Imperical College Lodon.
Urdahl, O., A. O. Fredheim and K.-P. Løken (1997). "Viscosity measurements of water-in-crude-oil
emulsions under flowing conditions: A theoretical and practical approach." Colloids and Surfaces A:
Physicochemical and Engineering Aspects 123–124: 623-634.
Urdahl, O., A. Lund, P. Mork and T. N. Nilsen (1995). "INHIBITION OF GAS HYDRATE FORMATION BY
MEANS OF CHEMICAL ADDITIVES .1. DEVELOPMENT OF AN EXPERIMENTAL SET-UP FOR
CHARACTERIZATION OF GAS HYDRATE INHIBITOR EFFICIENCY WITH RESPECT TO FLOW PROPERTIES
AND DEPOSITION." Chemical Engineering Science 50(5): 863-870.
Valle, A. (2000). Three Phase Gas-Oil-Water Pipe Flow Ph.D. Thesis, Imperial College of Science,
Technology and Medicine.
Yang, Z., J. Velthuis, J. Veltin and A. Twerda (2013). Cold restart of viscous multiphase flowline by hot
water flushing. 16th International Conference on Multiphase Production Technology. Cannes, BHR
Group.
ACCEPTED MANUSCRIPT
Water cut
Velocities
System pressure
AC
C
EP
TE
D
M
AN
U
SC
System temperature
Fluid system #1: synthetic natural gas, brine (5wt% NaCl), Crude Oil
(ρ=0.85kg/l, µ=7.5cP @ 20°C)
Fluid system #2: Nitrogen, Tap water, Exxsol D80 (ρ=0.8kg/l, µ=1.8cP @
25°C)
0.75 (Fluid system #1 and #2)
0.3 (Fluid system #1)
0.25 (Fluid system #2)
0.1m/s, 0.2m/s, 0.3m/s, 0.4m/s, 0.5m/s, 0.6m/s, 0.7m/s, 0.8m/s, .0.9m/s,
1.0m/s, 1.5m/s, 2m/s
Low pressure: 1barg
High pressure: ≈50barg
10°C, 25°C, 50°C
RI
PT
Fluid systems
ACCEPTED MANUSCRIPT
Pipe
Steel, d = 24.3mm
Steel, d = 56.3mm
Steel, d = 69mm
Fluid system
Exxsol D80, tap water
Exxsol D60, tap water
Exxsol D80, tap water
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
Data
Angeli (1996)
Elseth (2001)
Data owned by SINTEF
Petroleum Research
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
AC
C
velocity [m/s]
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
2
Fluid system #2, WC = 0.75, high pressure
velocity profile
1.5
1
0.5
0
0
100
200
300
400
500
600
700
time [min]
50C
25C
10C
torque [Nm]
6
4
1
2
0
0
100
200
300
400
time [min]
500
600
700
ACCEPTED MANUSCRIPT
Onset of dispersion
1.6
1.4
RI
PT
1
0.8
0.6
SC
0.4
0.2
0
10
20
30
40
50
60
Temperature [°C]
TE
D
FS1 - 30% WC, high pressure
FS2 - 75% WC, high pressure
FS2 - 25% WC, high pressure
EP
FS1 - 75% WC, high pressure
FS1 - 30% WC, low pressure
FS2 - 75% WC, low pressure
FS2 - 25% WC, low pressure
M
AN
U
0
AC
C
Velocity [m/s]
1.2
ACCEPTED MANUSCRIPT
2
1
0
2
1
0
0.5
1
1.5
Mixing velocity [m/s]
2
1
0
0
0.5
1
1.5
Mixing velocity [m/s]
FS1 - 30% WC, high P, 50C
FS1 - 30% WC, high P, 25C
FS1 - 30% WC, high P, 10C
FS1 - 30% WC, low P, 50C
FS1 - 30% WC, low P, 25C
FS1 - 30% WC, low P, 10C
2
0
0.5
1
1.5
Mixing velocity [m/s]
FS2 - 75% WC, high P, 50C
FS2 - 75% WC, high P, 25C
FS2 - 75% WC, high P, 10C
FS2 - 75% WC, low P, 50C
FS2 - 75% WC, low P, 25C
FS2 - 75% WC, low P, 10C
TE
D
M
AN
U
SC
FS1 - 75% WC, high P, 50C
FS1 - 75% WC, high P, 25C
FS1 - 75% WC, high P, 10C
EP
0
2
2
RI
PT
Torque ratio [-]
3
AC
C
Torque ratio [-]
4
3
Torque ratio [-]
3
5
ACCEPTED MANUSCRIPT
dispersed
dispersed
partly
partly
SC
partly
partly
RI
PT
dispersed
dispersed
1
0.1
M
AN
U
Energy dissipation rate [W/kg]
10
0.01
Wheel (52mm) - SINTEF
Pipe (69mm) - SINTEF
Pipe (56mm) - Elseth(2001)
Pipe (24mm) - Angeli(1996)
3
TE
D
2
EP
1
Velocity [m/s]
AC
C
0
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
Velocity profile
velocity [m/s]
2
1.5
1
0.5
0
0
200
400
600
time [min]
800
1000
ACCEPTED MANUSCRIPT
6
RI
PT
4
3
2
SC
1
0
1
1.5
2
2.5
Velocity [m/s]
Tap water
Empty wheel
TE
D
Exxsol D80
M
AN
U
0.5
EP
0
AC
C
Torque [Nm]
5
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
AC
C
velocity [m/s]
2
1.5
1
0.5
0
0
10
20
30
40
50
60
70
80
90
100
time [min]
Uncertainty [Nm]
2
water-gas
oil-gas
1.5
1
0.5
0
0
10
20
30
40
50
time [min]
60
70
80
90
100
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
velocity [m/s]
2
Fluid system #1, WC = 0.75, T = 25C, high pressure
1.5
1
0.5
torque [Nm]
0
6
0
100
200
300
0
100
200
300
400
500
600
70
400
500
600
70
4
2
0
time [min]
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
Fluid system #1, WC = 0.75, high pressure
EP
velocity profile
1.5
1
0.5
0
0
10
100
200
300
400
500
60
time [min]
50C
25C
10C
8
torque [Nm]
AC
C
velocity [m/s]
2
6
4
2
25
50C
0
0
100
200
300
time [min]
400
500
60
AC
C
velocity [m/s]
EP
TE
D
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
2
Fluid system #1, WC = 0.3, T = 50C
velocity profile
1.5
1
0.5
0
0
50
100
150
200
250
300
350
400
300
350
400
time [min]
torque [Nm]
2
high pressure
low pressure
1.5
low pressure
1
0.5
high pressure
0
0
50
100
150
200
250
time [min]
ACCEPTED MANUSCRIPT
A wheel flow loop was used to study oil-water dispersion characteristics.
-
Dispersion formation, stability and long term flow development were investigated.
-
The effect of temperature, pressure and water cut was studied.
-
Important differences between a crude oil and a model oil system were found.
-
Energy dissipation rate was found to be an appropriate geometry scaling parameter.
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
-